Medical abortion offered in pharmacy versus clinic‐based settings

Abstract

Background

Medical abortion is usually offered in a clinic or hospital, but could potentially be offered in other settings such as pharmacies. In many countries, pharmacies are a common first point of access for women seeking reproductive health information and services. Offering medical abortion through pharmacies is a potential strategy to improve access to abortion.

Objectives

To compare the effectiveness and safety of medical abortion offered in pharmacy settings with clinic‐based medical abortion.

Search methods

We searched CENTRAL, MEDLINE, Embase, four other databases, two trials registries and grey literature websites in November 2020. We also handsearched key references and contacted authors to locate unpublished studies or studies not identified in the database searches.

Selection criteria

We identified studies that compared women receiving the same regimen of medical abortion or post‐abortion care in either a clinic or pharmacy setting. Studies published in any language employing the following designs were included: randomized trials and non‐randomized studies including a comparative group.

Data collection and analysis

Two review authors independently reviewed both retrieved abstracts and full‐text publications. A third author was consulted in case of disagreement. We intended to use the Cochrane risk of bias tool, RoB 2, for randomized studies and used the ROBINS‐I tool (Risk Of Bias In Non‐randomized Studies of Interventions) to assess risk of bias in non‐randomized studies. GRADE methodology was used to assess the certainty of the evidence. The primary outcomes were completion of abortion without additional intervention, need for blood transfusion, and presence of uterine or systemic infection within 30 days of medical abortion.

Main results

Our search yielded 2030 records. We assessed a total of 89 full‐text articles for eligibility. One prospective cohort study met our inclusion criteria.

The included study collected data on outcomes from 605 women who obtained a medical abortion in Nepal from either a clinic or pharmacy setting. Both sites of care were staffed by the same auxiliary nurse midwives. Over all domains, the risk of bias was judged to be low for our primary outcome. During the pre‐intervention period, the study’s investigators identified a priori appropriate confounders, which were clearly measured and adjusted for in the final analysis.

For women who received medical abortion in a pharmacy setting, compared to a clinic setting, there may be little or no difference in complete abortion rates (adjusted risk difference (RD)) 1.5, 95% confidence interval (CI) ‐0.8 to 3.8; 1 study, 600 participants; low certainty evidence). The study reported no cases of blood transfusion, and a composite outcome, comprised mainly of infection complications, showed there may be little or no difference between settings (adjusted RD 0.8, 95% CI ‐1.0 to 2.8; 1 study, 600 participants; very low certainty evidence). The study reported no events for hospital admission for an abortion‐related event or need for surgical intervention, and there may be no difference in women reporting being highly satisfied with the facility where they were seen (38% pharmacy versus 34% clinic, P = 0.87; 1 study, 600 participants; low certainty evidence).

Authors' conclusions

Conclusions about the effectiveness and safety of pharmacy provision of medical abortion are limited by the lack of comparative studies. One study, judged to provide low certainty evidence, suggests that the effectiveness of medical abortion may not be different between the pharmacy and clinic settings. However, evidence for safety is insufficient to draw any conclusions, and more research on factors contributing to potential differences in quality of care is needed. It is important to note that this study included a care model where a clinician provided services in a pharmacy, not direct provision of care by pharmacists or pharmacy staff. Three ongoing studies are potentially eligible for inclusion in review updates. More research is needed because pharmacy provision could expand timely access to medical abortion, especially in settings where clinic services may be more difficult to obtain. Evidence is particularly limited on the patient experience and how the care process and quality of services may differ across different types of settings.

Plain language summary

Medical abortion offered in pharmacy versus clinic‐based settings

Why this review is important 

Medical abortion is offered routinely in clinics and hospitals, but could be offered in other settings such as pharmacies. In many countries, pharmacies are a first and common point of access for women seeking reproductive health information and services, including abortion. Expanding access to medical abortion through pharmacies is a potential strategy to promote safe abortion care.

How did we identify and evaluate the evidence?

We searched seven medical research databases for randomized controlled trials, and websites for grey literature (i.e. research produced by organizations outside of the traditional commercial and academic publishing and distribution channels). In addition, we handsearched key references and contacted authors to locate unpublished studies or studies not identified in the database searches.

We identified studies that compared women receiving the same medication and dosage for medical abortion or post‐abortion care in either a clinic or pharmacy setting. We included studies published in any language, including the following designs: randomized trials and non‐randomized studies that included a comparison group.

We read and evaluated all abstracts and full‐text articles, and rated our confidence in the evidence, based on factors such as study methods and sizes.

What did we find? 

We found 2030 records. We screened the retrieved abstracts, and applied exclusion criteria. We assessed a total of 89 full‐text articles for eligibility. One prospective cohort study met our inclusion criteria. In this study, 605 women in Nepal received medical abortions from the same health care providers (auxiliary nurse midwives) in either a clinic or pharmacy‐based setting. There was no difference in complete abortion rates between the two different abortion settings. We also examined rates of blood transfusion and infection within 30 days of medical abortion. These outcomes were rare and the evidence was limited for drawing conclusions about differences by site. Additional secondary outcomes included hospital admission for an abortion‐related event, additional surgical interventions needed (besides uterine aspiration), and measures of quality of care. No hospital admissions or additional surgical procedures occurred within either group, and information about quality of care was limited.

What does this mean? 

A single non‐randomized study provides us with low certainty that the effectiveness of medical abortion probably does not differ between the pharmacy or clinic setting when the care is provided by the same clinicians. Three ongoing studies are potentially eligible for inclusion in an update of this review. Conclusions about the effectiveness, safety and quality of care of pharmacist provision of medical abortion are limited by the lack of comparative studies. More research is needed because pharmacy provision could expand timely access to medical abortion, especially in settings where clinic services may be more difficult to obtain.

How up‐to‐date is this review?

The evidence in this Cochrane Review is current to November 2020.

Summary of findings

Summary of findings 1. Critical outcomes for medical abortion offered in pharmacy versus clinic‐based settings.

Outcomesa	Illustrative comparative risksb (95% CI)		Adjusted between group risk difference  (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE) Comments
	Corresponding risk	Assumed risk
	Pharmacy	Clinic
Complete abortion	98.6% (293/297)	97.3% (295/303)	1.5 (‐0.8 to 3.8)c	600	Lowd
Blood transfusione	0.0% (0/297)	0.0% (0/303)	‐‐‐	600	Very lowf
Complication requiring treatment (infection)e	1.7% (5/297)	0.7% (2/303)	0.8 (‐1.0 to 2.8)c	600	Very lowf
Quality of medical abortion (Highly satisfied with services at facility)	38.4% (114/297)g	35.4% (107/303)	Not reported P value for comparison between groups from adjusted model = 0.87	600	Lowd
Additional surgical interventions required	Not reported

^aAll outcomes were assessed within 30 days of medical abortion.
^bActual absolute event rates reported in the study groups.
^cMixed models accounted for clustering by site and provider, and adjusted for age, education, parity, prior contraceptive use and gestation.
^dWell‐designed observational study with no serious concerns for bias, downgraded to low for indirectness. Confounding factors were considered and adjusted for across all outcomes. As findings are from a single study in one country, 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.
^eDefined as any complication requiring treatment based on provider assessment. The study reported that this primarily consisted of antibiotics for possible infection. The composite outcome did not include serious adverse events such as haemorrhage needing blood transfusion or conditions requiring hospitalization.
^fDowngraded for imprecision given low number of events and indirectness. We are very uncertain about the estimate.
^g(n/N) are calculated.

