Effect of Mailed Human Papillomavirus Test Kits vs Usual Care Reminders on Cervical Cancer Screening Uptake, Precancer Detection, and Treatment

Key Points Question Do mailed human papillomavirus self-sampling kits increase detection and treatment of cervical precancers and screening uptake vs usual care (reminders for in-clinic screening)? Findings This randomized clinical trial included 19 851 women; 26% were screened after receiving a human papillomavirus kit vs 17% with usual care, a significant difference. There was no statistically significant difference in the number of cases of precancers detected or treated. Meaning This study indicates that mailing human papillomavirus kits to underscreened women can increase cervical cancer screening, and implementation efforts should strategize how to further increase kit uptake and follow-up of positive results to maximize detection and treatment of precancers in women at high risk.


INTERVENTIONS
The control group received usual care (annual patient reminders and ad hoc outreach from primary care clinics).The intervention group received usual care plus a mailed HPV selfsampling kit.
MAIN OUTCOMES AND MEASURES Two primary outcomes were (1) CIN2+ detection within 6 months of screening and (2) treatment within 6 months of CIN2+ detection.Screening uptake within 6 months of randomization was a secondary outcome.

Introduction
[13] In 2018, the US Preventive Services Task Force released updated cervical cancer screening guidelines 14 that include 3 recommended options for women aged 30 to 65 years: Papanicolaou testing alone, Papanicolaou and human papillomavirus (HPV) cotesting, and primary HPV-only screening (a new strategy).With primary HPV screening, home-based screening is an emerging option because HPV tests (unlike Papanicolaou tests) can be performed on clinician-or self-collected samples with comparable sensitivity. 15,16Furthermore, population-based randomized clinical trials in countries with organized screening programs (ie, centrally designed and managed, with standardized screening invitations and follow-up for a specific target population) demonstrated that mailing HPV self-sampling kits to underscreened women increased participation compared with invitations for clinic-based screening 15 and diagnostic follow-up compliance after HPV-positive selfsampling results was high, 15 yielding increased detection of cervical intraepithelial neoplasia grade 2 or higher (CIN2+). 15Consequently, Australia and the Netherlands-the first countries to implement primary HPV screening-have included HPV self-sampling options for underscreened women. 17,18ta from settings without organized screening programs-including opportunistic and/or health system-based screening programs in the United States-are needed to evaluate whether home-based HPV self-sampling increases screening participation and effectiveness.The Home-Based Options to Make Cervical Cancer Screening Easy (HOME) trial 19 compared a programmatic strategy of mailed HPV self-sampling kits with usual care (outreach via patient reminders to attend in-clinic screening) for increasing detection and treatment of cervical precancers and uptake of cervical cancer screening.The HOME study is the first US trial, to our knowledge, to evaluate effectiveness of HPV self-sampling as a screening method by assessing the entire cervical cancer prevention process (diagnostic follow-up and treatment) and measuring precancer outcomes.To our knowledge, it is also the first HPV self-sampling trial in any setting to evaluate treated precancer cases as an outcome.

