Interventions targeted at women to encourage the uptake of cervical screening

Abstract

Background

This is an update of the Cochrane review published in Issue 5, 2011.

Worldwide, cervical cancer is the fourth commonest cancer affecting women. High‐risk human papillomavirus (HPV) infection is causative in 99.7% of cases. Other risk factors include smoking, multiple sexual partners, the presence of other sexually transmitted diseases and immunosuppression. Primary prevention strategies for cervical cancer focus on reducing HPV infection via vaccination and data suggest that this has the potential to prevent nearly 90% of cases in those vaccinated prior to HPV exposure. However, not all countries can afford vaccination programmes and, worryingly, uptake in many countries has been extremely poor. Secondary prevention, through screening programmes, will remain critical to reducing cervical cancer, especially in unvaccinated women or those vaccinated later in adolescence. This includes screening for the detection of pre‐cancerous cells, as well as high‐risk HPV.

In the UK, since the introduction of the Cervical Screening Programme in 1988, the associated mortality rate from cervical cancer has fallen. However, worldwide, there is great variation between countries in both coverage and uptake of screening. In some countries, national screening programmes are available whereas in others, screening is provided on an opportunistic basis. Additionally, there are differences within countries in uptake dependent on ethnic origin, age, education and socioeconomic status. Thus, understanding and incorporating these factors in screening programmes can increase the uptake of screening. This, together with vaccination, can lead to cervical cancer becoming a rare disease.

Objectives

To assess the effectiveness of interventions aimed at women, to increase the uptake, including informed uptake, of cervical screening.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL), Issue 6, 2020. MEDLINE, Embase and LILACS databases up to June 2020. We also searched registers of clinical trials, abstracts of scientific meetings, reference lists of included studies and contacted experts in the field.

Selection criteria

Randomised controlled trials (RCTs) of interventions to increase uptake/informed uptake of cervical screening.

Data collection and analysis

Two review authors independently extracted data and assessed risk of bias. Where possible, the data were synthesised in a meta‐analysis using standard Cochrane methodology.

Main results

Comprehensive literature searches identified 2597 records; of these, 70 met our inclusion criteria, of which 69 trials (257,899 participants) were entered into a meta‐analysis. The studies assessed the effectiveness of invitational and educational interventions, lay health worker involvement, counselling and risk factor assessment. Clinical and statistical heterogeneity between trials limited statistical pooling of data.

Overall, there was moderate‐certainty evidence to suggest that invitations appear to be an effective method of increasing uptake compared to control (risk ratio (RR) 1.71, 95% confidence interval (CI) 1.49 to 1.96; 141,391 participants; 24 studies). Additional analyses, ranging from low to moderate‐certainty evidence, suggested that invitations that were personalised, i.e. personal invitation, GP invitation letter or letter with a fixed appointment, appeared to be more successful. More specifically, there was very low‐certainty evidence to support the use of GP invitation letters as compared to other authority sources' invitation letters within two RCTs, one RCT assessing 86 participants (RR 1.69 95% CI 0.75 to 3.82) and another, showing a modest benefit, included over 4000 participants (RR 1.13, 95 % CI 1.05 to 1.21). Low‐certainty evidence favoured personalised invitations (telephone call, face‐to‐face or targeted letters) as compared to standard invitation letters (RR 1.32, 95 % CI 1.11 to 1.21; 27,663 participants; 5 studies). There was moderate‐certainty evidence to support a letter with a fixed appointment to attend, as compared to a letter with an open invitation to make an appointment (RR 1.61, 95 % CI 1.48 to 1.75; 5742 participants; 5 studies).

Low‐certainty evidence supported the use of educational materials (RR 1.35, 95% CI 1.18 to 1.54; 63,415 participants; 13 studies) and lay health worker involvement (RR 2.30, 95% CI 1.44 to 3.65; 4330 participants; 11 studies). Other less widely reported interventions included counselling, risk factor assessment, access to a health promotion nurse, photo comic book, intensive recruitment and message framing. It was difficult to deduce any meaningful conclusions from these interventions due to sparse data and low‐certainty evidence. However, having access to a health promotion nurse and attempts at intensive recruitment may have increased uptake.

One trial reported an economic outcome and randomised 3124 participants within a national screening programme to either receive the standard screening invitation, which would incur a fee, or an invitation offering screening free of charge. No difference in the uptake at 90 days was found (574/1562 intervention versus 612/1562 control, (RR 0.94, 95% CI: 0.86 to 1.03).

The use of HPV self‐testing as an alternative to conventional screening may also be effective at increasing uptake and this will be covered in a subsequent review. Secondary outcomes, including cost data, were incompletely documented. The majority of cluster‐RCTs did not account for clustering or adequately report the number of clusters in the trial in order to estimate the design effect, so we did not selectively adjust the trials. It is unlikely that reporting of these trials would impact the overall conclusions and robustness of the results. Of the meta‐analyses that could be performed, there was considerable statistical heterogeneity, and this should be borne in mind when interpreting these findings. Given this and the low to moderate evidence, further research may change these findings. The risk of bias in the majority of trials was unclear, and a number of trials suffered from methodological problems and inadequate reporting. We downgraded the certainty of evidence because of an unclear or high risk of bias with regards to allocation concealment, blinding, incomplete outcome data and other biases.

Authors' conclusions

There is moderate‐certainty evidence to support the use of invitation letters to increase the uptake of cervical screening. Low‐certainty evidence showed lay health worker involvement amongst ethnic minority populations may increase screening coverage, and there was also support for educational interventions, but it is unclear what format is most effective. The majority of the studies were from developed countries and so the relevance of low‐ and middle‐income countries (LMICs), is unclear. Overall, the low‐certainty evidence that was identified makes it difficult to infer as to which interventions were best, with exception of invitational interventions, where there appeared to be more reliable evidence.

Plain language summary

Do invitations, lay health worker interventions and educational interventions increase the uptake of cervical screening?

The issue
Cervical cancer is the fourth most common cancer in women worldwide. At present, women are asked to attend cervical screening (also known as a 'smear' or 'Pap test') to detect the presence of high‐risk HPV and/or abnormal or pre‐cancerous cells. The uptake of cervical screening is low globally. The UK's Cervical Screening Programme has shown that screening can reduce mortality through early detection and treatment of pre‐cancerous changes before cancer develops. However, there is variation between and within countries in the availability and uptake of screening. There are also differences based on ethnic groups, age, education and socioeconomic status and this needs to be borne in mind when developing interventions to increase uptake.

The aim of the review
The aim of this review was to look at the methods used to encourage women to undergo cervical screening. These included invitations, reminders, education, message framing, counselling, risk factor assessment, procedures and economic interventions.

What are the main findings?
Seventy trials were included in this review, of which 69 trials (257,899 women) were entered into a meta‐analysis. Invitations, and to a lesser extent, educational materials probably increase the uptake of cervical screening (moderate‐certainty evidence). HPV self‐testing, as an alternative to Pap smears, may also increase screening coverage. However, self‐testing was not covered in this review and will be considered in a subsequent review. Lay health workers used to promote screening to ethnic minority groups may increase screening uptake (low‐certainty evidence).

It was difficult to deduce any meaningful conclusions for other less widely reported interventions such as counselling, risk factor assessment, access to health promotion nurse, photo comic book, intensive recruitment and message framing, due to sparse data and low‐certainty evidence. However, having access to a health promotion nurse and attempts at intensive recruitment may increase uptake.

Certainty of the evidence

The majority of the evidence was of a low to moderate certainty (quality) and further research may change these findings. For the majority of trials, the risk of bias was unclear, making it difficult to make firm assertions from their results.

What are the conclusions?
Invitation letters probably increase the uptake of cervical screening, and use of lay health worker involvement amongst ethnic minority populations may do so. Educational interventions may also increase screening; however, it is unclear what format is the most effective. These findings apply to developed countries and their relevance to low‐ and middle‐income countries is unclear.

Background

This is an update of the Cochrane review published in Issue 5, 2011.

Description of the condition

Cervical cancer is the fourth most common cancer among women (GLOBOCAN 2018). A woman's risk of developing cervical cancer by age 65 years ranges from 0.8% in developed countries to 1.5% in low‐ and middle‐income countries (LMICs). The management of cervical cancer varies depending on stage, healthcare resources and policy. If found at an early stage, normally through screening, localised excision of abnormal cells within the cervix using diathermy, laser or knife cone excision may be all that is needed. However, for anything more than the earliest stages, treatment most commonly involves either radical surgery, requiring hysterectomy, or chemoradiotherapy. Survival rates are stage dependent. In Europe and the USA, the 5‐year survival rate for all stages combined is between 60% and 72% (Cancer Statistics 2019; Jemal 2008) and, in England and Wales between 2010 and 2014, the 5‐year survival rate overall was 66.8% (ONS 2015).

Primary and secondary prevention

Human papillomavirus (HPV) infection is believed to be the primary cause of cancer of the cervix, with studies estimating the worldwide HPV prevalence in cervical cancers to be 99.7% (Eurogin 2016; Walboomers 1999). In particular, two subtypes of HPV (16 and 18) are present in over 80% of invasive cervical cancers. Other known risk factors for cervical cancer include smoking (Bosch 2011; Brinton 1986), the early onset of sexual activity, multiple sexual partners, the presence of other sexually transmitted diseases (STDs) (La Vecchia 1986; Mitra 2016) and the immunological status of the woman (Schneider 1983; Stanley 2015). Individuals who receive immunosuppressive therapy for organ transplants and those infected with human immunodeficiency virus (HIV) are therefore particularly at risk of developing pre‐invasive disease. Primary strategies to prevent the development of cervical cancer focus on reducing these known risk factors by HPV vaccination, encouraging a healthy lifestyle, smoking cessation and the adoption of 'safer' sexual behaviours aimed at reducing the risk of HPV infection (Shepherd 2011).

Understanding the role of HPV in cervical cancer has led to the development of prophylactic HPV vaccination and immunisation programmes are available in many countries. The WHO has also recommended that HPV vaccination become part of the routine schedule of vaccination in girls between the ages of nine and 13 years. Vaccines currently available are effective against two, four and nine different HPV subtypes and all three subtypes of vaccines protect against HPV 16 and 18. Currently, vaccinated women are still asked to continue with secondary prevention, as data demonstrating efficacy of the vaccine are only now being reported. Current reports suggest that, in those populations where vaccine uptake is over 80%, the effect on cervical cancer and pre‐cancers is substantial, with almost 90% reduction in cervical cancer rates in those vaccinated before the age of 17 years (Cruickshanks 2019; Lei 2020). However, such dramatic results are solely dependant on vaccine uptake and, with the rise of the 'anti‐vax' lobby, together with, at times, unbalanced reporting of vaccine safety fears, uptake in countries, such as the USA and Japan has been severely reduced. In Japan, uptake of the vaccine has fallen to less than 1% (Hanley 2015). Thus, until vaccine programmes have a much higher uptake and global reach, secondary prevention methods will need to be the mainstay of efforts to reduce cervical cancer. These secondary methods involve screening for the detection of high‐risk HPV DNA and abnormal (pre‐cancerous) cell changes (i.e. any changes which may precede, be associated with or carry a significant risk of developing cancer).

Description of the intervention

Screening

The Papanicolau, or Pap smear, cervical screening test is used worldwide and is primarily aimed at detecting pre‐cancerous changes within the cervix (i.e. abnormalities in the cells of the cervix known as dysplasia) before they have an opportunity to progress to invasive carcinoma. More than 90% of cervical cancers develop within a small area of the cervix known as the transformation zone. Disease progression from dysplasia to invasive cancer is usually slow, therefore, providing an opportunity to detect and treat pre‐cancerous disease. During a smear test, cells within the external and internal layers of the transformation zone (i.e. ecto‐ and endo‐cervical cells) are collected and subsequently examined for abnormal cytological changes. The reliability of the technique is dependent both on the expertise of the health professional taking the smear and the individual examining the smear. Even in the best laboratories, 5% to 15% of abnormal smears may be reported as normal (Nottingham 1998). More recently, the use of liquid based cytology (LBC) has reduced the number of inadequate smears and subsequent need for recalls (Moss 2004; NICE 2003).

Since the cervical screening programme in 1988, the associated mortality rate in females under 35 years in the United Kingdom has fallen (Peto 2004) and will continue to do so as the screening programme improves and the vaccinated population increases (Sasieni 2017). Worldwide, great variation exists between countries in terms of the coverage and uptake of cervical screening. In a number of countries such as Finland, Australia, Sweden and Spain, national cervical screening programmes exist, similar to the UK. Such screening programmes are usually aimed at those women most at risk of developing cervical cancer (i.e. usually women aged between 20 and 65 years). Recommendations vary between countries, but women are usually screened every one to five years. In many other countries, Pap smear services are provided on a much more local and opportunistic basis, if at all.

In recent years, conventional screening has undergone a conceptual change in many countries. Instead of first performing cytology examination on a cervical screening sample, the sample (still most commonly obtained by a healthcare professional during a vaginal examination to visualise the cervix) is tested for the presence of HPV DNA. This stems from the fact that 99.7% of cervical cancer is caused by HPV and, if a woman is uninfected, her risk of cervical cancer is negligible. Primary HPV screening has been shown to have a higher sensitivity to high‐grade lesions (Koliopoulos 2017), but at the expense of specificity. Thus, implementation of HPV screening will require additional triage tests together with a probable rise in colposcopy referrals (Rijkaart 2012). These additional resource burdens may make HPV screening only possible in high‐income countries while in LMIC, Pap smears may be a better approach. A recent cost‐benefit analysis demonstrated that three‐yearly cytology testing has a better cost‐benefit profile together with no loss in quality of life in comparison to primary HPV screening (Sawaya 2019).

Cervical screening (regardless of whether by Pap smear or LBC or HPV screening) uptake and coverage not only vary between countries, but differences also exist within countries between different sociodemographic groups, according to factors including ethnic origin, age, education and socioeconomic status. Lower uptake rates have been found to occur in those women who are older, less well‐educated, from lower socioeconomic groups or those who reside in rural locations (Tangka 2015). Certain ethnic groups have also been identified as having lower rates of cervical screening uptake, such as African‐American, Hispanic and Native American in the USA and Asian women in the UK (Moser 2009; Tangka 2015). Consequently, interventions have been aimed at trying to increase screening amongst these groups of women. Thus, there are a number of factors to consider when developing interventions to increase the uptake of Pap smear screening. These factors are likely to differ between developing and developed countries and between individual populations in a country.

Encouraging the uptake of screening

One of the major obstacles to the success of cervical screening worldwide is the uptake of the programme by women. Understanding the various reasons for women never attending for a smear test, or failing to continue in further rounds of screening, are difficult to assess. Much work across the world has been undertaken to determine contributing factors, such as cost, anxiety, embarrassment, and fear of cancer. Women from ethnic minorities and deprived subgroups in the population have shown consistently lower uptake over decades of screening in countries worldwide (Moser 2009; Webb 2004). This may be attributable to health literacy, especially since screening literature can include complex concepts. HPV is transmitted sexually and therefore screening and cervical cancer can be perceived as a consequence of promiscuity and thus have negative connotations. Encouragingly though, in a survey of Muslim Turkish women who mostly consider talking about sex as a taboo, the majority of women felt the recommendations from health workers was the major influence in attending screening and accepting the HPV vaccination for their daughters (Ilter 2010). Given the complex nature of the factors involved, a number of interventions have therefore been based on theoretical models of health behaviour, such as the Health Belief Model (Kreuter 1996; Marcus 1992) and the Transtheoretical Model (Rimer 1999). It is important to acknowledge that, due to differences between populations, interventions that are effective in one setting may not be as effective in another.

In the UK, websites, such as those provided by the National Health Service (NHS) Cervical Screening Programme and Jo's Cervical Cancer Trust, can go some way in trying to break down barriers to screening. These websites provide written, audio and visual resources aimed at answering common concerns, explaining the procedure and explanation of results. Key documents are provided in a translated format covering many languages spoken by the larger minority groups in the UK.

How the intervention might work

Informed consent

The main focus of attention of cervical screening programmes is to increase the uptake of cervical screening. However, this must be done in the context of informed consent and understanding of the screening tests. It is recognised that both informed uptake and consent is important, since screening can cause harm, with inevitable false negatives leading to women being wrongly reassured and false positives resulting in unnecessary anxiety and further investigations and possibly even treatment. In particular, informed uptake needs to be considered, especially when topical media coverage, exemplified by the cervical cancer sufferer Jade Goody in the UK, can result in such an increase in women attending screening whether required or not (an increase of 3.6 million women screened in 2008/09 compared to 3.2 million in the previous year) yet the numbers soon fall when the media interest settles (3.3. million women screened in 2009/10) (NHS Information Centre 2014).

Why it is important to do this review

The incidence of cervical cancer is reduced by 93.5%, 92.5%, 90.8%, 83.6% and 64.1% if women have screening every year, every 2 years, every 3 years, every five years and every 10 years, respectively; these screening intervals would mean women having 50, 25, 16, 10 and 5 smear tests, respectively, in their lifetime (IARC 1986). Through modelling analyses, it has been shown that extending the re‐screening interval from one year to every three years results in an average excess risk of about 3 per 100,000 (Sawaya 2003). In the UK, women aged 25 to 50 years are invited for screening every three years and those aged 50 to 64 every five years. Each year, around 3.5 million women accept screening (NHS Information Centre 2014) and this has been estimated to prevent up to 3900 cases of cervical cancer and save over 4500 lives annually in the UK (Peto 2004; Sasieni 1996; Sasieni 2017). However, despite its effectiveness, the uptake rate of cervical screening by eligible women remains stubbornly below 75% (NHS Information Centre 2014). Information is needed to establish what can be done to increase this uptake rate, particularly in the 20% of women who are missing out on screening.

Objectives

To assess the effectiveness of interventions aimed at women, to increase the uptake, including informed uptake, of cervical screening.

Methods

Criteria for considering studies for this review

Types of studies

This update of this review used the same methodology as the original and previously updated versions of this review (Everett 2011). Randomised controlled trials (RCTs) and cluster‐RCTs of universal, selective or opportunistic cervical screening were included. We excluded quasi‐randomised trials and other non‐RCTs in an attempt to minimise bias.

Types of participants

All women eligible to participate in a cervical screening programme as defined by the entry criteria for that programme. Women due or overdue were all considered for inclusion.

Types of interventions

All interventions targeted at women who are eligible for screening. Interventions aimed at communities, such as mass media campaigns (Grilli 2002), and those aimed at health professionals were excluded as they are considered in other Cochrane reviews. Interventions targeted at health professionals that are covered in other Cochrane reviews include: audit and feedback (Jamtvedt 2006); educational outreach visits (O'Brien 1997); printed educational materials (Freemantle 1997); computer‐generated paper reminders (Arditi 2010); manual paper reminders (Romero 2004); on‐screen computer reminders (Gordon 1998); and other interventions (Hulscher 2006).

For the subgroup analyses, the interventions were categorised as follows (Jepson 2000):

• Invitations

  • Invitations to women due for screening (either first round or second round). Did not include women who were overdue for screening. Included fixed or open appointments, letters, telephone calls, verbal recommendations, prompts and follow‐up letters.

• Reminders

  • Reminders to women who were overdue for screening and had not responded to the first round of screening. Included fixed or open appointments, letters, telephone calls, verbal recommendations, prompts and follow‐up letters.

• Education

  • Educational interventions aiming to increase knowledge of the screening programme or the disease being screened for, that did not contain a counselling component. Included printed educational materials, audiovisual materials, group and individual teaching and home visits.

• Message framing

  • Messages about screening (either verbal or written) that were framed either positively or negatively.

• Counselling

  • Counselling either face‐to‐face or on the telephone. Must have involved a discussion of barriers to screening as well as an educational component.

• Risk factor assessment

  • Risk factor questionnaires and computer programmes assessing a person's risk status.

• Procedures

  • Interventions to increase screening uptake by making the screening procedure easier or more acceptable to individuals undergoing screening. Included different screening tests for the same disease, or length of time that screening test took, and opportunistic testing and notification of results. HPV self‐testing was a procedure identified that may increase uptake by making screening more acceptable. Although, meeting the inclusion criteria for this review, we excluded this intervention from meta‐analyses and discussion as it will be examined in a separate Cochrane review.

• Economic

  • Removal of financial barriers or economic incentives. Included reduced cost or free screening tests, transport costs, free postage for returning tests and 'rewards' for completion of a screening test.

Controls

• Control groups were those with no intervention or no intervention other than that routinely undertaken by the local screening programme.

Types of outcome measures

Primary outcomes

• Uptake or non‐uptake of cervical screening, as recorded by health service records (such as screening administration system, hospital or primary care physician records)

• Uptake or non‐uptake of cervical screening as collected via self‐report (i.e. directly reported by the participant in a telephone interview or questionnaire)

Secondary outcomes

The following intermediate and other outcomes were considered, if reported:

• Booking of appointments;

• Reported intentions to attend screening;

• Attitudes to screening;

• Knowledge of screening;

• Satisfaction with screening service;

• Costs of the interventions.

We presented summary of findings tables reporting the following outcomes:

1. Uptake or non‐uptake of cervical screening, as recorded by health service records

2. Uptake or non‐uptake of cervical screening, as collected via self‐report

Search methods for identification of studies

Papers in all languages were sought and translations carried out, when necessary.

Electronic searches

See: Cochrane Gynaecological Cancer Group methods used in reviews.

For this update, the following electronic databases were searched on 8 June, 2020:

• The Cochrane Central Register of Controlled Trials (CENTRAL; Issue 6, 2020) in the Cochrane Library;

• MEDLINE via Ovid (1946 to June week 1 2020);

• Embase via Ovid (1980 to 2020 week 23);

The CENTRAL, MEDLINE and Embase search strategies used to identify RCTs comparing interventions targeted at women to encourage the uptake of cervical screening, are presented in Appendix 1, Appendix 2, and Appendix 3, respectively. All relevant articles found were identified on PubMed and, using the 'related articles' feature, a further search was carried out for newly published articles.

Searching other resources

Unpublished and grey literature

Metaregister, Physicians Data Query (PDQ), www.isrctn.com/rct, www.clinicaltrials.gov and www.cancer.gov/about-cancer/treatment/clinical-trials were searched for ongoing trials. The main investigators of the relevant ongoing trials were contacted for further information, as were the major co‐operative trials groups active in this area.

Published and unpublished studies were included if they met the inclusion criteria for the review.

Reference lists and correspondence

The citation lists of included trials were checked and experts in the field contacted to identify further reports of trials.

Data collection and analysis

Selection of studies

All titles and abstracts retrieved by electronic searching were downloaded to the reference management database (Endnote), duplicates were then removed and the remaining references examined by two review authors (a combination of HS, AS, NS, KG) independently. Those studies which clearly did not meet the inclusion criteria were excluded and copies of the full text of potentially relevant references were obtained. The eligibility of retrieved papers was assessed independently by all review authors. Disagreements were resolved by discussion between the review authors.

Data extraction and management

For included trials, the following data were abstracted:

• Author, year of publication and journal citation (including language)

• Country

• Setting

• Inclusion and exclusion criteria

• Study design, methodology

• Study population

  • Total number enrolled

  • Patient characteristics

  • Age

• Total number of intervention groups

• Intervention details

  • Type of intervention

  • Description of intervention

  • Frequency and duration of intervention

  • Type of healthcare professional who provided the intervention

• Control details

  • Any other reported information other than no active intervention was given

• Risk of bias in study (see below)

• Duration of follow‐up

• Outcomes – Uptake or non‐uptake of cervical screening, booking of appointments, reported intentions to attend screening, attitudes to screening, knowledge of screening, satisfaction with screening service, costs of the interventions. For each outcome:

• • Outcome definition;

  • Unit of measurement (if relevant);

  • For scales: upper and lower limits, and whether high or low score is good;

  • Results: number of participants allocated to each intervention group;

  • For each outcome of interest: sample size; missing participants.

Data on outcomes were extracted as below:

• For dichotomous outcomes (e.g. uptake or non‐uptake), we extracted the number of women in each treatment arm who underwent screening for cervical cancer and the number of women assessed at endpoint, in order to estimate a risk ratio.

Where possible, all data extracted were those relevant to an intention‐to‐treat analysis, in which participants were analysed in groups to which they were assigned.

The time points at which outcomes were collected and reported was noted.

Data were abstracted independently by all authors onto a data abstraction form specially designed for the review. Differences between reviewers were resolved by discussion.

Assessment of risk of bias in included studies

The risk of bias in included RCTs was assessed using the Cochrane Collaboration's tool. This included assessment of:

• sequence generation

• allocation concealment

• blinding

• incomplete outcome data: we recorded the proportion of participants whose outcomes were not reported at the end of the study and noted whether loss to follow‐up was not reported. We coded a satisfactory level of loss to follow‐up for each outcome as:

  • Yes, if fewer than 20% of patients were lost to follow‐up and reasons for loss to follow‐up were similar in both treatment arms

  • No, if more than 20% of patients were lost to follow‐up or reasons for loss to follow‐up differed between treatment arms

  • Unclear if loss to follow‐up was not reported

• selective reporting of outcomes

• other possible sources of bias

The risk of bias tool was applied independently by two review authors (HS and KG) to each study and differences resolved by discussion. Results are presented in both a risk of bias graph and a risk of bias summary (See Figure 1; Figure 2). Results of meta‐analyses were interpreted in light of the findings with respect to risk of bias.

1.

Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.

2.

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

We also used the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) (GRADE Working Group 2004) approach to assess the quality of evidence provided by the included studies and presented this in Table 1; Table 2; Table 3.

Summary of findings 1. Summary of findings.

