Effect of HPV self-collection kits on cervical cancer screening uptake among under-screened women from low-income US backgrounds (MBMT-3): a phase 3, open-label, randomised controlled trial

Summary

Background

Most cervical cancer in the USA occurs in under-screened women. The My Body, My Test-3 (MBMT-3) trial sought to assess the efficacy of mailed human papillomavirus (HPV) self-collection kits with appointment-scheduling assistance to increase uptake of cervical cancer screening among under-screened women from low-income backgrounds compared with scheduling assistance alone.

Methods

MBMT-3 is a phase 3, open-label, two-arm, randomised controlled trial. Participants were recruited from 22 counties in North Carolina state, USA, and we partnered with 21 clinics across these counties. Participants were eligible for inclusion if they were aged 25–64 years, had an intact cervix, were uninsured or enrolled in Medicaid or Medicare, had an income of 250% or less of the US Federal Poverty Level, were living within the catchment area of a trial-associated clinic, and were overdue for screening (ie, Papanicolaou test ≥4 years ago or high-risk HPV test ≥6 years ago). Participants were randomly assigned (2:1) to receive a mailed HPV self-collection kit and assistance for scheduling a free screening appointment (intervention group) or to receive scheduling assistance alone (control group). Randomisation was conducted by county using permuted blocks of nine patients and assignment to group was not masked. Participants in the intervention group were mailed HPV self-collection kits to collect a cervical-vaginal sample and return it by mail for testing. Samples were tested with the Aptima HPV assay (Hologic, San Diego, CA, USA), and participants were informed of high-risk HPV results by telephone call. Trial staff made up to three telephone call attempts to provide scheduling assistance for in-clinic screening for all participants. The primary outcome was cervical cancer screening uptake (ie, attending an in-clinic screening appointment or testing negative for high-risk HPV with a returned self-collected sample) within 6 months of enrolment in the intention-to-treat population. This trial is registered with ClinicalTrials.gov, NCT02651883, and has been completed.

Findings

Recruitment occurred between April 11, 2016, and Dec 16, 2019. 4256 women contacted the trial to participate, of whom 899 (21%) were eligible for inclusion and 697 (78%) returned consent forms. Of those who consented, 461 (66%) women were randomly assigned to the intervention group and 236 (34%) women were randomly assigned to the control group. We excluded 32 ineligible women post-randomisation, leaving 665 for primary analysis. Screening uptake was higher in the intervention group (317 [72%] of 438) than control group (85 [37%] of 227; risk ratio 1·93, 95% CI 1·62–2·31). Among intervention participants, 341 (78%) of 438 returned a self-collection kit. Three participants reported hurt or injury when using the self-collection kit; no participants withdrew due to adverse effects.

Interpretation

Among under-screened women from low-income backgrounds, mailed HPV self-collection kits with scheduling assistance led to greater uptake of cervical cancer screening than scheduling assistance alone. At-home HPV self-collection testing has the potential to increase screening uptake among under-screened women.

Funding

National Cancer Institute.

Introduction

Early detection and treatment of cervical precancerous lesions can reduce the burden of cervical cancer.¹ Cervical cancer incidence and mortality rates have substantially declined over the past three decades in the USA, largely attributed to Papanicolaou smear screening.¹ Screening guidelines recommend Papanicolaou testing alone every 3 years for individuals aged 21 years and older and have incorporated testing of high-risk (ie, oncogenic) human papillomavirus (HPV) infection every 5 years, either as primary testing or in conjunction with Papanicolaou testing (co-testing) for people aged 30 years and older.² Cervical cancer disproportionately affects Black and Hispanic women in North Carolina state, USA, and the rest of the USA, with the highest incidence among Hispanic women and the highest mortality among Black women.³ Most cervical cancer cases occur in under-screened women (ie, women who are not screened as frequently as the guidelines recommend).⁴ An estimated 15% of eligible adults in the USA are overdue for screening.⁵ Novel, scalable, and cost-effective screening approaches that target non-attendees of routine screening are needed to reduce morbidity and mortality related to cervical cancer.

Mailed at-home HPV self-collection testing has the potential to reduce barriers to screening access and increase uptake of cervical cancer screening among under-screened women.^4,6–8 HPV self-collection is a technique by which a simple collection device (eg, a brush or swab) is used to obtain cervicovaginal cell samples to test for high-risk HPV infection. Self-collection for high-risk HPV testing has similar sensitivity and specificity to physician collection for the detection of high-grade cervical precancerous lesions if highly sensitive detection assays are used.^6,8–10 An advantage of self-collection is that it eliminates the need for an initial in-clinic pelvic examination.¹¹ Self-collection can be performed at a location that is convenient for the individual, such as at home, and people with positive high-risk HPV results can be referred for follow-up care.