Methods
The HOME study was a parallel, investigator-blinded, randomized clinical trial comparing 2 programmatic strategies for improving screening effectiveness and uptake in women who were not adherent to routine Papanicolaou screening.The trial was fully embedded within Kaiser Permanente Washington (KPWA) (see the article by Winer et al 19 1).The trial protocol is available in Supplement 1.
We identified potentially eligible women using electronic medical record (EMR) data; all eligible women were enrolled under a waiver of informed consent.No incentives were offered for participation.The institutional review boards determined that enrolling women under a waiver of informed consent was minimal risk compared with the scientific benefit of reducing participation bias.Women became eligible 5 months after receiving an annual preventive care reminder letter indicating they were due or overdue for their Papanicolaou screening (ie, no prior screening or most recent screening >3 years prior) to allow women to achieve screening uptake without additional intervention and to ensure that the 6-month time frame for assessing screening uptake ended before the next year's preventive care reminder letter was sent.Pathology management guidelines. 22Standardized protocols were developed by clinical and systemlevel collaborators to educate primary care teams on recommended follow-up (published previously 19 ).
We used programmatic extraction and manual EMR review to identify 2 primary outcomes: histologically diagnosed CIN2+ and treated CIN2+.The 2 primary outcomes were selected based on clinical relevance.We did not adjust for multiple comparisons or have an order or gatekeeping strategy for testing given that these 2 outcomes are strongly correlated with each other (CIN2+ We identified 63 789 women who were between the ages of 30 and 64 years, had not had a hysterectomy, had a primary care clinician at Kaiser Permanente Washington (KPWA), and had been continuously enrolled in the health system for at least 3 years and 5 months.Of these women, 17 256 (27.1%) had not had a Papanicolaou test within 3 years and 5 months.Postrandomization exclusions were applied owing to delays in data warehouse update.The total included in the intervention group analysis (9960) equals the total included in the round 1 analysis group (8120) plus the total included in the round 2 analysis group (1631) plus the total included in the round 3 analysis group (209).
Similarly, the total included in the control group analysis (9891) equals the total included in the round 1 analysis group (8111) plus the total included in the round 2 analysis group (1585) plus the total included in the round 3 analysis group (195).
treatment cannot occur without a CIN2+ diagnosis), and it was highly unlikely that only 1 would be significant.To count as an outcome, CIN2+ had to be preceded by an abnormal screening result within 6 months after randomization or rerandomization.Diagnosed CIN2+ was captured within 6 months after an abnormal screening result, and treated CIN2+ was captured within 6 months after a CIN2+ diagnosis.With these windows, each woman was followed up for a maximum of 18 months after randomization.
Secondary outcomes included screening uptake and abnormal screening results, obtained from the EMR.Uptake was captured within 6 months after randomization or rerandomization and defined as (1)   24 ) were derived from EMR data.Outcome data were available for all participants, including women who did not return a kit or opted out of medical record review (details of data access and aggregation for these groups were published previously 19 ).
The HOME trial was powered on the primary outcomes of diagnosed and treated CIN2+ assuming a 2-sided α of .05 and used assumptions based on the literature [25][26][27][28][29][30] and 2012 preliminary data. 19We estimated we would identify approximately 17 600 eligible women over 2.5 years and would have 85% power to detect between-group differences in proportions with diagnosed CIN2+ (0.188% vs 0.036% [relative risk {RR}, 5.2] in the intervention and control groups, respectively) and 81% power for treated CIN2+ (0.170% vs 0.033% [RR, 5.2] in the intervention and control groups, respectively).We estimated 100% power to detect differences in abnormal screening results (1.09% vs 0.44% [RR, 2.5] in the intervention and control groups, respectively) and screening uptake (35.9% vs 8.8% [RR, 4.1] in the intervention and control groups, respectively).

Statistical Analysis
We analyzed data using the intention-to-treat principle.Denominators for each group included all women randomized, minus the small number of women identified as ineligible after randomization (Figure 1).We described the distribution of participant characteristics by randomization group to identify any imbalances warranting adjustment in regression models.Outcome proportions in the intervention group were compared with the usual care group and RRs estimated using log-binomial regression.Absolute risk differences were estimated using binomial regression with an identity link.
Robust variance estimates were used to account for within-participant correlation due to rerandomized participants contributing more than 1 observation period.Statistical significance was defined as a 2-sided P < .05.
As a prespecified exploratory analysis, positive predictive value of an abnormal screening test for detecting CIN2+ was estimated within each randomization group.Two different denominators were evaluated: women with abnormal screening results warranting colposcopic referral and women receiving colposcopy after an abnormal result.In a post hoc analysis, we used Kaplan-Meier methods and log-rank tests to compare time from randomization to screening uptake in the control vs intervention group.Time to screening uptake was also compared among the control group vs intervention group women who returned HPV kits vs intervention group women with in-clinic screening only.Analyses were conducted using SAS statistical software version 9.4 (SAS Institute).