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.

Background

Description of the condition

Unsafe abortion remains a significant threat to women’s lives and health (Alkema 2016; Ganatra 2017; WHO 2016). The World Health Organization (WHO) estimates that globally, 25 million unsafe abortions occur every year. Unsafe abortion is the fifth leading cause of maternal mortality (Ganatra 2017).

Improving access to medical abortion is one strategy to reduce unsafe abortion, particularly where trained surgical abortion providers are limited. A growing proportion of abortions globally are medical abortions (Jones 2017; United Nations Population Fund 1994). The WHO has published guidance on effective regimens for medical abortion, and interventions such as laboratory testing or ultrasound are not universally required (WHO 2012).

Description of the intervention

Medical abortion is offered routinely in clinics and hospitals, but could be offered in other settings such as pharmacies. The safety and effectiveness of medical abortion provision by non‐physician clinicians, such as nurses and auxiliary nurse midwives, has been established (Olavarrieta 2015; Warriner 2011). In many countries, pharmacies are a first and common point of access for women seeking reproductive health information and services, including abortion (Billings 2009; Footman 2018; Sneeringer 2012). Expanding access to medical abortion through pharmacies is a potential strategy to promote safe abortion care.

The safety and effectiveness of services obtained from pharmacies relative to other clinical sites are not known.  Pharmacies are not routinely equipped to be able to offer physical exam services, including a pelvic exam or ultrasound. Screening criteria may vary between the two types of sites. It is possible that lower‐quality information or products may be supplied in pharmacies compared with clinics, increasing the rates of incomplete abortion or other complications.

Currently, medical abortion is usually offered in clinics and hospitals. Research to date has shown that a medical abortion regimen combining mifepristone with misoprostol is most effective. However, there is some variation in recommendations related to dose, timing and route of administration of the two drugs. A large body of evidence, and recommendations by the WHO, support the efficacy of a 200 mg dose of mifepristone followed by 800 µg of misoprostol in pregnancies up to 63 days’ gestational age (Raymond 2013; WHO 2014). Recent data support extending its use up to 70 days’ gestation (Abbas 2015). These protocols are highly effective and safe, with unsuccessful abortion resulting in approximately 2% to 5% of cases (Kulier 2011; Raymond 2013). In settings where mifepristone is not available, medical abortion is carried out using only misoprostol.

The recommended misoprostol regimen is 800 μg administered vaginally or sublingually (under the tongue), and repeated at intervals of no less than three hours but no more than 12 hours, for up to three doses. This regimen is 75% to 90% effective in completing abortions up to 84 days’ gestation. Gestational age is known to affect the efficacy of all regimens, with decreasing efficacy after nine weeks’ gestation (Winikoff 2008), which is why regimens for these gestations recommend repeating misoprostol doses. This review will focus on medical abortion provided with mifepristone and misoprostol or misoprostol‐alone regimens. 

How the intervention might work

Pharmacies may be able to improve access to safe and effective medical abortion care. Pharmacies are utilized for their convenience, anonymity and low cost, compared to a traditional health clinic or hospital (Ahmed 2007; Footman 2018). They may improve access for women with limited autonomy, or those living in rural areas, where clinical access is remote (Rocca 2018). Trained pharmacists and pharmacy workers deliver care related to a range of reproductive health conditions, including sexually transmitted infections, emergency contraception and provision of other family planning methods, such as birth control pills (Sneeringer 2012). Pharmacists have been successful in delivering reproductive health care because of their ability to provide quick access to necessary information, medications and referrals, while maintaining confidentiality (Gonsalves 2017). Clinicians could also provide services within pharmacies.

It is not known how pharmacist provision of medical abortion may impact important safety or efficacy outcomes, compared with the clinical setting. Pharmacists may have less training than clinicians in accurate usage of the medications to achieve a complete abortion. Pharmacists may have less access to routine physical exam services, altering the eligibility criteria for women to receive care. It is possible that pharmacists working in retail settings would have less time than clinicians to counsel women on known side effects or possible complications, increasing the risk of infection or heavy bleeding leading to hospital attendance. It is important to explore how the setting of care provision (pharmacy versus clinic) impacts key safety and efficacy outcomes, including complete abortion, blood transfusion or hospital admission.

Why it is important to do this review

Globally, pharmacies play a key role in the formal and informal distribution of information and medications for abortion (Billings 2009; Footman 2018; Lara 2011; Reiss 2016; Reiss 2017; Sneeringer 2012; Tamang 2015; Tamang 2018). Existing data on the safety and efficacy of this practice are limited, and have demonstrated mixed results on the accuracy of information and medical abortion regimens provided by pharmacy workers (Ahmed 2007; Billings 2009; Footman 2018; Reiss 2016; Rocca 2018). Safe and effective abortion can reduce complications associated with unsafe abortion, and maternal mortality (Ganatra 2017; WHO 2012). Pharmacy provision of medical abortion may have the potential to reduce morbidity associated with unsafe abortion. However, evidence is needed to establish whether the safety and effectiveness of care is equivalent to that offered in a clinic.

Objectives

To compare the effectiveness and safety of medical abortion offered in pharmacy settings with clinic‐based medical abortion.

Methods

Criteria for considering studies for this review

Types of studies

We sought studies that compared women receiving the same regimen of medical abortion or post‐abortion care in either a clinic or pharmacy setting. We included studies published in any language employing the following designs: randomized trials (clustered or individually randomized); quasi‐experimental designs, such as non‐randomized controlled studies or stepped‐wedge design experiments; and cohort studies with a control group comparing services provided in clinical versus pharmacy settings.

For safe abortion, programmes will incorporate the woman’s right to choose her preferred mode of abortion. In other settings, studies may be carried out in countries where abortion access is restricted, and a randomized controlled trial (RCT) not possible. Thus, observational studies comparing pharmacy and clinic settings were included in this review to assess programmatic implementation and detect serious and uncommon harms.

Types of participants

Pregnant women of any age, seeking medical abortion care in pharmacies or traditional clinics, are eligible for inclusion.

Types of interventions

The intervention is pharmacy delivery of any component of medical abortion services. This includes dispensing medical abortion medications. Only studies that provided medical abortion using mifepristone and misoprostol or misoprostol alone were included. The only administration for mifepristone is oral. Different administration routes and dosing regimens for misoprostol may be used, and we included studies using any route (oral, sublingual, buccal and vaginal) or regimen (e.g. repeat dosing). We considered all types of providers in the intervention group (including clinicians, pharmacists and pharmacy workers) and in the comparison group (physician, nurse midwifes, auxiliary nurse midwifes, and nurses in the clinic setting). We did not consider surgical abortion outcomes in this review.

Comparison

Women receiving medical abortion in clinical health care settings compared with pharmacy settings.

Types of outcome measures

Primary outcomes

• Complete abortion, defined as completion of abortion within 30 days of taking the first medication, and not requiring surgical intervention for completion

• Blood transfusion within 30 days of medical abortion

• Uterine or systemic infection within 30 days of medical abortion

Secondary outcomes

• Hospital admission for an abortion‐related event

• Quality of medical abortion care (any study‐reported measures of technical or interpersonal quality of care were considered ‐ priority was placed on synthesis of validated measures if available; Darney 2018; Darney  2019)

• Additional surgical interventions (besides uterine aspiration)

Search methods for identification of studies

The Fertility Regulation Group Information Specialist conducted a search for all published, unpublished and ongoing studies, without restrictions on language or publication status. Database search strategies are available in Appendix 1. We considered adverse effects described in included studies only.