An alternative method that is recommended by the US Community Preventive Services Task Force to increase cervical cancer screening is multicomponent interventions, which combine two or more intervention approaches to reduce logistical and structural barriers to cancer screening through strategies that increase community demand and access.¹² Reviews from the task force recommended enhanced client reminders as one component of a multicomponent intervention.^12–14 These reminders often include information on the benefits of cervical cancer screening and provision of logistical support, such as appointment scheduling assistance.^12–14 Randomised controlled trials and nationally representative studies from the USA and Europe have shown that providing enhanced reminders is an efficacious method to increase in-clinic screening uptake,^8,13,15 yet data are scarce for under-screened and never-screened women from low-income backgrounds. Additionally, data are scarce for the benefits of including HPV self-collection kits with scheduling assistance.

To address the scarcity of data for the efficacy of HPV self-collection among under-screened women from low-income backgrounds in the USA, we aimed to identify whether mailing HPV self-collection kits to women’s homes in conjunction with providing appointment scheduling assistance resulted in increased uptake of cervical cancer screening compared with offering scheduling assistance alone. The inclusion of women from low-income backgrounds is important given that low socioeconomic status is associated with low screening rates;¹⁶ thus, evidence is crucial for the effect of HPV self-collection on screening uptake among this group at high risk of cervical cancer.

Methods

Study design and participants

The My Body, My Test-3 (MBMT-3) trial is a phase 3, open-label, two-arm, randomised controlled trial.¹³ We partnered with 21 clinics and recruited participants from 22 counties in North Carolina state, USA. Counties were selected on the basis of the ability of trial-affiliated clinics to provide services and the travel distance for participants.

Eligible individuals were aged 25–64 years, were not pregnant, had an intact cervix (ie, no history of hysterectomy), had an income of 250% or less of the US Federal Poverty Level, were uninsured or enrolled in Medicaid or Medicare, and were living within the catchment area of a trial-associated clinic and eligible to receive screening from a trial-associated clinic.^13,17 The US Census Bureau sets annual household income thresholds, which vary by family size and composition, to define who is living in federal poverty.¹⁸ Additionally, women were eligible only if they self-reported not having a Papanicolaou test for 4 years or more or not having a physician-collected HPV test in 6 years or more, which is considered overdue for screening according to US guidelines.^2,13,17 If the participant was an established patient at one of the trial-affiliated clinics, the date of their last cervical cancer screening was confirmed by use of the participant’s medical records at the time of appointment scheduling. Trial eligibility was assessed by use of a telephone questionnaire administered by trial staff. Women who did not meet inclusion criteria received information on clinics in their county where they might be eligible for a free or low-cost cervical cancer screening. Recruitment occurred through printed advertisements, radio announcements, and online postings (appendix pp 1–9).^13,17 Women were also recruited at community events and referred through community organisations and the United Way 2–1–1 social assistance helpline.^13,17 Recruitment materials referred to participants as “women”, but gender identity was not an exclusion criterion. Therefore, we do not assume that all participants identified as women.

Written consent was sought and obtained from eligible women, and Heath Information Portability and Accountability Act forms were sent via mail. Before enrolment, all required forms had to be completed and returned. The University of North Carolina Institutional Review Board provided ethical approval for this trial (number 14–3042). The protocol has previously been published.¹³

Randomisation and masking

All enrolled women were randomly assigned (2:1) to receive a mailed HPV self-collection kit and assistance in scheduling appointments for in-clinic cervical cancer screening (intervention group) or to receive scheduling assistance alone (active control group). Separate randomisation lists were generated per county, using permuted blocks of nine patients (6:3 intervention to control) through the randomisation function of the web programming language hypertext processor. Stratified randomisation was used to ensure that the proportion of individuals assigned to the intervention was similar across counties and to avoid an imbalance in county distribution of potentially predictive covariates between trial groups. Participating women and trial staff were not masked to randomised assignment.

Procedures

Participants in the intervention group were mailed HPV self-collection kits with instructions in English or Spanish for collecting a cervicovaginal sample and returning it by mail for testing.^13,17 Each kit included a Viba-Brush (Rovers Medical Devices, Oss, Netherlands) and a vial containing 4·3 mL of Aptima sample transport media (Hologic, San Diego, CA, USA), which is approved for mailing in the USA and validated for sample stability for HPV testing up to 60 days at room temperature.¹⁷ Once the participant had collected their sample, a Rover’s Removers tube (Rovers Medical Devices, Oss, Netherlands) was used to eject the Viba-Brush into the specimen transport media. Self-collection kits were mailed back to the University of North Carolina, Chapel Hill, NC, USA, de-identified, and sent to Labcorp in Burlington, NC, USA, for high-risk HPV testing.¹⁷ Samples were tested with the Aptima HPV assay (Hologic, San Diego, CA, USA), a transcription-mediated amplification assay that qualitatively detects high-risk HPV E6 and E7 mRNA of the 14 high-risk HPV genotypes that are most prevalent in people with cervical cancer.¹⁷ All intervention participants were informed of high-risk HPV results by a telephone call from a trained interviewer.^13,17 Women who did not return a self-collected sample within 3 weeks were sent a reminder letter and, after another 2 weeks, were given a reminder telephone call.