Primary Outcomes
Twelve patients with CIN2+ (0.12%) were detected in the intervention group (including 2 cases in women who returned a kit) vs 8 (0.08%) in the control group (Figure 2 and Table 2).The RR for CIN2+ detection was 1.49 (95% CI, 0.61-3.64) in the intervention vs control group.All women with CIN2+ (0.12%) were treated in the intervention group vs 7 in the control group (0.07%); the RR for CIN2+ treatment was 1.70 (95% CI, 0.67-4.32) in the intervention vs control group.

Additional Analyses
Four women returned nonviable HPV kits, which were not tested (including 2 received beyond maximum specimen stability of 21 days after the collection date and 2 with visible stool in the collection tube).All 4 were notified and offered a replacement kit; 3 returned a second kit within 6 months of randomization, received HPV-negative results, and were counted in the screening uptake numerator.Additionally, 47 of the 1440 intervention group women (3.3%) who went directly to in-clinic screening subsequently returned an HPV kit; these HPV kit results did not affect primary or secondary outcomes.
Time to screening uptake was shorter in the intervention group vs the control group (log-rank P < .001)(Figure 3).Restricting to women who completed screening uptake, the median restricting to those who received colposcopy.
a Self-sampling other hrHPV+ only or unsatisfactory results required in-clinic follow-up to complete screening uptake.
b Of these 20 women, 15 went directly to colposcopy and 5 went to Papanicolaou test or cotest before colposcopy.
c These 5 women completed screening uptake by receiving colposcopy in the 6 month screening window.
d This 1 woman did not complete screening uptake because colposcopy was completed outside the 6-month screening window.
e Includes 2 women with home kit other hrHPV+ and Papanicolaou test with unknown result.Because unknown Papanicolaou test results were coded as normal for analysis, the final result is considered other hrHPV+ and Papanicolaou test negative.One of these women received a colposcopy.
f Follow-up recommendations were per national guidelines.a Robust variance estimates were used to account for within-participant correlation due to rerandomized participants contributing more than 1 observation period.
b Screening uptake is defined as completion of screening episode; therefore, women who tested positive for human papillomavirus types other than 16 or 18 only or unsatisfactory on the mailed kit must have completed the additional step of in-clinic follow-up (Papanicolaou, cotest, or colposcopy).
c Abnormal screening result defined as a result that warrants repeated testing, surveillance, or immediate colposcopy (per national guidelines 22 ) before returning to a routine screening schedule.