Electronic searches

We searched these databases from their inception to 14 May 2020:

• Cochrane Central Register of Controlled Trials (CENTRAL) via EBM Reviews (Ovid), including ClinicalTrials.gov and WHO ICTRP records

• MEDLINE ALL (Ovid)

• Embase.com (www.embase.com)

• CINAHL (EBSCOhost)

• LILACs (Latin American & Caribbean Health Sciences Literature; lilacs.bvsalud.org/en/)

• Global Health (Ovid)

• Scopus

Searching other resources

We checked the bibliographies of included studies and any relevant systematic reviews identified for further references to relevant studies. We contacted experts and organizations in the field to obtain additional information on relevant studies. 

We searched these grey literature websites:

• Guttmacher Institute (www.guttmacher.org/united-states/abortion)

• International Planned Parenthood Federation (www.ippf.org/)

• Ibis Reproductive Health (ibisreproductivehealth.org/)

• Women on Waves (www.womenonwaves.org/)

• Marie Stopes International (www.mariestopes.org/)

• Population Council (www.popcouncil.org/)

• Population Services International (www.psi.org/)

• Ipas (www.ipas.org/)

• Google Scholar (scholar.google.com/)

Data collection and analysis

Selection of studies

We downloaded all titles and abstracts retrieved by electronic searching to a reference management database, and removed duplicates (Covidence). Two review authors independently screened titles and abstracts for inclusion. We then retrieved the full‐text study reports or publications. Two review authors independently screened the full‐text publications, identified studies for inclusion, and identified and recorded reasons for exclusion of the ineligible studies. We resolved any disagreement through discussion.

We list studies that initially appeared to meet the inclusion criteria, but that we later excluded, in the 'Characteristics of excluded studies' section. We collated multiple reports of the same study, so that each study, rather than each report, is the unit of interest in the review. We also provide any information we could obtain about ongoing studies. We recorded the selection process in a PRISMA flow diagram (Liberati 2009).

Data extraction and management

Two review authors independently screened and extracted data from eligible studies, using a data extraction form designed and pilot‐tested by the review authors. We resolved any disagreements through discussion. We described study design, participant characteristics and outcome data in a Characteristics of included studies table. We recorded the drugs used, dose and route of administration, as well as the included study’s inclusion and exclusion criteria.

Assessment of risk of bias in included studies

For randomized trials of interventions, we planned to use the Cochrane Risk of Bias Assessment tool (Higgins 2019) to assess selection bias (random sequence generation and allocation concealment); performance bias (blinding of participants and personnel); detection bias (blinding of outcome assessors); attrition bias (incomplete outcome data); reporting bias (selective reporting); and other biases. Based on these assessments, we would have rated studies as at low risk, high risk or unclear risk of bias using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

For the included non‐randomized study (NRS), we conducted dual, independent assessment of risk of bias using the ROBINS‐I tool (Risk Of Bias In Non‐randomized Studies of Interventions; Sterne 2016). The domains assessed with this tool are bias at the pre‐, at‐, and post‐intervention stages of the study. Specifically, the domains are: pre‐intervention bias due to confounding (prognostic variables predict outcome of interest) and selection (inclusion or exclusion of participants related to outcome of interest); at‐intervention information bias (misclassification of intervention status); and, post‐intervention confounding, selection bias, information bias, and reporting bias. Use of the ROBINS‐I tool facilitated assessment of risk of bias for each domain. For the comparisons we evaluated, we expected selection bias and confounding pre‐ and post‐intervention are likely to be of particular importance. Based on responses to signalling questions for each domain, we made an overall judgment of low, moderate, serious or critical risk of bias. Gestational age is known to be an important confounder in medical abortion (Kahn 2000). We examined the presence of appropriate adjustment for gestational age in evaluating the risk of bias.

We used the GRADE approach to assess the certainty of the evidence for the primary outcomes from the included study. For the synthesis of outcomes drawn from non‐randomized studies, the evidence begins with a rating of low certainty, given the risks of bias from selection and confounding inherent in non‐randomized study designs. With the ROBINS‐I tool, however, certainty may be upgraded if effects are particularly strong and the risks of confounding and selection are judged to be particularly well‐mitigated.

Measures of treatment effect

For dichotomous trial data (e.g. complete abortion, yes/no), we would have used the number of events in the clinic and pharmacy groups of each study to calculate relative risks (RR) or Mantel‐Haenszel odds ratios (ORs), depending on the most commonly reported effect estimates across the body of evidence. Similarly, we would have extracted reported means and standard deviations for continuous outcomes, either as reported in the primary study or calculated from reported estimates of variance to calculate mean difference with 95% confidence intervals. For the included non‐randomized study, we prioritized study‐reported adjusted effect estimates in our synthesis and selected the estimate judged to most minimize the risk of bias due to confounding, study design, and selection. We present 95% confidence intervals (CI) as the measure of precision for all outcome estimates. Where data to calculate ORs, RRs or mean differences were not available, we used the most detailed numerical data available to facilitate synthesis across included studies (e.g. test statistics, P values). We assessed whether the estimates we calculate in the review for the included study were consistent with the available estimates of effects reported in the study publication.

Unit of analysis issues

The primary unit of analysis will be per woman randomized for RCTs and per woman who undergoes medical abortion (classified as pharmacy or clinic‐administered) for non‐randomized studies. For any cluster‐RCTs included in the review, we planned to report trial outcome data adjusted for the hierarchical study design (i.e. within cluster correlations among observations that lead to underestimation of standard errors) whenever available and use these estimates in meta‐analysis. We planned to describe studies that did not report data with appropriate adjustments for study design in tables with the potential for overestimating effect precision noted. For meta‐analysis of unadjusted cluster‐RCT outcomes, we planned to use intracluster correlation coefficients (ICC) based on observations from similar studies to estimate adjustments to the standard errors following recommended procedures (Higgins 2019).

Dealing with missing data

We planned to analyze the data on an intention‐to‐treat (ITT) basis as far as possible and to contact authors to obtain missing data as needed. When unobtainable, we conducted analyses on the available data.

Assessment of heterogeneity

We planned to evaluate whether included studies were similar with respect to study participants, interventions, comparators and outcomes, and whether effects were relatively consistent with moderate statistical heterogeneity. We planned to assess statistical heterogeneity using Review Manager Web for meta‐analysis (Review Manger 2020), using Chi² tests and I² statistics and following recommendations in the Cochrane Handbook for Systematic Reviews of Interventions for interpreting I² values when estimable (Higgins 2019).

Assessment of reporting biases

Considering the difficulties in detecting and correcting for publication bias and other reporting biases, we aimed to minimize their potential impact by ensuring a comprehensive search for eligible studies, and by being alert to duplication of data. We planned to use a funnel plot to explore the possibility of small‐study effects if at least 10 studies were available for pooled analyses.

Data synthesis

Had we obtained multiple, comparable studies, we would have synthesized intervention effectiveness in a meta‐analysis using a random‐effects model, to produce pooled OR, RR, or mean difference effect estimates with 95% confidence interval (CI). We selected this model a priori to address the likely effects of trial heterogeneity and diversity in the evidence on this topic. We planned to conduct quantitative synthesis separately for randomized and non‐randomized study evidence using the DerSimonian and Laird random‐effects model (DerSimonian 1986). These summary effects are grounded in the assumption that the pooled estimate is an average effect from an underlying distribution of true effects. For meta‐analysis of non‐randomized studies, we would have sought to pool adjusted effect estimates using the generic inverse variance approach. If only unadjusted estimates were available, they would be pooled separately from adjusted effects. We planned, if possible, to present forest plots showing the pooled estimates and 95% CI for each outcome suitable for meta‐analysis.