All participants were offered assistance in scheduling appointments for in-clinic cervical cancer screening, irrespective of trial group or self-collection results.^13,17 During the scheduling assistance call, which occurred 1 week after trial enrolment, women received information on screening and scheduling assistance for a free Papanicolaou–HPV co-testing appointment at a trial-affiliated clinic.^13,17 Trial staff made up to three telephone call attempts to reach participants to provide scheduling assistance. Trial-affiliated clinics offered co-testing using ThinPrep liquid-based cytology (Hologic, San Diego, CA, USA) with brush samples preserved in PreservCyt solution (Hologic, San Diego, CA, USA).¹⁷ In the active control group, hereafter referred to as the control group, women received scheduling assistance alone to schedule an appointment for in-clinic screening.

All participants completed telephone-based questionnaires at four stages: before enrolment to assess eligibility, at randomisation, 1 week after delivery of self-collected HPV result in the intervention group or appointment scheduling call in the control group, and 2 weeks after an in-clinic screening appointment or 6 months after enrolment for those who did not attend an appointment during this time. Questionnaires captured sociodemographic and behavioural data, including knowledge and attitudes about cervical cancer screening. Participants received financial compensation for completing questionnaires regardless of whether they were screened.

Outcomes

The primary outcome was uptake of cervical cancer screening (also termed screening completion) within 6 months of trial enrolment, defined as attending a screening appointment at any clinic or testing negative for high-risk HPV on self-collected samples, which represents how self-collection methods would be implemented if approved for primary screening in the USA. Participants who were randomly assigned to the intervention group who tested positive or invalid for high-risk HPV, or did not return a self-collection kit, were considered to have attained screening uptake only if they attended an in-clinic screening appointment. The primary outcome was assessed in the intention-to-treat (ITT) population. We also assessed the uptake of screening among the per-protocol and complete case populations.

Other trial outcomes include the prevalence of HPV mRNA detection in self-collected samples (intervention group only); prevalence of abnormal cytology detected in clinic samples; number and percentage of patients referred to and completing colposcopy, defined as attending a follow-up colposcopy; prevalence of high-grade lesions (cervical intraepithelial neoplasia of grade 2 or worse [CIN2+]) as detected in follow-up colposcopy (when indicated); and number and percentage of people referred to and completing treatment.

Secondary trial outcomes were the levels of risk appraisal with regard to cervical cancer screening, costs to payers and incremental cost to payer (public or private) per additional woman screened, level of intention to complete cervical cancer screening, level of self-efficacy to complete cervical cancer screening, and percentage of participants who scheduled an in-clinic cervical cancer screening appointment. Additionally, attitudes towards HPV, cervical cancer, and cervical cancer screening are other trial outcomes. These secondary outcomes are measured among the ITT population and will be reported in future publications.

In-clinic cervical cancer screening appointment attendance and colposcopy completion were tracked by use of medical records from trial-affiliated clinics. Study staff also requested medical records from clinics for participants who reported attending a non-trial affiliated clinic and contacted other clinics in the surrounding area to establish in-clinic screening attendance and colposcopy completion among participants whose in-clinic cervical cancer screening appointment attendance status could not be confirmed. Participants for whom the in-clinic screening appointment status could not be identified were considered unscreened for analyses.

Information regarding adverse events while using self-collection kits was collected from intervention participants during the follow-up questionnaire taken 1 week after delivery of self-collected HPV results.¹⁹ Participants were asked “Did you hurt or injure yourself in any way when using the self-test?” Those who responded “Yes” were then asked, “Could you please explain what happened?”¹⁹ The trial did not collect information on adverse events in the control group or during in-clinic screening appointments, and thus their prevalence is unknown.

For both trial groups, the median time from enrolment to screening uptake and the median time from in-clinic screening attendance to colposcopy completion were included as post-hoc outcomes.

Statistical analysis

Power analysis identified that 510 participants in a 2 to 1 randomisation ratio of intervention to control would provide 88–94% power to detect a 15% or greater difference between groups for the primary outcome, assuming a screening uptake of 60–80% in the intervention group, which is consistent with previous trials using a similar recruitment approach.^6,13,19 The trial was not powered to conduct subgroup comparisons.

For the comparison of screening uptake between trial groups, we applied an ITT principle, such that outcomes were analysed according to trial group assigned regardless of whether the participants complied with intervention components of each trial group. Risk ratios (RRs) comparing uptake of cervical cancer screening between the trial groups were estimated using log-binomial regression modelling with corresponding 95% CIs based on binomial approximation.

Secondary per-protocol and complete case analyses were conducted to examine the primary trial outcome (appendix p 10). Per-protocol analysis included all participants who received the self-collection kit and scheduling assistance (intervention group) or scheduling assistance alone (control group), per the trial protocol. Complete case analysis excluded individuals whose screening status could not be established. Inverse probability of censoring weights was used to control for possible selection bias induced by excluding individuals who were randomly assigned to a group from analysis.²⁰ The censoring weights were estimated by logistic regression modelling of the probability of inclusion into the per-protocol and complete case analyses. In particular, stabilised censoring weights were computed for each individual, with the denominator of the weight equal to the estimated probability of inclusion in the analysis conditional on trial group, age, marital status, and Medicaid or Medicare coverage and the numerator calculated similarly but conditional only on trial group. The primary trial outcome was examined using log-binomial regression modelling to calculate weighted RRs in both the per-protocol and complete case analyses.