Discussion
In this randomized clinical trial, there were no significant differences in CIN2+ detection or treatment among underscreened women who received a mailed HPV self-sampling kit compared with usual care alone.However, mailing HPV kits significantly increased screening uptake and reduced time to screening compared with usual care.
We are unaware of other HPV self-sampling trials that evaluated treated CIN2+ cases as an outcome.However, many international trials have evaluated the effect of mailing HPV kits to underscreened women on CIN2+ detection, with mixed results. 15High-risk HPV-positive selfsampling triage strategies vary across settings and influence CIN2+ detection rates. 15Detection of CIN2+ was higher comparing HPV self-sampling with control groups in trials in which women were referred directly to colposcopy (pooled RR including 6 trials, 3.0; 95% CI, 1.9-4.8)compared with trials that triaged women to additional screening (usually cytology) (pooled RR including 7 trials, 1.8; 95% CI, 0.8-4.0). 15One Australian trial triaged HPV-16 and HPV-18-positive results directly to colposcopy and other high-risk-positive results to Papanicolaou testing (similar to this trial); 76% received appropriate follow-up after an HPV-positive self-sampling result, and no significant between-group differences in CIN2+ detection were reported (although the trial was not powered on this outcome). 15,31Although our trial was powered on CIN2+ detection and treatment outcomes, power calculations were based on assumptions of higher kit uptake (30%), lower control group screening uptake (9%), and higher follow-up adherence (particularly for HPV-positive kit results [89%]) than was observed. 19Additionally, HPV positivity (11.3%) was higher than estimated (7.5%), but comparable to that reported in the most recent meta-analysis (pooled positivity, 11%; 95% CI, 15 Among the intervention women who self-sampled, most with nonnegative results attended in-clinic follow-up (70%); however, only 59% with HPV- self-sampling, additional resources to support screening, diagnosis, and treatment completion may be warranted. 32 a meta-analysis of 19 trials in countries with organized screening programs, 25% of underscreened women were screened (HPV kit or Papanicolaou test) after receiving an unsolicited HPV self-sampling kit in the mail, 15 similar to the 26% uptake in the intervention group.However, in this trial, slightly more women who screened in the intervention group chose in-clinic screening (14%) over self-sampling (12%), whereas in the meta-analysis, more women who were mailed a kit chose self-sampling (19%) over in-clinic screening (6%). 15The absolute (9%) and relative (1.5)   increases observed in screening uptake in the mailed HPV kit vs usual care alone group in our trial were within the lower range of estimates from international trials.The pooled absolute difference in the meta-analysis was 13% (95% CI, 10%-15%), and the pooled RR was 2.3 (95% CI, 1.9-2.9). 15Further investigation of US clinician and patient knowledge and attitudes are needed to understand why uptake of mailed HPV kit screening was lower than that observed internationally.
The HOME trial was strengthened by the pragmatic design, with direct integration of the intervention into existing clinical protocols.Data from EMRs were used to identify and randomize women who would be eligible for the intervention if it were adopted into clinical practice, thereby enhancing generalizability of the results.Results from the post hoc time-to-screen analysis add context to the findings; the rapid time to screening uptake associated with mailed kits is an additional benefit from a health system resource perspective.

Limitations
This study has several limitations.A key difference between HOME and similar international trials was the need to consider the impact of the HPV self-sampling intervention on Healthcare Effectiveness Data and Information Set quality metrics. 2 During the trial, only completed Papanicolaou tests counted toward the Healthcare Effectiveness Data and Information Set cervical cancer screening outcome.This directly affected the trial design in 2 important ways.First, to allow enough time for the delivery system to activate overdue women, it was necessary to delay mailing HPV kits for 5 months after the most recent annual preventive services reminder letter.Second, kit materials recommended all women attend Papanicolaou screening, even with HPV-negative self-sampling results, which may have negatively affected kit uptake.Furthermore, mailing kits as part of a research study may have had either a positive or negative effect on kit uptake.Results may have been different if (as in most international studies) HPV kits had been mailed to underscreened women as part of standard care and with the reassurance that negative results would not require in-clinic follow-up.An additional limitation was that kit materials were developed in English only, necessitating the exclusion of non-English-speaking women.

Conclusions
Mailing HPV kits to underscreened women increased screening uptake compared with usual care alone, with no significant differences in clinical outcomes.Nearly three-quarters of women who received the intervention remained underscreened, and follow-up among self-testing HPV-positive women has room for improvement.Health care systems in the United States considering implementing primary HPV screening and outreach strategies with HPV self-sampling should focus on approaches to increase kit uptake and follow-up of positive results.Additional research efforts should consider a variety of strategies to increase cervical cancer screening.