Invitation compared with control or alternative invitation for uptake of cervical screening
Patient or population: women for cervical screening Settings: any Intervention: invitation Comparison: alternative invitation or control
Outcomes	Illustrative comparative risks* (of uptake) (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Corresponding risk (uptake)	Assumed risk (uptake)
	Invitation	Other
Invitation versus control: uptake of screening	139 per 1000 (121 to 159)	81 per 1000	RR 1.71 (1.49 to 1.96)	141,391 (24 RCTs)	⊕⊕⊕⊝ moderate	Downgraded to moderate quality. Although studies were randomised controlled trials with large numbers and did include studies of adequate selection bias and blinding, there were still some trials where the method of randomisation was unclear or had high attrition bias.
GP invitation letter versus invitation letter from other authority sources: uptake of screening	Trial results reported separately due to heterogeneity in comparisons; 1: GP invitation letter vs health clinic invitation letter; 2: GP invitation letter vs letter from programme coordinator		1: RR 1.69 (0.75 to 3.82) & 2: RR 1.13 (1.05 to 1.21)	86 & 4028 (2 RCTs)	⊕⊝⊝⊝ very low	Only two trials, which were reported separately. Both had concerns regarding risk of bias, hence quality downgraded. One trial had low numbers.
Personal invitation versus invitation letter: uptake of screening	142 per 1000 (120 to 168)	108 per 1000	RR 1.32 (1.11 to 1.56)	27,663 (5 RCTs)	⊕⊕⊝⊝ low	Randomised controlled trials with a large number of participants
Letter with fixed appointment versus letter with open invitation to make appointment: uptake of screening	364 per 1000 (335 to 396)	226 per 1000	RR 1.61 (1.48 to 1.75)	5742 (5 RCTs)	⊕⊕⊕⊝ moderate	Majority of trials had adequate randomisation and a low risk of attrition bias.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk Ratio; RCT: Randomised controlled trial
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

Invitation vs. Control:

Eleven trials had adequate allocation of study participants (Abdullah 2013a; Broberg 2013; Buehler 1997; Burack 1998; Haguenoer 2015; Hunt 1998; Mullins 2009; Racey 2016; Radde 2016; Rashid 2013; Stein 2005). The method of randomisation was unclear in the remaining trials. Only five trials reported the outcome assessor as being blinded to treatment allocation (Bowman 1995; Del Mar 1998; Hunt 1998; Peitzmeier 2016; Rashid 2013). This was not clear in the remaining trials. Loss to follow‐up was low in 15 of the trials (Abdullah 2013a; Acera 2017; Binstock 1997; Buehler 1997; Burack 1998; Del Mar 1998; Haguenoer 2015; Heranney 2011; Lancaster 1992; McDowell 1989; Mullins 2009; Peitzmeier 2016; Pierce 1989; Racey 2016; Somkin 1997); whereas attrition bias was high in seven trials (Bowman 1995; Broberg 2013; Burack 1998; Radde 2016; Rashid 2013; Stein 2005; Vogt 2003). In the trial of Racey 2016, there was a change in the study protocol and some participants in the Pap test group received a reminder phone call which may have influenced the overall outcome.

Personal invitation vs. invitation letter:

One trial (Jensen 2009) failed to fulfil any of the risk of bias criteria adequately. Only one trial (Hunt 1998) used an adequate method of randomisation and reported on the outcome assessor being blinded.

Letter with fixed appointment vs. letter with open invitation to make appointment:

Three trials used an adequate method of randomisation (Lonnberg 2016; Pritchard 1995; Segnan 1998). Only one trial (Bowman 1995) reported adequate blinding of outcome assessor. Only one trial (Bowman 1995) had a high risk of attrition bias; the remainder had low numbers lost to follow‐up (Lonnberg 2016; Pritchard 1995; Segnan 1998; Wilson 1987).

Summary of findings 2. Summary of findings.

Counselling compared with other interventions for uptake of cervical screening
Patient or population: women for cervical screening Settings: any Intervention: counselling Comparison: other
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Corresponding risk (uptake)	Assumed risk (uptake)
	Counselling	Other
Counselling versus control: uptake of screening	656 per 1000 (554 to 773)	553 per 1000	RR 1.23 (1.04 to 1.45)	393 (2 RCTs)	⊕⊝⊝⊝ very low	Downgraded for risk of bias and small trials. Unclear risk of bias for both trials. Two trials assessed different type of counselling (face‐to‐face and telephone), therefore heterogeneity was present. Relative results not presented in abstract and merely listed as to not mislead as data were deemed too sparse to make any meaningful inferences
Counselling versus other: uptake of screening	638 per 1000 (508 to 796)	564 per 1000	RR 1.13 (0.90 to 1.41)	208 (1 RCT)	⊕⊝⊝⊝ very low	Only one trial with small numbers. Mainly at unclear risk of bias with high attrition bias. Relative results not presented in abstract and merely listed as to not mislead as data were deemed too sparse to make any meaningful inferences
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk Ratio; RCT: Randomised controlled trial
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

Summary of findings 3. Summary of findings.

Education compared with control or other interventions for uptake of cervical screening
Patient or population: women for cervical screening Settings: any Intervention: education Comparison: other
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Corresponding risk (uptake)	Assumed risk (uptake)
	Education	Other
Education versus control: uptake of screening	211 per 1000 (184 to 240)	156 per 1000	RR 1.35 (1.18 to 1.54)	63,415 (13 RCTs)	⊕⊕⊝⊝ low	Large number of randomised controlled trials with varying participant numbers. Likely to be some heterogeneity between trials as comparison included different types of education (printed, face‐to‐face/verbal, other) vs control. Lay health outreach worker vs control also assessed 4330 participants across 11 RCTs in separate analysis (RR 2.30, 95% CI 1.44 to 3.65).
Education versus other: uptake of screening	Incompletely and inconsistently reported. Comparisons were narratively reported across seven trials and included lay health outreach workers and media education vs media education alone (two small trials), educational material vs other (one small multi‐arm trial, two small trials and one trial assessing > 2500 participants) and one trial comparing individual education vs social support group (n = 475).				⊕⊝⊝⊝ very low	Small numbers and uncertainty regarding risk of bias and ability to combine results from various trials
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk Ratio; RCT: Randomised controlled trial
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

Education vs. control:

Two trials (Rimer 1999; Sung 1997) failed to fulfil any of the criteria adequately. Only four trials used an adequate method of randomisation (Decker 2013; McAvoy 1991; Mullins 2009; Radde 2016), and the outcome assessor was only blinded in three trials (Bowman 1995; Mishra 2009; Thompson 2016). There was high attrition bias in four trials (Bowman 1995; Radde 2016; Rimer 1999; Sung 1997).

Education vs. other:

Two trials (Allen 2001; Greene 1999) failed to fulfil any of the risk of bias criteria adequately. Three trials (Allen 2001; Bowman 1995; Larkey 2012) had a high risk of attrition bias. In the Mock 2007 trial, women were selected from the social networks of the lay health outreach workers and therefore may have been more motivated to comply.

Measures of treatment effect

We used the following measures of the effect of treatment:

• For dichotomous outcomes (e.g. uptake or non‐uptake), we used the risk ratio as a means of assessment. However, in some cases where results were reported using odds ratios, this was reported as such.

Unit of analysis issues

For cluster‐randomised controlled trials, if the analysis accounted for the cluster design, then a direct estimate of the desired treatment effect was extracted e.g. risk ratio (RR) plus 95% confidence interval (CI). If the analysis did not account for the cluster design, we planned to extract the number of clusters randomised to each intervention, the average cluster size in each intervention group and the outcome data, ignoring the cluster design, for all women in each group. We then planned to use an external estimate of the intracluster coefficient (ICC) to estimate a design effect to inflate the variance of the effect estimate. It would then have been possible to enter the data into RevMan 5 and combine the cluster randomised trials with individually randomised trials in the same meta‐analysis, using the generic inverse variance method of meta‐analysis.

Dealing with missing data

We did not impute missing outcome data for the primary outcome. If data were missing or only imputed data were reported, we contacted trial authors to request data on the outcomes only among participants who were assessed.

Assessment of heterogeneity

Heterogeneity between trials was assessed by visual inspection of forest plots, by estimation of the percentage heterogeneity between trials which could not be ascribed to sampling variation (Higgins 2003), by a formal statistical test of the significance of the heterogeneity (Deeks 2001) and, if possible, by subgroup analyses (see below). If there was evidence of substantial heterogeneity, the possible reasons for this were investigated and reported.

Assessment of reporting biases

Funnel plots corresponding to meta‐analysis of the primary outcome were examined to assess the potential for small study effects. When there was evidence of small‐study effects, publication bias was considered as only one of a number of possible explanations. If these plots suggested that treatment effects may not be sampled from a symmetric distribution, as assumed by the random‐effects model, sensitivity analyses were performed using fixed‐effects models.

Data synthesis

If sufficient clinically similar studies were available, their results were pooled in meta‐analyses.

• For dichotomous outcomes, the risk ratios were pooled.   

For trials with multiple treatment groups, the ‘shared’ comparison group was divided according to the number of treatment groups and comparisons between each treatment group and the split comparison group were treated as independent comparisons.

A random‐effects model with inverse variance weighting was used for all comparisons (DerSimonian 1986).

Where interventions differed to any degree or there was other substantial heterogeneity, the results were reported in a narrative synthesis.

Subgroup analysis and investigation of heterogeneity

Subgroup analyses were performed according to the different categories of intervention type.

Sensitivity analysis

We intended to repeat the meta‐analyses excluding trials at high risk of bias, and examine very large trials separately to determine their overall influence. However, all trials were of unclear or high risk of bias so, consequently, we did not perform sensitivity analyses.

Summary of findings and assessment of the certainty of the evidence

We have presented the overall certainty of the evidence for each outcome according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, which takes into account issues not only related to internal validity (risk of bias, inconsistency, imprecision, publication bias) but also to external validity such as directness of results (Langendam 2013; Schünemann 2011). We created summary of findings tables (Table 1, Table 2, Table 3) based on the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and using GRADEPro GDT 2014. We used the GRADE checklist and GRADE Working Group certainty of evidence definitions (Meader 2014). We downgraded the evidence from 'high' certainty by one level for serious (or by two for very serious) concerns for each limitation:

• High‐certainty: We are very confident that the true effect lies close to that of the estimate of the effect.

• Moderate‐certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

• Low‐certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.

• Very low‐certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

Results

Description of studies

Results of the search

Comprehensive literature searches in June 2020 yielded the following results: CENTRAL (814 references), MEDLINE (581 references), and Embase (1202 references). We screened titles and abstracts (where available) of 2597 references for inclusion in the review. For the latest review, 97 full‐text articles were assessed for inclusion and updated versions of relevant trials. The full‐text screening excluded 65 trials for reasons described in Characteristics of excluded studies, leaving an additional 32 studies that met our inclusion criteria. These, plus the 38 studies included previously, were included in this review, making a total of 70 completed trials. Due to the complexity of one trial (Kitchener 2016), with interventions relevant to subcategories considered in and outside of this review, it was not included in meta‐analyses and results from this will be discussed narratively as a single trial.  Therefore, 69 completed trials, randomising a total of 257,899 women, were included in the meta‐analyses (Figure 3). All included studies are described in the table of Characteristics of included studies.

3.

Included studies

See Characteristics of included studies

Seventy RCTs were included and 69 in the meta‐analysis. Thirty‐three of the trials were performed in the USA (Allen 2001; Binstock 1997; Burack 1998; Burack 2003; Byrd 2013; Clementz 1990; Dietrich 2006; Fang 2017; Greene 1999; Jandorf 2008; Kreuter 1996; Krist 2012; Larkey 2012; Ma 2015; Mishra 2009; Mock 2007; Navarro 1995; Nuno 2011; O'Brien 2010; Ornstein 1991; Paskett 2011; Peitzmeier 2016; Peterson 2012; Rimer 1999; Rivers 2005; Somkin 1997; Studts 2012; Sung 1997; Taylor 2002; Taylor 2010; Thompson 2016; Vogt 2003; Wright 2012), ten in Australia (Bowman 1995; Byles 1994; Byles 1995; Byles 1996; Del Mar 1998; Hunt 1998; Morrell 2005; Mullins 2009; Pritchard 1995; Ward 1991), seven in the UK (Adab 2003; Lancaster 1992; McAvoy 1991; Pierce 1989; Robson 1989; Stein 2005; Wilson 1987), four in Sweden (Alfonzo 2016; Broberg 2013; Eaker 2004; Oscarsson 2007), four in Canada (Buehler 1997; Decker 2013; McDowell 1989; Racey 2016), two in France (Haguenoer 2015; Heranney 2011), one in Italy (Segnan 1998), two in Malaysia (Abdullah 2013a; Rashid 2013), one in Denmark (Jensen 2009); one in Norway (Lonnberg 2016); one in South Africa (Risi 2004); one in Spain (Acera 2017); one in Germany (Radde 2016); and one in Kenya (Rosser 2015).

The majority of the trials were set in community clinics and primary care practices. Fourteen trials were aimed at specific ethnic populations including Asian women (McAvoy 1991), Afro‐American women (Sung 1997), Vietnamese‐American women (Ma 2015; Mock 2007; Taylor 2010), Korean‐American women (Fang 2017), Chinese women (Taylor 2002) and Latinas (Byrd 2013; Jandorf 2008; Larkey 2012; Navarro 1995; Nuno 2011; O'Brien 2010; Thompson 2016). Two trials were aimed at increasing cervical screening uptake in Appalachian women (Paskett 2011; Studts 2012).

Twenty‐eight trials had more than two arms (Acera 2017; Binstock 1997; Bowman 1995; Burack 1998; Byles 1994; Byles 1995; Byles 1996; Byrd 2013; Greene 1999; Hunt 1998; Kreuter 1996; McAvoy 1991; McDowell 1989; Mullins 2009; Ornstein 1991; Peitzmeier 2016; Pierce 1989; Pritchard 1995; Radde 2016; Rashid 2013; Rimer 1999; Rivers 2005; Segnan 1998; Somkin 1997; Stein 2005; Taylor 2002; Thompson 2016; Vogt 2003). Fourteen trials were cluster randomised (Abdullah 2013a; Acera 2017; Allen 2001; Burack 2003; Byles 1994; Byles 1995; Byles 1996; Decker 2013; Jensen 2009; Kitchener 2016; Mock 2007; Navarro 1995; Ornstein 1991; Wright 2012).

Invitations

Twenty‐nine trials (167,980 participants) evaluated the effectiveness of invitation letters (Abdullah 2013a; Acera 2017; Binstock 1997; Bowman 1995; Broberg 2013; Buehler 1997; Burack 1998; Burack 2003; Del Mar 1995; Haguenoer 2015; Heranney 2011; Hunt 1998; Jensen 2009; Lancaster 1992; Lonnberg 2016; McDowell 1989; Morrell 2005; Mullins 2009; Peitzmeier 2016; Pierce 1989; Pritchard 1995; Racey 2016; Radde 2016; Rashid 2013; Segnan 1998; Somkin 1997; Stein 2005; Vogt 2003; Wilson 1987). The trials were subdivided according to the invitation type (i.e. general practitioner/primary care physician (GP) letter, letter from another authority source, face‐to‐face invitation, open invitation and invitation with fixed appointment). Comparison groups included different types of invitation or a control group (usually consisting of usual care or no intervention).

One trial was complex including multiple interventions incorporated within two phases in an attempt to increase uptake amongst those receiving their first invitation to cervical screening (intervention ‐ invitation) (Kitchener 2016): (1) a pre‐routine invitation leaflet (intervention ‐ education), with or without access to online booking versus control; (2) non‐ attenders were further randomised into interventions designed to address possible barriers to screening: vaginal HPV self‐sampling kit sent unrequested or offered on request (intervention‐ procedure), timed appointments, access to a nurse navigator, choice between a nurse navigator or HPV self‐sampling.

Eight trials (22,735 participants) looked at telephone invitations (Binstock 1997; Broberg 2013; Heranney 2011; McDowell 1989; Peitzmeier 2016; Rashid 2013; Stein 2005; Vogt 2003), and one trial looked at face‐to‐face invitations from a health worker or GP (Hunt 1998). Four trials (2998 participants) looked at the use of a letter invitation which was then followed by a telephone/email reminder/invitation (Abdullah 2013a; Acera 2017; Peitzmeier 2016; Vogt 2003). Five trials (27,663 participants) looked at the use of a personalised invitation compared to a standard invitation (Binstock 1997; Heranney 2011; Hunt 1998; Jensen 2009; McDowell 1989). Two trials (4114 participants) evaluated invitations from different authority sources (Bowman 1995; Segnan 1998), and the use of letters with appointments to attend for screening; three additional studies (5742 participants) also evaluated the use of letters with appointments (Lonnberg 2016; Pritchard 1995; Wilson 1987). Four trials (2998 participants) examined the use of letter with open invitations to make appointments versus control (usual care) (Bowman 1995; Pritchard 1995; Somkin 1997; Vogt 2003). One trial compared the use of letters, telephone, email and multimodal invitation reminders (Peitzmeier 2016).

Education

Eight trials (61,182 participants) evaluated educational printed materials (Acera 2017; Bowman 1995; Decker 2013; Eaker 2004; McAvoy 1991; Mullins 2009; Radde 2016; Rimer 1999). Three trials (1318 participants) assessed face‐to‐face educational home visit trials (McAvoy 1991, Sung 1997; Taylor 2002) and one trial did not report in detail the type of educational intervention used (Greene 1999). One cluster‐randomised trial adequately accounted for the clustering in its analyses, but the data reported were not suitable for calculating risk ratios (Navarro 1995). Two trials looked at the use of electronic health maintenance reminders (Krist 2012; Wright 2012). Thirteen trials (4783 participants) looked at the use of health outreach worker programmes in various communities (Byrd 2013; Fang 2017; Jandorf 2008; Ma 2015; Mock 2007; Nuno 2011; O'Brien 2010; Paskett 2011; Peterson 2012; Rosser 2015; Studts 2012; Taylor 2010; Thompson 2016). One trial compared individual education versus education given within a social support group (Larkey 2012).

Counselling

Three trials (601 participants) examined the use of counselling. One compared face‐to‐face counselling by a GP as compared to no counselling (Ward 1991); the other compared telephone counselling and patient prompts, versus patient prompts alone (control group) and provider prompts alone (Rimer 1999). One study compared a series of telephone calls providing motivational support and counselling versus a single telephone reminder (Dietrich 2006).

Risk factor assessment

Two trials (145 participants) evaluated risk factor assessment (Greene 1999; Kreuter 1996). Both used an enhanced risk factor assessment that involved a personally tailored assessment and discussion with the healthcare provider about the woman's personal risk factors for developing cervical cancer. Both interventions were based on theoretical models of behaviour, the Social Cognitive Theory and Motivational Interviewing Methods (Greene 1999) and the Health Belief Model (Kreuter 1996), with a view to changing behaviour to increase the uptake of Pap smears. One trial (Greene 1999) compared the intervention to usual care whereas the other trial (Kreuter 1996) compared the intervention to a no intervention control group (Kreuter 1996). Similarly, this trial also compared enhanced risk factor assessment with a less intense 'typical' risk factor assessment. The typical risk factor assessment involved supplying the participant with personal risk factor information but not discussing the information provided.

Secondary outcomes

A summary of the data relating to secondary outcomes is presented in Table 4.

1. Details of secondary outcomes.

Study details	Interventions	Secondary outcome(s)	Results
Binstock 1997	1. Telephone call n = 1526 (1526 analysed) 2. Letter n = 1526 (1526 analysed) 3. Memo to woman's primary provider n = 1526 (1526 analysed) 4. Chart reminder affixed to outside of woman's medical record n =1526 (1526 analysed) 5. Control group n = 1526 (1526)	Costs	Total estimated costs ($US) per intervention: 1. $4282; 2. $1918; 3. $8933; 4. $1090; 5. Not stated. Estimated cost ($US) per additional Pap smear performed: 1. $7.99; 2. $4.76; 3. $22.96; 4. $2.99; 5. Not applicable
Byles 1995	1. Personally addressed letter with simple information about Pap smears n = ? (1128 analysed), 2. Personally addressed letter combined with a series of targeted behavioural prompts (e.g. prompt cards) designed to address aspects believed to be associated with poor screening rates n = ? (1098 analysed), 3. Control n = ? (1414 analysed)	Acceptability of the intervention	Number (%) of responding women receiving the intervention: 1. 154 (72%); 2. 134 (78%) letter, 100 (58%) card, 109 (64%) pamphlet; 3. Not applicable. Number (%) of women responders who said they had read the material sent: 1. 147 (69%); 2. 128 (75%) letter, 7 (4%) card, 101 (59%) pamphlet; 3. Not applicable. For intervention 1: 118/151 (78%) of the women said that they were pleased to have the intervention personally addressed to them, only 1/151 (1%) said they were displeased and the remainder were not sure. In intervention 2: 89/132 (68%) were pleased, 3/132 (2%) were displeased and the remainder were unsure. In intervention 1: 152/155 (98%) of the women thought that the intervention should be sent to all women, 2/155 (1.3%) did not and the remainder were unsure. In intervention 2: 124/130 (95%) of women thought the intervention should be sent to all women, 1/130 (1%) did not and the remainder were unsure.
Greene 1999	1. Usual care, n = 79 (? analysed) received general dietary and health information 2. Cancer education, n = 97 (? analysed) received general information about cervical cancer risk factors and screening recommendations 3. Cognitive behavioural intervention, n = 97 (? analysed) received feedback about personal risk for cancer and engaged in a clinical interview to enhance self‐efficacy for preventive behaviour	Booking of appointments	Women in group 1 were more likely to schedule an appointment for a Pap smear than those in group 3 (group 1 = 79.4% versus group 3 = 36.7%, P </= 0.0001). Women in group 1 were also more likely to attend without rescheduling the appointment (group 1 = 63.9% versus group 3 = 35.4%, P </= 0.001). Group 2 did not differ from group 3 on these measures.
Heranney 2011	1. Telephone reminder, n = 5310 2. Mail reminder, n = 5352	Cost‐effectiveness of each intervention	The global budget was higher for the telephone strategy. An investment of 46 euro was needed to convince a women by telephone to have a smear, versus 33 euro by mail.
Jandorf 2008	1. Breast and cervical cancer education programme, n = 308 2. Diabetes education programme, n = 179	Cervical cancer knowledge & screening	Those receiving the breast and cervical education programme were more knowledgeable in this topic. Those receiving the diabetes education programme were knowledgeable in this topic.
Larkey 2012	Community‐based cancer screening/prevention classes led by lay health workers delivered: 1. Individually, n = 192 2. Social support group, n = 283	Comparison of cost of achieving screening with the two interventions	Cost to achieve one cancer screening ranged from $263 to $527 in the social support group and from $862 to $1716 in the individual group. The cost per participant and the cost per participant screened were more than three times higher in the individual group than the social support group.
McDowell 1989	1. GP letter and reminder letter after 21 days, n = 367 (367 analysed) 2. Physician reminder, n = 332 (332 analysed) 3. Telephone call, n = 377 (377 analysed) 4. Control group, n = 330 (330 analysed)	Costs	The costs for the GP letter were $14.23 per screening gained, compared with $11.75 assuming a salary of $60 per hour (or $5.88 at $30 per hour) per screening gained.
Mishra 2009	1. Control, no intervention, n = 197 2. English and Samoan language cervical cancer education booklets, skill building and behavioural exercises, interactive group discussion, n = 201	Cervical cancer knowledge and attitudes	The only difference between the intervention and control groups was the mean knowledge score, which was higher in the intervention group (P < 0.05)
O'Brien 2010	1. Control group (usual care), n = 36 2. Two 3‐hour workshops giving information on risk factors for cervical cancer, screening procedures and recommendations, including 4‐10 women, led by lay health workers, n = 34	Cervical cancer knowledge and self‐efficacy to undergo screening	Cervical cancer knowledge and self‐efficacy was higher among the intervention group at 6 months (P < 0.001). Only post‐intervention cervical cancer knowledge and group assignment were predictive of receiving cervical screening.
Oscarsson 2007	1. Control. No intervention, n = 400 2. Intervention included invitation letters, telephone interviews and promotive efforts for having a cervical smear taken, n = 400	Costs	Cost of extra Pap smear gained was calculated, 151.36€. The cost of a smear in the intervention group was calculated at 66.87€ each and 16.63€ in the control group.
Rosser 2015	1. Control, n = 212 2. 10‐minute health talk delivered by community health workers on cervical cancer knowledge and attitudes, n = 207	Cervical cancer knowledge and attitudes	Increase in cervical cancer knowledge of 26.4% in the intervention group compared to 17.4% in the control group (P < 0.01)
Stein 2005	1. Control, no intervention, n = 285 2. Telephone call. Telephone call from experienced research nurse using a prepared script. Maxiumum of three attempts were made on consecutive days, n = 285 3. Letter from Health Authority District Cervical Screening Commisioner on behalf of National Cervical Screening Programme, n = 285 4. Letter from a well known journalist and broadcaster (Claire Rayner) who was also Chair of the Patients Association, n = 285	Costs	Average cost per attender was £145.12 for telephone call, £14.29 for letter from commissioner and £37.14 for letter from celebrity
Thompson 2016	1. Control, n = 147 2. Spanish language video, n = 150 3. Spanish language video and visit by a lay health worker, n = 146	Cervical cancer knowledge and attitudes Costs	There was increase in knowledge in all study groups. The Spanish video only intervention cost an additional $15 per participant and the Spanish video and lay health worker intervention cost and additional $82 per participant. Neither were cost effective.
Vogt 2003	1. Usual‐care control 2. Letter/letter intervention: subjects were sent a letter and relevant brochure. Women who had not attended for screening within 6 weeks were sent a further letter emphasising the importance of screening and providing a number to call. 3. Letter/phone intervention: letter and brochure as above. Women who had not attended for screening within 6 weeks received a telephone call by study interventionist who offered to schedule appointments, answer questions, address barriers and concerns and discuss the importance of screening. 4. Phone/phone intervention: subjects in this group received two sequential telephone calls, the second coming 6 weeks after the first if they had not been screened in the interim. Contents of the initial letter and phone scripts were similar. Follow‐up telephone calls were by study interventionist, as above.	Costs	The letter/letter intervention produced one additional Pap smear for $185. The phone/phone intervention cost $305 and the letter/phone intervention cost $1117 for each additional Pap smear.

One trial used the booking of appointments for screening as an outcome measure (Greene 1999). Six trials examined participants knowledge of and attitudes to Pap smear screening (Byles 1995; Jandorf 2008; Mishra 2009; O'Brien 2010; Rosser 2015; Thompson 2016). Eight trials presented cost data (Binstock 1997; Heranney 2011; Larkey 2012; McDowell 1989; Oscarsson 2007; Stein 2005; Thompson 2016; Vogt 2003). Many of the trials used multiple intervention groups but only those groups that used an intervention aimed at women (and not healthcare providers) were included in this review.

