HPV MES began in 2015 with a pilot study (N = 202) offering HPV self-collection testing to evaluate if this would be an acceptable and impactful method of primary cervical cancer screening in multi-ethnic, rural areas of Guatemala. The pilot cross-sectional study demonstrated high acceptability of self-collection [19, 20], so we proceeded to recruit a cohort in 2016 to investigate the impact of a HPV self-collection program on rates of future screening and follow-up. Women in the cohort study were offered HPV self-collection testing at baseline and then were re-contacted at two annual interviews to identify continued interactions with existing screening programs, creating three waves of data collection: baseline (survey and optional HPV testing), follow-up in 2017 (FU1: survey), and follow-up in 2018 (FU2: survey). Surveys were specifically developed for this study and were based on the pilot survey which was created using the STEPwise Approach to Surveillance (STEPS) survey [27] and the University of Michigan’s Michigan HPV and Oropharyngeal Cancer study [28]. Blank copies of the baseline, FU1, and FU2 questionnaires can be found in the Supplementary Material (Survey A1, A2, A3).
HPV MES has been previously described in detail along with baseline acceptability results [19, 20]. Briefly, the cohort included 956 women from two distinct communities in rural Guatemala: Santiago Atitlan, Sololá (N = 500) and Livingston, Izabal (N = 456). In Santiago Atitlan, over 95% of participants identified as Tz’utujil Mayan, while in Livingston, 25% identified as Ladino (of Spanish or mixed descent), 32% as Garífuna (of Afro-Caribbean descent), and 42% as Q’eqchi Mayan [19]. For these analyses, the data was subset to participants between the ages of 25 and 54 (N = 438 in Santiago, N = 322 in Livingston), which is the age group eligible for cervical cancer screening according to guidelines in Guatemala.
Data
At baseline, women completed an in-home comprehensive questionnaire facilitated by community health workers (CHWs) and were offered a self-collection HPV test. After completion of the questionnaire, regardless of decision to self-collect, all participants received a voucher for free screening in their local private clinic. This was in addition to the free screening available in their local public clinics. Notably, while nearly all age-eligible Santiago participants chose to self-collect (N = 410, 94%), only 169 (53%) of age-eligible Livingston participants self-collected [19]. However, among those who collected, over 80% in both communities found the test comfortable and easy to use, and 95% were willing to continue using HPV testing as cervical cancer screening, demonstrating a high acceptability of the intervention [19]. Test results were returned to 347 (85%) and 113 (67%) of participants in Santiago and Livingston, respectively, within 3 months of recruitment, along with recommendations for next steps. Positive results were returned in-person by a physician and negative results were returned over the phone by a CHW. Of the participants who were not immediately contacted, 40 and 34 in Santiago and Livingston, respectively, had inconclusive test results that were retested throughout follow-up, and the remaining 23 and 22, respectively, were unable to be reached. At study conclusion in 2018, participants with initially inconclusive results were contacted and attempts were made to contact those who were previously unreachable.
At baseline, women were consented to participate in the baseline survey and HPV testing and separately consented to follow-up contact. Women who consented to follow-up were re-contacted by phone at FU1 and FU2. During follow-up waves, all consenting participants (regardless of decision to self-collect) were asked to complete a short questionnaire about any cervical cancer screening or follow-up care that they received that follow-up year.
Variable creation and statistical analysis
Any reported screen over follow-up was considered an initial screen (i.e. not a follow-up test in accordance with known prior screening results). At the time of the study, HPV testing was not approved for clinical use in Guatemala, so all women, regardless of HPV status, were referred for screening to the national screening program. It is possible that women who were HPV positive provided these results to their health care worker during screening, but, as the data is self-reported, we have no way to verify this. In practice, initial screening and follow-up testing would appear identical, with both involving either cytology or VIA.
We measured the prevalence of ever-screened for cervical cancer, which is an established measure of performance of screening programs [29], as well as the proportion of women compliant with country-specific screening guidelines [8, 30]. Screening compliance was defined as screening in the past 3 years, in accordance with the Guatemalan Ministry of Health recommendation. All data was self-reported [29], and we were unable to confirm receipt of screening after study participation through health clinics. Ever screened and screening compliance were calculated at each study wave for age (25–39/40–54), location (Santiago/Livingston), ethnicity (Tz’utujil or Ladino in Santiago; Q’echchi, Ladino, or Garifuna in Livingston), literacy (yes/no), education level (less than primary, primary or secondary, more than secondary), and the following, non-exclusive, six groupings: 1) those who completed self-collection, 2) those who did not complete self-collected, 3) those who received the results of their HPV test, 4) those who did not received results, and, of those who had received results, 5) those who tested positive and 6) those who tested negative.
We then investigated changes in screening behavior between that reported at baseline (for the 3 years prior to study participation) and that reported at FU1 (for the year post-entry) and FU2 (for the 2 years post-entry). We stratified the population at baseline by the above-described six HPV test collection-related groupings and location (Santiago/Livingston) and calculated the percent who reported screening prior to and post-baseline entry into the study. We used McNemar tests for paired proportions to assess statistical changes between the percent of women who had screened in the 3 years before baseline and the percent who screened during the first year and first 2 years after study entry for each group. This analysis was repeated after stratifying by participants who self-identified as literate (“I can read and write.”) versus those who identified as illiterate (“I cannot read and/or write.”).
To assess the changes in ever-screening and compliance at an individual level, we explored the flow across screening groups (ever-screened versus never, and compliant versus not) for all waves, using alluvial plots, which group categorical data into flows that can be traced across time points. Alluvial plots were constructed with memory, which tracked groups of individuals who had the same trajectories across the three waves. Plots were created for the overall population as well as for only those who had data at all three time points and were additionally stratified by location.
Finally, we investigated whether self-collection versus not, receipt of test result versus no receipt, and testing positive versus negative were associated with gaining or losing screening compliance. To investigate gaining compliance, the data was subset to baseline incompliant women. Those who screened during either follow-up wave were considered to have become compliant. Analogously, for loss of compliance, we subset to baseline compliant women. Those who had their last screen more than a year prior to enrollment and did not screen during the follow-up period were considered to have lost compliance. We ran multivariate logistic models to calculate odds ratios for each scenario. All models were adjusted for age (continuous), lifetime screening before study enrollment (yes/no), urbanicity (yes/no), ethnicity (Ladino, Tz’utujil, Q’eqchi’, Garífuna, other), and literacy (yes/no), in parallel with prior HPV MES analyses [19, 20].