Question
Do mailed human papillomavirus self-sampling kits increase detection and treatment of cervical precancers and screening uptake vs usual care (reminders for in-clinic screening)?Findings This randomized clinical trial included 19 851 women; 26% were screened after receiving a human papillomavirus kit vs 17% with usual care, a significant difference.There was no statistically significant difference in the number of cases of precancers detected or treated.Meaning This study indicates that mailing human papillomavirus kits to underscreened women can increase cervical cancer screening, and implementation efforts should strategize how to further increase kit uptake and follow-up of positive results to maximize detection and treatment of precancers in women at high risk.
captured up to 6 mo after CIN2+ results Diagnosis: CIN2+ captured up to 6 mo after screening resultsScreening: Screening uptake captured up to 6 mo after randomization n

JAMA Network Open | Obstetrics and Gynecology
Effect of Mailed HPV Test Kits on Cervical Cancer Screening and Treatment JAMA Network Open.2019;2(11):e1914729. doi:10.1001/jamanetworkopen.2019.14729(Reprinted) November 6, 2019 3/14 Downloaded From: https://jamanetwork.com/ on 09/28/2023 1631 Included in round 2 analysis group 3 Excluded from analysis, recent disenrollment from KPWA 7 Excluded from the analysis, recent Papanicolaou test prior to randomization date 1585 Included in round 2 analysis group 2 Excluded from analysis, recent disenrollment from KPWA 3 Excluded from analysis, recent Papanicolaou test prior to randomization date 3473 Reassessed for eligibility 1 y after round 1 randomization 8283 Assigned to receive intervention 8283 Received intervention 8307 Assigned to receive usual care only 8307 Received usual care only 1641 Assigned to receive intervention 1641 Received intervention 1590 Assigned to receive usual care only 1590 Received usual care only 17 256 Assessed for eligibility 666 Excluded 541 Language interpreter needed 83 On do-not-contact list for research 42 Pregnant 242 Excluded 199 Language interpreter needed 36 On do-not-contact list for research 7 Pregnant 66 Excluded 53 Language interpreter needed 12 On do-not-contact list for research 1 Pregnant 16 590 Randomized (round 1) 3231 Rerandomized (round 2) 475 Reassessed for eligibility 1 y after round 2 randomization 209 Included in round 3 analysis group 1 Excluded from analysis, recent disenrollment from KPWA 4 Excluded from analysis, recent Papanicolaou test prior to randomization date 195 Included in round 3 analysis group 0 Excluded from analysis, recent disenrollment from KPWA 0 Excluded from analysis, recent Papanicolaou test prior to randomization date 214 Assigned to receive intervention 214 Received intervention 195 Assigned to receive usual care only 195 Received usual care only 9960 Total included in intervention group analysis 9891 Total included in control group analysis 409 Rerandomized (round 3) Effect of Mailed HPV Test Kits on Cervical Cancer Screening and Treatment JAMA Network Open.2019;2(11):e1914729. doi:10.1001/jamanetworkopen.2019.14729(Reprinted) November 6, 2019 5/14 Downloaded From: https://jamanetwork.com/ on 09/28/2023 From March 23, 2015, to August 29, 2016, 3231 women who were initially randomized to the control group and still eligible 1 year after randomization were rerandomized; between March 21, 2016, and August 29, 2016, 409 additional control group women were rerandomized a second time.Each rerandomization was treated as a distinct observation in the analyses.Therefore, the total number randomized was 20 230; 379 were retroactively excluded, and the remaining 19 851 (mean [SD] age, 50.1 [9.5] years) were included in the intention-to-treat analysis (9960 in the intervention group, 9891 in the control group).Baseline characteristics were similar between groups (Table1).

Table 1 .
Baseline Characteristics of Intervention and Control Group Participants (continued) Effect of Mailed HPV Test Kits on Cervical Cancer Screening and Treatment Figure 2. Diagram of Cervical Cancer Screening, Diagnosis, and Treatment Outcomes a Based on electronic medical record data.b Baseline characteristics are not available for 117 participants who opted out of electronic medical record review.c Calculated as weight in kilograms divided by height in meters squared.JAMA Network Open | Obstetrics and Gynecology JAMA Network Open.2019;2(11):e1914729. doi:10.1001/jamanetworkopen.2019.14729(Reprinted) November 6, 2019 8/14 Downloaded From: https://jamanetwork.com/ on 09/28/2023 22