Excluded studies

The 129 references excluded after assessing full‐text paper copies are listed in the Characteristics of excluded studies table, with reasons for their exclusion:

• Seventeen were quasi‐randomised (Baele 1998; Chumworathayi 2007; Hicks 1997; Hou 2002; Hou 2005; Lantz 1995; Lantz 1996; Levine 2003; Love 2009; Love 2012; Marcus 1992; Margolis 1998; Maxwell 2003; Park 2005; Paskett 1990; Ward 1999; Yancey 1995),

• Three studies included participants who may have been screened before receiving the intervention (Dignan 1996; Dignan 1998; Gotay 2000),

• Thirteen studies used an intervention aimed at either the physician, both the physician and patient or the community (Boissel 1995; Bonevski 1999; Campbell 1997; Cecchini 1989; Hillman 1998; Litzelman 1993; Maddocks 2011; Manfredi 1998; Nguyen 2000; Rock 2014; Roetzheim 2004; Roetzheim 2005; Thompson 2006),

• Twenty‐one studies used an intervention or an outcome that was not strictly concerned with increasing uptake (Atlas 2013; Bebis 2012; Benard 2014; Brewster 2002; Burger 2017; Darlin 2013; Del Mar 1995; Holloway 2003; Katz 2007; Luszczynska 2011; Miller 2017; Mutyaba 2009; Paul 2003; Philips 2006; Pirzadeh 2012; Sankaranarayanan 2003; Shastri 2014; Shojaeizadeh 2011; Wichachai 2016; Wood 2014; Wright 2010),

• Ten studies concerned interventions aimed at improving follow‐up of an abnormal smear result rather than Pap screening uptake (Cofta‐Woerpel 2009; Engelstad 2005; Lauver 1990; Marcus 1998; Miller 1999; Paskett 1995; Peters 1999; Stewart 1994; Takacs 2004; Tomlinson 2004),

• One trial (Corkrey 2005) did not adequately report outcomes suitable for inclusion in the review.

• One study (Hancock 2001) did not meet the inclusion criteria as it was not a randomised trial.

• One trial (Lynch 2004) did not look at interventions aimed at routine cervical screening, rather at an overdue population and women attending both mammography and cervical screening.

• One trial (Newell 2002) was an RCT with public health record books delivered to the intervention arm town versus control. However, in the intervention arm, in addition to record books, media campaign and recruitment of family practitioners to promote the books was carried out, making it not suitable to assess the impact of intervention.

• Three studies were excluded as they included women over the upper age limit of most routine cervical screening programmes (Mayer 1992; Ruffin 2004; Valanis 2003),

• Two studies did not separate attendance for cervical screening from other screening tests (Mitchell 1991; Powers 1992)

• Sixteen studies did not use a randomised design (Al Sairafi 2009; Blomberg 2011; Chigbu 2017; Duke 2015; German 1995; Hiatt 2008; Jenkins 1999; Karwalajtys 2007; Lyimo 2012; Miller 2007; Mitchell 1997; Paskett 1999; Perkins 2007; Shelley 1991; Shenson 2011Torres‐Mejia 2000).

• Fourteen papers (Abdullah 2013b; Acera 2014; Broberg 2012; Duggan 2012; Kobetz 2017; Krok‐Schoen 2016; Lairson 2010; Lam 2003; Mbah 2015; Scoggins 2010; Smith 2013; Tranberg 2016; Williams 2013; Wright 2008) described a study protocol and/or preliminary/duplicate results, and were available in another paper; the full results have already been included in this review (Abdullah 2013a; Acera 2017; Broberg 2013; Byrd 2013; Mock 2007; Paskett 2011; Taylor 2010; Wright 2012).

• One trial was a cluster‐RCT examining anxiety among women with mild dyskaryosis and the aim of the educational intervention was to reduce anxiety so the scope differed to that of this review (Peters 1999).

• Twenty papers discussed trials which were excluded from current analyses despite meeting the inclusion criteria, as they assessed HPV self‐sampling as a method to increase cervical screening and this will be addressed in a separate Cochrane review being written in parallel with this review (Broberg 2014; Cadman 2015; Del Mistro 2017; Enerly 2016; Giorgi Rossi 2011; Giorgi Rossi 2015; Gok 2012; Haguenoer 2015; Piana 2011; Racey 2016; Sancho‐Garnier 2013; Sewali 2015; Sultana 2014; Sultana 2016; Szarewski 2011; Virtanen 2011a; Virtanen 2011b; Viviano 2017; Wikstrom 2011; Zehbe 2016).

• Five references obtained were for abstracts only. It was not possible to obtain data or details regarding the studies that would be usable in this review (Elder 2016; Lopez‐Torres Hidalgo 2016; Murphy 2015; Okeke 2013; Wedisinghe 2013).

Risk of bias in included studies

(See Risk of Bias tables in Characteristics of included studies; Figure 1; Figure 2)

The risk of bias in the majority of trials was unclear. The majority of trials, at most, were judged as at low risk of bias for only one or two of the criteria. Eleven trials failed to fulfil any of the criteria adequately (Allen 2001; Byles 1994; Byles 1995; Byles 1996; Greene 1999; Jensen 2009; Kreuter 1996; Krist 2012; Navarro 1995; Rimer 1999; Sung 1997).

Allocation

Thirty‐four trials used an adequate method of generation of the sequence of random numbers to allocate women to treatment arms (Abdullah 2013a; Adab 2003; Alfonzo 2016; Broberg 2013; Buehler 1997; Burack 1998; Byrd 2013; Clementz 1990; Dietrich 2006; Fang 2017; Haguenoer 2015; Hunt 1998; Jandorf 2008; Kitchener 2016; Larkey 2012; Lonnberg 2016; McAvoy 1991; Mullins 2009; Nuno 2011; O'Brien 2010; Oscarsson 2007; Paskett 2011; Peterson 2012; Pritchard 1995; Racey 2016; Radde 2016; Rashid 2013; Rivers 2005; Robson 1989; Rosser 2015; Segnan 1998; Stein 2005; Taylor 2010; Wright 2012). The method of randomisation was unclear in the remaining trials.

The allocation was adequately concealed in only 11 trials (Abdullah 2013a; Decker 2013; Eaker 2004; Haguenoer 2015; Jandorf 2008; Kitchener 2016; Mishra 2009; Paskett 2011; Rashid 2013; Studts 2012; Wilson 1987), with it being unclear in all remaining trials, with exception of the trials of Clementz 1990 and Peitzmeier 2016 where the authors reported an inadequate method of allocation concealment.

Blinding

Blinding was only reported in nine trials (Bowman 1995; Del Mar 1998; Hunt 1998; Jandorf 2008; Mishra 2009; Peitzmeier 2016; Rashid 2013; Rosser 2015; Thompson 2016). In the trial of Wright 2012, it was reported that there was no blinding and in the trial of Nuno 2011, there was the possibility of contamination between groups. Blinding was unclear in the remaining trials as it was not reported.

Incomplete outcome data

Loss to follow‐up was low in 43 of the trials, with at least 80% of women being assessed at the end of the study (Abdullah 2013a; Acera 2017; Adab 2003; Binstock 1997; Buehler 1997; Burack 2003; Byrd 2013; Clementz 1990; Decker 2013; Del Mar 1998; Dietrich 2006; Eaker 2004; Fang 2017; Haguenoer 2015; Heranney 2011; Hunt 1998; Lancaster 1992; Lonnberg 2016; Ma 2015; McAvoy 1991; McDowell 1989; Mishra 2009; Mock 2007; Morrell 2005; Mullins 2009; Nuno 2011; Ornstein 1991; Oscarsson 2007; Paskett 2011; Peitzmeier 2016; Pierce 1989; Pritchard 1995; Racey 2016; Risi 2004; Robson 1989; Segnan 1998; Somkin 1997; Studts 2012; Taylor 2002; Thompson 2016; Ward 1991; Wilson 1987; Wright 2012). Risk of attrition bias was high in 18 trials, as less than 80% of women were assessed at endpoint in at least one of the outcomes (Allen 2001; Bowman 1995; Broberg 2013; Burack 1998; Jandorf 2008; Kreuter 1996; Larkey 2012; Navarro 1995; O'Brien 2010; Peterson 2012; Radde 2016; Rashid 2013; Rimer 1999; Rivers 2005; Rosser 2015; Stein 2005; Sung 1997; Vogt 2003). In the remainder of trials, it was unclear.

Selective reporting

In most trials, it was unclear as to whether outcomes had been selectively reported as there was insufficient information to permit judgement.

Other potential sources of bias

In the majority of trials, there was insufficient information to assess whether any important additional risk of bias existed. The trial of Robson 1989 was potentially biased as it stopped early because participating doctors were not prepared to continue excluding half the practice from access to the health promotion nurse. In Rivers 2005, the women received a telephone call at six months to ascertain screening uptake and this may itself have acted as a prompt for non‐attenders to attend for screening and as such influenced the 12‐month data. Women were selected from the social networks of the lay health outreach workers in Mock 2007 and, therefore, may be more motivated to comply. In the trial of Racey 2016, there was a change in the study protocol and some participants in the Pap test group received a reminder phone call which may have influenced the overall outcome.

Effects of interventions

See: Table 1; Table 2; Table 3

Pooling of data was restricted because of clinical heterogeneity. The majority of cluster‐RCTs did not account for clustering or adequately report the number of clusters in the trial in order to estimate the design effect, so we did not selectively adjust the trials on this basis. Several RCTs were not included in any of the meta analyses because it was either not possible to extract binary data or interventions or/and outcomes differed sufficiently. We obtained values of the ICC that ranged from 0.02 to 0.29 (Hade 2010) and sensitivity analyses were not plausible, given the high number of included trials and multiple thresholds of the ICC that could be used. Furthermore, when 0.02 was used as the ICC for the trial of Byles 1996, we got an estimated design effect of 109 due to the average cluster size being large, which considerably decreased the effective sample size. Since the unit of randomisation was postal codes, we would not expect a large ICC, but do not have estimates from any pilot studies so any assumptions in these sort of cases are dubious. There were also trials where reducing the effective sample size would have made a modicum of difference to the results; for example, the trial of Jensen 2009 had a design effect of 12.5 using an ICC value of 0.02 and this would have been applied to a very large sample size. To a much lesser extent, a design effect of 8.2 in the trial of Wright 2012 would have made little difference. However, it is highly unlikely that selectively adjusting the trials with adequate reporting of clustering would make much difference to the overall conclusions and robustness of the results. Additionally, the results should not be biased by arm, but just the precision would be a little greater than it should be, thus 95% CIs would be wider in reality. Of the meta‐analyses that could be performed, there was quite often significant statistical heterogeneity present, and this should be borne in mind when interpreting these findings.

The Kitchener 2016 trial included two phases; the first accessed the impact of a pre‐routine invitation leaflet and online booking on the uptake of screening in women receiving their first invitation. Non‐attenders were then further randomised into various interventions that might address possible screening barriers: vaginal HPV self‐sampling sent unrequested or offered on request, timed appointments, access to a nurse navigator, choice between a nurse navigator, or HPV self‐sampling. Due to the complexity of this trial, with interventions relevant to various subcategories considered in and outside of this review, it was discussed narratively as a single trial.

Uptake of screening

Invitations

Invitation versus control

Meta‐analysis of 24 trials (Acera 2017; Binstock 1997; Bowman 1995; Broberg 2013; Buehler 1997; Burack 1998; Burack 2003; Del Mar 1998; Haguenoer 2015; Hunt 1998; Lancaster 1992; McDowell 1989; Morrell 2005; Mullins 2009; Peitzmeier 2016; Pierce 1989; Pritchard 1995; Racey 2016; Radde 2016; Rashid 2013; Somkin 1997; Stein 2005; Vogt 2003) comparing invitation versus control to increase uptake of screening, found that women who received invitations were more likely to attend cervical screening than women who received usual care or no invitation (RR 1.71, 95 % CI: 1.49 to 1.96; moderate‐certainty evidence). A funnel plot corresponding to the invitation versus control subgroup in the above meta‐analysis showed no evidence of bias in small studies (see Figure 4) (Analysis 1.1). The percentage of the variability in effect estimates that was due to heterogeneity rather than sampling error (chance) may represent considerable heterogeneity (I² = 92%) (Analysis 1.1,, Comparison 1.1.1).

4.

Funnel plot of comparison: 1 Invitation vs control, outcome: 1.1 Uptake of screening

1.1. Analysis.

Comparison 1: Invitation versus control, Outcome 1: Uptake of screening

Invitation letter versus control

Meta‐analysis of 18 trials (Abdullah 2013a; Acera 2017; Binstock 1997; Bowman 1995; Buehler 1997; Burack 1998; Burack 2003; Del Mar 1998; Haguenoer 2015; Heranney 2011; Hunt 1998; Lancaster 1992; McDowell 1989; Morrell 2005; Mullins 2009; Peitzmeier 2016; Pierce 1989; Racey 2016; Radde 2016; Stein 2005), assessing 118,911 participants, found that women who received invitation letters to attend cervical screening programmes were more likely to attend screening than women who received usual care or no invitation (RR 1.56, 95% CI: 1.32 to 1.83).
The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 92%) (Analysis 1.1,, Comparison 1.1.1).

Telephone invitation versus control

Meta‐analysis of seven trials (Binstock 1997; Broberg 2013; McDowell 1989; Peitzmeier 2016; Rashid 2013; Stein 2005; Vogt 2003), assessing 10,174 participants, found that women who received a telephone invitation had a higher uptake of screening than those in the control group (RR 1.95, 95% CI: 1.65 to 2.30). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent moderate heterogeneity (I² = 46%) (Analysis 1.1,, Comparison 1.1.2).

Face‐to‐face invitation versus control

The trial of Hunt 1998, which assessed 121 participants, found no difference in the uptake of screening between women who received a face‐to‐face invitation and those in the control group (RR 9.15, 95% CI: 0.50 to 166.30) (Analysis 1.1,, Comparison 1.1.3). However, only four out of the 121 women attended for screening.

Letter with open invitation to make appointment versus control

Meta‐analysis of four trials (Bowman 1995; Pritchard 1995; Somkin 1997; Vogt 2003), assessing 2998 participants, found that women who received letters with an open invitation to attend a cervical screening programme had a higher uptake of cervical screening than women in the control group (RR 1.61, 95% CI: 1.15 to 2.26). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent moderate heterogeneity (I² = 50%) (Analysis 1.1,, Comparison 1.1.4).

The Byles 1995 cluster‐RCT assessed the effectiveness of two direct mail strategies to encourage women to have Pap smears compared to a control. The two strategies were assessed in two geographically separated postal regions in Australia. The first intervention consisted of a personally addressed letter providing basic information about Pap smears, information on screening, advice, lists of local providers and an invitation to enrol with a free Pap reminder service. The second intervention, in addition to the letter, included a multi‐faceted intervention package designed to address a number of factors associated with screening behaviour. The effectiveness of each strategy was assessed using a multiple group time‐series design involving three postal regions. Both interventions resulted in increased attendances for screening over the post intervention period (42.2% in the region receiving the simple prompt and 39.6% in the region receiving the multi‐faceted approach).

Similarly, the cluster‐RCT of Byles 1996 reported results of a mass mailing campaign to promote screening. The intervention letter provided information on the screening services available in the local area and included an invitation to enrol with the Pap smear reminder service. In the intervention arms, women were either sent a personalised letter with the initial invitation or, three years later, a reminder invitation. A control group received no letter at all. Significant increases in screening rates were observed in those receiving a personalised letter with the initial invitation.

The cluster‐RCT of Ornstein 1991 assessed the effectiveness of three interventions that aimed to encourage uptake of various screening/vaccine sessions, including a Pap smear. Patients and their physicians were randomly assigned by the practice group into one of four groups which included physician reminders, patient reminders, patient and physician reminders and a control group. The authors concluded that computer‐based physician and patient reminder systems improved adherence to preventive services in primary care settings. In this one year study, a decline in Pap smear adherence was confined to the physician reminder group. Small, inconsistent declines in adherence occurred in all four groups. These declines were significant only for white women.

Letter with fixed appointment versus control

In the trial of Pritchard 1995, which assessed 177 participants, women who received letters with a fixed appointment to attend a cervical screening programme had a higher uptake of screening than the control group (RR 1.80, 95% CI: 1.04 to 3.11) (Analysis 1.1, Comparison 1.1.5).

Letter invitation with telephone/email follow‐up versus control

Meta‐analysis of four trials (Abdullah 2013a; Acera 2017; Peitzmeier 2016; Vogt 2003), assessing 8059 participants, found that women who received a letter invitation to attend for cervical screening with a telephone follow‐up had higher attendance rates for screening than women receiving usual care (RR 2.19, 95% CI: 1.39 to 3.44). The percentage of variability in effect estimates that was due to heterogeneity rather than sampling error (chance) may represent considerable heterogeneity (I² = 83%) (Analysis 1.1,, Comparison 1.1.6).

Celebrity invitation versus control

The trial of Stein 2005, which assessed 316 participants, found no difference in the uptake of screening between women who received a celebrity‐endorsed letter of invitation and those in the control group (RR 2.15, 95% CI: 0.25 to 18.15) ((Analysis 1.1,, Comparison 1.1.7).

SMS text invitation versus control

The trial of Rashid 2013, which assessed 376 participants, found a greater uptake of screening in women who received a SMS text invitation compared to a control group (RR 2.24, 95% CI: 1.67 to 3.00) (Analysis 1.1,, Comparison 1.1.8).

Email versus control

The trial of Peitzmeier 2016, which assessed 259 participants, found no difference in the uptake of screening between women who received an email invitation compared to the control group (RR 1.18, 95% CI 0.67 to 2.09) (Analysis 1.1,, Comparison 1.1.9).

Overall, these trials were of moderate quality (Table 1), providing moderate‐certainty evidence that invitations appear to be effective at increasing the uptake of cervical screening.

GP invitation letter versus invitation letter from other authority sources

The trial of Bowman 1995, which assessed 86 participants, found little difference between GP invitation letters and health clinic invitation letters in the uptake of cervical screening (RR 1.69, 95% CI: 0.75 to 3.82). In the trial of Segnan 1998, which assessed 4028 participants, women who received GP letters to attend a cervical screening programme had a significantly higher uptake of screening than those who received invitation letters from programme coordinators (RR 1.13, 95% CI: 1.05 to 1.21) (Analysis 2.1). However, these two trials provided very low‐certainty evidence (Table 1).

2.1. Analysis.

Comparison 2: GP invitation letter versus invitation letter from other authority sources, Outcome 1: Uptake of screening

Personal invitation versus invitation letter

Meta‐analysis of five trials (Binstock 1997; Heranney 2011; Hunt 1998; Jensen 2009; McDowell 1989), assessing 27,663 participants, found that personal invitations (telephone call, face‐to‐face or targeted letter) to attend cervical screening had higher uptake as compared to standard invitation letters (RR 1.32 95 % CI: 1.11 to 1.21). The trials provided low‐certainty evidence (Table 1).

Meta‐analysis of three trials (Binstock 1997; Heranney 2011; McDowell 1989), assessing 12,561 participants, found that women who received telephone invitations to attend cervical screening had a higher uptake of screening than women given invitation letters (RR 1.21, 95% CI: 1.05 to 1.40). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance was not important (I² = 39%) (Comparison 3.1.1). The trial of Hunt 1998, which assessed 123 participants, found no difference between face‐to‐face invitations and invitation letters in the uptake of cervical screening (RR 2.10, 95% CI: 0.40 to 11.05) (Analysis 3.1).

3.1. Analysis.

Comparison 3: Personal invitation versus invitation letter, Outcome 1: Uptake of screening

The cluster trial of Jensen 2009 had an intervention aimed at both the individual women and GPs. Cluster randomisation was performed at the level of the GP. GPs in the intervention group received a visit from a facilitator who facilitated quality enhancement of procedures related to screening. Women in this intervention group also received a targeted invitation letter. The control group received a standard invitation letter. After nine months, this trial found it was possible to increase the coverage rate in a screening programme by using targeted invitations combined with a visit to GPs. There was a difference of 1.97% (95% CI: 0.03 to 3.91) in the change of coverage rate at nine months.

These trials provided low‐certainty evidence (Table 1).

Letter with fixed appointment versus letter with open invitation to make an appointment

Meta‐analysis of five trials (Bowman 1995; Lonnberg 2016; Pritchard 1995; Segnan 1998; Wilson 1987), assessing 5742 participants, found that women who were given letters with a fixed appointment to attend a cervical screening programme had a higher uptake of screening than women who received letters with an open invitation (RR 1.61, 95% CI 1.48 to 1.75). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance was not important (I² = 0%) (Analysis 4.1). These trials provided moderate‐certainty evidence that letters with a fixed appointment may be effective at increasing cervical screening (Table 1).

4.1. Analysis.

Comparison 4: Letter with fixed appointment versus letter with open invitation to make an appointment, Outcome 1: Uptake of screening

Three‐way comparison of television media, television media combined with invitation letter, and television media combined with GP‐based recruitment

The cluster‐RCT of Byles 1994 assessed the effectiveness of three community‐based strategies to promote screening for cervical cancer. A trial of each television media intervention was carried out in three postal regions in New South Wales ‐ a rural locality, a country town and a major rural centre. Three control regions were selected to be demographically similar to the corresponding regions. Television media alone was associated with a significant increase in attendances for screening in the rural centre. The media/letter‐based campaign was associated with a significant increase in attendances in the rural locality and rural centre. The media/GP‐based campaign was associated with significant increases in attendances in all three regions. All three interventions were associated with significant increases in the number of women attending for screening above those observed in the control regions. Furthermore, these increases were not restricted to women at low risk. They were also found for older women (aged 50 to 69 years) and women who had not had a Pap smear within the past three years.

Pre‐routine invitation leaflet versus no pre‐invitation leaflet at first screening invitation

The Kitchener 2016 trial was embedded in the NHS Cervical Screening Programme. Participants were cluster‐randomised by general practice. In its first phase, 20,879 women were randomised over an 18‐month period into those who would receive either a pre‐invitation leaflet six weeks before the standard invitation, or the routine standard invitation alone. A focus group of young women played a role in the development of a concise factual pre‐invitation leaflet as it was thought that 20 to 25 year olds would lack interest in the lengthy factual national information sheet. It was designed to target issues that the group of women within the focus group thought about themselves when thinking about attending for cervical screening. This trial took part in two UK areas ‐ Manchester and Scotland. The women in Manchester were also offered access to an online booking system in an attempt to overcome an assumed barrier of inconvenience by the need to book an appointment. This was irrespective of whether or not they had received a pre‐invitation leaflet. There was no difference in screening uptake between those who received a pre‐invitation leaflet and those who did not at three months (18.8% versus 19.2%, respectively, [OR 0.96, 95% CI: 0.88 to 1.06]) and six months (31.1% versus 30.6%, respectively, [OR 1.01 95% CI 0.93 to 1.10]). Access to online booking also had no impact on the uptake of screening with 28.8% of those offered online booking participating in screening compared to 26.6% of those who were offered routine booking of screening at six months (OR 1.09, 95% CI: 0.94 to 1.28). The booking of appointments only was only used by 6% of those who were offered it.

Pre‐routine invitation leaflet versus no pre‐invitation leaflet at first screening invitation plus further intervention

In the second phase of the Kitchener 2016 trial, in addition to the initial pre‐invitation leaflet and routine invitation or routine invitation only, non‐attenders (n = 10,126) at six months were further randomised into another one of six interventions that may overcome potential barriers to screening: (1) HPV self‐sampling kit sent unrequested or (2) offered on request, (3) timed appointments, (4) access to a nurse navigator, (5) a choice between access to a nurse navigator or HPV self‐sampling kit, or (6) no further intervention (control). The HPV self‐sampling kit would address the dislike of having a gynaecological exam. The HPV self‐test or the mailing of a timed appointment would address the inconvenience of having to book an appointment. Nurse navigators offered support and guidance and could allay fears regarding what screening involves and address poor understanding of the reasons for screening and its potential benefits and harms. They could also address any other barriers to screening. At 12 months following the initial invitation for screening, only the group who were sent the HPV self‐test unrequested (OR 1.51, 95% CI 1.20 to 1.91) and the group who were sent timed appointments (OR 1.4, 95% CI 1.14 to 1.74) had a higher rate of screening uptake. This increase in screening uptake was only maintained in the group who were sent the HPV self‐test unrequested at 18 months, with 30% of those sent the HPV self‐test participating in screening compared to 27.1% of controls at 18 months (OR 1.29, 96% CI 1.06 to 1.57).

The Kitchener 2016 trial also suggested that those who received the HPV vaccination as part of a national programme in Scotland may have a sense of immunity from cervical cancer and therefore refrain from screening. This was estimated to be 65% to 70% of the women participating in this trial from Scotland as the result of the national catch‐up vaccination campaign of 2008 to 2010. This trial found that the uptake of screening at three months was higher amongst previously HPV‐vaccinated women (23.7% in HPV‐vaccinated versus 11% in non‐vaccinated, OR 2.07, 95% CI: 1.69 to 2.53).

Counselling

Counselling versus control

Meta‐analysis of two trials (Rimer 1999; Ward 1991), assessing 393 participants, found that women given counselling to encourage attendance of a cervical screening programme had a higher uptake of screening than those given no counselling or patient prompts alone (RR 1.23, 95% CI 1.04 to 1.45; very low‐certainty evidence). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance was not important (I² = 0%) (Analysis 5.1). The trial of Rimer 1999 found little difference in the uptake of cervical screening between women who received telephone counselling aimed at increasing awareness of cervical screening programmes and women who received provider prompts (RR 1.13, 95% CI: 0.90 to 1.41; very low‐certainty evidence). The cluster trial by Dietrich 2006 evaluated the effect of counselling via a series of telephone support calls compared to a single call advising participants to participate in preventive care. Women assigned to the intervention group received an average of four calls and, in this group, screening rates increased by 0.07, compared to no change in the control group (P < 0.001). These trials provided very low‐certainty evidence that counselling may increase the uptake of cervical screening (Table 2).

5.1. Analysis.

Comparison 5: Counselling versus control, Outcome 1: Uptake of screening

Education

Education versus control

Meta‐analysis of 13 trials (Acera 2017; Bowman 1995; Decker 2013; Eaker 2004; Greene 1999; McAvoy 1991; Mishra 2009; Mullins 2009; Radde 2016; Rimer 1999; Taylor 2002; Thompson 2016; Wright 2012), assessing 63,415 participants, showed an increase in the uptake of screening between women who received printed material as a form of education compared to those in the control group (RR 1.35, 95% CI 1.18 to 1.54). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 91%) (Analysis 7.1).