Subgroup ITT analyses used the Wald test of interaction to examine differences in the percentage of participants reaching the primary outcome of screening uptake in different demographic categories including age, race or ethnicity, education, and annual household income. Time since last cervical cancer screening and Medicaid or Medicare coverage were added post hoc to subgroup analyses. No adjustments were made for multiple comparisons.

Log-binomial regression modelling was used to calculate RRs comparing proportions of other outcomes, including the prevalence of abnormal cytology detected in clinic samples, proportion of patients referred to and completing colposcopy, prevalence of CIN2+ as detected in follow-up colposcopy (when indicated), and proportion of people referred to and completing treatment between the study groups of the ITT analysis sample. Post-hoc sensitivity analysis, using the ITT sample, assessed how the primary trial RR estimate would change if intervention participants who had positive results for HPV types 16, 18, or 45 from self-collected samples were classified as having attained screening uptake regardless of their in-clinic screening attendance, as done in a previous US study,²¹ given that clinical guidance recommends that people with positive results for HPV types 16, 18, or 45 should be directly referred to diagnostic colposcopy.²¹ SAS statistical software version 9.4 was used to perform data analyses.

This trial did not use a data safety monitoring board, as it was considered a low-risk study. This study is registered with ClinicalTrials.gov, NCT02651883.

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Trial recruitment was conducted between April 11, 2016 and Dec 16, 2019; participant follow-up ended on March 27, 2020. Of the 4256 women who contacted the trial to participate, 2770 (65%) did not meet the inclusion criteria and 587 (14%) did not fully complete eligibility screening, resulting in a total of 899 women who were eligible for inclusion in the trial (figure 1). Of the 791 (88%) who completed the baseline questionnaire, 697 women returned consent forms and were randomly assigned to the intervention group or the control group in a 2 to 1 ratio. A total of 461 women were randomly assigned to receive an HPV self-collection kit with scheduling assistance (intervention group), and 236 women were randomly assigned to receive scheduling assistance alone (control group). 32 (5%) of 697 women were excluded after they were randomly assigned to a group because they were ineligible for screening (n=24) or unable to obtain an in-clinic screening appointment due to the COVID-19 pandemic (n=8).²² Screening status could not be confirmed for 82 (12%) of 665 participants overall (35 [8%] of 438 participants in the intervention group and 47 [21%] of 227 participants in the control group). The final analytical sample was 665 women (438 women in the intervention group and 227 women in the control group) for primary ITT analyses.

Figure 1: Trial profile.

HPV=human papillomavirus. *Technical error is defined as database programming error or data entry error (eg, entering the wrong birth date of a participant during the screening, making the person ineligible).

Median age of trial participants was 42 (IQR 33–51) years (table 1). Most participants with data available were Black non-Latina or non-Hispanic (311 [47%] of 664), uninsured (518 [78%] of 663), single or never married (342 [52%] of 661), and unemployed (373 [57%] of 660).

Table 1:

Characteristics of MBMT-3 trial participants at baseline

	Intervention group: HPV self-collection and scheduling assistance (n=438)	Control group: scheduling assistance only (n=227)
Age, years	42 (32–51)	41 (33–51)
Time since last cervical cancer screening, years	5 (4–8)	5 (4–8)
Not sure, but >4 years ago	32 (7%)	22 (10%)
Never screened	19 (4%)	10 (4%)
Number of doctor’s appointments or clinic visits in past year	2 (0–3)	2 (0–3)
Race and ethnicity
Black non-Latina or non-Hispanic	200 (46%)	111/226 (49%)
White non-Latina or non-Hispanic	175 (40%)	82/226 (36%)
Latina or Hispanic	38 (9%)	16/226 (7%)
Other*	25 (6%)	17/226 (8%)
Medicaid or Medicare coverage
Medicaid	86 (20%)	45/225 (20%)
Medicare only or Medicaid and Medicare†	8 (2%)	6/225 (3%)
Uninsured	344 (79%)	174/225 (77%)
Marital status
Single or never married	218/435 (50%)	124/226 (55%)
Married or living with a partner	98/435 (23%)	40/226 (18%)
Divorced, separated, or widowed	119/435 (27%)	62/226 (27%)
Employment
Unemployed	242/434 (56%)	131/226 (58%)
Employed	192/434 (44%)	95/226 (42%)
Education
High school, GED, or less	188 (43%)	93 (41%)
Some college‡ or more	250 (57%)	134 (59%)
Annual household income, US$	$15 000 ($8400–25 000)	$15 000 ($3282–25 000)
Current smoker
No	256/434 (59%)	125/226 (55%)
Yes	178/434 (41%)	101/226 (45%)
BMI	29·5 (24·4–35·5)	29·3 (25·1–36·0)
Rurality§
Urban setting	403 (92%)	201 (89%)
Rural setting	35 (8%)	26 (11%)
Number of live births	2 (0–3)	2 (0–3)

Data are median (IQR) or n (%) and are calculated from available data. Data were missing for some participants for race and ethnicity (n=1), health insurance (n=2), marital status (n=4), employment (n=5), annual household income (n=33; n=22 in the intervention group, n=11 in the control group), current smoker (n=5), BMI (n=12; n=9 in the intervention group, n=3 in the control group), and number of live births (n=4; n=3 in the intervention group, n=1 in the control group). GED=General Education Development or Diploma. HPV=human papillomavirus.