We conducted a narrative synthesis for outcomes lacking adequate data to combine studies. This synthesis considered the consequences of possible incomplete reporting on the outcomes of interest, and the strengths and limitations of available studies for evaluating the review questions.

Subgroup analysis and investigation of heterogeneity

There may be differences in the effectiveness and safety of pharmacy‐provided medical abortion depending on several factors, including the following, which we defined in advance of the review to be important to evaluate in subgroup comparisons: type of health workers providing abortion care in pharmacies (physicians, midwives, nurses, pharmacists, medical assistants), client characteristics (e.g. gestation less than nine weeks, parity), abortion regimen (e.g. dosage, administration route), and the Human Development Index category (hdr.undp.org/en/content/human-development-index-hdi) of the country where the study was conducted.

In the presence of adequate data for subgroup testing, we planned to produce stratified forest plots with pooled subgroup effect estimates and statistical tests for interaction using meta‐regression. Additionally, if substantial statistical heterogeneity was not explained by these factors, we planned to explore additional possible explanations in post hoc comparisons derived from the available evidence and its synthesis.

Sensitivity analysis

As needed, to more fully understand and evaluate the body of evidence, we planned to perform sensitivity analyses to assess the effect of risk of bias pooled estimates for primary outcomes by removing included studies rated as at high or critical risk of bias.

Summary of findings and assessment of the certainty of the evidence

We used GRADEpro and Cochrane methods to prepare a ‘Summary of findings’ table (GRADEpro GDT; Higgins 2019). The table was used to evaluate the overall certainty of the body of evidence for the review outcomes on effectiveness and safety of medical abortion provided in pharmacy settings. We used the GRADE criteria (e.g. risk of bias, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of the evidence (Guyatt 2008). As noted above, initially, we rated the non‐randomized study evidence as low certainty and the rating was further refined based on the ROBINS‐I risk of bias assessments.

Two review authors worked independently to judge the evidence certainty (e.g. high, moderate, low or very low) and resolved inconsistencies through discussion. The reviewers recorded notes to justify, document and incorporate their judgments into reporting the results of each outcome.

Results

Description of studies

Results of the search

Our search study yielded 2032 papers from 2030 records (Figure 1). After removal of duplicates, 1240 studies remained. We screened these abstracts and applied exclusion criteria. We then assessed a total of 89 full‐text articles for eligibility. One study met our inclusion criteria.

1.

Literature flow diagram

Included studies

A non‐inferiority analysis of a non‐randomized comparison study enrolled 605 women in Nepal who were seeking medical abortion in six public clinics or six pharmacies across two regions of Nepal (Rocca 2018). All participants received 200 mg of mifepristone orally, and 800 µg of misoprostol (sublingually or intravaginally) 24 hours later. Follow‐up occurred at 14 to 21 days. The intervention was provision of abortion medications by six auxiliary nurse midwives at pharmacies, as compared with the standard clinic setting. The study was funded by the Society of Family Planning Research fund.

Ongoing studies 

We identified three ongoing studies that are potentially eligible, but have not been incorporated into the review. One is a mixed methods study enrolling 260 participants in two states in the USA to determine the feasibility, acceptability and effectiveness of dispensing mifepristone in a pharmacy rather than the clinic. The primary outcome is the number of pharmacists who refuse to dispense mifepristone (Grossman ongoing a). A second in‐process clinical trial was identified. This prospective cohort study plans to recruit 425 women requesting medical abortion in the USA, and compare satisfaction with utilizing a mail order pharmacy versus clinic for dispensing of medications (Grossman ongoing b). A third ongoing study of 4000 participants is being conducted in Ghana and Cambodia. Women who source medications for abortion in pharmacies are being compared to women who receive medical abortions in clinics, and followed to determine whether there is a difference in subsequent interventions needed to complete the abortion (Ipas ongoing).

Excluded studies

We reviewed 88 full‐text study publications, and excluded these from the review (Figure 1). Of these, we excluded 48 studies due to being the wrong study design. We excluded an additional 24 studies because they focused on a population different than our review’s focus. These studies typically focused on pharmacists' attitudes or practices. We excluded seven studies because they did not have the correct intervention, and four because they had the wrong comparator. We excluded three studies because they had the wrong setting and a final report because it consisted of preliminary findings from the included full study.

Risk of bias in included studies

We used the ROBINS‐I tool to assess the risk of bias for the primary outcomes in the non‐randomized study identified. We evaluated risk of bias at distinct study time points: pre‐intervention (confounding and selection bias); at intervention (information bias, misclassification of intervention); and post‐intervention (confounding, selection and reporting bias). Our ROBINS‐I assessments spreadsheet is stored on the review group cloud storage and can be made available upon request by emailing CochraneFRG@ohsu.edu.

Over all domains, we judged the risk of bias to be low for primary outcomes. During the pre‐intervention period, appropriate confounders were identified a priori and were clearly measured and adjusted for in the final analysis (Table 2).

1. Summarizing data without meta‐analysis.

Outcome1	Pharmacy events/total n (%)	Public health facility events/total n (%)	Adjusted between group risk difference (95% CI)
Complete abortion	293/297 293 (99%)	295/303 295 (97%)	1.5 (‐0.8 to 3.8)
Complication requiring treatment (infection)2	5/297 5 (1.7%)	2/303 2 (0.7%)	0.8 (‐1.0 to 2.8)
Abdominal cramping/pain	277/301 277 (92%)	277/304 277 (91%)	NR3 P = 0.82
Nausea	186/301 186 (62%)	183/304 183 (60%)	NR3 P = 0.68
Highly satisfied with services at facility	105 (38%)	107 (34%)	NR3 P = 0.87
Preferred to come to same facility type for future services if needed	284 (94%)	297 (97%)	NR3 P < 0.05

¹All outcomes were assessed within 30 days of medical abortion.

² Mixed models accounted for clustering by site and provider, and adjusted for age, education, parity, prior contraceptive use, and gestation.Defined as any complication requiring treatment based on provider assessment. The study reported that this primarily consisted of antibiotics for possible infection. The composite outcome did not include serious adverse events such as haemorrhage needing blood transfusion or conditions requiring hospitalization.

³Results did not report adjusted RD, instead P values for the differences from adjusted analysis are provided.

Effects of interventions

See: Table 1

We identified one study from the published literature (Rocca 2018). A total of 605 women were enrolled, with 297 in the pharmacy arm, and 303 in the clinical arm obtaining medical abortion. Services were provided by a clinician (auxiliary nurse midwives) in the pharmacy setting. Follow‐up data were available for 600 of the study participants (99%). The primary outcome was complete abortion within 30 days of taking medication, without surgical intervention. The study computed mixed effects models to estimate the adjusted risk differences between study groups that accounted for the study design (site and provider clustering) and potential confounding by participant age, education, parity, prior contraceptive use, and gestation at the time of abortion.

Abortion completion

The proportion with complete abortion was not associated with the abortion setting (Table 2; Table 1). The proportion of complete abortion was similar in the pharmacy setting (98.7%) and the clinical setting (97.4%) with adjusted estimates of the risk difference crossing null (adjusted RD 1.5, 95% CI ‐0.8 to 3.8; 600 participants; low certainty of evidence).