7.1. Analysis.

Comparison 7: Education versus control, Outcome 1: Uptake of screening

Education (printed material) versus control

Meta‐analysis of eight trials (Acera 2017; Bowman 1995; Decker 2013; Eaker 2004; McAvoy 1991; Mullins 2009; Radde 2016; Rimer 1999), assessing 61,182 participants, showed an increase in the uptake of screening between women who received printed material as a form of education compared to those in the control group (RR 1.23, 95% CI 1.05 to 1.44). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 94%) (Analysis 7.1, Comparison 7.1.1).

Education (miscellaneous) versus control

Meta‐analysis of four trials (Greene 1999; Mishra 2009; Taylor 2002; Thompson 2016), which assessed 915 participants, showed a higher uptake of screening in women in the education group compared to women in the control group (RR 1.50, 95% CI 1.17 to 1.93). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance was low (I² = 25%). Women were given an educational exercise (format unknown) (Greene 1999) or a variety of educational materials (Taylor 2002), aimed at increasing awareness of the cervical screening programme. The trial of Mishra 2009 consisted of three parts: (1) education booklets; (2) behavioural exercises; and (3) interactive group discussion sessions (Analysis 7.1, Comparison 7.1.2).

Education (face‐to‐face home visits) versus control

Meta‐analysis of three trials (McAvoy 1991; Sung 1997; Taylor 2002), assessing 1318 participants, showed an increase in screening in women who received face‐to‐face home visits as a form of education compared to those in the control group (RR 2.33, 95% CI: 1.04 to 5.23). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 79%) (Analysis 7.1, Comparison 7.1.3).

Education (electronic) versus control

The trial of Wright 2012 investigated the use of an electronic journal (eJournal) to increase uptake of cervical screening. An eJournal is an interactive electronic communication and information‐sharing tool. Those assigned to the intervention group received health maintenance reminders via the eJournal. Patients in the intervention group were more likely to have a Pap smear (25/38 versus 7/60, P < 0.001). However, when these results were validated using on‐treatment analysis to test whether the differences observed were associated with having viewed an eJournal, there was no longer an improvement in the Pap smear uptake (adjusted OR 1.74, P = 0.342). Krist 2012 also looked at the use of electronic health records to increase screening uptake, and also included an educational component. Patients enrolled onto the intervention group of this study had access to an interactive preventive health record (IPHR) that was linked to their electronic health record. The IPHR gave individualised preventive service recommendations. Alongside this, the IPHR explained the reasoning for the preventive service and referenced relevant details in the patient's history and included links to evidence‐based educational resources and decision aids. The difference between the intervention and control group in the increase in being up‐to‐date with cervical screening was 13.4% (P < 0.02) at 16 months.

The trials assessing any form of education versus control were of low quality (Table 3), therefore, there is low‐certainty evidence that education may increase the uptake of cervical screening.

Education versus other

The trial of Bowman 1995, which assessed 99 participants, showed little difference in the uptake of screening between women who received printed material as a form of education and those who received a health clinic invitation letter (RR 1.08, 95% CI: 0.45 to 2.61). Similarly there was no difference in the uptake of screening between printed material and GP invitation letters (RR 0.64, 95% CI: 0.31 to 1.32). The trial of Greene 1999, which assessed 98 participants, found little difference in the uptake of cervical screening between education (format unknown) aimed at increasing awareness of the cervical screening programme and enhanced risk assessment (RR 0.87, 95% CI: 0.63 to 1.21). The trial of McAvoy 1991, which assessed 482 participants, showed little difference in the uptake of screening between women who received printed material as a form of education and those shown educational videos or slides (RR 0.86, 95% CI: 0.64 to 1.14). The trial of Allen 2001, which assessed 2944 participants, showed no difference in the uptake of screening between women who worked in work sites with workshops aimed at increasing cervical screening led by peer health advisors and those in the non‐intervention group (RR 1.02, 95% CI: 0.99 to 1.05). The trials assessing this comparison provided very low‐certainty evidence.

Lay health outreach worker

Lay health outreach worker versus control

Meta‐analysis of eleven trials (Byrd 2013; Fang 2017; Jandorf 2008; Ma 2015; Nuno 2011; O'Brien 2010; Paskett 2011; Peterson 2012; Rosser 2015; Studts 2012; Taylor 2010), assessing 4330 participants, found that a lay outreach programme can increase the uptake of cervical screening (RR 2.30, 95% CI 1.44 to 3.65; low‐certainty evidence). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 96%) (Analysis 7.2).

7.2. Analysis.

Comparison 7: Education versus control, Outcome 2: Lay health outreach worker vs control

Lay health outreach worker workshop versus control

Meta‐analysis of seven trials (Byrd 2013; Fang 2017; Jandorf 2008; Ma 2015; Nuno 2011; O'Brien 2010; Rosser 2015), assessing 3413 participants, found that workshops run by lay health outreach workers can increase the uptake of cervical screening (RR 3.00, 95% CI 1.54 to 5.83). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 97%) (Analysis 7.2,, Comparison 7.2.1). The cluster trial of Larkey 2012 investigated two delivery methods of lay health workers to increase screening behaviour. Participants were randomised to cancer screen/prevention classes delivered individually or to social support groups over eight weeks. The trial concluded that screening behaviour was not different between the two groups, and that both produced a robust achievement of screening for previously non‐adherent participants.

Lay health outreach worker home visit versus control

One trial looked at the use of home visits to inform and educate women about cervical screening to increase its uptake (Studts 2012). Accounting for the random effect of church involvement, there was an increase in screening uptake in those who had home visits (19/173) compared to the control group (8/169), where the intervention group had over twice the odds of being screened (OR 2.73, 95% CI 1.08‐6.89). 

Lay health outreach worker home visit with telephone support versus control

Meta‐analysis of three trials (Paskett 2011; Peterson 2012; Taylor 2010), assessing 575 participants, found that if home visits by a lay health worker were followed with telephone support when compared with controls, the uptake of cervical screening was greater (RR 1.66, 95% CI 1.03 to 2.68). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent substantial heterogeneity (I² = 57%) (Analysis 7.2,, Comparison 7.2.3).

Lay health outreach worker and media education versus media education alone

In the cluster‐RCT of Mock 2007, 1005 Vietnamese‐American women were randomised into either a lay health worker (LHW) outreach plus media‐based education combined intervention or a media‐based education only (control). Each LHW used her social network to recruit 20 women before they were randomised. Over the programme period, 16% of women in the combined intervention group obtained a Pap test (increasing from 65.8% to 81.8%; P < 0.001) compared with 5.4% in the media‐only group (increasing from 70.1% to 75.5%; P < 0.001). The increase in the combined intervention group was significantly greater than that in the media‐only group (Z test: P = 0.001). Among women who at baseline had never had a Pap test, 46% of those in the combined intervention group obtained one during the programme period (P < 0.001) compared with 27% of those in the media‐only group (P < 0.001). Again, the increase was greater in the combined intervention group (Z test: P = 0.001). In the combined intervention group, 21.6% became up‐to‐date during the programme period (increasing from 45.7% to 67.3%; P < 0.001) compared with 4.8% in the media‐only group (increasing from 50.9% to 55.7%; P = 0.035). The increase in being up‐to‐date was also greater in the combined intervention group (Z test: P = 0.001). The benefit of a combination of LHW and media‐based education was also demonstrated by Thompson 2016, who randomised 443 Latina‐American women in one of three groups: (1) usual control, (2) intervention with a Spanish language video sent to participants homes to inform them of the importance of cervical screening (low‐intensity intervention), (3) intervention with a Spanish language video and home visit by a lay health worker (high‐intensity intervention). After a seven‐month follow‐up period, more women in the high‐intensity intervention had cervical screening compared to usual control (78/146 versus 20/147, P < 0.001), or compared to the low‐intensity intervention (78/146 versus 58/150, P < 0.01). There was no difference in screening uptake between the low‐intensity and control group.

The trials assessing the use of lay health outreach workers were inconsistently reported and comparisons were narratively reported. They provided low‐certainty evidence (Table 3) that education provided in various formats by lay health outreach workers may increase the uptake of cervical screening.

12‐week cancer screening education versus control

In the cluster‐RCT of Navarro 1995, 36 lay community workers (consejeras) were recruited and trained to conduct educational group sessions. Each consejera recruited approximately 14 peers from the community to participate in the programme. The consejeras were randomly assigned to either a 12‐week cancer screening intervention group or a control group. The authors reported that, although both groups increased Pap smear use, the increase was higher for the cancer intervention group than the control group. The difference approached statistical significance using participants as the unit of analysis (P = 0.10), but not when the consejera was the unit of analysis (P = 0.37).

Risk Factor Assessment

Enhanced risk assessment versus control

Meta‐analysis of two trials (Greene 1999; Kreuter 1996), assessing 145 participants, showed no significant difference in the uptake of screening between women who had an enhanced risk assessment and those in the control group (RR 1.52, 95% CI: 0.58 to 3.95; very low‐certainty evidence) (Analysis 9.1). The percentage of the variability in effect estimates that was due to heterogeneity rather than chance may represent considerable heterogeneity (I² = 87%). The results of the two trials differed markedly; the trial of Greene 1999 showed a significant increase in the number of women screened who received the enhanced risk assessment compared to control (RR 2.53, 95% CI: 1.42 to 4.51), whereas the Kreuter 1996 trial showed no significant difference between the two groups (RR 0.95, 95% CI: 0.64 to 1.42; very‐low certainty evidence).

9.1. Analysis.

Comparison 9: Enhanced risk assessment versus control, Outcome 1: Uptake of screening

Enhanced risk assessment versus other

The trial of Kreuter 1996, which analysed 70 participants, found no difference in the uptake of cervical screening between women who had an enhanced risk assessment and women who received a 'typical' risk assessment (RR 1.20, 95% CI: 0.79 to 1.81; very low‐certainty evidence) (Analysis 10.1).

10.1. Analysis.

Comparison 10: Enhanced risk assessment versus other, Outcome 1: Uptake of screening

Procedures

Access to health promotion nurse versus control

The trial of Robson 1989, which assessed 1407 participants, found an increase in the uptake of screening in women who had access to a health prevention nurse compared to those who did not (RR 1.18, 95% CI: 1.10 to 1.26; very low‐certainty evidence). There was substantial heterogeneity between trials (I² = 76%) (Analysis 11.1).

11.1. Analysis.

Comparison 11: Access to health promotion nurse versus control, Outcome 1: Uptake of screening

Photo comic book

Photo comic book verus placebo comic book

Only one trial (Risi 2004), assessing 658 participants, studied photo comic book use as a tool to promote cervical screening uptake. It found no difference in the uptake of cervical screening between women who had been exposed to the photo comic book aimed at promoting cervical screening uptake compared with women who been exposed to a placebo photo comic book (RR 0.96, 95% CI: 0.53 to 1.73; very low‐certainty evidence) (Analysis 12.1).

12.1. Analysis.

Comparison 12: Photo comic versus placebo comic, Outcome 1: Uptake of screening

Intensive recruitment

Intensive recruitment attempts versus control

One trial (Oscarsson 2007), with 800 participants, studied intensive recruitment. An increase in cervical screening uptake was found in women in the intensive recruitment intervention group compared with those in the control group (RR 1.59, 95% CI: 1.24 to 2.06; very low‐certainty evidence) (Analysis 13.1).

13.1. Analysis.

Comparison 13: Intensive recruitment attempts versus control, Outcome 1: Uptake of screening

Message framing

Only one trial (Rivers 2005), assessed message framing in the uptake of cervical screening in 441 participants. No differences in uptake were seen. Loss‐framed messages, whether prevention or detection‐phrased, and gain‐framed detection compared to loss‐framed detection messages both showed a non‐significant decrease in uptake (RR 0.79, 95% CI: 0.48 to 1.30). Gain‐framed messages, whether prevention or detection‐phrased, and gain‐framed detection compared to loss‐framed prevention messages both showed a non‐significant increase in uptake (RR 1.07, 95% CI 0.62 to 1.83) (Analysis 14.1). There was very low‐certainty evidence for this outcome comparison.

14.1. Analysis.

Comparison 14: Message framing, Outcome 1: Gain versus loss message framing (detection)

Economic

One trial assessed whether waiving the usual cervical screening fee would increase uptake. Alfonzo 2016 randomised 3124 participants within the national screening programme to either receive the standard screening invitation which would incur a fee, or a invitation offering screening free of charge. After a time period of 90 days, no difference in the uptake of screening was found (574/1562 intervention versus 612/1562 control, (RR 0.94, 95% CI: 0.86 to 1.03; very low‐certainty evidence) (Analysis 15.1).

15.1. Analysis.

Comparison 15: Economic: free screening test versus standard fee, Outcome 1: Uptake of screening

Informed uptake of cervical screening

None of the trials identified in this review reported informed consent to cervical screening.

Secondary outcomes

A summary of the data relating to secondary outcomes is presented in Table 4. These outcomes provided very low‐certainty evidence.

Booking of appointments

One trial, which assessed 273 participants, used the booking of appointments for screening as an outcome measure (Greene 1999). The trial population was randomly divided into three groups: usual care (women received general dietary and health information), cancer education (women received general information about cervical cancer risk factors and screening recommendations), and cognitive behavioural intervention (women received feedback about personal risk for cancer and engaged in a clinical interview to enhance self‐efficacy for preventive behaviour). Women in the usual care group were more likely to schedule an appointment for a Pap smear than those who received the cognitive behavioural intervention (usual care [79%] versus cognitive behavioural intervention [37%], P < 0.0001). Women in the usual care group were also more likely to attend without rescheduling the appointment (usual care [64%] versus cognitive behavioural intervention [35%], P < 0.001). The booking of appointments did not differ significantly between the women who received cancer education and those who received the cognitive behavioural intervention. It was difficult to assess the quality of this study as it was only published as an abstract and no further details were available.

Attitudes to screening

Six trials examined participants attitudes to Pap smear screening (Byles 1995; Jandorf 2008; Mishra 2009; O'Brien 2010; Rosser 2015; Thompson 2016). It was not possible to extract comparable data for meta‐analysis.

The Byles 1995 trial assessed the impact of two letter based interventions on the uptake of cervical screening in 3094 participants. Alongside this they also examined their attitudes to Pap smear screening and specifically their attitudes towards reminder letters. To assess the latter 1300 questionnaires were distributed between each intervention group. Of these only 384 questionnarires were returned with 213 from intervention Group 1 (simple letter) and 171 from Group 2 (letter, prompt cards and pamphlet ‐ designed to address aspects believed to be associated with poor screening rates). Within Group 1, 154 (72%) women reported receiving the letter and 147 (69%) women reported reading it. Within Group 2, 134 (78%) women reported receiving the letter and 128 (75%) of women reported having read it. Of the women receiving prompt cards a 100 (58%) women reported receiving them and of these only 7 (4%) women reported using them. Finally, within Group 2, 109 (64%) women reported receiving the pamphlet with 101 (59%) of women reporting reading them. Within these groups when attitudes to screening was assessed the majority of women felt pleased about receiving such a letter (Group 1: 118/151 women, 78%; Group 2: 89/132 women, 68%) with only 1 woman in Group 1 and 3 women in Group 2 being displeased about receiving such a letter. Moreover, when these women were asked if all women should receive such a letter, a resounding 98% (152/155) of women in Group 1 and 95% (124/130) of women in Group 2 said this should be done.

Mishra 2009 tested the effectiveness of a theory‐guided, culturally‐tailored cervical cancer education programme to increase Pap smear usage amongst Samoan women. This programme had three parts: (1) education booklets, (2) skill‐building and behavioural exercises, and (3) interactive group discussion sessions led by health educators; they were executed over three sessions/modules. The 398 participants were given a pre‐test and post‐test questionnaire to assess cervical cancer knowledge and attitudes. This was measured by items that addressed knowledge of risk factors, cervical cancer‐related beliefs, knowledge of aetiology, prevention and treatment, perceived efficacy of Pap smears and the perceived susceptibility to and severity of cervical cancer. The only significant difference between the intervention and control groups was the mean knowledge score, which was higher in the intervention group (0.44 versus 0.04, P < 0.05).

The O'Brien 2010 trial that assessed 120 participants was an RCT of a lay health worker‐led educational intervention in a Hispanic community. The intervention was two three‐hour workshops including between four and ten women‐led by lay health workers which covered content related to cervical cancer. Cervical cancer knowledge and self‐efficacy were measured by a questionnaire pre‐ and post‐intervention. Knowledge was measured on a 0‐6 scale and had questions covering the role of HPV, the epidemiology of cervical cancer, methods for prevention, screening recommendations and the meaning of a positive Pap smear. Cancer knowledge was greater in the intervention group at six months post‐intervention (P < 0.001). Exploratory multivariate logistic regression was applied, and only post‐intervention cervical cancer knowledge and assignment to intervention group were predictive of receiving a Pap smear during the six‐month follow‐up period.

The Jandorf 2008 trial assessed 847 participants and aimed to assess the effectiveness of a culturally customised programme to increase breast and cervical cancer screening in Latino immigrants in the United States. Participants were randomised to either a breast and cervical cancer education programme or a diabetes (control) education programme. Pre‐ and post‐test knowledge surveys were completed. Those receiving the intervention were significantly more likely to have increased knowledge scores (P < 0.0001).

Rosser 2015 assessed 419 participants and evaluated the impact of a 30‐minute health talk delivered by community health workers in rural Kenya on cervical cancer knowledge and attitudes in addition to the uptake of screening. Knowledge scores based on a pre‐talk survey and post‐talk survey at three months showed an increase in cervical cancer knowledge of 26.4% in the intervention group compared to 17.4% in the control group (P < 0.01).

The Thompson 2016 trial randomised 443 Latina‐American women into one of three groups: (1) usual control, (2) low‐intensity intervention with a Spanish language video sent to participants' homes to inform them of the importance of cervical screening, (3) high‐intensity intervention with a Spanish language video and visit by a lay health worker. In addition to the uptake of screening, this trial investigated whether these interventions would change participants' knowledge and attitudes towards cervical screening, and this was assessed by surveys before and after the intervention. Overall, there was an increase in knowledge in all study groups. This trial also performed a cost‐effectiveness analysis comparing the three arms. The low‐intensity intervention (video) cost an additional $15 per participant and the high‐intensity intervention cost an additional $82 per participant. Neither were cost‐effective.

Costs of the interventions

Eight trials presented cost data (Binstock 1997; Heranney 2011; Larkey 2012; McDowell 1989; Oscarsson 2007; Stein 2005; Thompson 2016; Vogt 2003).

The first trial used five different intervention groups (Binstock 1997). However, only those groups that used an intervention aimed at women (and not healthcare providers) were included in this review: telephone invitation, invitation letter, and a control group. The total estimated costs ($US) per intervention group were as follows: telephone invitation: $4282, invitation letter: $1918, memo to primary provider: $8933, medical record reminder: $1090 and control group (not stated). In terms of the uptake of screening tests, telephone intervention produced a greater increase (18.8%) compared with invitation letters (7.6%) or the control group.

The Heranney 2011 trial compared the efficacy and cost‐effectiveness of two different reminder strategies. Participants (N = 10,662) were randomised into either an invitation‐to‐screening letter or a telephone reminder. There was no significant difference in the uptake of cervical screening between the two groups. Furthermore, the global budget was higher for the telephone strategy (Euro 15,464 versus Euro 10,216) and investment of Euro 46 was necessary to convince a women by telephone to have a Pap smear compared to only Euro 33 by mail.

The Larkey 2012 trial assessed 1006 participants and compared lay health worker prevention classes delivered individually or within a social support group over a eight‐week period. The intervention costs (US$) per study arm were compared. Larkey 2012 found that screening behaviours were not different between the two groups. The costs to achieve one cancer screening in each group were: individual class: $862 to $1716 and social support group class: $263 to $517. The study concluded that although both interventions achieved screening, group‐based led interventions were more cost‐effective.

The McDowell 1989 trial used four different intervention groups but, again, not all of the interventions were aimed solely at women: some were aimed at healthcare providers. The following groups were considered in this review: GP letter invitation, telephone invitation, and control (usual care). The estimated costs ($US) per additional Pap smear performed as compared with usual care were: GP invitation letter: $14.23, telephone intervention: $11.75 (assuming a salary of $60 per hour) or $5.88 (assuming a salary of $30 per hour).

The trial of Stein 2005 used four different intervention groups: no intervention (control); telephone intervention; letter from Health Authority District Cervical Screening Commisioner on behalf of the National Cervical Screening Programme and letter from a well‐known journalist and broadcaster. Cost‐effectiveness analysis was performed. Average cost per attender was £145.12 for a telephone call, £14.29 for a letter from the commissioner and £37.14 for a letter from a celebrity.

The trial of Vogt 2003 had examined the costs of each smear gained in the intervention groups above the cost of a smear in the control (usual care) group. Cost‐effectiveness analysis showed that, for each additional Pap smear, the letter/additional reminder letter intervention cost $185, the phone/additional phone reminder intervention cost $305 and the letter/additional phone reminder intervention cost $1117.

Oscarsson 2007 compared intensive recruitment using multiple methods with a control group. The cost per smear was €66.87 and €16.63, respectively. Each additional smear obtained in the trial cost €151.36.

Discussion

Summary of main results

Overall, invitations, education, and lay health worker involvement were probably effective methods for increasing the uptake of cervical screening. Evidence regarding the effectiveness of other interventions such as economic incentives, procedural interventions (i.e. revealing the gender of the smear taker and using a health promotion nurse), counselling and risk factor assessment was limited.

The majority of cluster‐RCTs did not account for clustering or adequately report the number of clusters in the trial in order to estimate the design effect, so we did not selectively adjust the trials on this basis. It is unlikely reporting of these few trials would impact on the overall conclusions and robustness of the results.

Fourteen trials (Egawa‐Takata 2018; Erwin 2019; Firmino‐Machado 2018; Firmino‐Machado 2019; Han 2017; Kiran 2018; Kitchener 2018; Koc 2019; Kurt 2019; Nicolau 2017; Okuhara 2018; Tanjasiri 2019; Valdez 2018; Wong 2019) have been identified since the data collection and analyses were completed for this updated review and are awaiting classification. It is unlikely that their addition would affect the findings and quality of this review. They are referred to, where relevant, in this discussion and will be included in full in future updates of this review, if they meet the inclusion criteria.

Invitations

In general, there was low‐certainty evidence to suggest that invitation letters were effective at encouraging women to attend for cervical screening. Cervical screening programmes in the UK, Italy, Sweden and other countries already invite women to attend via a letter, with or without appointments, as part of their national call/recall system. However, the use of such systems in LMICs may be difficult to implement where issues of migration, literacy and access to remote areas may be of concern.

There was moderate‐certainty evidence that telephone invitations increased screening uptake, but it was unclear whether this practice was more effective than invitation letters. Telephone invitations are not routinely used in organised screening programmes such as that in the UK and would be even more difficult to implement in LMICs, where access to telephones may be an issue. Also, telephone invitations are more costly than invitation letters ($4,282 versus $1,918 (Binstock 1997); €4,306 versus €2,670 (Broberg 2014)). Heranney 2011 found that an extra investment of €46 was needed to convince a woman by telephone to have cervical screening versus €33 by mail. There was moderate‐certainty evidence to suggest that invitation letters with fixed appointments were more effective than invitations with open appointments.

Current practice in the UK and a number of other countries involves sending invitation letters both from GPs or Health Authorities, or both (NHS Information Centre 2010). The effectiveness of sending letters from different authority sources was evaluated in three trials (Bowman 1995; Segnan 1998; Stein 2005). These trials provided very low‐certainty evidence favouring GP or local authority letters over other sources. It is not possible to say definitively which approach was more effective, due to the limited evidence from good‐quality trials. No trial showed a cost‐effectiveness benefit of any alternative invitation method over the standard care (usually a letter). Given the effectiveness of invitation letters in encouraging uptake and given that no other intervention has been shown to be more cost‐effective, it remains appropriate to use existing invitation approaches. It must be recognised that for this approach to be effective, robust administrative procedures must be in place and a comprehensive, accurate, up‐to‐date register maintained.

The Okuhara 2018 trial was the only trial to examine the effect of improving written materials in terms of the ease at which people process the information and its effect on encouraging women to attend for cervical screening. The control group was sent a standard leaflet for cancer screening, whereas the intervention group received materials with improved perceptual fluency (e.g. legibility), linguistic fluency (e.g. readability), retrieval fluency (e.g. reducing the amount of information) and imagery fluency (e.g. having recipients imagine future behaviour and events).  It found that the screening rates in the intervention group were significantly greater than those of the control group (29.4% versus 14.2%, P = 0.007).

The advancement of technology has led to the possibility of the standard invitation letter being replaced by SMS messages, emails and other forms of electronic communication via social media. Only two studies included in this review (Peitzmeier 2016; Rashid 2013) addressed these effects on the uptake of cervical screening. Peitzmeier 2016 randomised women to receive invitations to cervical screening via different forms of communications including an email reminder and found that it did not increase the uptake of cervical screening (RR 1.18, 95% 0.67 to 2.09). Rashid 2013 also randomised women to receive invitations to cervical screening via different forms of communication including a SMS message. In this trial, those women who received an SMS message were more likely to attend for cervical screening (RR 2.24, 95% CI 1.67 to 3.00). Since the trials of Peitzmeier 2016 and Rashid 2013, Erwin 2019 randomised women into three groups: (1) to receive 15 SMS promoting behaviour change, (2) to receive 15 SMS promoting behaviour change and an eVoucher for transportation to the clinic, or (3) to receive a single SMS informing them of the nearest clinic (control); they found that at 60 days there was a greater uptake of cervical screening in both intervention groups (SMS promoting behaviour change OR 3.0, 95% CI: 1.5 to 6.2; SMS and eVoucher OR 4.7, 95% CI: 2.9 to 7.4). Furthermore, a trial by Firmino‐Machado 2018 investigated the effect of a three‐step invitation to screening where participants randomised to the intervention group received automated text messages/phone call (step 1), manual phone calls (step 2) and face‐to‐face interviews (step 3). This trial found that the adherence to screening was higher in those assigned to the intervention (51.2% versus 34%, P < 0.001). This suggests that SMS messages may play a role in increasing the uptake of cervical screening and further research is needed for the use of this intervention.