^*Includes non-Latina or non-Hispanic American Indian or Alaska Native (n=9), non-Latina or non-Hispanic Native Hawaiian or Pacific Islander (n=2), non-Latina or non-Hispanic Asian (n=6), and other ethnicities that are non-Latina or non-Hispanic (n=25).

^†Ten participants were enrolled in both Medicare and Medicaid, and four participants were enrolled in Medicare only.

^‡Some college indicates the participant took college-level courses but did not earn a degree.

^§Rurality was defined by use of the participant’s Zoning Improvement Plan code and corresponding Rural-Urban Commuting Area code.

Uptake of cervical cancer screening was higher in the intervention group than in the control group in the ITT analysis (table 2). Screening uptake was also higher in the intervention group than in the control group in both per-protocol and complete case analyses.

Table 2:

Effects of self-collection intervention on uptake of cervical cancer screening

	Intervention group	Control group	Risk ratio (95% CI)
Intention-to-treat analysis* (n=665)
Participants analysed	438	227	··
Screening uptake	317 (72%)	85 (37%)	1·93 (1·62–2·31)
Not screened	121 (28%)	142 (63%)	··
Confirmed	86 (20%)	95 (42%)	··
Not confirmed	35 (8%)	47 (21%)	··
Per-protocol analysis† (n=549)
Participants analysed	364	185	··
Screening uptake	302 (83%)	81 (44%)	1·90 (1·60–2·26)‡
Not screened	62 (17%)	104 (56%)	··
Confirmed	45 (12%)	71 (38%)	··
Not confirmed	17 (5%)	33 (18%)	··
Complete case analysis§ (n=583)
Participants analysed	403	180	··
Screening uptake	317 (79%)	85 (47%)	1·67 (1·42–1·97)‡
Not screened	86 (21%)	95 (53%)	··

Data are n or n (%), unless otherwise stated. Screening uptake was defined as having received a self-collected high-risk HPV-negative test result or attending an in-clinic cervical cancer screening appointment. Not screened was defined as participants who were due for screening (ie, all participants in the control group and participants in the intervention group who received a high-risk HPV-positive or invalid self-collected result or did not return their self-collection kit) and confirmed that they did not attend a screening appointment. Screening was considered as not confirmed when a participant was due for screening and the participant did not provide a response and no clinic records confirmed that screening was conducted (82 [12%] of 665 overall; 35 [8%] of 438 in the intervention group and 47 [21%] of 227 in the control group).

^*Intention-to-treat analysis included all participants who were randomly assigned to a trial group apart from those who were withdrawn due to ineligibility after being assigned or those who had early termination of the trial due to COVID-19.

^†Per protocol analysis included all participants who received the self-collection kit and scheduling assistance in the intervention group or scheduling assistance in the control group, per the study protocol.

^‡Risk ratio estimate from log-binomial regression model using inverse probability of censoring weights based on age (ie, continuous), marital status (ie, single or never married; married or living with partner; divorced, separated, or widowed), and Medicaid or Medicare coverage.

^§Complete case analysis excluded individuals whose screening status was not confirmed.

Analyses showed a similar intervention effect among the ITT sample for strata of the prespecified subgroups age (p=0·18 for interaction), race and ethnicity (p=0·69 for interaction), and education (p=0·35 for interaction) and among the subgroups added post-hoc (p=0·75 for interaction for time since last cervical cancer screening and p=0·15 for interaction for Medicaid or Medicare coverage). Results from the interaction test suggested a differential intervention effect across strata of annual household income (p=0·04 for interaction), with a stronger effect among participants with an annual household income of US$25 000 or higher (table 3).

Table 3:

Subgroup analysis of intervention effect on uptake of cervical cancer screening in intention-to-treat analysis

	Intervention group (n=438)	Control group (n=227)	Risk ratio, (95% CI)
Age group, years
25–34	83/133 (62%)	19/76 (25%)	2·50 (1·65–3·77)
35–44	88/117 (75%)	20/54 (37%)	2·03 (1·41–2·92)
45 or older	146/188 (78%)	46/97 (47%)	1·64 (1·31–2·05)
Time since last cervical cancer screening, years
≤5	153/199 (77%)	43/114 (38%)	2·04 (1·59–2·61)
>5	130/188 (69%)	32/81 (40%)	1·75 (1·32–2·33)
Not sure, but >4 years ago	22/32 (69%)	8/22 (36%)	1·89 (1·04–3·45)
Never screened	12/19 (63%)	2/10 (20%)	3·16 (0·87–11·43)
Race and ethnicity
Black non-Latina or non-Hispanic	143/200 (72%)	41/111 (37%)	1·94 (1·50–2·51)
White non-Latina or non-Hispanic	126/175 (72%)	28/82 (34%)	2·11 (1·54–2·89)
Latina or Hispanic	29/38 (76%)	8/16 (50%)	1·53 (0·91–2·57)
Other	19/25 (76%)	8/17 (47%)	1·62 (0·93–2·80)
Medicaid or Medicaid coverage
Medicare only, Medicaid only, or Medicaid and Medicare*	63/94 (67%)	13/51 (26%)	2·63 (1·61–4·29)
Uninsured	254/344 (74%)	72/174 (41%)	1·79 (1·48–2·15)
Education
High school, GED, or less	135/188 (72%)	31/93 (33%)	2·15 (1·59–2·91)
Some college† or more	182/250 (73%)	54/134 (40%)	1·78 (1·48–2·15)
Annual household income, US$
$0–9999	92/135 (68%)	34/82 (42%)	1·64 (1·24–2·18)
$10 000–24 000	125/166 (76%)	33/71 (47%)	1·62 (1·24–2·11)
$25 000 or higher	85/115 (74%)	15/63 (24%)	3·10 (1·97–4·89)