Post‐abortion complications

No serious adverse events were reported in either the pharmacy or clinical settings within 30 days of medical abortion. The study reported a composite variable that captured complications defined as those warranting treatment based on provider assessment. These were noted by the study authors to be primarily antibiotics for fever and potential infection. The adjusted risk difference between groups was small and ranged from 1% percent fewer to 3% more complications in the pharmacy setting (adjusted RD 0.8, 95% CI ‐1.0 to 2.8; 600 participants; very low certainty of evidence).

Additional secondary outcomes included abortion symptoms such as cramping and pain, hospital admission for an abortion‐related event and additional surgical interventions needed (besides uterine aspiration). Proportions were similar, and no differences between groups were observed for any of these outcomes in adjusted analyses (Table 1).

Quality of abortion care

Quality of care outcomes reported in the study included receipt of post‐abortion contraception, satisfaction with services at the facility (highly satisfied, satisfied, not satisfied), and preference for returning to the same facility for future services as needed. No statistical differences in satisfaction with services or receipt of contraceptive care were found between settings, with 38% of women obtaining care in the pharmacy setting highly satisfied and 35% of women seen in the clinical setting reporting this level of satisfaction (P = 0.82).

Certainty of results

The overall direction of effect suggests that providing medical abortion in a pharmacy, as compared to a clinic setting, does not reduce effectiveness or safety of abortion. The certainty of the evidence was maintained at low for our primary outcome, complete abortion, because it was a well‐designed observational study with no serious sources of bias beyond the observational design and some indirectness owing to the availability of evidence from only one country and the use of the same providers at both types of settings. The quality of the evidence was downgraded one additional level (from low to very low) for other outcomes due to additional concerns regarding imprecision since there were very few or no events reported.

Discussion

The safety and effectiveness of medical abortion provision through non‐physician clinicians, such as nurses and auxiliary nurse midwives, has been established (Olavarrieta 2015; Warriner 2011). The safety and effectiveness of services obtained from pharmacies relative to other clinical sites are not known. Expanding access to medical abortion through pharmacies is a potential strategy to promote safe abortion care. In many countries, pharmacies are a first and common point of access for women seeking reproductive health information and services, including abortion (Billings 2009; Footman 2018; Sneeringer 2012). In this review, we assessed the evidence regarding the safety and efficacy of medical abortion offered to pregnant women at any gestational age in pharmacy settings as compared with clinic‐based medical abortion.

Summary of main results

We identified one study of 605 women conducted in Nepal that met inclusion criteria for our main outcome analysis, with 297 women in the pharmacy arm, and 303 in the clinical arm (Rocca 2018). Our primary outcome was complete abortion within 30 days of taking medication, without surgical intervention. We found no difference in abortion effectiveness or complications by setting based on limited evidence (see Table 1).

Overall completeness and applicability of evidence

We identified only one study which reported on primary outcomes that were judged to provide low and very low certainty results, but with no serious quality concerns related to risk of bias (Rocca 2018). This study reported on all three of our primary outcomes: complete abortion, blood transfusion and abortion complication requiring hospitalization. Data on additional secondary outcomes are described in this review (quality of medical care and additional surgical interventions), but not included in the GRADE summary of findings.

Quality of the evidence

We identified one study meeting inclusion criteria. We reviewed the certainty of the evidence for each of our critical outcomes using the GRADE process; these findings are summarized in Table 1. Only one non‐randomized intervention study was included for our main outcome, complete abortion. As a non‐randomized trial, the certainty is rated 'low' by GRADE methodology. Although this was a very well‐designed observational study with no serious concerns for bias, we did not upgrade the certainty due to indirectness. Because the data are from one country, they reflect outcomes in a  particular health care system and population health context. The results may not translate well to other settings. In addition, the included study was designed to evaluate whether the same clinicians providing care in health clinics could provide similarly effective and safe care through pharmacies. Thus, the study provides results that are somewhat indirect for answering the broader question with regard to differences in care provided at pharmacies without necessarily adding clinic‐level staffing, versus health clinic settings.

All study outcomes were measured at no later than 30 days after obtaining a medication abortion. This follow‐up time is reasonable for measuring key effectiveness and safety outcomes.

For all secondary outcomes (blood transfusion; infection; hospital admission; need for additional surgical interventions) the evidence began at 'low' due to the non‐randomized design. We further downgraded this assessment one level due to the significant concerns for imprecision of secondary outcomes (rare outcomes) in addition to the concerns outlined above.

Data on general satisfaction with care at the facilities where the participants obtained their abortion were reported in the study, but more specific aspects of the quality of care in terms of information, patient support and accessibility were not obtained in our review of the evidence. We did not downgrade this patient‐reported outcome because the reporting was sufficiently complete and the outcome was not a rare event.

Potential biases in the review process

Our review of the literature adhered to robust methods including the use of the ROBINS‐I instrument for assessing risk of bias in non‐randomized studies. We employed independent dual review at all stages of the review process. We adhered to the original protocol in the selection of studies and abstraction of reported study outcomes. We are not aware of any biases in the review process but recognize the possibility that potentially relevant studies published in languages other than English or published in journals not indexed in the databases searches could have been missed.

Agreements and disagreements with other studies or reviews

We did not identify any existing systematic reviews on this topic.

Authors' conclusions

Implications for practice.

The evidence for the success of medical abortion administered in pharmacies, rather than clinic settings, is supported by one non‐randomized comparative study.

The interpretation of this study, as it applies broadly to medical abortion provision, is limited by the study setting (one country) and having the same provider type provide the services in both intervention and comparison settings. While this study suggests that medical abortion may be effectively provided in the pharmacy setting, more information is needed on the safety of this practice and how efficacy may vary when diverse cadres of healthcare workers provide the counseling (e.g. pharmacy staff versus clinical staff).

Conclusions about the effectiveness and safety of pharmacy provision of medical abortion are limited by the lack of comparative studies. More research is needed because pharmacy provision could expand timely access to medical abortion, especially in rural settings where clinic services may be more difficult to obtain.

Implications for research.

The need for further research on this topic is readily apparent given the low yield of our systematic review. Studies from diverse settings that evaluate the effectiveness and safety of medical abortion provided in pharmacies as compared to clinics would strengthen the certainty of our evidence base and improve generalizability of our findings.

It is important to note that the one study included in our review used the same trained healthcare providers (auxiliary nurse midwives) to provide medical abortions in both settings: the pharmacies as well as the clinics. Replication of this study in different countries, where other types of healthcare providers routinely staff pharmacies (e.g. pharmacy staff, pharmacists, nurses), and comparing effectiveness and safety of medical abortion with clinic‐based care would be valuable. A well‐designed non‐randomized study or randomized trial would add meaningfully to the evidence base.

Randomized intervention studies can be highly challenging, especially in the context of abortion provision where individuals may have strong preferences about their care. However, it is possible that a future randomized study could be conducted in a country where pharmacists routinely provide a range of medical services and counselling. Participants in these communities might be willing to be randomized to the clinic or pharmacy setting.