The Egawa‐Takata 2018 trial was the only trial to assess the effectiveness of maternal education to increase the uptake of cervical screening. Participants were sent either a personalised daughter‐directed reminder leaflet for cervical screening or a daughter‐and‐mother combination package which contained the same daughter‐directed reminder leaflet and an additional leaflet for the mother, which requested that she recommend that her daughter undergo cervical screening. A higher uptake of screening was found in the group receiving the daughter‐and‐mother combination package (11% versus 9%, P = 0.0049). This trial was the only trial to use maternal education and further research is needed to assess whether education through family is effective.

Education

The overall consensus from the trials examining educational interventions was in favour of the intervention over the no intervention or usual care control. However, heterogeneity between the trials limited the statistical pooling of data. Amongst ethnic minority groups, there was low‐certainty evidence to support the use of lay members of the community in presenting culturally‐tailored information. This was effective as either workshops in the community, home visits or telephone support, and may be of relevance in LMICs where remote areas and literacy may be an issue. One trial did look at the use of community health workers to deliver health talks on cervical cancer knowledge, attitudes and screening in rural Kenya, and although this was successful at increasing participants knowledge of cervical cancer, it did not increase screening uptake (Rosser 2015). This differs from the majority of the other trials making use of community/lay health workers within ethnic minorities which were mainly based within the US.

More recent trials (Han 2017; Koc 2019; Kurt 2019; Nicolau 2017; Tanjasiri 2019) also add to the consensus that screening uptake is increased by any educational intervention when compared to no intervention. The Han 2017 trial found that an individually tailored cancer screening brochure followed by community health worker‐led health literacy training and monthly telephone counselling with navigation assistance increased the uptake of cervical screening amongst Korean‐American women (OR 13.3, 95% CI 7.9 to 22.3). The Kurt 2019 trial also favoured education over no education and found higher rates of screening uptake in participants who received one‐to‐one training accompanied by an educational brochure (66/118 participants) or an educational brochure only (38/119 participants) when compared to a control group who received an invitation only (30/119 participants), P < 0.001. The Nicolau 2017 trial also examined educational interventions and was unique in that it assessed the efficacy of behavioural and educational interventions undertaken by telephone (educational telephone call versus telephone reminder versus control). Higher rates of screening were found in the intervention group. The Valdez 2018 trial differed in its conclusion and found that educational interventions, although increasing participants knowledge of cervical cancer, did not necessarily increase uptake of screening.

Educational materials are likely to be important in increasing informed uptake, providing they cover all aspects of the screening process. No trials have attempted to measure the effectiveness of interventions at increasing the informed uptake of Pap smears

Counselling

Women given counselling to encourage attendance at a cervical screening programme had a higher uptake of screening. This finding was the result of three trials (Dietrich 2006; Rimer 1999; Ward 1991), and further research in this area is needed as these trials provided only very low‐certainty evidence.

Overall completeness and applicability of evidence

Future intervention trials should aim to minimise barriers to screening uptake and to increase understanding of the likely benefits, limitations and potential harm of screening. Trials should include a measure of knowledge and whether the information provided is used in the decision‐making process. Just as an intervention to increase uptake may be ineffective, an intervention to increase informed uptake might also be ineffective. For example, it should not be assumed that giving a leaflet on the risks and benefits of screening will necessarily increase informed uptake. It may be that some interventions, which are effective for increasing uptake (such as appointments), are not effective at increasing informed uptake, and the opposite may also be true. Similarly, interventions which are effective in developed countries may not be as effective in LMICs or may present problems in terms of their implementation. At present, the evidence regarding the effectiveness of interventions is dominated by studies set in higher income countries (HICs) and there is a need for research which is likely to be more applicable to LMICs. Research into screening uptake including the uptake of Pap smears is still expanding with new studies being published each year. However, at present, there is very little research relevant to LMICs and it is difficult to state with any degree of certainty how effective the interventions discussed in this review will be in such settings. Lay health worker involvement appears to increase the uptake of screening in ethnic minorities in developed countries (although the level of evidence is of low quality); whether this is able to be translated to LMICs needs further research.

Quality of the evidence

This review and its findings are very much dependent on the validity and quality of the trials reported. The risk of bias of the individual trials included in the review was assessed independently by two review authors using pre‐defined checklists. Although a number of the trials were of reasonable quality, a number were at high risk of bias and suffered from methodological problems and inadequate reporting. We downgraded the quality of evidence because of an unclear or high risk of bias with regards to allocation concealment, blinding, incomplete outcome data, and other biases.

The certainty of the evidence (GRADE Working Group 2004) for uptake of cervical screening ranged from very low to moderate across all comparisons. In most cases, evidence was downgraded due to small numbers of participants, resulting in imprecision; and high or unclear risk of bias, particularly around whether it was confirmed that trials were truly randomised and conducted adequately. There was also substantial or considerate statistical heterogeneity in many of the analyses, which made it difficult to get a feel for the overall pooled magnitude of effects, especially when there was inconsistency across trials.

There was moderate‐certainty evidence to suggest that invitations appear to be effective methods of increasing uptake compared to control. Additional analyses, ranging from low‐ to moderate‐certainty evidence, suggested that invitations that were more personalised in the form of personal invitation, GP invitation letter or letter with fixed appointment appeared to be the more successful invitational approaches in terms of uptake of screening.

There was also low‐certainty evidence to support the use of educational materials and lay health worker involvement. Other less widely reported interventions included counselling, risk factor assessment, access to health promotion nurse, photo comic book, intensive recruitment and message framing. It was difficult to deduce any meaningful conclusions from comparisons involving these interventions due to sparse data and low‐certainty evidence. However, having access to a health promotion nurse and attempts at intensive recruitment may have increased uptake.

Other less widely reported primary outcomes and all secondary outcomes provided low‐ or very low‐certainty evidence for all comparisons.

Of note, several excluded trials would have been eligible for inclusion, if proper randomisation procedures had been followed, rather than using quasi‐randomisation techniques. A number of trials randomised women without first assessing their eligibility, so leading to the exclusion of large numbers of women post‐randomisation. Many trials also failed to use appropriate analyses such as intention to intervene analyses and the appropriate consideration of the effects of clustering in cluster‐randomised trials.

Even though risk ratios were calculated in most of the RCTs included in this review, the pooling of data was restricted because of clinical heterogeneity. Of the meta‐analyses that could be performed, there was quite often significant statistical heterogeneity present also. The conclusions and implications for practice are primarily based on those interventions for which there was evidence from several RCTs, i.e. invitations, educational materials, and the use of lay health outreach workers. However, issues of heterogeneity and study quality should be borne in mind when interpreting these findings.

Potential biases in the review process

The comprehensive search strategy used in the review is likely to have located most of the published trials and our thorough search of the grey literature meant that every attempt had been made to obtain data from unpublished trials. Decisions on the relevance of trials were made by two reviewers in a two‐stage sifting process. Titles and abstracts of the search results were initially searched, then full‐text articles were sifted from the potentially relevant papers identified from the initial sift. In cases of disagreement, a third reviewer was called to decide on disputed trials. We restricted the included studies to RCTs as they provide the strongest level of evidence available. We excluded quasi‐randomised trials and other non‐RCTs, hence, we have attempted to reduce bias in the results of this review.

The greatest threat to the validity of the review is likely to be the possibility of publication bias, i.e. studies that did not find the treatment to have been effective may not have been published. We were unable to assess this possibility for most outcomes, as most comparisons of the interventions were restricted to either a meta‐analysis of only a low number of trials or single‐trial analyses. It is acknowledged that although abstracts, full‐text articles and unpublished reports were found (through contacting experts in the field and searching the grey literature and reference lists), some may have been missed, but this risk of publication bias is likely to be minimal.

The majority of cluster‐RCTs did not account for clustering or adequately report the number of clusters in the trial in order to estimate the design effect. We planned to use an external estimate of the ICC to estimate a design effect to inflate the variance of the effect estimate in cluster‐RCTs that did not account for clustering. We obtained values of the ICC that ranged from 0.02 to 0.29 (Hade 2010), but sensitivity analyses were not plausible given the high number of included trials and multiple thresholds of the ICC that could be used. However, it is highly unlikely that selectively adjusting the trials with adequate reporting of clustering would make much difference to the overall conclusions and robustness of the results. 

Agreements and disagreements with other studies or reviews

A meta‐analysis performed by Tseng 2001 showed that women who received a reminder letter to attend for cervical screening were significantly more likely to attend than those who received no intervention. This is in agreement with the findings of this updated Cochrane review. A meta‐analysis by Yabroff 2003 agrees with the findings of this review, that letter reminders and telephone reminders are effective at increasing cervical screening uptake.

More recent reviews, which have focused on subgroups of the general population, have less conclusive results. Albrow 2014 performed a systematic review of interventions to improve cervical screening uptake in women less than 35 years. This review included four studies and concluded that there was insufficient evidence to determine which interventions are effective in this age group. It did suggest that letters may increase uptake; which is in agreement with this current updated Cochrane review.

Lu 2012 looked at interventions to increase breast and cervical cancer screening uptake among Asian women. This review included 11 studies that reported on the uptake of cervical screening following an intervention. The review were not able to arrive at a conclusive or generalisable conclusion on the effectiveness of any one particular intervention because of the heterogeneity between studies. There was evidence to suggest that community‐based group education with culturally sensitive educational materials and outreach materials, media campaigns, home visits and mail campaigns with follow‐up phone calls may increase the uptake of screening. This is all in agreement with this updated Cochrane review.

Authors' conclusions

Implications for practice.

There was low‐ to moderate‐certainty evidence to support the use of invitation letters, lay health worker involvement and educational interventions to increase the uptake of cervical screening. It is unclear which educational methods (i.e. printed, video/slide or face‐to‐face presentations) are most effective. There was no evidence on which to base implications for practice regarding the informed uptake of cervical screening. Overall, these findings relate to screening in developed countries and their relevance to LMICs is unclear.

Overall, the low‐ and very low‐certainty evidence identified for many of the comparisons make it difficult to make inferences as to which interventions were best, with exception of invitational interventions where there appeared to be more reliable evidence.

Implications for research.

The following implications are likely to be relevant to screening in developed countries:

• Invitations and educational materials appear to be effective at increasing uptake of cervical screening. Further research into the relative effectiveness and cost‐effectiveness of these interventions would help to inform decision‐making. In particular, it is unclear which types of educational intervention are the most effective.

• Further research is required to determine the effectiveness of promising interventions, such as revealing in an invitation letter the gender of the smear taker.

• Future research is needed to address how screening programmes have developed as technology has advanced. The Internet and social media may be a great source of health promotion, though research into this may be limited by data protection and privacy laws. Further research is needed to investigate the effect of SMS message invitations/reminders.

• When designing and reporting future trials, researchers should pay particular attention to the following issues: the use of an adequate method of randomisation, the blinding of those assessing study outcome measures, adequate concealment of treatment allocation, adequate follow‐up of all participants included in the initial randomisation process, selective reporting of outcomes, and the use of appropriate analyses, particularly in the case of cluster‐RCTs. Researchers should also try to ensure the enrolment of adequate numbers of eligible participants and interventions should be reported in sufficient detail.

• A group of women who attend for smears that are particularly at risk are those with inadequate or abnormal smears. Further research should examine the effectiveness of various methods to ensure adequate attendance at follow‐up for abnormal smears.

Implications for research in LMICs:

• A concentrated effort should be made to conduct good‐quality trials in LMICs. In these areas, there is usually no national screening programme, cervical screening is often non‐existent, and Pap smear or liquid‐based cytology are not practical. In such situations, other screening modalities such as visual inspection with acetic acid have been shown to be acceptable, feasible and safe (Sankaranarayanan 2004) and should be considered. The effectiveness of lay health worker involvement should also be assessed in these communities.

What's new

Date	Event	Description
13 May 2021	New citation required but conclusions have not changed	Thirty‐one additional studies included which support the original findings of the review
8 June 2020	New search has been performed	Search updated

History

Protocol first published: Issue 4, 2000
Review first published: Issue 3, 2002

Date	Event	Description
29 March 2011	New citation required but conclusions have not changed	New authors added and text amendments.
29 March 2011	New search has been performed	Changes to text and new authors added.
15 May 2002	New citation required and conclusions have changed	Substantive amendment

Appendices

Appendix 1. Updated CENTRAL search strategy

CENTRAL 

1. MeSH descriptor Vaginal Smears explode all trees

2. vagina* near/5 smear*

3. pap* near/5 (test* or smear*)

4. cervi* near/5 (smear* or screen*)

5. (cytology or cytobrush) and cervi*

6. (#1 OR #2 OR #3 OR #4 OR #5)

7. satisf* or dropout* or drop out

8. compliance or complie* or comply*

9. encourage* or improve* or improving or increas* or promot*

10. uptake or particip* or nonattend*

11. accept* or attend* or attitude* or utilisation or utilization

12. refus* or respond* or reluctan* or nonrespond*

13. (#7 OR #8 OR #9 OR #10 OR #11 OR #12)

14. (#6 AND #13)

Appendix 2. Updated MEDLINE search strategy

Medline via Ovid

1. Vaginal Smears/

2. (vagina* adj5 smear*).mp.

3. (pap* adj5 (test* or smear*)).mp.

4. (cervi* adj5 (smear* or screen*)).mp.

5. ((cytology or cytobrush) and cervi*).mp.

6. 1 or 2 or 3 or 4 or 5

7. exp "Patient Acceptance of Health Care"/

8. (satisf* or dropout* or drop out).mp.

9. (compliance or complie* or comply*).mp.

10. (encourage* or improve* or improving or increas* or promot*).mp.

11. (uptake or particip* or nonattend*).mp.

12. (accept* or attend* or attitude* or utilisation or utilization).mp.

13. (refus* or respon* or reluctan* or nonrespon*).mp.

14. 7 or 8 or 9 or 10 or 11 or 12 or 13

15. 6 and 14

16. randomized controlled trial.pt.

17. controlled clinical trial.pt.

18. randomized.ab.

19. randomly.ab.

20. trial.ab.

21. groups.ab.

22. 16 or 17 or 18 or 19 or 20 or 21

23. 15 and 22

key: mp=title, original title, abstract, name of substance word, subject heading word, pt=publication type, fs=floating subheading

Appendix 3. Updated Embase search strategy

Embase via Ovid

1. exp Vagina Smear/

2. (vagina* adj5 smear*).mp.

3. (pap* adj5 (test* or smear*)).mp.

4. (cervi* adj5 (smear* or screen*)).mp.

5. ((cytology or cytobrush) and cervi*).mp.

6. 1 or 2 or 3 or 4 or 5

7. exp Patient Attitude/

8. (satisf* or dropout* or drop out).mp.

9. (compliance or complie* or comply*).mp.

10. (encourage* or improve* or improving or increas* or promot*).mp.

11. (uptake or particip* or nonattend*).mp.

12. (accept* or attend* or attitude* or utilisation or utilization).mp.

13. (refus* or respon* or reluctan* or nonrespon*).mp.

14. 7 or 8 or 9 or 10 or 11 or 12 or 13

15. 6 and 14

16. exp controlled clinical trial/

17. randomized.ab.

18. randomly.ab.

19. trial.ab.

20. groups.ab.

21. 16 or 17 or 18 or 19 or 20

22. 15 and 21

key: mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer name, ab=abstract

Data and analyses

Comparison 1. Invitation versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Uptake of screening	24	141391	Risk Ratio (IV, Random, 95% CI)	1.71 [1.49, 1.96]
1.1.1 Invitation letter vs control	18	118911	Risk Ratio (IV, Random, 95% CI)	1.56 [1.32, 1.83]
1.1.2 Telephone invitation vs control	7	10174	Risk Ratio (IV, Random, 95% CI)	1.95 [1.65, 2.30]
1.1.3 Face to face invitation vs control	1	121	Risk Ratio (IV, Random, 95% CI)	9.15 [0.50, 166.30]
1.1.4 Letter with open invitation to make appointment vs control	4	2998	Risk Ratio (IV, Random, 95% CI)	1.61 [1.15, 2.26]
1.1.5 Letter with fixed appointment vs control	1	177	Risk Ratio (IV, Random, 95% CI)	1.80 [1.04, 3.11]
1.1.6 Letter invitation with telephone/email follow up vs control	4	8059	Risk Ratio (IV, Random, 95% CI)	2.19 [1.39, 3.44]
1.1.7 Celebrity letter invitation vs control	1	316	Risk Ratio (IV, Random, 95% CI)	2.15 [0.25, 18.15]
1.1.8 SMS vs control	1	376	Risk Ratio (IV, Random, 95% CI)	2.24 [1.67, 3.00]
1.1.9 Email vs control	1	259	Risk Ratio (IV, Random, 95% CI)	1.18 [0.67, 2.09]

Comparison 2. GP invitation letter versus invitation letter from other authority sources.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Uptake of screening	2		Risk Ratio (IV, Random, 95% CI)	Totals not selected
2.1.1 GP invitation letter vs health clinic invitation letter	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
2.1.2 GP invitation letter vs invitation letter from programme coordinator	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected

Comparison 3. Personal invitation versus invitation letter.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Uptake of screening	5	27663	Risk Ratio (IV, Random, 95% CI)	1.32 [1.11, 1.56]
3.1.1 Telephone invitation vs invitation letter	3	12561	Risk Ratio (IV, Random, 95% CI)	1.21 [1.05, 1.40]
3.1.2 Face‐to‐face invitation vs invitation letter	1	123	Risk Ratio (IV, Random, 95% CI)	2.10 [0.40, 11.05]
3.1.3 Targeted letter vs invitation letter	1	14979	Risk Ratio (IV, Random, 95% CI)	1.51 [1.40, 1.63]

Comparison 4. Letter with fixed appointment versus letter with open invitation to make an appointment.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Uptake of screening	5	5742	Risk Ratio (IV, Random, 95% CI)	1.61 [1.48, 1.75]

Comparison 5. Counselling versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Uptake of screening	2	393	Risk Ratio (IV, Random, 95% CI)	1.23 [1.04, 1.45]
5.1.1 Face‐to‐face counselling vs control	1	184	Risk Ratio (IV, Random, 95% CI)	1.23 [0.98, 1.55]
5.1.2 Telephone counselling vs control	1	209	Risk Ratio (IV, Random, 95% CI)	1.22 [0.97, 1.55]

Comparison 6. Counselling versus other.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
6.1 Uptake of screening	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
6.1.1 Telephone counselling vs provider prompts	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected

6.1. Analysis.

Comparison 6: Counselling versus other, Outcome 1: Uptake of screening

Comparison 7. Education versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
7.1 Uptake of screening	13	63415	Risk Ratio (IV, Random, 95% CI)	1.35 [1.18, 1.54]
7.1.1 Education (printed material) vs control	8	61182	Risk Ratio (IV, Random, 95% CI)	1.23 [1.05, 1.44]
7.1.2 Education (miscellaneous) vs control	4	915	Risk Ratio (IV, Random, 95% CI)	1.50 [1.17, 1.93]
7.1.3 Education (face‐to‐face home visits) vs control	3	1318	Risk Ratio (IV, Random, 95% CI)	2.33 [1.04, 5.23]
7.2 Lay health outreach worker vs control	11	4330	Risk Ratio (IV, Random, 95% CI)	2.30 [1.44, 3.65]
7.2.1 Lay health outreach worker workshop vs control	7	3413	Risk Ratio (IV, Random, 95% CI)	3.00 [1.54, 5.83]
7.2.2 Lay health outreach worker home visit vs control	1	342	Risk Ratio (IV, Random, 95% CI)	0.79 [0.55, 1.14]
7.2.3 Lay health outreach worker home visit and telephone support vs control	3	575	Risk Ratio (IV, Random, 95% CI)	1.66 [1.03, 2.68]

Comparison 8. Education versus other.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
8.1 Lay health outreach worker and media education vs media education	2		Risk Ratio (IV, Random, 95% CI)	Subtotals only
8.2 Uptake of screening	4		Risk Ratio (IV, Random, 95% CI)	Totals not selected
8.2.1 Education (printed material) vs health clinic invitation letter	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
8.2.2 Education (printed material) vs GP invitation letter	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
8.2.3 Education (format unknown) vs enhanced risk assessment	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
8.2.4 Education (printed material) vs education (video/slide)	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
8.2.5 Intensive peer health advice vs other	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected
8.3 Individual education vs social support group	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

8.1. Analysis.

Comparison 8: Education versus other, Outcome 1: Lay health outreach worker and media education vs media education

8.2. Analysis.

Comparison 8: Education versus other, Outcome 2: Uptake of screening

8.3. Analysis.

Comparison 8: Education versus other, Outcome 3: Individual education vs social support group

Comparison 9. Enhanced risk assessment versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
9.1 Uptake of screening	2	145	Risk Ratio (IV, Random, 95% CI)	1.52 [0.58, 3.95]

Comparison 10. Enhanced risk assessment versus other.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
10.1 Uptake of screening	1		Risk Ratio (IV, Random, 95% CI)	Totals not selected

Comparison 11. Access to health promotion nurse versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
11.1 Uptake of screening	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

Comparison 12. Photo comic versus placebo comic.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
12.1 Uptake of screening	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

Comparison 13. Intensive recruitment attempts versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
13.1 Uptake of screening	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

Comparison 14. Message framing.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
14.1 Gain versus loss message framing (detection)	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only
14.2 Gain versus loss message framing (prevention)	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only
14.3 Gain message framing: prevention vs detection	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only
14.4 Loss message framing: prevention vs detection	1		Risk Ratio (IV, Random, 95% CI)	Subtotals only

14.2. Analysis.

Comparison 14: Message framing, Outcome 2: Gain versus loss message framing (prevention)

14.3. Analysis.

Comparison 14: Message framing, Outcome 3: Gain message framing: prevention vs detection

14.4. Analysis.

Comparison 14: Message framing, Outcome 4: Loss message framing: prevention vs detection

Comparison 15. Economic: free screening test versus standard fee.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
15.1 Uptake of screening	1	3124	Risk Ratio (IV, Fixed, 95% CI)	0.94 [0.86, 1.03]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Abdullah 2013a.

Study characteristics
Methods	Design ‐ cluster RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 24 weeks
Participants	Country ‐ Malaysia Setting ‐ secondary schools in Kuala Lumpa Initial screening status ‐ overdue Ten secondary schools in Kuala Lumpa were randomised to either the intervention or control arm.
Interventions	1. Intervention ‐ personally distributed invitation letter and information pamphlet, followed by a telephone reminder 2. Control ‐ standard care
Outcomes	Uptake of Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation method
Allocation concealment (selection bias)	Low risk	All teachers at the same school assigned to the same group. Randomisation revealed after recruitment of final school to ensure concealment of allocation
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unclear if blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 98.7% (398/403)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether a risk of bias existed

Acera 2017.

Study characteristics
Methods	Design ‐ the CRICERVA study was a cluster‐randomised community‐based clinical trial Follow‐up ‐ not specified
Participants	Country ‐ Spain Setting ‐ primary healthcare services in Barcelona Initial screening status ‐ overdue Women were recruited from 4 different participating centres. Each participating centre was assigned one study arm. Inclusion criteria ‐ 30‐70 year olds
Interventions	1. Routine protocol 2. Personalised letter 3. Personalised letter and information leaflet 4. Personalised letter, information leaflet and a telephone reminder
Outcomes	Uptake in cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Unknown if assessors were blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	16% lost to follow‐up in each group
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether a risk of bias existed

Adab 2003.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ significantly higher proportion of non‐white women in the intervention group. Otherwise no significant differences between study groups Follow‐up ‐ Nil
Participants	Country ‐ UK Setting ‐ 3 general practices in Birmingham Initial screening status ‐ any 300 women attending their GP practice Inclusion criteria ‐ aged 20 to 64 years Exclusion criteria ‐ incomplete questionnaire
Interventions	1. Control leaflet based on that produced by National Health Service Cervical Screening Programme, though with references to "cervical" cancer, "cervical" screening or "smear test" removed 2. Intervention leaflet. As above, with additional information on average individual risk of cervical cancer, possibility of false positive/negative results, uncertainties attached to screening process, the absolute benefit associated with the screening and the cost of the process to the NHS
Outcomes	Self‐reporting of "willingness to have study screening test."
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Computer‐generated list of random numbers"
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Participants were blinded. It is unclear if outcome assessors were blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 91% (274/300)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Alfonzo 2016.

Study characteristics
Methods	Design ‐ single‐blinded randomised controlled trial within the regular cervical cancer screening programme Baseline characteristics ‐ no statistically significant difference in age Follow‐up ‐ 90 days
Participants	Country ‐ Sweden Setting ‐ cervical cancer screening programme Initial screening status ‐ due, overdue Inclusion criteria ‐ all women aged 23‐63 years consecutively invited for planned screening according to the standard routine
Interventions	1. Screening invitation stating that the test was free (intervention) 2. Screening invitation stating that it cost 100 SEK (control)
Outcomes	Uptake of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer randomisation
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Single‐blinded, "the midwives performing the screen were unaware". It was unclear as to whether the participants were blinded.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Unclear
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether a risk of bias existed

Allen 2001.