Data are n/N (%) unless otherwise stated, where n is the number of participants screened and N is the total number of participants in the subgroup. Data are shown only for participants with available data, as indicated in table 1. Screening uptake is defined as attending cervical cancer screening at any clinic or testing negative for HPV on self-collected samples. GED=General Education Development or Diploma. HPV=human papilloma virus. Where values do not add to the total, there were missing data in subgroup variables.

^*Ten participants were enrolled in both Medicare and Medicaid, four participants were enrolled in Medicare only, and 131 participants were enrolled in Medicaid only.

^†Some college indicates the participant took college-level courses but did not earn a degree.

Among 438 intervention participants, 341 (78%) returned a self-collection kit (figure 2). Of those who returned a self-collection kit, 11 (3%) participants had final invalid self-collected results and one (<1%) participant’s sample was received damaged, leaving a total of 329 (96%) participants with valid HPV self-collection results. Among these 329 participants, 52 (16%) tested positive for high-risk HPV. All brush head sample transfers from the Viba-Brush to collection liquid using the Rover’s Removers tube were successful, and no participants returned samples without the brush head.

Figure 2: Detailed diagram of trial outcomes.

Three call attempts were made to contact the participant to provide them with scheduling assistance. Participants who could not be reached were classified as scheduling assistance not completed. Participants who attended an in-clinic screening appointment but were classified as scheduling assistance not completed would have scheduled the appointment on their own. HPV=human papillomavirus. CIN2+=cervical intraepithelial neoplasia 2 or worse. *During the tracking period, one participant obtained scheduling assistance but did not complete results delivery. †During the tracking period, two participants obtained scheduling assistance but did not complete results delivery. ‡Two participants did not receive a self-collection kit but completed scheduling assistance. §Two participants did not receive a self-collection kit and did not complete scheduling assistance.

Among the 665 participants included in the primary ITT analysis, abnormal cytology was detected in eight (2%) of 438 intervention participants and five (2%) of 227 control participants (RR 0·83, 95% CI 0·27–2·51). Follow-up colposcopies were recommended for five (1%) intervention and two (1%) control participants (RR 1·30, 95% CI 0·25–6·63). Colposcopies were completed in three (1%) participants in the intervention group compared with none in the control group. Screening detected CIN2+ in two (<1%) intervention participants and no control participants. Treatment for CIN2+ was referred and completed for the two (<1%) participants in the intervention group and no control participants. Trial staff confirmed the date of last cervical cancer screening for 351 (53%) of 665 participants through medical records at trial-affiliated clinics.

Only three (1%) of 438 intervention participants reported experiences of hurt or injury when using the self-collection kit. When asked to describe these incidents, one participant said that they scratched themselves and two participants did not respond.

The post-hoc sensitivity analysis showed that, if our definition for screening uptake was revised to include positivity for HPV types 16, 18, or 45 in self-collected samples, we would have a slightly higher proportion of screening uptake among intervention participants (321 [73%] of 438 participants), resulting in a similar intervention effect (RR 1·96, 95% CI 1·64–2·34). Among participants who attained screening uptake, additional post-hoc analyses identified that the median time from enrolment to screening uptake was 31 days (IQR 24–52) in the intervention group and 43 days (IQR 30–73) in the control group. The median time from in-clinic screening attendance to colposcopy completion was 70 days (IQR 66–183) among participants in the intervention group. No participants in the control group completed colposcopy.

Discussion

Mailed HPV self-collection kits with scheduling assistance resulted in greater uptake of cervical cancer screening among under-screened US women from low-income backgrounds than did scheduling assistance alone. The benefit of mailed HPV tests was consistent across ITT, per-protocol, and complete case analyses and across demographic subgroups, showing the robustness of our findings.

Most studies evaluating the effectiveness of HPV self-collection on cervical screening uptake have been conducted outside of the USA in countries with national screening registries and universal health care. A meta-analysis of international HPV self-collection randomised controlled trials identified that mailing self-collection kits directly to participants using mail-to-all scenarios was highly effective in reaching adults who were under-screened for cervical cancer versus invitation or reminder letters for in-clinic screening (RR 2·33, 95% CI 1·86–2·91).⁸ One trial in the meta-analysis used community campaign invitation scenarios (eg, community-supported actions and mass media outreach), similar to the recruitment method adapted in our trial.^8,23 This Canadian trial was conducted among Indigenous women aged 25–69 years and reported notably higher screening uptake in the HPV self-collection group than in the Papanicolaou test group (RR 2·58, 95% CI 1·67–3·99).^8,23 Screening uptake in the trial was defined as providing a self-collected sample or undergoing Papanicolaou testing.