History

Protocol first published: Issue 3, 2020

Appendices

Appendix 1. Search Strategies

Cochrane Central Register of Controlled Trials via EBM Review (Ovid) April 2020
Date searched: 14 May 2020

1 (abortifacient* or abortion* or (menstrua* adj3 regulat*) or pre‐abortion or preabortion or post‐abortion or postabortion or post‐abortum or postabortum or feticid* or foeticid* or ((medical* or medication or medicin* or trimester* or gestation* or pregnan*) adj5 (post‐terminat* or postterminat* or pre‐terminat* or preterminat* or terminat*))).ti,ab. (4799)
2 (Mifepristone or Misoprostol or Abo‐pill or Colestone or Cytotec or Elmif or Epostane or Fenprostalene or GyMiso or Korlym or Medabon or Mefeprin or Mefipil or Mifebort or Mifegest or Mifegyne or Mifeprex or Miferiv or Mifty or Mtpill or Nalador or RU‐38486 or RU38486 or RU‐486 or RU486 or T‐Pill or Termipil).ti,ab. (3958)
3 or/1‐2 (7580)
4 (apothecar* or chemist* or dispens* or druggist* or drugstore* or pharmacy or pharmacies or pharmacist* or OTC or over‐the‐counter or ((drug or medicine) adj4 (retail* or seller* or shop* or store* or vendor*))).ti,ab. (15392)
5 and/3‐4 (71)

MEDLINE ALL (Ovid) 1946 to 13 May 2020
Date searched: 14 May 2020

1 Abortion, Induced/ or Abortion, Eugenic/ or Abortion, Legal/ or Abortion, Therapeutic/ or Abortion, Incomplete/ or Abortion Applicants/ or Abortion, Criminal/ or Abortifacient Agents/ or Abortifacient Agents, Nonsteroidal/ or Abortifacient Agents, Steroidal/ or Menstruation‐Inducing Agents/ (42592)
2 (abortifacient* or abortion* or (menstrua* adj3 regulat*) or pre‐abortion or preabortion or post‐abortion or postabortion or post‐abortum or postabortum or feticid* or foeticid* or ((medical* or medication or medicin* or trimester* or gestation* or pregnan*) adj5 (post‐terminat* or postterminat* or pre‐terminat* or preterminat* or terminat*))).tw,kf. (73817)
3 Mifepristone/ or Misoprostol/ (9344)
4 (Mifepristone or Misoprostol or Abo‐pill or Colestone or Cytotec or Elmif or Epostane or Fenprostalene or GyMiso or Korlym or Medabon or Mefeprin or Mefipil or Mifebort or Mifegest or Mifegyne or Mifeprex or Miferiv or Mifty or Mtpill or Nalador or RU‐38486 or RU38486 or RU‐486 or RU486 or T‐Pill or Termipil).tw,kf. (11546)
5 or/1‐4 (96099)
6 Pharmacies/ or Pharmacists/ or Pharmaceutical Services/ or Pharmacy Technicians/ or Community Pharmacy Services/ (29285)
7 (apothecar* or chemist* or dispens* or druggist* or drugstore* or pharmacy or pharmacies or pharmacist* or OTC or over‐the‐counter or ((drug or medicine) adj4 (retail* or seller* or shop* or store* or vendor*))).ti,ab,kf. (289955)
8 or/6‐7 (295354)
9 and/5,8 (656)

Embase.com
Date searched: 18 May 2020
#1 'induced abortion'/mj OR 'medical abortion'/mj OR 'illegal abortion'/mj OR 'legal abortion'/mj OR 'abortive agent'/mj OR 'menstruation‐inducing agent'/mj (17,215)
#2 abortifacient*:ti,ab,kw OR abortion*:ti,ab,kw OR ((menstrua* NEAR/3 regulat*):ti,ab,kw) OR 'pre abortion':ti,ab,kw OR preabortion:ti,ab,kw OR 'post abortion':ti,ab,kw OR postabortion:ti,ab,kw OR 'post abortum':ti,ab,kw OR postabortum:ti,ab,kw OR feticid*:ti,ab,kw OR foeticid*:ti,ab,kw OR (((medical* OR medication OR medicin* OR trimester* OR gestation* OR pregnan*) NEAR/5 ('post terminat*' OR postterminat* OR 'pre terminat*' OR preterminat* OR terminat*)):ti,ab,kw) (88,248)
#3 'mifepristone'/exp/mj OR 'misoprostol'/mj OR 'mifepristone plus misoprostol'/mj (8,998)
#4 mifepristone:ti,ab,kw OR misoprostol:ti,ab,kw OR 'abo pill':ti,ab,kw OR colestone:ti,ab,kw OR cytotec:ti,ab,kw OR elmif:ti,ab,kw OR epostane:ti,ab,kw OR fenprostalene:ti,ab,kw OR gymiso:ti,ab,kw OR korlym:ti,ab,kw OR medabon:ti,ab,kw OR mefeprin:ti,ab,kw OR mefipil:ti,ab,kw OR mifebort:ti,ab,kw OR mifegest:ti,ab,kw OR mifegyne:ti,ab,kw OR mifeprex:ti,ab,kw OR miferiv:ti,ab,kw OR mifty:ti,ab,kw OR mtpill:ti,ab,kw OR nalador:ti,ab,kw OR 'ru 38486':ti,ab,kw OR ru38486:ti,ab,kw OR 'ru 486':ti,ab,kw OR ru486:ti,ab,kw OR 't pill':ti,ab,kw OR termipil:ti,ab,kw (14,795)
#5 #1 OR #2 OR #3 OR #4 (104,662)
#6 'pharmacy shop'/mj OR 'hospital pharmacy'/mj OR 'pharmacy technician'/mj OR 'pharmacist'/mj OR 'clinical pharmacist'/mj OR 'community pharmacist'/mj OR 'hospital pharmacist'/mj (69,720)
#7 apothecar*:ti,ab,kw OR chemist*:ti,ab,kw OR dispens*:ti,ab,kw OR druggist*:ti,ab,kw OR drugstore*:ti,ab,kw OR pharmacy:ti,ab,kw OR pharmacies:ti,ab,kw OR pharmacist*:ti,ab,kw OR otc:ti,ab,kw OR 'over the counter':ti,ab,kw OR (((drug OR medicine) NEAR/4 (retail* OR seller* OR shop* OR store* OR vendor*)):ti,ab,kw) (427,942)
#8 #6 OR #7 (449,770)
#9 #5 AND #8 (851)