Study characteristics
Methods	Design ‐ cluster RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 3 years
Participants	Country ‐ US Setting ‐ workplace Inclusion criteria ‐ aged 50 years or older
Interventions	1. Workplace at worksites led by trained peer health advisors, n = 1512 2. No workshops, n = 1431
Outcomes	Pap smear uptake ‐ self‐reported
Notes	Intervention lasted 16 months.  Non‐intervention group were provided with skills and resources to replicate intervention programme.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 66% (2795/4253)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Binstock 1997.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 1 year
Participants	Country ‐ USA Setting ‐ HMO Initial screening status ‐ overdue Eligible participants were identified from the medical records of the Kaiser Permanente Health Plan, South California Region (HMO). Half of those eligible (n = 7630) were included in the final analysis. Inclusion criteria ‐ aged 25 to 49 years; enrolled in HMO for at least 3 years; likely to seek outpatient care at one of the three medical centres Exclusion criteria ‐ Pap smear within the last 3 years
Interventions	1. Telephone call, n = 1526 2. Letter, n = 1526 3. Memo to woman's primary provider, n = 1526 4. Chart reminder affixed to outside of woman's medical record, n = 1526 5. Control group, n = 1526
Outcomes	Pap smear uptake and costs determined by administrative records
Notes	No details were provided as to the selection criteria for half of the women who were entered into the study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 7630/7630 (100%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Bowman 1995.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 6 months
Participants	Country ‐ Australia Setting ‐ General practice Initial screening status ‐ overdue Over 7000 potentially eligible women in an Australian community were identified by a random household survey (developed by the Australian Bureau of Statistics) Inclusion criteria ‐ aged 18 to 70 years Exclusion criteria ‐ not sexually active; could not speak English; infirm; not at home when contacted; hysterectomy
Interventions	1. GP reminder letter, n = 255 2. Women's health clinic invitation, n = 220 3. Pamphlet, n = 219 4. Control group (not stated), n = 219
Outcomes	Pap smear uptake determined by administrative records
Notes	Comparison of self‐reported uptake and administrative records of uptake indicated that women were very accurate in their self‐report of screening when it had actually taken place, but inaccurate in almost a quarter of instances when they stated that it had not occurred
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Low risk	"Interviewers were unaware of which group in the study women had been assigned to".
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 659/913 (72%) GP reminder letter: 178/255 (70%) women were analysed ("Thirty‐five women initially assigned to this group were excluded because of the non‐participation of their GPs"). Women's health clinic invitation: 164/220 (75%) women were analysed. Pamphlet: 162/219 (74%) women were analysed. Control group: 155/219 (71%) women were analysed.
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Broberg 2013.

Study characteristics
Methods	Design ‐ RCT Baseline characteristics ‐ distribution of age and Pap smear history was similar between groups Follow‐up ‐ 12 months
Participants	Country ‐ Sweden Setting ‐ conducted in the context of the Swedish population‐based screening programme (71 antenatal health clinics in Western Sweden) Initial screening status ‐ overdue
Interventions	1. Telephone call to offer appointment for a smear test (up to 10 attempts made) 2. HPV self‐test (not reported)) 3. Routine care (control)
Outcomes	Uptake of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Statistical analysis software 9.2; the Plan procedure was used to select and randomise women into parallel groups.
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	Telephone arm: % analysed: 55.7% (1176/2110)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Buehler 1997.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 2 months and 6 months
Participants	Country ‐ Canada Setting ‐ family medicine clinic Initial screening status ‐ due Random sample of 441 women listed as patients of two clinics (one urban and one rural) affiliated with the Memorial University of Newfoundland Inclusion criteria ‐ 18 to 69 years Exclusion criteria ‐ Pap smear in past 3 years; hysterectomy; moved or had records with clerical errors
Interventions	1. Personal letter and reminder letter 4 weeks later, n = 221 2. Control group received no letter, n = 220
Outcomes	Pap smear uptake determined by administrative records
Notes	Sample size calculations did not take into account the lag time between taking tests and registering tests, which could and did cause the loss of participants.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Of the 1302 ... eligible women, we randomly selected 650 using computer‐generated numbers". From the CONSORT flow diagram, it appeared that the 441 women participating in the trial after exclusions were randomised in a similar way.
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 88% (386/441) By treatment arm: Personal letter and reminder letter: 178/221 (81%) Control: 208/220 (95%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Burack 1998.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 1 year Before randomisation, women were excluded if their last smear was abnormal/insufficient for cytology (n = 4708); 140 were excluded from the patient reminder intervention as they discontinued HMO membership; 2055 were excluded from the physician reminder intervention as they did not attend their physician appointment.
Participants	Country ‐ USA Setting ‐ HMO Initial screening status ‐ due Women were recruited from five HMO sites enrolled in year 1. Only three of these sites enrolled in year 2. 5801 women were randomised to physician reminder/no reminder. During a second later round of randomisation (patient reminder vs no reminder), further women were excluded. Inclusion criteria ‐ at least 40 years old; HMO member; visited one of the primary care study sites in Detroit, Michigan, USA Exclusion criteria ‐ previous abnormal or insufficient Pap smear
Interventions	1. An invitation letter reminding women that they were due for a Pap smear (964 analysed) 2. Reminders for both physician and participants (960 analysed) 3. Reminders for the physicians (960 analysed) 4. Control (no reminder to either physicians or participants (964 analysed)) 5801 women were randomised but the publication only gave a breakdown of women analysed by treatment arm.
Outcomes	Pap smear uptake determined by administrative records
Notes	Unclear methodology. Two‐stage randomisation and large numbers of exclusions after first randomisation. Not clear how many women were originally randomised to each of the four study groups
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"A two–stage randomization procedure was used ... At the beginning of the study period, the 5801 women ... were randomly assigned using a site specific, stratified randomization procedure to receive or not receive the physician reminder intervention. Strata were defined by age, previous Pap smear use, and number of HMO visits in the preceding year. To avoid overloading the clinics, a separate randomization was carried out to assign women to patient reminder intervention. Women who remained eligible were selected and randomized to patient reminder intervention on a weekly basis in groups of 156".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 66% (3848/5801)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Burack 2003.

Study characteristics
Methods	Design ‐ Cluster‐RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 1 year
Participants	Country ‐ US, Detroit. Setting ‐ HMO Initial screening status ‐ due Inclusion criteria ‐ aged 40 years or older and had visited a primary car provider at one of the study sites during the two years preceding the intervention period Exclusion criteria ‐ not reported
Interventions	Cluster‐RCT
Outcomes	1. Sent reminders for Pap smear and mammogram: n = 1243 2. Sent reminders for mammogram only: n = 1228
Notes	Pap smear uptake? from study site records
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (2471/2471) However, loss of participants who may have moved out of area, not received reminder etc. is not clear.
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Byles 1994.

Study characteristics
Methods	Design ‐ RCT (cluster) Baseline comparability ‐ study regions matched on census data Follow‐up ‐ 3 months (TV media and letter), 6 months (GP intervention)
Participants	Country ‐ Australia Setting ‐ community Initial screening status ‐ due and overdue Nine geographically discrete, regions were selected within three adjacent TV broadcasting areas. The regions were randomly assigned to the study groups and data gathered on eligible women through administrative records pre‐and post‐intervention. Inclusion criteria ‐ aged 18 to 70 years; English‐speaking Exclusion criteria ‐ physically/intellectually impaired
Interventions	1. TV media campaign: n = n/a 2. TV media combined with invitation letter: n = n/a 3. TV media combined with GP‐based recruitment through workshops: n = n/a 4. Control: n = n/a n/a = not applicable, as data were gathered from administrative records for the regions giving overall Pap smear attendances during the pre‐ and post‐intervention periods. In the letter intervention group (using information gathered from electoral registers (registration was mandatory)), all eligible women were sent a letter.
Outcomes	Pap smear uptake determined by administrative records
Notes	Analysis limited by the 3‐ and 6‐month post‐intervention follow‐up periods; a longer period was prevented from contamination by a state‐wide media campaign. Differential effects of interventions on outcome for the different regions may reflect different baseline screening rates that could not be assessed during matching. Units of allocation different from units of analysis and no appropriate account was taken of this is the analysis. 1. TV media campaign: n = n/a 2. TV media combined with invitation letter: n = n/a 3. TV media combined with GP‐based recruitment through workshops: n = n/a 4. Control: n = n/a
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not reported
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Byles 1995.

Study characteristics
Methods	Design ‐ RCT (cluster) Baseline comparability ‐ study regions were matched as closely as possible using census data Follow‐up ‐ 3 months 28% of the letter intervention group did not recall ever receiving the intervention; not clear how many women were followed‐up
Participants	Country ‐ Australia Setting ‐ community Initial screening status ‐ due Three geographically separate postal regions were randomly allocated to different interventions. Data on eligible women within the regions was gathered via administrative records pre‐ and post‐intervention. Inclusion criteria ‐ aged 18 to 70 years; no Pap smear in previous 3 years; Australian or British citizenship Exclusion criteria ‐ not stated
Interventions	1. Personally addressed letter with simple information about Pap smears; N = 1128 (959 analysed) (99 attended for screening) 2. Personally addressed letter combined with a series of targeted behavioural prompts (e.g. prompt cards) designed to address aspects believed to be associated with poor screening rates; N = 1098 (933 analysed) (95 attended for screening) 3. Control: N = 1414 (1202 analysed) (97 attended for screening)
Outcomes	Pap smear uptake determined by administrative records
Notes	Timescale of the intervention was not stated and the 3‐month follow‐up period was short and may have limited the results. Units of allocation different from units of analysis and no appropriate account was taken of this in the analysis. Unclear how many women were followed‐up
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not reported
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Byles 1996.

Study characteristics
Methods	Design ‐ RCT (cluster) Baseline comparability ‐ regions were matched as closely as possible using census data Follow‐up ‐ 3 months A 15% adjustment of the denominator was made to account for the estimated hysterectomy rate.
Participants	Country ‐ Australia Setting ‐ community Initial screening status ‐ due and overdue Nine geographically distinct postal regions were randomly allocated to one of the intervention groups. Data about the women within the regions were gathered pre‐ and post‐intervention using administrative records. Inclusion criteria ‐ aged 18 to 70 years; no Pap smear in the previous 3 years Exclusion criteria ‐ not stated
Interventions	1. Personalised letter advising women to attend screening and providing simple information. Followed up by a second mailing campaign 3 years later; n = ? (? analysed) 2. No letter in the first mailing but letter sent during second mailing 3 years later; n = ? (? analysed) 3. Control, no letter on either occasion; n = ? (? analysed)
Outcomes	Pap smear uptake determined by administrative records
Notes	Previous campaigns may have had an unknown influence on the current campaign. The iterative process used to provide estimates of expected and observed may be affected by the limited follow‐up period, questioning the reliability of the analysis. Participants were only partially randomised (to initial letter). Units of allocation different from units of analysis and no appropriate account taken of this in the analysis
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not reported
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Byrd 2013.

Study characteristics
Methods	Design ‐ RCT Follow‐up ‐ 6 months
Participants	Country ‐ US Setting ‐ 3 diverse sites: El Paso, Texas, on the US/Mexico border; Houston, Texas, an urban centre; and Yakima Valley, Washington, a rural farming community Initial screening status ‐ due and overdue Eligible participants were of Mexican origin aged > 21 years.
Interventions	The AMIGAS intervention was designed for delivery by trained lay health workers. 1. Video discussing barriers to screening and a flip chart reviewing the information in the video 2. Flip chart only 3. Video only 4. Usual care control
Outcomes	Self‐report of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation scheme
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 83.6% (513/613)
Selective reporting (reporting bias)	Unclear risk	Results based largely on self‐report
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Clementz 1990.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no differences in any of the variables examined Follow‐up ‐ 4 months
Participants	Country ‐ USA Setting ‐ University family medicine clinic Initial screening status ‐ due 220 female patients attending ambulatory clinic Inclusion criteria ‐ aged 50 to 69 years Exclusion criteria ‐ symptomatic for cervical cancer; previously had cancer
Interventions	1. Personalised GPs letter, one month before due date of tests with an educational component; n = 116 2. Control group received usual care (not described); n = 104
Outcomes	Pap smear uptake determined by administrative records
Notes	Authors offered no explanation as to why the recall intervention had an adverse effect on people attending cervical screening, i.e. why such an intervention would make people less likely to attend. The low power of the study was attributed to imbalances between the intervention and control groups. There was an additional imbalance as a result of excluding patients post‐randomisation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Patients were then assigned by a computer‐generated random number to two groups".
Allocation concealment (selection bias)	High risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	"The physicians remained blinded to the individual patient's status throughout the study". However it was unclear whether or not the outcome assessors were blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 178/220 (81%) By treatment arms: Personalised GPs letter: 102/116 (88%) Control: 76/104 (73%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Decker 2013.

Study characteristics
Methods	Design ‐ Cluster‐RCT Baseline characteristics ‐ significant difference between intervention and control group with regards to visible minority status, education, income, area of residence, opportunity to be screened and continuity of care Follow‐up ‐ 6 months
Participants	Country ‐ Canada Setting ‐ Manitoba Initial screening status ‐ never had screening Unscreened women aged 30 to 69 years
Interventions	1. Invitation letter and brochure mailed to participant 2. Control
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Not reported
Allocation concealment (selection bias)	Low risk	Cluster‐randomised by forward sortation area to reduce contamination between intervention and control groups
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 96.5% (30358/31451)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Lack of stratification before randomisation. These factors compared between groups after randomisation and the practical differences were small.

Del Mar 1998.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no differences in age and postcode area Follow‐up ‐ 1 year
Participants	Country ‐ Australia Setting ‐ community Initial screening status ‐ due and overdue 689 women on the electoral roll in South Brisbane Inclusion criteria ‐ aged 18 to 67 years; Vietnamese Exclusion criteria ‐ not stated
Interventions	Media campaign on cervical screening introduced for whole region 2 months before letters sent 1. Personal letter (in Vietnamese) informing them about screening and its benefits; n = 359 2. Control group did not receive a letter; n = 330
Outcomes	Pap smear uptake determined by administrative records
Notes	Women in both groups were drawn from the Vietnamese community resident in one area, so there was a possibility of contamination.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Low risk	"After each pathology service had supplied the dates and results for each woman, the dataset was stripped of its fields of names and dates of birth, and the order changed randomly so that no individual woman's results could be identified. This file was then analysed".
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 689/689 (100%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Dietrich 2006.

Study characteristics
Methods	Design ‐ RCT Baseline characteristics ‐ primary language was Spanish for 63%. Ethnicity and income not known for 38% Follow‐up ‐ 18 months
Participants	Country ‐ US Setting ‐ 11 community and migrant health centres in New York City Initial screening status ‐ overdue Women aged 50‐69 years
Interventions	1. Intervention: series of telephone support calls over 18 months. Screening facilitated and barriers addressed 2. Control group: Single telephone call where trial staff answered questions about preventive care
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Sealed randomisation forms
Allocation concealment (selection bias)	Unclear risk	Informed of allocation individually by telephone
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 98.2% (1323/1346)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Eaker 2004.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ 5 months
Participants	Country ‐ Sweden Setting ‐ Uppsala county, Sweden Initial screening status ‐? due Inclusion criteria ‐ aged 25 to 59 years, no Pap smear within past 3 years Exclusion criteria ‐ had asked to be excluded from call‐recall system
Interventions	1. Standard invitation: n = 6140 2. Modified invitation: n = 6100
Outcomes	Pap smear uptake determined by national register
Notes	Study included multiple sequential randomised interventions. In view of sequential effects on subsequent interventions, only primary intervention included in forest plot. Other interventions were reminder letter versus no reminder to women who had not had smear at 5 months and then phone reminder versus no phone reminder at 2 months to women who had still not had smear following reminder letter. Total follow‐up for whole study: 12 months
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Low risk	"Collaborators in the study were blinded to the women's group assignment".
Blinding (performance bias and detection bias) All outcomes	Unclear risk	However, "Becasue the end‐point was whether a woman had a Pap smear or not, none of the collaborators, except the research assistants conducting the phone reminder, could influence the decision".
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 99% (12,157/12,240)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Fang 2017.

Study characteristics
Methods	Design ‐ two‐arm group randomised study Baseline characteristics ‐ significant difference between study groups in age, marital status, having health insurance and a regular physician and history of prior screening. The intervention group was younger, more likely to be married, but less likely to have health insurance or a regular physician. A greater number of the control group participants had no history of cervical screening. Follow‐up ‐ 6 months
Participants	Country ‐ Pennsylvania & New Jersey, US Setting ‐ Korean‐American church communities Initial screening status ‐ not adherent to cervical screening guideline Inclusion criteria: aged 21 years or above
Interventions	1. Intervention: participants met in small groups and received a single two‐hour educational session conducted by bilingual community health educators. Each educational session was held at church sites and focused on cervical cancer risk factors, screening guidelines and procedures, discussion of barriers to screening, follow‐up and clinical management if abnormal result identified. Navigation assistance for screening offered 2. Control: Two‐hour education session on general health and cancer education
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computerised randomisation
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	83.4% retention rate, with no difference between groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Greene 1999.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ not stated Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ rural primary care in low income, minority population Initial screening status ‐ due 273 women presenting for outpatient care who did not have a Pap test during the preceding year Inclusion criteria ‐ not stated Exclusion criteria ‐ not stated
Interventions	Based on Social Cognitive Theory and Motivational Interviewing methods 1. Usual care: n = 79 (? analysed) received general dietary and health information 2. Cancer education: n = 97 (? analysed) received general information about cervical cancer risk factors and screening recommendations 3. Cognitive behavioural intervention: n = 97 (? analysed) received feedback about personal risk for cancer and engaged in a clinical interview to enhance self‐efficacy for preventive behaviour
Outcomes	Pap smear uptake and booking of appointments determined by administrative records
Notes	Standard clinical procedures to advocate for and provide Pap tests were not withheld from any of the participants; all study participants received attention in addition to usual preventive care.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not reported
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Haguenoer 2015.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ stratified by age Follow‐up ‐ 9 months
Participants	Country ‐ France Setting ‐ regional cervical cancer screening programme Initial screening status ‐ women aged 30‐65 years who were overdue for screening
Interventions	1. HPV self‐test (not reported in this review ‐ this will meet the inclusion criteria of a HPV self‐testing intervention to increase uptake of cervical cancer screening in another Cochrane review) 2. Recall invitation for Pap smear (intervention) 3. No intervention (control)
Outcomes	Uptake of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random sampling and randomisation handled by independent computer programme management software
Allocation concealment (selection bias)	Low risk	Allocation method concealed to study coordinator
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Local ethics committee waived requirement for informed consent. In the "recall" letters, participants were informed that participation was part of a research programme about screening, but women were not aware that they were in a randomised trial. Women in the control group received no information and were unaware of the study.
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 94.6% (5678/5998) Because of address errors, recall letters could not be delivered to 156 women.
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Heranney 2011.

Study characteristics
Methods	Design ‐ RCT Follow‐up ‐ 8 months
Participants	Country ‐ France Setting ‐ organised cervical cancer screening Initial screening status ‐ overdue
Interventions	1. Letter invitation for a Pap smear 2. Telephone reminder that Pap smears were necessary. Ten attempts were made to contact the women.
Outcomes	Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	Telephone reminder: appropriate dialogue established with 82.8% of those reached (742 could not be reached, 785 refused the dialogue) Mail reminder: 99.6% of participants reached (20 letters returned as incorrect address)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Hunt 1998.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between the study groups in terms of the factors investigated Follow‐up ‐ 3 months 97/119 (81.5%) of women in the personal approach group and 37/125 (30%) of in the letter group were not contacted. These women were included in the final analysis.
Participants	Country ‐ Australia Setting ‐ community Initial screening status ‐ overdue 372 women identified from files at a women's clinic staffed by Aboriginal health workers in Danila Bilba Inclusion criteria ‐ resident in the Darwin area; overdue for screening Exclusion criteria ‐ not stated
Interventions	1. Personal approach. Women approached by Aboriginal health workers and invited for screening (119 analysed) 2. Letter. Designed by Aboriginal workers stating individuals overdue for smear and inviting them to attend (125 analysed) 3. Control. Usual care with reminder tags for clinic staff attached to medical records (122 analysed) 372 women were randomised but the study only gave the breakdown of women analysed by treatment arm.
Outcomes	Pap smear uptake determined by administrative records
Notes	Women were included in the final analysis even though, in many cases, particularly in the personal approach group, they had not received the intervention. The 3‐mth follow‐up period was relatively short.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Women were randomly allocated to one of three groups by matching a list of the women’s file numbers to a list of computer‐generated random numbers designating the group number (1, 2 or 3)".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Low risk	"The person reviewing the files (JH) was not aware of the women’s group allocation, and was not involved in sending letters or contacting women in the personal approach group".
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 98% (366/372)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Jandorf 2008.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no difference between groups Follow‐up ‐ 2 months
Participants	Country ‐ USA Setting ‐ Latinas in New York and Arkansas Initial screening status ‐ overdue
Interventions	1. Breast and cervical cancer prevention health education programme 2. Diabetes education programme (control) Both educational programmes conducted by staff and peer volunteers
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Cluster‐randomisation. Random assignment lists generated for each site separately using SPSS
Allocation concealment (selection bias)	Low risk	Coordinators and site host blinded
Blinding (performance bias and detection bias) All outcomes	Low risk	Cluster‐randomisation by site
Incomplete outcome data (attrition bias) All outcomes	High risk	Approximately half of the women were reachable for follow‐up (49%). The majority who were not reachable did not provide a telephone number.
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Jensen 2009.

Study characteristics
Methods	Design ‐ cluster‐RCT (cluster‐randomisation by GP) Follow‐up ‐ 9 months
Participants	Country ‐ Denmark Setting ‐ local screening programme for cancer Initial screening status ‐ overdue Women aged 23‐59 years
Interventions	1. Standard invitation letter 2. Personalised targeted invitation letter from GP. GP also received visit from facilitator to discuss ways to increase uptake of screening.
Outcomes	Coverage rate (uptake of cervical cancer screening)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	3 GPs and their patients excluded from analysis because of termination of GP practice (unknown figures)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Kitchener 2016.

Study characteristics
Methods	Design ‐ cluster‐RCT based on general practices performed in two phases Follow‐up ‐ 6 months (phase 1), 12 months (phase 2)
Participants	Country ‐ UK Setting ‐ within national screening programme Initial screening status ‐ due for first invitation for cervical smear (phase 1) & non‐attender to first invitation for smear at 7.5 months (phase 2)
Interventions	Phase 1: 1. specially designed pre‐invitation leaflet sent 4‐6 weeks before the initial routine invitation. The invitation was designed by young women in focus groups. 2. standard initial invitation Phase 2: 1. vaginal HPV self‐sampling kit sent unrequested 2. vaginal HPV self‐sampling kit offered on request 3. timed appointment for smear 4. access to nurse navigator to help overcome barriers to screening 5. offer of nurse navigator or a HPV self‐sampling kit 6. no further action (control)
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Allocations of interventions to practices were generated using a random number generator.
Allocation concealment (selection bias)	Low risk	Allocation was carried out separately for each PCT with the names and location of the practices concealed.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Only the trial team was blinded from the screening agency.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Unclear
Selective reporting (reporting bias)	Low risk	Outcomes did not appear to have been selectively reported.
Other bias	Low risk	No reason to suspect any other source of bias

Kreuter 1996.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between the study groups in terms of demographic variables Follow‐up ‐ 6 months 186/1317 failed to complete the 6‐mth follow‐up questionnaire; 457/1131 were not considered to be at risk or did not want to change and so were not included in the final analysis.
Participants	Country ‐ USA Setting ‐ family medical practice Initial screening status ‐ unclear 1317 adult patients from eight family medical practices in North Carolina, USA Inclusion criteria ‐ aged 18 to 75 years; completed baseline survey Exclusion criteria ‐ not stated
Interventions	Based on Health Belief Model 1. Typical HRA‐computerised assessment of participants' health risks and provision of individualised feedback as to their calculated mortality risks, n = 427 2. Enhanced HRA‐as previous but also assessed benefits, barriers and other psychosocial factors influencing the individuals' health‐related behaviour in order to provide individualised feedback designed to facilitate self‐change in health behaviours, n = 427 3. Control ‐ no feedback given to participants, n = 463
Outcomes	Pap smear uptake determined by self‐report via questionnaire
Notes	Also mentioned the Precaution Adaption Model. Absolute values for the original number of individuals eligible to receive the tests at baseline not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 126/1317 (10%) By treatment arm: Typical HRA‐computerised assessment: 46/427 (11%) Enhanced HRA: 48/427 (11%) Control: 32/463 (7%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Krist 2012.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ patients stratified equally by age, sex and practice. Therefore, control and intervention participants had similar demographics. Follow‐up ‐ 16 months
Participants	Country ‐ USA Setting ‐ 8 primary care practices within a private medical group Initial screening status ‐ overdue
Interventions	1. Invitation to use interactive preventive health record that was linked to participants electronic health record. Preventive services were recommended and the interface offered links to detailed personal messages that explained preventive services and its rationale. 2. Usual care (control). Received no mailings about an interactive health preventive record and were unable to access it
Outcomes	Uptake of screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomised by strata, but no other details given
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Not reported
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Lancaster 1992.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups in terms of mean age Follow‐up ‐ not stated
Participants	Country ‐ UK Setting ‐ general practice Initial screening status ‐ due 2131 women registered with general practices in North Manchester Inclusion criteria ‐ aged 50 to 64 years; resident in study area Exclusion criteria ‐ hysterectomy
Interventions	1. Cervical screening invitation sent with breast screening invitation, n = 965 2. Breast screening invitation only sent (control), n = 947
Outcomes	Pap smear uptake determined by administrative records
Notes	Eligibility criteria for participation in the study and for breast and cervical screening were not explicit. Ineligible women were included in the initial randomisation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 1794/1912 (94%) By treatment arm: Cervical screening invitation: 908/965 (94%) Control: 886/947 (94%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Larkey 2012.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no difference between groups Follow‐up ‐ 15 months
Participants	Country ‐ USA Setting ‐ Latino population in Phoenix, Arizona Initial screening status ‐ overdue
Interventions	Trial testing two methods of delivering cancer screening and prevention curriculum taught by Promotoras de Salud. The intervention was delivered in six 80‐minute sessions, either 1. Individually 2. Group‐based education classes
Outcomes	Cancer screening uptake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomisation
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	49.4% attrition rate
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Lonnberg 2016.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ age, number of previous tests and time since previous test were similar in the control and intervention arms. Follow‐up ‐ 6 months
Participants	Country ‐ Norway Setting ‐ study performed within national screening programme Initial screening status ‐ overdue Inclusion criteria ‐ aged 25‐69 years with no smear within the past ~4 years
Interventions	1. standard open reminder (control) 2. reminder letter with a scheduled appointment in 2‐4 weeks time. There were limited possibilities for rescheduling, and appointments were primarily during normal office hours. Women were not required to confirm their attendance in advance (intervention).
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer randomised 1:1
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	No incomplete data
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Ma 2015.