Only one other trial in the USA, to our knowledge, has assessed the effectiveness of mailed HPV self-collection tests on uptake of cervical cancer screening compared with usual care (ie, receiving annual screening reminders): the HOME trial.²¹ Notably, this trial was conducted among a sample of privately insured patients within Kaiser Permanente Washington, Renton, WA, USA, an integrated health-care delivery system.²¹ HOME trial participants had a median household income of approximately $65 000, and the trial excluded non-English speakers.²¹ By contrast, our trial sample consisted of only publicly insured (eg, Medicaid or Medicare) or uninsured women, with a median annual income of $15 000, and accommodated for the inclusion of Spanish-speaking individuals. Our findings are consistent with the HOME trial, which reported higher screening uptake in the intervention group than in the control group (RR 1·51, 95% CI 1·43–1·60).²¹ The HOME trial defined screening uptake as attending a Papanicolaou test or co-test appointment; self-sampling positive for HPV 16 or HPV 18, regardless of subsequent in-clinic follow-up; self-sampling HPV-negative; or self-sampling positive for other types of HPV or unsatisfactory self-sampling result and attending an in-clinic screening appointment.²¹ By contrast, participants in our trial with a self-collected HPV-positive result (regardless of type) were considered unscreened until they attended an in-clinic appointment. Changing our definition for uptake of cervical cancer screening to include self-collected samples positive for HPV 16, 18, or 45 resulted in a slightly higher proportion of screened intervention participants and a slightly larger effect size than observed in the HOME trial.

In our study, subgroup ITT analyses showed comparable efficacy across strata, including among Black non-Latina or non-Hispanic women and Latina or Hispanic women, which is important given that these populations suffer a greater burden of cervical cancer in North Carolina state, USA, and the USA overall, than women of other races and ethnicities.^3,24 A trial among under-screened Hispanic, Haitian, and non-Hispanic Black women in Florida state, USA, compared the efficacy of delivery methods on self-collection uptake.²⁴ HPV self-collection delivered via US mail resulted in similar self-collection kit return as HPV self-collection kits delivered in-person by a community health-care worker, which is important from a resource-use perspective.²⁴ In terms of results interpretation, we chose to use scheduling assistance for a free in-clinic screening appointment in both study groups, which might have increased appointment attendance compared with when people do not receive this support in a usual care scenario.

In terms of trial limitations, the recruitment of under-screened women from low-income backgrounds was limited to those who decided to respond to our outreach efforts, who might have been more motivated to participate in screening than their counterparts in the general population. Although this outreach approach might oversample motivated women and somewhat limit the study’s generalisability, recruitment from the general population allowed us to reach medically underserved women who do not regularly use clinic services. Our intensive community outreach methods resulted in recruiting and retaining a large sample of individuals who are hard to reach and at high risk of cervical cancer, with only a small proportion of participants whose screening status could not be identified. We acknowledge that mailed HPV self-collection kits does not meet the needs of all under-screened, hard-to-reach populations, as this intervention requires access to a telephone, a location to receive a self-collection kit, a place to conduct the collection, and access to transportation if in-clinic screening or follow-up colposcopy is needed. The logistics associated with attending in-clinic follow-up appointments, such as transportation costs, create additional barriers for already hard-to-reach populations. Analyses of the secondary outcomes of the MBMT-3 trial will aim to identify barriers to in-clinic cervical cancer screenings and cost considerations of the two trial groups. Although a negative high-risk HPV self-collection result allows for convenient screening uptake, it is important to note that, among intervention participants with positive high-risk HPV self-collected results, less than half attended an in-clinic screening appointment, which was similar to that of the control group. These findings are consistent with other self-collection studies among under-screened women in the USA, emphasising the importance of further efforts to ensure continuity of care among women with positive self-test results.²¹ Participants identified to be high-risk HPV-positive by self-collection were referred to in-clinic co-testing. In practice, it might be preferrable to refer women with positive HPV 16 or HPV 18 results directly to colposcopy and those with high-risk HPV types other than HPV 16 or HPV 18 to undergo repeat HPV testing with clinician-collected samples rather than co-testing. In terms of the HPV assay used, we conducted two studies before the development of the MBMT-3 trial that showed similar sensitivity of self-collected samples resuspended in Aptima standard transport media and tested with the Aptima assay compared with clinician-collected samples for CIN2+.^6,25,26 In 2022, a systematic review and meta-analysis concluded that Aptima had lower CIN2+ sensitivity for self-collected samples than for clinician-collected samples using cross-sectional data from five studies,²⁷ although four of the studies cited that used validated specimen collection devices showed similar sensitivity of Aptima for CIN2+ detection in self-collected compared with clinician-collected specimens. Given the small number of people identified with CIN2+ in these studies, further data are needed for Aptima self-collection performance using validated methods. The open-label trial design allows for the possibility of compensatory rivalry or resentful demoralisation, but nothing appeared to threaten the construct validity of the intervention during the trial.²⁸ The median time since last cervical cancer screening among participants was 5 years, possibly indicating that participants were not at notably increased risk given that 5 years is the recommended screening interval in some countries in Europe.²⁹ Because the USA does not have a national screening registry, we had to rely on the self-reported time since last screening for participants whose cervical cancer screening history could not be confirmed through medical records. We tracked the number of successful brush head sample transfers from the Viba-Brush to collection liquid using the Rover’s Removers tube and observed that no participants returned samples without the brush head. We could not measure sampling integrity because the Aptima HPV test does not include an internal control. Participants’ HPV vaccination status was not collected as part of this trial and should be considered in future research. Our trial was also conducted before the COVID-19 pandemic; as such, we cannot conclude the effect of self-collection on uptake of cervical cancer screening in the post-pandemic era.