CINAHL Plus withFulltext(EBSCOhost)
Date searched: 18 May 2020
S1 (MH "Abortion, Criminal") OR (MH "Abortion, Induced") OR (MH "Abortion, Incomplete") (11,186)
S2 TI ( abortifacient* OR abortion* OR (menstrua* W3 regulat*) OR pre‐abortion OR preabortion OR post‐abortion OR postabortion OR post‐abortum OR postabortum OR feticid* OR foeticid* OR ((medical* OR medication OR medicin* OR trimester* OR gestation* OR pregnan*) W5 (post‐terminat* OR postterminat* OR pre‐terminat* OR preterminat* OR terminat*)) ) OR AB ( abortifacient* OR abortion* OR (menstrua* W3 regulat*) OR pre‐abortion OR preabortion OR post‐abortion OR postabortion OR post‐abortum OR postabortum OR feticid* OR foeticid* OR ((medical* OR medication OR medicin* OR trimester* OR gestation* OR pregnan*) W5 (post‐terminat* OR postterminat* OR pre‐terminat* OR preterminat* OR terminat*)) ) (13,419)
S3 (MH "Mifepristone") OR (MH "Misoprostol") (2,492)
S4 TI ( Mifepristone OR Misoprostol OR Abo‐pill OR Colestone OR Cytotec OR Elmif OR Epostane OR Fenprostalene OR GyMiso OR Korlym OR Medabon OR Mefeprin OR Mefipil OR Mifebort OR Mifegest OR Mifegyne OR Mifeprex OR Miferiv OR Mifty OR Mtpill OR Nalador OR RU‐38486 OR RU38486 OR RU‐486 OR RU486 OR T‐Pill OR Termipil ) OR AB ( Mifepristone OR Misoprostol OR Abo‐pill OR Colestone OR Cytotec OR Elmif OR Epostane OR Fenprostalene OR GyMiso OR Korlym OR Medabon OR Mefeprin OR Mefipil OR Mifebort OR Mifegest OR Mifegyne OR Mifeprex OR Miferiv OR Mifty OR Mtpill OR Nalador OR RU‐38486 OR RU38486 OR RU‐486 OR RU486 OR T‐Pill OR Termipil ) (2,446)
S5 S1 OR S2 OR S3 OR S4 (19,199)
S6 (MH "Pharmacy, Retail") OR (MH "Pharmacy Technicians") OR (MH "Pharmacy Service") OR (MH "Pharmacists") (25,831)
S7 TI ( apothecar* OR chemist* OR dispens* OR druggist* OR drugstore* OR pharmacy OR pharmacies OR pharmacist* OR OTC OR over‐the‐counter OR ((drug OR medicine) W4 (retail* OR seller* OR shop* OR store* OR vendor*)) ) OR AB ( apothecar* OR chemist* OR dispens* OR druggist* OR drugstore* OR pharmacy OR pharmacies OR pharmacist* OR OTC OR over‐the‐counter OR ((drug OR medicine) W4 (retail* OR seller* OR shop* OR store* OR vendor*)) ) (54,312)
S8 S6 OR S7 (62,997)
S9 S5 AND S8 (275)

Global Health (Ovid) 1973 to 2020 Week 18
Date searched: 14 May 2020       

1 (abortifacient* or abortion* or (menstrua* adj3 regulat*) or pre‐abortion or preabortion or post‐abortion or postabortion or post‐abortum or postabortum or feticid* or foeticid* or ((medical* or medication or medicin* or trimester* or gestation* or pregnan*) adj5 (post‐terminat* or postterminat* or pre‐terminat* or preterminat* or terminat*))).ti,ab. (11889)
2 (Mifepristone or Misoprostol or Abo‐pill or Colestone or Cytotec or Elmif or Epostane or Fenprostalene or GyMiso or Korlym or Medabon or Mefeprin or Mefipil or Mifebort or Mifegest or Mifegyne or Mifeprex or Miferiv or Mifty or Mtpill or Nalador or RU‐38486 or RU38486 or RU‐486 or RU486 or T‐Pill or Termipil).ti,ab. (869)
3 or/1‐2 (12418)
4 (apothecar* or chemist* or dispens* or druggist* or drugstore* or pharmacy or pharmacies or pharmacist* or OTC or over‐the‐counter or ((drug or medicine) adj4 (retail* or seller* or shop* or store* or vendor*))).ti,ab. (28484)
5 and/3‐4 (160)
        

LILACS
Date searched: May 19, 2020

ABSTRACT (abortion OR abortions OR abortifacient OR abortifacients OR Mifepristone OR Misoprostol OR Abo‐pill OR Colestone OR Cytotec OR Elmif OR Epostane OR Fenprostalene OR GyMiso OR Korlym OR Medabon OR Mefeprin OR Mefipil OR Mifebort OR Mifegest OR Mifegyne OR Mifeprex OR Miferiv OR Mifty OR Mtpill OR Nalador OR RU‐38486 OR RU38486 OR RU‐486 OR RU486 OR T‐Pill OR Termipil) AND ABSTRACT (apothecary OR apothecaries OR chemist OR chemists OR dispensary OR dispensaries OR druggist OR druggists OR drugstore OR drugstores OR pharmacy OR pharmacies OR pharmacist OR pharmacists OR technician OR technicians OR OTC OR over‐the‐counter OR retailer OR retailers OR seller OR sellers OR shop OR shops OR store OR stores OR vendor OR vendors) (19)

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Rocca 2018.

Study characteristics
Methods	Observational non‐inferiority study
Participants	N = 605 "Eligible women were ≤ 63 days pregnant, aged 16‐45, and had no medical contraindications."
Interventions	"200 mg mifepristone orally and 800 μg misoprostol sublingually or intravaginally 24 hours later, and followed‐up 14‐21 days later."
Outcomes	Complete abortion without manual vacuum aspiration Complication requiring treatment
Notes

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Adinma 2012	Wrong study design
Akici 2017	Wrong patient population
Alam 2012	Wrong patient population
Alhamadani 2009	Wrong study design
Aniteye 2016	Wrong study design
Ankomah 1997	Wrong study design
Anonymous 1958	Wrong study design
Anonymous 1993	Wrong study design
Appiah‐Agyekum 2018	Wrong study design
Averbach 2016	Wrong study design
Averbach 2018	Wrong study design
Bancsi 2019	Wrong study design
Banu 2017	Wrong study design
Beckett 1991	Wrong study design
Berry‐Bibee 2015	Wrong study design
Biggs 2019	Wrong patient population
Billings 2009	Wrong patient population
Black 2015	Wrong intervention
Blackwell 1999	Wrong patient population
Brahmi 2014	Wrong patient population
Chaturachinda 2017	Wrong study design
Chin 2012	Wrong patient population
Coelho 1991	Wrong study design
Coeytaux 2014	Wrong study design
Cohen 2005	Wrong study design
Costa 1998	Wrong study design
Creanga 2018	Wrong patient population
Damalie 2014	Wrong intervention
De Leeuw 2015	Wrong study design
Diamond‐Smith 2019a	Wrong study design
Diamond‐Smith 2019b	Wrong patient population
Diniz 2012a	Wrong study design
Diniz 2012b	Wrong study design
Fetters 2017	Wrong study design
Footman 2017	Wrong intervention
Footman 2018a	Wrong comparator
Footman 2018b	Wrong study design
Frye 2015	Wrong study design
Ganle 2019a	Wrong patient population
Ganle 2019b	Wrong patient population
Gonsalves 2017	Wrong study design
Grossman 2010	Wrong intervention
Grossman 2015	Wrong study design
Gupta 2018	Wrong comparator
Hendrickson 2016	Wrong patient population
Huda 2014	Wrong patient population
Huda 2017	Wrong patient population, duplicate
Huda 2018	Wrong study design
Jewkes 2005	Wrong study design
Kapp 2019	Wrong study design
Kapp 2020	Wrong study design
Keenan 2019	Wrong intervention
Krayacich 1991	Wrong study design
Kumar 2013	Wrong intervention
Kunwar 2014	Wrong study design
Lara 2006	Wrong patient population
Lara 2011	Wrong patient population
Levandowski 2012	Wrong patient population
London 2014	Wrong study design
Mascret 2011	Wrong study design
Miller 2005	Wrong study design
Mohd 2014	Wrong study design
Ngo 2012	Wrong patient population
Nivedita 2015	Wrong comparator
Njuma 2015	Wrong patient population
Osur 2009	Wrong study design
Owens 2014	Wrong setting
Oyebode 2009	Wrong study design
Oyebode 2015	Wrong study design
Panda 2016	Wrong study design
Powell‐Jackson 2015	Wrong study design
Puri 2020	Wrong study design
Raifman 2019	Wrong patient population
Reiss 2016	Wrong patient population
Rocca 2015	Preliminary report from included full study
Sarojini 2017	Wrong study design
Sherris 2005	Wrong patient population
Shrivastava 2017	Wrong study design
Sneeringer 2012	Wrong study design
Stillman 2020	Wrong study design
Tamang 2015	Wrong patient population
Tamang 2018	Wrong comparator
Ujah 2009	Wrong study design
Uma 2017	Wrong study design
Zavier 2019	Wrong intervention

Characteristics of ongoing studies [ordered by study ID]

Grossman ongoing a.