Study characteristics
Methods	Design ‐ RCT, using community organisations as the unit of randomisation Baseline comparability ‐ no significant difference in baseline outcome of ever having had a Pap test Significant difference in education, having been born outside the US and years living in the US Follow‐up ‐ 12 months
Participants	Country ‐ US Setting ‐ Vietnamese community organisations Inclusion criteria ‐ Vietnamese, aged 21 to 70 years who were not adherent to Pap test guidelines
Interventions	1. Intervention participants met in small groups with a Community Health Educator and went over information about the female body such as cervical cancer and its risks, particularly in the Vietnamese population, and procedures for Pap testing. Visual aids included pictures of Vietnamese women and doctors and were available in English and Vietnamese. The intervention participants had access to: supplementary visual aids, multimedia educational material in Vietnamese, client‐physician communication via videotaping, patient navigation, referral to Pap test sites, six‐monthly reminders. 2. Control participants were given information in Vietnamese for general health issues including mention of a routine health exam. There was no specific mention of a Pap test.
Outcomes	Cervical screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	95% completed 12‐month follow‐up. No significant difference in incomplete follow‐up between groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

McAvoy 1991.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ not stated Follow‐up ‐ 2 months and 4 months The overall response and consent rate was 73%.
Participants	Country ‐ UK Setting ‐ National screening programme Initial screening status ‐ overdue 737 randomly selected women from the Asian community in Leicester Inclusion criteria ‐ resident of Leicester; aged 18 to 52 years; not recorded as having had a smear test Exclusion criteria ‐ not stated
Interventions	1. Home visit and a multilingual video, n = 263 2. Home visit, multilingual leaflet and fact sheet, n = 219 3. Posted multilingual leaflet and fact sheet, n = 131 4. Control group received no intervention, n = 124.
Outcomes	Pap smear uptake determined by administrative records
Notes	The sample may not be representative of the general UK population as it used only Asian participants and originated from a previous study on the use of health services. The sample had an over‐representation of one ethnic group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The study appeared to have used a minimisation technique, where an attempt was made to minimise the imbalance between the number of patients in each treatment group over a number of factors; "The final sample was stratified by age, religion, postcode area, and by participation in the previous study (either as respondents or non‐respondents) and then divided into four groups".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 737/737 (100%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

McDowell 1989.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups in terms of marital status and age Follow‐up ‐ 1 year 447/2034 women who were not due for screening were excluded pre‐randomisation.
Participants	Country ‐ Canada Setting ‐ hospital Initial screening status ‐ due and overdue 2034 female patients attending a hospital‐based family medical centre in Ottawa Inclusion criteria ‐ aged 18 to 35 years; no previous smear in past year Exclusion criteria ‐ not stated
Interventions	1. GP letter and reminder letter after 21 days, n = 367 2. Physician reminder, n = 332 3. Telephone call, n = 377 4. Control group, n = 330
Outcomes	Pap smear uptake and costs determined by administrative records
Notes	Study also incorporated 628/2034 women who were assigned to a practice control group, but these women were not randomly assigned. By not assessing the eligibility of women (i.e. whether they had a smear in the preceding year) a number of women were excluded from the study post‐ randomisation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (1406/1406)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Mishra 2009.

Study characteristics
Methods	Design ‐ RCT Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ Samoan churches Initial screening status ‐ overdue
Interventions	1. Theory‐guided, culturally tailored cervical cancer education programme consisting of three parts: (1) education booklets in English and Samoan; (2) skill‐building and behavioural exercises; (3) interactive group discussion sessions (3 sessions lasting 2 hours each) 2. Control, no intervention
Outcomes	Self‐reported receipt of Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation of study locations (Samoan churches)
Allocation concealment (selection bias)	Low risk	Stratification by church
Blinding (performance bias and detection bias) All outcomes	Low risk	Interviewers blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 96% (16 women left the study site and 2 women too ill to participate)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Participants were paid.

Mock 2007.

Study characteristics
Methods	Design ‐ cluster‐RCT Baseline comparability ‐ no significant differences between study groups Follow‐up ‐ “3 to 4 months”
Participants	Country ‐ US, Santa Clara County, California Setting ‐ Vietnamese‐American community Initial screening status ‐ due Inclusion criteria ‐ aged 18 years or older living in Santa Clara county, California Exclusion criteria ‐ not reported
Interventions	1. Lay health worker intervention involving presentations about cervical cancer, question‐and‐answer sessions, and regular contact with participants to explain access to and scheduling of appointments. Also had media intervention (n = 491) 2. Media intervention only. Television, radio and print advertisements targeted at the Vietnamese‐American female population (n = 477)
Outcomes	Pap smear uptake ‐ self‐reported post‐intervention questionnaire
Notes	Women selected from the social networks of the lay health outreach workers and therefore may be more motivated to comply
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 96.3% (968/1005)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	High risk	Women selected from the social networks of the lay health outreach workers and therefore may be more motivated to comply

Morrell 2005.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ not reported Follow‐up ‐ 90 days
Participants	Country ‐ Australia Setting ‐ community Initial screening status ‐ due 90,000 women who had not had a Pap smear for ≥ 48 months Inclusion criteria ‐ age 20 to 69 years, not had smear in past 48 months, on New South Wales Pap Test Register Exclusion criteria ‐ not stated
Interventions	1. No intervention 2. Letter identical to that usually sent out to women at 27 months after latest Pap smear or letter giving a similar message, but phrased in a tone more sympathetic to other factors going on in the woman's life that might have stopped her from having the test to date
Outcomes	Pap smear as recorded on New South Wales Pap Test Register
Notes	Both letter styles were analysed together.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 99% (89,699/90,247)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Mullins 2009.

Study characteristics
Methods	Design ‐ RCT Follow‐up ‐ 11 weeks
Participants	Country ‐ Australia Setting ‐ national screening programme Initial screening status ‐ aged 65‐69 years & overdue
Interventions	1. Targeted reminder letter 2. General reminder letter 3. Control, no intervention
Outcomes	Pap test
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation using a database function
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	17% of letters were returned.
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Navarro 1995.

Study characteristics
Methods	Design ‐ RCT (cluster) Baseline comparability ‐ only one statistically significant difference found, the proportion of women who were employed (17.5% control versus 8.9% intervention), but not regarded as a threat to internal validity Follow‐up ‐ 6 months Outcome measure ‐ self‐report via interview
Participants	Country ‐ USA Setting ‐ community Initial screening status ‐ unclear 500 Latinas in groups of 10 to 15 were recruited through 'consejeras' (traditional lay health workers in the Latino community) and randomly assigned according to their consejeras to either the intervention or control. Inclusion criteria ‐ not stated Exclusion criteria ‐ not stated
Interventions	Based on Cognitive Social Learning Theory 1. Por La Vida (PLV) programme with consejeras (n = 18) taking 12 weekly educational sessions with the groups of women, n = 274 2. Control, no PLV programme; instead consejeras (n = 18) participated in a 'Community Living Skills' programme, n = 238
Outcomes	Pap smear uptake determined by self‐report via interview
Notes	The generalisability may be limited as the study focused on US Latinas of low socioeconomic status who have a low level of acculturation. The differences between the control (Community Living Skills) and intervention (PLV) programmes were not very clear. Units of allocation different from units of analysis, but appropriate analysis using clusters not individuals was performed. The results were presented using both the women and the Consejera as the units of analysis. The authors stated that the results were limited as the test completion rates for both the pre‐ and post‐test were lower than desired.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 71% (361/512) PLV programme: 199/274 (73%) Control: 162/238 (68%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Nuno 2011.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ participant age was statistically different between the two groups. Follow‐up ‐ 2 years
Participants	Country ‐ USA Setting ‐ Latinas in Yuma County, Arizona Initial screening status ‐ postmenopausal and overdue
Interventions	1. Promotora administered educational intervention: the intervention was administered in Spanish in participants' homes in small groups of 3‐12 women. Participants were asked to invite other female peers. The small groups were designed to be interactive and encouraged discussion and participation. It included messages embedded in ice‐breakers, games and activities, and prizes in the form of educational materials were distributed. Participants attended one or two classes. 2. Usual care: letter and telephone reminder to attend screening. No educational component
Outcomes	Self‐reported Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Assigned via odd and even numbers
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	High risk	Possibility of contamination between groups. It is possible that usual care participants had access to educational materials through acquaintances in the intervention group.
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 91.3% (348/381)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

O'Brien 2010.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no difference between two groups Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ South Philadelphia Hispanic community Initial screening status ‐ overdue
Interventions	1. Series of two workshops (3 hours each) with community members covering content related to cervical cancer led by four promotoras 2. Usual care
Outcomes	Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random number sequence
Allocation concealment (selection bias)	Unclear risk	Promotoras were not blinded.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	6‐month follow‐up was 58% (participants could not be located).
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Participants received gift card.

Ornstein 1991.

Study characteristics
Methods	Design ‐ RCT (cluster) Baseline comparability ‐ study groups differed significantly (P = 0.0001) in terms of race, type of insurance and visit frequency Follow‐up ‐ 1 year 818/3833 in the letter intervention groups (letter only; letter + physician reminder) did not receive the letters.
Participants	Country ‐ USA Setting ‐ family medicine clinic Initial screening status ‐ due 7397 participants and 49 physicians from a university‐based medical centre participated in the study. Inclusion criteria ‐ aged 18 years and over; not screened in previous 2 years; 'active' patient of the family medicine centre (i.e. had visited clinic in previous 2 years) Exclusion criteria ‐ not stated
Interventions	1. Physicians received computerised reminders, n = 1988 participants, 14 physicians 2. Participants were sent an invitation to attend followed by another personalised reminder letter (6 months later), n = 1925 participants, 12 physicians 3. Both physician and participant reminders, n = 1908 participants, 13 physicians 4. Control group, no intervention, n = 1576 participants, 10 physicians
Outcomes	Pap smear uptake determined by administrative records
Notes	A number of biases were reported. The study was limited to analyses of attending participants; physicians in the 4 groups were in the same building (blinding was not possible and the Hawthorne effect may have contributed to some of the improvements); there were baseline differences in participant characteristics; the unit of allocation (practice group) was different from the unit of analysis (participant)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (7397/7397)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Oscarsson 2007.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ not reported Follow‐up ‐ 1 year
Participants	Country ‐ Sweden Setting ‐ community Initial screening status ‐ unclear 800 women selected at random Inclusion criteria ‐ aged 28 to 65 years, resident in Kalmar County, Sweden, no registered cervical smear during the last 5 years Exclusion criteria ‐ not stated
Interventions	1. Control. No intervention, n = 400 2. Intervention included invitation letters, telephone interviews and promotive efforts for having a cervical smear taken, n = 400.
Outcomes	Pap smear uptake taken from National Population Register
Notes	Cost of extra Pap smear gained was calculated as 151.36€. Smears cost 66.87€ each in the intervention group and 16.63€ in the control group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"A computer randomly selected 400 women to serve as a study group and another 400 women to serve as a control group".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed:100% (800/800)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Paskett 2011.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no statistical difference in baseline characteristics Follow‐up ‐ 12 months
Participants	Country ‐ USA Setting ‐ Ohio Appalachian clinics Initial screening status ‐ overdue
Interventions	1. Lay health advisor intervention group: participants received two in‐person visits, two telephone calls, and four postcards from lay health advisors over 10 months. Information given about cervical cancer, screening and its importance and barriers to screening addressed 2. Routine care ‐ standard invitation for Pap smear
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomisation by clinic
Allocation concealment (selection bias)	Low risk	Investigators blinded
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Lay health advisors and participants not blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 83.2% (233/280)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Peitzmeier 2016.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant difference Follow‐up ‐ 17 months
Participants	Country ‐ US Setting ‐ urban federally qualified community health centre Initial screening status ‐ overdue
Interventions	Four intervention groups each received a maximum of three outreach attempts at one‐monthly intervals: 1. Letter outreach: standard letter with the patient's name and signed by the patient's provider was sent to them stating that their Pap test was overdue. An educational flyer was also included. This was sent up to three times. 2. Email outreach: standard email sent from the patient's providers email account stating that their Pap test was overdue. An educational flyer was also attached. This was sent up to three times. 3. Telephone outreach: a script was read to the patient stating that their Pap test was overdue, and an appointment could be made for a Pap test at the same time if the patient agreed. The telephone call was made up to three times. 4. Multimodal outreach: (1) standard letter and educational flyer, (2) standard email with educational flyer attached, (3) telephone outreach as per above. 5. Usual care ‐ review screening history with patients during clinic visits and suggestion made of need for screening
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	High risk	Group assignment was not blinded to investigators.
Blinding (performance bias and detection bias) All outcomes	Low risk	Patients were not aware that they were taking part in a study.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No incomplete data
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Peterson 2012.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ Oregon Initial screening status ‐ overdue women aged 35‐64 years with mobility impairment
Interventions	1. 90‐120 minute participatory workshop (mean of 4 participants per workshop) and structured telephone support for 6 months. Intervention led by women with mobility disabilities and included education about breast and cervical cancers, susceptibility to these conditions, screening benefits, overcoming barriers to screening, procedures and recommendations. Participants received an activity workbook and information brochures and a copy of the training presentation. 2. No intervention
Outcomes	Pap test
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Randomly assigned by alternating assignment
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 73.9% (156/211)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to assess whether an important risk of bias existed
Other bias	Unclear risk	Relied on eligible participants contacting study if wanting to take part. Participants given monetary incentive

Pierce 1989.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences were identified between the study groups for any of the characteristics examined Follow‐up ‐ 1 year 27% (38/142) of women in tagged group did not receive the intervention as they did not consult their doctor during the study period. 61 women were removed from practice list during the study: screening group (n = 24), tagged notes group (n = 20), control group (n = 17); n = 3 died and n = 58 left the practices.
Participants	Country ‐ UK Setting ‐ general practice Initial screening status ‐ due 146/1232 women registered with a general practice Inclusion criteria ‐ eligible for a smear test Exclusion criteria ‐ smear in past 5 years; hysterectomy; already on call‐recall list
Interventions	1. Letter asking women to have a smear, n = 140 2. Physician reminder, n = 142 3. Control group, n = 134
Outcomes	Pap smear uptake determined by administrative records
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (416/416)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Pritchard 1995.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no statistically significant differences between study groups and all women who attended the practice during the study period for age, country of birth, marital status and education Follow‐up ‐ 1 year 22 women in the intervention groups had hysterectomies but were retained in the analyses; 60% of women in the tagged notes group did not receive the intervention.
Participants	Country ‐ Australia Setting ‐ general practice Initial screening status ‐ due 757/2139 women at a university general practice in a socioeconomically disadvantaged area of Perth Inclusion criteria ‐ women aged 36 to 69 years Exclusion criteria ‐ Pap smear in past 2 years; hysterectomy; no attendance at practice for 3 years or more; known to attend another practice; terminally ill
Interventions	1. Physician reminder (tagged notes) group, n = 198 2. Letter with invitation to make an appointment, n = 206 3. Letter with fixed appointment, n = 168 4. Control group (usual care), n = 185
Outcomes	Pap smear uptake determined by administrative records
Notes	Follow‐up period was 1 year and recommended screening interval 2 years, so some women may have been screened after study period but within recommended interval.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Eligible women were randomly allocated to one of four groups using a table of random numbers".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (757/757)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Racey 2016.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no difference in age
Participants	Country ‐ Canada Setting ‐ opportunistic screening within primary care in Ontanrio Initial screening status ‐ overdue/underscreened
Interventions	1. HPV self‐test (not reported in this review ‐ this will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another Cochrane review) 2. Invitation letter for Pap test 3. No intervention ‐ opportunistic screening
Outcomes	Uptake of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation using random number generator
Allocation concealment (selection bias)	Unclear risk	Final eligibility determined post‐randomisation
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	No incomplete data
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	High risk	Change in protocol and some participants in the Pap test group received a reminder phonecall.

Radde 2016.

Study characteristics
Methods	Design ‐ RCT Follow‐up ‐ 3 years
Participants	Country ‐ Germany Setting ‐ participants selected via population registry
Interventions	1. Invitation letter for Pap test 2. Invitation letter for Pap test and information brochure 3. No invitation
Outcomes	Cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Stratified by age and counties and randomised using statistical software
Allocation concealment (selection bias)	Unclear risk	High rates of participation before invitation letters sent. Underestimation of effect of invitation on participation among less health conscious women
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Underestimation of effect of invitation on participation among less health conscious women
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 67.8%(5265/7758). Incomplete data because of lack of information on screening
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Rashid 2013.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant difference in age and ethnic group Follow‐up ‐ 8 weeks
Participants	Country ‐ Malaysia Setting ‐ community clinics Initial screening status ‐ due
Interventions	Invitation to Pap smear by: 1. Postal letter 2. Registered letter 3. Short message by phone (SMS) 4. Telephone invitation
Outcomes	Attendance for Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated
Allocation concealment (selection bias)	Low risk	Research staff blinded
Blinding (performance bias and detection bias) All outcomes	Low risk	Research staff blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed 74.5% (745/1000) (moved out/wrong numbers/did not receive letter)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Rimer 1999.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ not stated Follow‐up ‐ 16 months 37/889 women died before the follow‐up interview, and a further 24% could not be reached due to disconnected phones, 2% were not eligible for follow‐up interview due to health reasons and 2% refused to participate.
Participants	Country ‐ USA Setting ‐ community health centre Initial screening status ‐ unclear Adult users (over the age of 18 years) of the Lincoln Community Health Centre (which serves 30% of the black population and is the most important provider of care for low‐income people) Inclusion criteria ‐ aged 18 years or over; client of medical centre who had visited centre in previous 18 months Exclusion criteria ‐ not stated
Interventions	Based on Transtheoretical Model 1. Provider prompting intervention only (202 analysed) 2. Provider prompting and tailored educational print communications (204 analysed) 3. Provider prompting, tailored educational print communications and tailored telephone counselling (213 analysed) 1318 participants (men and women) of whom 889 eligible women were randomised, but study only gave breakdown of women analysed by treatment arm
Outcomes	Pap smear uptake determined by self‐report via questionnaire
Notes	The information presented seemed to be part of a larger study looking at the uptake of cancer screening in general, although only data on female participants attending mammography, Pap smear and CBE were presented. The use of a telephone to collect information about participants, as well as part of the interventions, may not have been appropriate as the study looked at screening behaviour among low‐income participants, many of whom had to be excluded because their telephone line had been disconnected. Difficult to assess which part of the invention was effective
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 70% (619/889)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Risi 2004.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ not reported Follow‐up ‐ 6 months
Participants	Country ‐ South Africa Setting ‐ community. Peri‐urban squatter community near Cape Town 658 women Initial screening status ‐ unclear Inclusion criteria: women aged 35 to 65 years resident in Khayelitsha who gave verbal consent Exclusion criteria ‐ not stated
Interventions	1. Photo‐comic with a storyline including scenarios based on cervical screening and common reasons for not participating in screening programme, n = 289 2. Photo‐comic containing no healthcare messages, n = 389
Outcomes	Self‐reported Pap smear uptake
Notes	After photo‐comic intervention, a radio intervention involving similar storylines to the intervention comic was broadcast. This part of the study was not randomised and was not included in the analysis in this review.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	"Sealed envelope containing a randomly allocated photo‐comic was provided". No comment regarding opacity of envelope made
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	> 99.8% analysed. One subject lost to follow‐up
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Rivers 2005.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ baseline demographics of all women reported, but not broken down into intervention arms Follow‐up ‐ 12 months
Participants	Country ‐ USA Setting ‐ urban community health clinic 441 women participated. Women were approached in the waiting room when attending the clinic. Women attending for obstetric or gynaecological reasons were not approached. Initial screening status ‐ unclear Inclusion criteria: women aged 18 to 65 years Exclusion criteria ‐ not stated
Interventions	1. 10‐minute video focusing on prevention of cervical cancer. Message gain‐framed 2. 10‐minute video focusing on prevention of cervical cancer. Message loss‐framed 3. 10‐minute video focusing on detection of cervical cancer. Message gain‐framed 4. 10‐minute video focusing on detection of cervical cancer. Message loss‐framed
Outcomes	Self‐reported uptake of Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated table of randomly sorted combinations of conditions
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 78% (343/441)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	High risk	Telephone call at 6 months may have acted as a prompt to uptake and influence 12‐month data. Number of recruits assigned to each intervention was not stated. We have assumed a 1:1:1:1 ratio with 110 in each intervention.

Robson 1989.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences were found between the study groups in terms of the variables examined. Follow‐up ‐ 2 years Women with hysterectomies were excluded from analyses. Trial discontinued after 2 years (versus 3 years), as GPs were no longer willing to exclude half the patients from accessing the health promotion nurse.
Participants	Country ‐ UK Setting ‐ general practice Initial screening status ‐ due Men and women registered with a general practice in inner London (UK) Inclusion criteria ‐ aged 30 to 65 years; registered with practice and living in area Exclusion criteria ‐ hysterectomy
Interventions	1. Patients had open access to a health promotion nurse and had their risk factors assessed and followed up by both their GP and the nurse, n = 799 2. Control, usual care (i.e. managed by GP alone), n = 806
Outcomes	Pap smear uptake determined by administrative records
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Patients ... were included and randomly allocated (with random number tables) to control or intervention groups".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 1605/1605 (100%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	High risk	"Although the trial had been designed to last three years, it was stopped after two years because participating doctors were not prepared to continue excluding half the practice from access to the health promotion nurse".

Rosser 2015.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ demographic characteristics did not differ significantly. Follow‐up ‐ 3 months
Participants	Country ‐ Kenya Setting ‐ government clinics Initial screening status ‐ never been screened Inclusion criteria ‐ non‐pregnant women aged 23 years and over
Interventions	1. Intervention: Thirty‐minute interactive talk about cervical cancer. The talk reviewed the basic health facts about cervical cancer, risk factors, how screening is performed, what screening results mean, and treatment options. The talk also included a guided discussion about barriers to screening and fears or stigma associated with screening. For standardisation, each session was guided by a flip chart and script. The talk was administered by community health workers. 2. Control: standard care
Outcomes	Cervical cancer screening uptake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was done in computer‐generated blocks of eight.
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Low risk	In order to minimise contamination, the health talk was given to groups of women in the intervention arm in a private area of the health facility.
Incomplete outcome data (attrition bias) All outcomes	High risk	Lost to follow‐up > 20 % in each arm
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to permit judgement

Segnan 1998.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences were found between the study groups in terms of the variables examined. Follow‐up ‐ 1 year
Participants	Country ‐ Italy Setting ‐ GP practice in national screening programme Initial screening status ‐ due 8385 women attending GPs in Turin who were part of the population‐based screening programme ('Prevenzione Serena') Inclusion criteria ‐ aged 25 to 64 years; resident of Turin Exclusion criteria ‐ previously diagnosed cervical cancer; suffering from terminal illness or severe psychiatric symptoms
Interventions	1. Personal letter signed by GP with prefixed appointment (control), n = 2100 2. Personal letter, signed by GP prompting appointment, n = 2093 3. Personal letter signed by programme co‐ordinator with prefixed appointment, n = 2094 4. Personal letter with extended text signed by GP with prefixed appointment, n = 2098
Outcomes	Pap smear uptake determined by administrative records
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Assignment to the different groups was performed automatically, following a randomised block design (block = GP)".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (8385/8385)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Somkin 1997.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no differences between study groups in terms of age Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ HMO Initial screening status ‐ due 7077 female HMO members Inclusion criteria ‐ aged 20 to 64 years; no prior Pap smear in the previous 36 months; residents of study area; were continuously enrolled as a member of the HMO for the previous 36 months Exclusion criteria ‐ not stated
Interventions	1. Letter inviting women to make an appointment, n = 1188 2. Physician reminder and letter to patient inviting appointment, n = 1188 3. Usual care (required a referral from physician), n = 1188
Outcomes	Pap smear uptake determined by administrative records
Notes	The authors listed the following study limitations: smears obtained outside the HMO were not recorded; the chart reminder intervention required the health provider to review the chart; the study had insufficient power to detect interaction effects between interventions and covariates, and within strata.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 100% (3564/3564)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Stein 2005.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no differences between study groups in terms of age Follow‐up ‐ 3 months
Participants	Country ‐ UK Setting ‐ community. Devon, UK Initial screening status ‐ overdue. 1140 women. No record of screening in past 15 years Inclusion criteria ‐ aged 39 to 64 years; no record of screening in past 15 years Exclusion criteria ‐ deceased, not resident in Devon, undergone hysterectomy, severe learning disability
Interventions	1. Control. No Intervention, n = 285 2. Telephone call. Telephone call from experienced research nurse using a prepared script. Maxiumum of three attempts were made on consecutive days, n = 285. 3. Letter from Health Authority District Cervical Screening Commisioner on behalf of National Cervical Screening Programme, n = 285 4. Letter from a well known journalist and broadcaster (Claire Rayner) who was also Chair of the Patients Association, n = 285
Outcomes	Pap smear uptake as recorded on the Devon Patient and Practitioners Services Agency database
Notes	Cost‐effectiveness analysis also performed. Average cost per attender was £145.12 for telephone call, £14.29 for letter from commissioner and £37.14 for letter from celebrity.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The sample was drawn randomly from the sampling frame using Microsoft Excel".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 73% (304/1140)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Studts 2012.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant difference in baseline characteristics Follow‐up ‐ 4 months
Participants	Country ‐ USA Setting ‐ four Appalachian Kentucky counties Initial screening status ‐ overdue
Interventions	All participants received an educational lunch session where cervical cancer screening was discussed; and then had either: 1. Home visits by lay health advisor to address specific barriers by each participant, and then a tailored newsletter was produced to address the barrier 2. Control
Outcomes	Uptake of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Non‐random sample of churches as recruitment site
Allocation concealment (selection bias)	Low risk	Investigators were blinded.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Participants were not blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 95.9% (331/345), 14 participants lost to follow‐up
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Participants were paid for their involvement.