The results of the MBMT-3 trial and previous research support recommendations for multicomponent interventions (ie, HPV self-collection plus appointment scheduling assistance) to increase cervical cancer screening and reduce structural barriers that are often faced by under-screened women from low-income backgrounds.¹² Mailed HPV self-collected samples have the potential for even greater effects if the US Food and Drug Administration approve them for primary screening, as only individuals with positive high-risk HPV self-collection results would need to be referred for in-clinic screening. Individuals with negative high-risk HPV self-collection results would not be required to attend in-clinic screening until the next screening round, thus alleviating barriers for many people in the USA who have trouble regularly accessing clinics. For future scalability, efforts are ongoing to obtain US approval for HPV self-collection. The Last Mile Initiative has been proposed by the National Cancer Institute to validate self-collection as compared with clinician-collected samples for primary HPV testing and advance progress towards Food and Drug Administration approval. Once approved, partnerships with programmes and community organisations, such as Medicaid, Federally Qualified Health Centers, or the Breast and Cervical Cancer Control Program, could be used to identify under-screened people and implement mailed HPV self-collection methods in a non-organised setting. However, it will be important to examine the effects of mailed HPV self-collection methods on uptake of cervical cancer screening across different social gradients and continuity of care to ensure that HPV self-collection does not increase inequities in cervical cancer screening. Our trial findings suggest that self-collection has potential to increase uptake of cervical cancer screening among under-screened adults in the USA—a group at high risk of invasive disease.

Research in context.

Evidence before this study

We searched PubMed for studies published between June 1, 2014 and Nov 8, 2022 using the search terms “human papillomavirus”, “self-sample”, “self-collection”, and “randomized trial”. Only randomised controlled trials were included in our literature review. We did not apply any language restrictions. A 2018 meta-analysis of international randomised controlled trials evaluating self-collection participation among under-screened women observed consistently higher response rates when human papillomavirus (HPV) self-collection kits were mailed or offered directly to women than when in-clinic screening invitation or reminder letters were sent, with substantial heterogeneity of participant response depending on invitation scenarios and local settings impeding universal recommendations. Therefore, local trials and pilot studies are most useful to characterise the probable effectiveness of mailing HPV self-collection kits in a given region or country before incorporating them into local strategies for cervical cancer screening. Among the few randomised controlled trials that delivered HPV self-collection kits via mail in the USA, participant recruitment occurred through community health-worker outreach, or eligible participants were identified through electronic medical records. None of these trials, however, evaluated the uptake of cervical cancer screening as their primary outcome or used population-based recruitment methods to reach medically underserved populations. Additionally, recommendations from the US Community Preventive Services Task Force to increase cervical cancer screening include multicomponent interventions that combine two or more intervention approaches as a method for reducing logistic and structural barriers. Yet, data evaluating multicomponent interventions are scarce among under-screened and never-screened women from low-income backgrounds.

Added value of this study

The primary goal of the My Body, My Test-3 trial was to evaluate the effectiveness of mailed at-home HPV self-collection kits with scheduling assistance as a multicomponent intervention for improving uptake of cervical cancer screening among under-screened women from low-income backgrounds in the USA—a population and region with little data. This goal is of particular importance given that low socioeconomic status is associated with lower screening rates. Additionally, we used intensive community outreach campaigns to recruit participants from hard-to-reach, under-screened populations. Screening uptake was defined as attending a screening appointment at any clinic or testing negative for high-risk HPV on self-collected samples. This definition differs from that of previous trials that defined uptake as self-collection participation (ie, returning a self-collection kit) or considered self-collected results that were positive for HPV 16 or HPV 18 as screening uptake regardless of in-clinic follow-up.

Implications of all the available evidence

Mailed HPV self-collection testing has potential to increase uptake of cervical cancer screening among under-screened women.

Data sharing

Investigators interested in accessing these data for the purposes of future studies can do so under the following conditions: Institutional Review Board approval has been obtained from the University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, the institution covering the investigator; data security procedures ensuring patient privacy have been shown by the investigator; and a Data Use Agreement has been filled. Final datasets for analysis will not include any identifying information.