Study name	Medication abortion via pharmacy dispensing
Methods	Mixed methods study
Participants	N = 260 Women seeking medication abortion through 70 days gestation; at least 15 years of age; English or Spanish speaking
Interventions	Intervention arm: patients will receive Mifeprex® (mifepristone) by pharmacy rather than standard care at clinic visit Experimental arm: oral mifepristone 200 mg, followed by misoprostol 800 µg administered buccally (at 24 to 48 hours following mifepristone) or vaginally (as soon as 6 hours following mifepristone)
Outcomes	Primary outcomes: Number of pharmacists who refuse to dispense Mifeprex at least once during the study period. Number of pharmacists who report being "somewhat satisfied" or "very satisfied" when asked "Overall, how satisfied are you with mifepristone dispensing at your pharmacy?" Number of pharmacists who refuse to dispense Mifeprex [ Time frame: end of the study, month 24 ] Number of pharmacists who report being satisfied with pharmacy dispensing of Mifeprex [ Time frame: end of the study, month 24 ] Secondary outcome: Number of participants who report being "somewhat satisfied" or "very satisfied" when asked "Overall, how satisfied were you with your experience at the pharmacy when you got the abortion pill?" Number of participants who report that their "abortion is now complete and they are no longer pregnant" and who did not "end up having a suction procedure (or vacuum aspiration or dilation and curettage procedure) to complete the abortion" Number of participants who had a medical problem that required them to go to the hospital, emergency department or a doctor's office (other than regularly scheduled follow‐up visit) since receiving the abortion pill Change in proportion of knowledge questions related to medication abortion answered correctly by pharmacists (out of 12 possible knowledge questions, with 0 questions correct meaning lowest knowledge and 12 questions correct meaning highest knowledge) Number of participants who report being satisfied with obtaining Mifeprex in the pharmacy [ Time frame: day 2 following initial medication abortion visit ] Number of participants with a complete abortion with medication alone and who do not require a surgical procedure to complete the abortion [ Time frame: up to 6 weeks after initial visit ] Number of participants with an adverse event [ Time frame: up to 6 weeks after initial visit ] Change in proportion of knowledge questions related to medication abortion answered correctly by pharmacists [ Time frame: before and after the intervention, approximately month 1 and month 24 of the study to assess change in knowledge ]
Starting date	January 2020
Contact information	Daniel Grossman, Professor in Dept ObGyn and Director of Advancing New Standards in Reproductive Health, University of California, San Francisco
Notes

Grossman ongoing b.

Study name	Mail order mifepristone study
Methods	Prospective cohort study
Participants	N=425 Women seeking medication abortion through 63 days gestation, 15 years and older, English or Spanish speaking
Interventions	Intervention arm: patients will have MifeprexⓇ (mifepristone) shipped to a preferred address via an online mail order pharmacy rather than standard care at clinic visit. Experimental arm: Oral Mifepristone 200 mg, followed by misoprostol 800 µg administered buccally (at 24 to 48 hours following mifepristone) or vaginally (as soon as 6 hours following mifepristone)
Outcomes	Primary outcomes: Proportion of patients reporting whether they would use the mail order service again if they needed another abortion. Proportion of patients who report positive experience. [ Time frame: day 14 following initial medication abortion visit ] Secondary Outcomes: Proportion of patients who receive medications by Day 2 and Day 3 and proportion of patients who indicate their confidentiality was maintained when receiving medications by mail. The investigators will also estimate the proportion of patients who report that medications were lost, stolen or damaged. Using electronic health record data related to abortion, the investigators will estimate the proportion of patients who experience a complete abortion and the proportion of patients who experience complications, including adverse or severe adverse events. Proportion of patients who report timely delivery of medication. [ Time frame: day 3 following initial medication abortion visit ] Proportion of patients who have a complete abortion [ Time frame: up to 6 weeks after initial medication abortion visit. ]
Starting date	April 2019
Contact information	Daniel Grossman, Professor and Director of ANSIRH, University of California, San Francisco
Notes

Ipas ongoing.

Study name	Study of clinic‐based versus self‐use of medical abortion pills (MOC)
Methods	Observational model: cohort Time perspective: prospective
Participants	Estimated enrollment (still recruiting): 4000 Women who are pregnant and seeking abortion in study sites
Interventions	One cohort using medical abortion pills sourced from pharmacies versus one cohort using medical abortion pills sourced from health clinic
Outcomes	Primary outcomes: The need for additional treatment to complete the abortion (either aspiration or repeated misoprostol) following a woman taking the medical abortion pills. Need for additional treatment to complete abortion [ Time frame: final assessment at 30 days following mifepristone administration ] Secondary outcomes: Number of participants who have a complication such as hemorrhage requiring a blood transfusion, hospitalization, serious infection and undiagnosed (at the time of mifepristone) ectopic pregnancy. Serious complications/morbidity [ Time frame: final assessment at 30 days following mifepristone administration ] Other outcomes: Women's reported use of contraception (yes or no and method type) following medical abortion Uptake of post‐abortion contraception [ Time frame: 30 days ]
Starting date	October 2018
Contact information	Ipas
Notes

Differences between protocol and review

The original protocol did not specify the priority to be placed on extracting adjusted estimates from included non‐randomized studies. These estimates are more appropriate to report since they account for important confounding and study design factors. Had more studies been included, the use of unadjusted or calculated estimates from trials would be considered to allow studies to be combined in meta‐analysis.  

Contributions of authors

Maria Rodriguez drafted the review with input from all authors, and signed off the final review.

Alison Edelman provided input for the draft review and signed off the final review.

Alyssa Hersh provided input for the draft review and signed off the final review.

Pragya Gartoulla provided input for the draft review and signed off the final review.

Jillian Henderson provided input for the draft review and signed off the final review.

Sources of support

Internal sources

• No internal sources of support, Other

External sources

• No external sources of support, Other

Declarations of interest

The authors do not have any interests to declare.

Maria I Rodriguez: reports FDA mandated trainings on contraceptive implant provision with Merck through winter 2019, and contracptive trainings on IUD insertions with Bayer through summer 2021. Dr. Rodriguez also reports participanting in data safety monitoring board for a clinical study on treatment of postpartum depression through winter 2021. Dr. Rodriguez reports working as a full scope obstetrician and gynecoogist in an academic setting at Oregon Health & Science University. 

Alison Edelman: reports working as a health professional and providing reproductive health care at Oregon Health & Science University. Dr. Edelman also reports an affiliation with an organization that has declared an opinion or position on the topic. Dr. Edelman provided training materials, trainings, and support for MOH or other clinical guidelines for Ipas, SugarPalm Foundation, and the American College of OB/GYN. 

Alyssa Hersh: has declared that they have no conflict of interest.

Pragya Gartoulla: has declared that they have no conflict of interest.

Jillian Henderson: reports working as a health professional at the University of California, San Francisco. Dr. Henderson is a published author on peer‐reviewed studies in public health and medical journals related to medical abortion safety and access. 

New