Sung 1997.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups in terms of those variables examined Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ community Initial screening status ‐ due 321 low‐income African‐American women from an inner‐city community health centre Inclusion criteria ‐ African‐American; aged 18 years or older Exclusion criteria ‐ hysterectomy; history of cervical cancer
Interventions	1. Lay health workers visited women three times to provide a culturally sensitive educational programme emphasising need for screening through printed material and video, n = 163. 2. Control group received educational information on completion of follow‐up, n = 158.
Outcomes	Pap smear uptake determined by self‐report via interview
Notes	Loss to follow‐up and Hawthorne effect may have biased the effects of the intervention, however, an intention‐to‐intervene analysis was also carried out with the aim of providing a conservative estimate of the effect size.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 61% (195/321) Lay health workers: 93/163 (57%) Control: 102/158 (65%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Taylor 2002.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups in terms of those variables examined Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ Chinese community in Seattle and Washington Initial screening status ‐ due Inclusion criteria ‐ 20 to 69 years, spoke Cantonese, Mandarin or English; no history of cervical cancer; not had hysterectomy; not had smear in last 2 years and/or did not intend to have Pap smear in the next 2 years
Interventions	1. Outreach worker intervention: home visit including videos, motivational pamphlet, tailored counselling, fact sheet and educational brochure. Follow‐up with telephone call with tailored counselling. Assistance also provided to arrange and attend appointment 2. Direct mail intervention: packet including video, motivational pamphlet, fact sheet and educational brochure 3. Usual care
Outcomes	Pap smear within previous 2 years. Self‐reported by questionnaire and cross‐checked with medical records where possible
Notes	"Groups of women were randomly assigned every month over the 6‐month period." Our understanding is that this referred to accrual date rather than referring the quasi‐randomisation. Intention to have smear in next 2 years also recorded
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 83% (402/482) for all outcomes
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Taylor 2010.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no difference in demographics Follow‐up ‐ 6 months
Participants	Country ‐ USA Setting ‐ Vietnamese community in Washington Initial screening status ‐ overdue
Interventions	1. Lay health worker home visit using educational materials ‐ Vietnamese language DVD & pamphlets providing information on cervical cancer and Pap tests within the context of Vietnamese traditional beliefs. Emphasised the importance of screening. Lay health workers were bilingual ethnic Vietnamese women. 2. Control group was mailed physical activity materials and a pedometer with instructions for use.
Outcomes	Pap test
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	% analysed: 74.3% (174/234) Incomplete data ‐ refused to participate in follow‐up, could not be contacted/moved address
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Thompson 2016.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no statistics reported, but characteristics within a table appeared similar Follow‐up ‐ 7 months
Participants	Country ‐ Lower Yakima Valley of Washington State, US Setting ‐ study performed in partnership with federally qualified health centre in area, with a state programme that provided cancer screening Inclusion criteria ‐ Latinas aged 21‐64 years who were nonadherent to screening Pap guidelines
Interventions	1. Control ‐ usual care 2. Low‐intensity intervention ‐ Spanish‐language video was sent to participants' homes and informed women of the importance of cervical cancer screening. 3. High‐intensity intervention ‐ Spanish‐langage video was sent to participants' homes and informed women of the importance of cervical cancer screening, and a home visit was made by a promotora who presented information to the participant on the importance of cervical screening.
Outcomes	Uptake of cervical cancer screening
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	The authors speculated that a relatively high proportion of women in the control group received Pap testing, which was not significantly different to that of the low‐intensity group. They noted that all participants, including the control group, underwent baseline assessment that focused on their reasons for not being screened. This many have encouraged all women to think about their reasons for not having a Pap test and may have motivated them to undergo testing.
Blinding (performance bias and detection bias) All outcomes	Low risk	Investigators and statisticians were blinded to randomisation status of participants.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Incomplete follow‐up: control 12/147 (8%), low‐intensity 10/150 (7%), high‐intensity 18/146 (12%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Participants were paid for their involvement.

Vogt 2003.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between study groups in terms of those variables examined Follow‐up ‐ 12 weeks
Participants	Country ‐ USA Setting ‐ managed care organisation (NWKP) in Portland, Oregon Initial screening status ‐ due Inclusion criteria ‐ 18 to 70 years; ⩾3 years continuous membership of NWKP; no history of cervical cancer or cervical dysplasia; not had hysterectomy; not had Pap smear in past 3 years Exclusion criteria ‐ women who no longer had a valid local address or phone
Interventions	1. Usual‐care control 2. Letter/letter intervention: subjects were sent a letter and relevant brochure. Women who had not attended for screening within 6 weeks were sent a further letter emphasising the importance of screening and providing a number to call. 3. Letter/phone intervention: letter and brochure as above. Women who had not attended for screening within 6 weeks received a telephone call by study interventionist who offered to schedule appointments, answer questions, address barriers and concerns and discuss the importance of screening. 4. Phone/phone intervention: subjects in this group received two sequential telephone calls, the second coming 6 weeks after the first if they had not been screened in the interim. Contents of the initial letter and phone scripts were similar. Follow‐up telephone calls were by study interventionist, as above.
Outcomes	Pap smear uptake as recorded on care organisation pathology database
Notes	Cost‐effectiveness analysis also performed. The letter/letter intervention produced one additional Pap smear for $185. The phone/phone intervention cost $305 and the letter/phone intervention cost $1117 for each additional Pap smear.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	High risk	% analysed: 72% (866/1200)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Ward 1991.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between the study groups in terms of factors studied Follow‐up ‐ 1 month
Participants	Country ‐ Australia Setting ‐ general practice Initial screening status ‐ due 204 female patients of 16 GPs in the inner metropolitan region of Sydney Inclusion criteria ‐ women: aged 20 to 65 years; provided consent; physicians: provided consent; complied with study procedures Exclusion criteria ‐ women: pregnant; had smear in past year; attending for smear that day; hysterectomy; never sexually active with male partner; insufficient command of English to complete questionnaire; physicians: worked < 20 hrs/week; were on leave/sick leave at time or recruitment; were expected to take leave during the study period; did not have the equipment to take smears
Interventions	1. Minimal intervention: GP advised eligible women of need for smear and offered to perform it immediately. Those not consenting advised to make appointment for smear within a week, n = 99 2. Maximal intervention: GP advised women of need for smear and offered to perform it immediately; GP attempted to persuade those not consenting during that consultation by exploring barriers and reasons for self‐exclusions. If still did not consent, GP advised making an appointment for smear within a week, n = 103
Outcomes	Pap smear uptake determined by administrative records
Notes	Fidelity of intervention implementation could not be checked; audiotapes were available for only a few consultations. One of the audiotapes recorded a time of 6 sec taken to give the maximal intervention (shortest time in minimal intervention was 10 sec).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 91% (184/202) By treatment arm: Minimal intervention: 95/99 (96%) Maximal intervention: 89/103 (86%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Wilson 1987.

Study characteristics
Methods	Design ‐ RCT Baseline comparability ‐ no significant differences between the study groups in terms of mean age Follow‐up ‐ 3 weeks from final invitation letter
Participants	Country ‐ UK Setting ‐ National Screening Programme Initial screening status ‐ due, recorded as never having a smear 250 randomly selected women from five general practices (50 women per practice) in the Nottingham Health Authority area Inclusion criteria ‐ aged 45 to 65 years; no record of having a previous smear Exclusion criteria ‐ hysterectomy or other medical condition
Interventions	1. Letter of invitation to make an appointment + two reminders, n = 125 2. Sent an appointment + two reminders, n = 125
Outcomes	Pap smear uptake determined by administrative records
Notes	Only published as a letter. Final numbers of study participants was small compared to the initial study population (588 women who fulfilled the study criteria were not included).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported
Allocation concealment (selection bias)	Low risk	Centralised allocation was used.
Blinding (performance bias and detection bias) All outcomes	Unclear risk	Not reported
Incomplete outcome data (attrition bias) All outcomes	Low risk	% analysed: 96% (240/250) By treatment arm: Letter invitation: 122/125 (98%) Appointment: 118/125 (94%)
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Insufficient information to assess whether an important risk of bias existed

Wright 2012.

Study characteristics
Methods	Design ‐ cluster‐RCT Baseline comparability ‐ compared to intervention group, the control arm participants were older, had higher income and less private insurance. Follow‐up ‐ 60 days
Participants	Country ‐ USA Setting ‐ primary care practices Initial screening status ‐ due/overdue
Interventions	1. Participants had access to their electronic patient health records. They received health maintenance reminders including the need for a Pap smear if it was due. Participants were given an explanation of the Pap test and why it was being recommended. 2. Active control. Participants also had access to their electronic patient health records. They did not get health maintenance reminders but could access eJournals that would allow them to update medications and diabetes management information.
Outcomes	Pap smear
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation by statistician
Allocation concealment (selection bias)	Unclear risk	Patients not aware which study arm they were in until after consent process
Blinding (performance bias and detection bias) All outcomes	High risk	No blinding
Incomplete outcome data (attrition bias) All outcomes	Low risk	No incomplete data
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement
Other bias	Unclear risk	Low penetration of patient health record. Given the significant differences between patient health record users and non‐users, the low penetration in the study and passive nature of the intervention, it is uncertain whether this intervention could have affected health maintenance screening rates.

CBE: Clinical Breast Exam
CRICERVA: Economic Evaluation of Three Population Screening Strategies for Cervical Cancer
DVD: Digital Video Disc
GP: General Practitioner
HMO: Health Maintenance Organisation
HPV: Human Papillomavirus
HRA: Human Research Authority
n/a: not applicable
NWKP: Northwest Kaiser Permanente
PCT: Post Coital Test
PLV: Por La Vida (For Life)
RCT: Randomised Control Trial
SEK: Swedish Krona

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Abdullah 2013b	Duplicate data; see Abdullah 2013a
Acera 2014	Duplicate trial; see Acera 2017
Al Sairafi 2009	Not RCT. Questionairre responses only
Atlas 2013	Intervention was aimed at helping practices identify who needed screening.
Baele 1998	Quasi‐RCT
Bebis 2012	Outcome measure was knowledge and awareness of cervical cancer, not uptake of screening.
Benard 2014	Outcome measure was not uptake of cervical cancer screening. Study examined provider and patient acceptance of cervical cancer screening with Pap test, whilst co‐testing for HPV with longer screening intervals.
Blomberg 2011	Face‐to‐face and internet‐based focus group discussions. Not an RCT
Boissel 1995	Interventions targeted at General Practitioners rather than woman
Bonevski 1999	Interventions targeted at General Practitioners rather than woman
Brewster 2002	Comparisons of study recruitment strategies, not of screening uptake
Broberg 2012	Abstract, see Broberg 2013
Broberg 2014	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Burger 2017	HPV self‐sampling. Did not meet the inclusion criteria for HPV self‐testing interventions to increase uptake of cervical cancer screening in another review because no data for uptake of cervical screening
Cadman 2015	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Campbell 1997	Intervention aimed at both the participants and the physician and data did not allow effects of the two components to be examined independently. Interventions aimed at physicians are excluded from this review.
Cecchini 1989	Interventions targeted at "promoting general practitioners' cooperation" and not at women
Chigbu 2017	Not RCT
Chumworathayi 2007	Quasi‐RCT
Cofta‐Woerpel 2009	Study protocol only
Corkrey 2005	Number of uptake cases in each group was not reported. Increased uptake of 0.43% reported in intervention group but no P value or 95% CI reported
Darlin 2013	Outcome was not uptake of cervical cancer screening.
Del Mar 1995	Intervention more concerned with obtaining more up‐to‐date addresses for participants rather than strictly increasing the uptake of screening
Del Mistro 2017	HPV self‐sampling. Did not meet inclusion criteria for HPV self‐testing intervention to increase uptake of cervical cancer screening in another review because continuation of trial for a subsequent round of screening.
Dignan 1996	Attendance for screening over the previous year was measured, but data were gathered only 6 months post‐intervention. Therefore, it was unclear how the intervention affected uptake as participants may have been screened prior to receiving the intervention.
Dignan 1998	Attendance for screening over the previous year was measured, but data were gathered only 6 months post‐intervention. Therefore, it was unclear how the intervention affected uptake as participants may have been screened prior to receiving the intervention.
Duggan 2012	Study protocol. For results, see Byrd 2013
Duke 2015	Not an RCT
Elder 2016	Conference abstract
Enerly 2016	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Engelstad 2005	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Erwin 2012	Abstract only
German 1995	The study examined the effect of the intervention on the uptake of overall preventive visits and the data were not specifically broken down into individual screening tests and procedures.
Giorgi Rossi 2011	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Giorgi Rossi 2015	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Gok 2012	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Gotay 2000	Outcome reported as "Ever had Pap test". This only captured new uptake rather than total uptake at follow‐up.
Hancock 2001	Not possible to extract relevant data for purposes of this review
Hiatt 2008	Not an RCT
Hicks 1997	Quasi‐RCT
Hillman 1998	Interventions targeted at GP practices and not at women
Hitzeman 2012	Summary of the previous Cochrane review on interventions targeted at women to encourage the uptake of cervical cancer screening
Holloway 2003	Outcomes reported at timing interval of smear uptake rather than number of women reporting uptake
Hou 2002	Quasi‐RCT
Hou 2005	Quasi‐RCT
Jenkins 1999	Not an RCT
Karwalajtys 2007	Not an RCT
Katz 2007	Intervention was aimed at increasing mammography uptake.
Kobetz 2017	Study protocol
Krok‐Schoen 2016	No data for the control group. Same study as Paskett 2011
Lairson 2010	Abstract only. Same study as Byrd 2013
Lam 2003	Initial data only. Full data included in Mock 2007
Lantz 1995	Quasi‐RCT. Impact of individual intervention (either letter or interview) could not be assesed separately.
Lantz 1996	Quasi‐RCT
Lauver 1990	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Levine 2003	Quasi‐RCT
Litzelman 1993	Interventions targeted at physicians to improve uptake and not at women
Lopez‐Torres Hidalgo 2016	Abstract only
Love 2009	Quasi‐experimental design. Outcome measure was improvement in knowledge of cervical cancer.
Love 2012	Quasi‐experimental design. Outcome measure was improvement in knowledge of cervical cancer.
Luszczynska 2011	Uptake of cervical cancer screening was not an outcome measure.
Lyimo 2012	Cross‐sectional study
Lynch 2004	Cost‐effectiveness study. Data reported as uptake of both cervical screening and mammography and not reported separately.
Maddocks 2011	Intervention aimed at General Practitioners
Manfredi 1998	Interventions targeted at physicians to improve uptake and not at women
Marcus 1992	Cluster‐quasi‐RCT
Marcus 1998	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Margolis 1998	Quasi‐RCT
Maxwell 2003	Quasi‐RCT. "....[women] wanted to attend the same group session, 5‐10 women were randomised to one study condition and the next group of 5‐10 women to the other".
Mayer 1992	Age range was 65 years and older and therefore did not represent the generally accepted age group for cervical screening programmes
Mbah 2015	Trial study design
Miller 1999	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Miller 2007	Not RCT. Qualitative study
Miller 2017	Uptake of screening not an outcome
Mitchell 1991	Not an RCT: the educational campaign was not randomly assigned and 2000 women were only randomly selected within each of the campaign study groups to receive the personal invitation letter.
Mitchell 1997	Not an RCT
Murphy 2015	Conference abstract
Mutyaba 2009	Outcome measure was not uptake of cervical cancer screening. Outcome measure was attendance to colposcopy following an abnormal Pap test.
Newell 2002	RCT. However, data were not presented in a way that was interpretable for purposes of the review.
Nguyen 2000	Interventions targeted at physicians to improve uptake and not at women
Okeke 2013	Cervical cancer screening was not available within the community outside of the programme.
Park 2005	LIkely quasi‐RCT. "...randomly assigned to the experimental or control group in order of contact..."
Paskett 1990	Quasi‐RCT
Paskett 1995	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Paskett 1999	Not RCT. "..mixed cohort/cross‐sectional design"
Paul 2003	Outcome was enrolment to Pap smear reminder service.
Perkins 2007	Not an RCT. Cross‐sectional study
Peters 1999	Cluster‐RCT examining anxiety among women with mild dyskaryosis and the aim of the educational intervention was to reduce anxiety so scope differed to that of this review.
Philips 2006	Outcome was "willingness to pay" rather than screening uptake.
Piana 2011	HPV self‐testing study. This intervention has been excluded from this review.
Pirzadeh 2012	Investigated effect of education on knowledge of cervical cancer and pap smear. Uptake of screening not an outcome measure
Powers 1992	Study examined attendance for a number of screening tests and did not separate data according to the type of test.
Rock 2014	Health promotion was not to the individual. Did not give absolute figures of those who had cervical screening in each group.
Roetzheim 2004	Interventions targeted at clinic staff and not at women
Roetzheim 2005	Interventions targeted at attempting "to change systematically the behaviours of all office staff"
Ruffin 2004	Age range was 50 years and older and, therefore, did not represent a majority of the generally accepted age group for cervical screening programmes. Loss to follow‐up was reported by practice, but not in terms of numbers of recruits or treatments arms. Percentage uptake rates of 2‐ and 3‐year follow‐up did not use baseline as a reference point.
Sancho‐Garnier 2013	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Sankaranarayanan 2003	Outcome was "effectiveness of VIA screening" not uptake of screening.
Scoggins 2010	Duplicate data. See Taylor 2010
Sewali 2015	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Shastri 2014	No cervical cancer screening programme in place
Shelley 1991	Not an RCT
Shenson 2011	Telephone survey. Not an RCT
Shojaeizadeh 2011	Evaluated health beliefs and practices before and after an educational programme. Quasi‐experimental “before‐and‐after” study. There was no control group.
Smith 2013	Described development of AMIGAS programme and did not give complete results. For results, see Byrd 2013
Stewart 1994	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Sultana 2014	Described proposed iPap cervical screening trial in Australia. For results, see Sultana 2016
Sultana 2016	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Szarewski 2011	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Takacs 2004	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Thompson 2006	Intervention was aimed at communities.
Tomlinson 2004	Intervention aimed at improving follow‐up following abnormal results rather than initial screening uptake
Torres‐Mejia 2000	Recruits were "systematically assigned." Not an RCT
Tranberg 2016	Study protocol
Valanis 2003	Age range was 50‐69 years and therefore did not represent a majority of the generally accepted age group for cervical screening programmes.
Virtanen 2011a	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Virtanen 2011b	Duplicate data. See Virtanen 2011a
Viviano 2017	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Ward 1999	Quasi‐RCT
Wedisinghe 2013	Abstract only
Wichachai 2016	Uptake of screening not an outcome
Wikstrom 2011	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.
Williams 2013	Study design only
Wood 2014	Described community engagement project to compare two different methods of cervical cancer screening. No intervention to increase uptake of cervical cancer screening
Wright 2008	Abstract only. For results, see Wright 2012
Wright 2010	Outcome measure was knowledge and awareness of cervical cancer, not uptake of screening.
Yancey 1995	Quasi‐RCT
Zehbe 2016	This will meet the inclusion criteria of HPV self‐testing interventions to increase uptake of cervical cancer screening in another review.

AMIGAS: Ayudando a Las Mujeres con Información, Guía y Amor para su Salud (Helping Women with Information, Guidance, and Love for Their Health)
CI: Confidence Interval
HPV: Human Papillomavirus
RCT: Randomised Control Trial

Characteristics of studies awaiting classification [ordered by study ID]

Egawa‐Takata 2018.

Methods	Randomised study
Participants	Study based in Japan. Participants were eligible for cervical screening (20 year olds who had just become eligible)
Interventions	Participants were randomised to receive either a stand‐alone invitation letter or an invitation letter for mother and daughter.
Outcomes	In the group where letters were sent to both mother and daughter, the uptake of cervical cancer screening was significantly higher in comparison to the groups where a stand‐alone letter was received.
Notes

Erwin 2019.

Methods	Randomised study
Participants	Study based in Tanzania. Participants were women eligible for cervical screening.
Interventions	Participants randomised to 3 groups ‐ educational SMS, educational SMS + transport voucher, SMS reminder
Outcomes	Both the educational SMS and educational SMS + transport groups improved cervical cancer screening uptake. However, the combined SMS and transport groups saw the largest improvement in uptake in comparison to the control group.
Notes

Firmino‐Machado 2018.

Methods	Randomised study
Participants	Study based in Portugal. Participants were women eligible for a cervical screening in a primary care setting.
Interventions	Participants were randomised either to SMS + telephone reminders versus standard letter reminders.
Outcomes	The participants in the SMS/telephone reminder group were more likely to attend for cervical screening in comparison to the control group.
Notes

Firmino‐Machado 2019.

Methods	RCT
Participants	Eligible patients for screening in Portugal
Interventions	The tested intervention was a 3‐step invitation to screening, based on automated text messages/phone calls (step 1), manual phone calls (step 2) and face‐to‐face interviews (step 3), applied sequentially to non‐adherent women after each step.
Outcomes	The primary outcome was the proportion of women screened, with adherence to cervical cancer screening being significantly higher among women assigned to the intervention.
Notes

Han 2017.

Methods	Randomised study
Participants	Study based in Baltimore, USA. Participants were eligible for cervical cancer or breast screening programmes.
Interventions	Participants were randomised to either receive education by community health workers vs standard reminders for screening.
Outcomes	Participants who received education on the benefits of screening were more likely to take up cervical and breast cancer screening.
Notes

Kiran 2018.

Methods	Randomised study
Participants	Study based in Toronto. Participants were women or men eligible for cervical, breast or colorectal screening.
Interventions	Participants were randomised to either receive a reminder phone call vs the standard reminder letter.
Outcomes	Participants who received a reminder phone call were significantly more likely to attend for cervical/breast/colorectal screening.
Notes

Kitchener 2018.

Methods	Cluster‐RCT
Participants	Eligible screening patients in the UK
Interventions	A two‐phase cluster randomised trial conducted in general practices in the NHS Cervical Screening Programme. In Phase 1, women in practices randomised to the intervention due for their first invitation to cervical screening received a pre‐invitation leaflet and, separately, access to online booking. In Phase 2, non‐attenders at six months were randomised to one of: vaginal self‐sample kits sent unrequested or offered; timed appointments; nurse navigator; or the choice between nurse navigator or self‐sample kits.
Outcomes	Primary outcome was uplift in intervention vs control practices, at 3 and 12 months post‐invitation, with self‐sample kits and timed appointments leading to an increase in screening uptake.
Notes

Koc 2019.

Methods	RCT
Participants	Turkish women eligible for cervical cancer screening
Interventions	Education programme on cervical cancer and screening in the interventional group using the PRECEDE education model
Outcomes	Improvement in cervical screening uptake in interventional group
Notes

Kurt 2019.

Methods	Randomised study
Participants	Study based in Turkey. Participants were eligible for cervical screening and were previous non‐attenders.
Interventions	Participants were randomised to either brochure + education on screening, brochure alone & standard reminder only.
Outcomes	Participants who were randomised to the brochure + education group were significantly more likely to attend for cervical screening.
Notes

Nicolau 2017.

Methods	Randomised study
Participants	Study based in Brazil. Participants were eligible for cervical screening.
Interventions	Participants were randomised to either an educational phone call on screening, telephone reminder for screening, or standard reminders (posters in clinics).
Outcomes	Both interventional groups (educational telephone call/reminder phone call) had a significantly higher uptake of cervical screening.
Notes

Okuhara 2018.

Methods	Leaflet‐based RCT
Participants	Women eligible for breast and cervical cancer screening
Interventions	Examined the effect of improving written materials in terms of processing fluency with regard to encouragement for obtaining breast and cervical cancer screening
Outcomes	The screening rate of the intervention group was significantly higher (nearly double) than that of the control group.
Notes

Tanjasiri 2019.

Methods	RCT
Participants	Pacific Island women based in the USA eligible for cervical screening
Interventions	A single‐session educational intervention was designed and tailored for ethnic‐ and gender‐specific groups to increase men's social support for their female wives/partners to receive a Pap test, and for women to receive a Pap test.
Outcomes	Ethnic‐ and gender‐tailored community interventions can successfully increase Pap test behaviours for Pacific Island women.
Notes

Valdez 2018.

Methods	RCT
Participants	Low‐income Latino women based in the USA
Interventions	Intervention group was exposed to a multimedia kiosk on cervical cancer and screening while control group had usual care.
Outcomes	While the intervention group participants had more knowledge and were less averse to a smear, there was no difference in the uptake of screening between the 2 groups.
Notes

Wong 2019.

Methods	RCT pilot study
Participants	Asian women eligible for cervical screening
Interventions	This study aimed to investigate the feasibility and preliminary effects of a CHW‐led multimedia intervention on cervical cancer screening.
Outcomes	No significant difference was noted in screening uptake and screening intention between arms.
Notes

CHW: Community Health Worker
NHS: National Health Service
PRECEDE: Predisposing, Reinforcing, and Enabling Constructs in Educational Diagnosis and Evaluation
RCT: Randomised Control Trial
SMS: Short Message Service

Differences between protocol and review

For cluster‐randomised controlled trials, if the analysis did not account for the cluster design, we had planned to extract the number of clusters randomised to each intervention, the average cluster size in each intervention group and the outcome data, ignoring the cluster design, for all women in each group. We had then planned to use an external estimate of the ICC to estimate a design effect to inflate the variance of the effect estimate, making it possible to combine the cluster‐randomised trials with individually randomised trials in the same meta‐analysis, using the generic inverse variance method of meta‐analysis. However, the majority of cluster‐RCTs did not account for clustering or adequately report the number of clusters in the trial in order to estimate the design effect. We therefore did not selectively adjust the trials on this basis. The results should not be biased by arm; just the precision will be greater than it should be, thus, 95% CIs would be wider in reality. It is unlikely that selectively adjusting the trials with adequate reporting of clustering would make much difference to the overall conclusions and robustness of the results.

Contributions of authors

In the original review, Carol Forbes and Ruth Jepson selected the studies, assessed study quality, extracted and analysed the data, and wrote the protocol and final review. Pierre Martin‐Hirsch commented on the protocol and will be responsible for updating the review.

In the previous version of the review, Thomas Everett, Pierre Martin‐Hirsch and Michelle Griffin drafted the clinical sections of the review and were responsible for sifting of studies and data extraction; Andrew Bryant drafted the methodological and statistical sections of the review and was responsible for carrying out the analyses.

In this updated review, all authors were involved with the sifting process, Aslam Shiraz, Norman Shreave and Helen Staley were responsible for data extraction; and any disagreements were discussed with Ketan Gajjar. Aslam Shiraz was responsible for updating the introduction and methods and Helen Staley carried out the analyses and drafted the results and conclusions. Andrew Bryant reviewed the statistical sections of the review. All authors agreed on the final version.

Sources of support

Internal sources

• NHS Centre for Reviews & Dissemination, UK

External sources

• NHS R&D Health Technology Assessment (HTA) Programme, UK

• Department of Health, UK

  NHS Cochrane Collaboration programme Grant Scheme CPG‐506

Declarations of interest

Helen Staley ‐ none known
Aslam Shiraz ‐ none known
Norman Shreeve ‐ none known 
Andrew Bryant ‐ none known
Pierre PL Martin‐Hirsch ‐ none known 
Ketankumar Gajjar ‐ none known

New search for studies and content updated (no change to conclusions)