Participation and Follow‐Up Compliance in Cervical Cancer Screening According to Delivery Strategy for HPV Self‐Collection: A Community Trial in Underserved Colombian Areas

ABSTRACT

Low‐ and middle‐income countries (LMIC) continue to bear a high burden of cervical cancer (CC). Women's participation in CC screening is challenging in LMICs, but self‐collection of HPV tests has shown increased participation in controlled trials. However, data from real‐world studies and data on follow‐up compliance after positive screening test results are scant. A two‐arm community trial was conducted in four municipalities from two different territories in Colombia (HPV self‐ and clinician‐collected tests). A comparison of participation and follow‐up compliance was done between arms, both delivered through campaigns and routine screening. Women aged 30–50 without screening records within the previous 30 months were eligible. HPV‐positive women were recalled for triage/treatment in a second visit. A follow‐up was done at 6 months. In total, 1400 women were expected, but only 532 attended the screening visit. Participation rates were 49% (95% CI: 46–53) in the self‐collected and 26% (95% CI: 22.8–29.5) in the clinician‐collected arm. Campaigns resulted in higher participation than routine screening for both collection methods (p < 0.05), and significant differences by territory were observed. Follow‐up compliance was 89.6% and 71% at triage/treatment and 6‐month visits, respectively, without significant differences between study arms. HPV self‐collection resulted in higher participation than clinician collection; however, it did not directly translate to screening coverage. Therefore, delivery strategies and contextual barriers should be carefully analyzed for successful implementation. Anyhow, the increased participation and high compliance with the triage/treatment visit support the relevance and suitability of alternative programmatic approaches for HPV‐based screening.

Abbreviations

CC

cervical cancer

DNA

deoxyribonucleic acid

HPV

human papillomavirus

ITS

intention to screen

LMIC

low‐ and middle‐income countries

NGOs

Non‐governmental Organizations

VIA‐VILI

Visual Inspection with Acetic Acid and Lugol's Iodine

1. Introduction

Low‐ and middle‐income countries (LMIC) continue to bear a high burden of cervical cancer, accounting for about 90% of cases and deaths around the world. Latin America and the Caribbean is the world region with the second highest incidence and mortality rate after Africa, but large disparities are observed within countries. In Colombia, cervical cancer is the third leading cause of cancer incidence and the fourth cause of cancer mortality among women, with age‐standardized rates of 13.7 and 6.9 per 100,000, respectively [1]. However, incidence rates among the different Colombian territories range from 3.1 to 28.3 per 100,000 [2], the latter corresponding to regions with low access to regular healthcare and, consequently, high cervical cancer mortality.

Similarly, cervical cancer screening participation varies greatly between urban centers and rural regions in Colombia, as in most LMICs, both settings lacking organized screening [3]. The implementation of Human papillomavirus (HPV)‐based screening in the country has progressed slowly. The technology has been available in the benefit package of the health system since 2011 (insurance coverage), but it was formally introduced in the screening programs only in 2018; yet, HPV‐based screening coverage remains very low (8.2% by 2023) [4]. In addition, the reports on follow‐up of women who screened positive show a loss to follow‐up of about 30% at 6 months, hindering the potential benefit of cervical cancer early detection [5, 6].

The positive effect of HPV self‐collection on women's participation has been extensively described [7, 8]; however, delivery for self‐collection has been restricted mainly to opt‐out strategies such as home visits and postal services, both of which require highly organized programs. Moreover, follow‐up compliance of HPV‐positive women shows high variability depending on the programmatic approach. Since the highest incidence and mortality rates are usually observed in underserved regions with low health system capacity, assessing the impact of alternative programmatic approaches in the implementation of HPV self‐collection is highly desirable.

To evaluate the effect of HPV self‐collection for cervical cancer screening on the participation rates of women with limited access to healthcare in Colombia, we developed a community‐based intervention in two Colombian departments (territories) by implementing two opt‐in delivery strategies (campaign and routine screening in health centers) and a combined follow‐up algorithm based on HPV genotyping. In this article, we report screening participation rates and follow‐up compliance among HPV‐positive women as primary outcomes.

2. Methods

2.1. Setting

Choco and Putumayo are Colombian departments (territories) with known barriers to regular healthcare and large vulnerable populations. These departments are characterized by limited transportation and road infrastructure, as well as a high percentage of women affected by armed conflict and living in rural and distant areas [9, 10].

We carried out the study in two municipalities from each territory: the Sibundoy Valley and Puerto Leguizamo in Putumayo, and Istmina and Nuqui in Choco. The Sibundoy Valley is in the foothills of the Andean Mountains (southwest Colombia), with accessible roads. In contrast, Puerto Leguizamo is in the Colombian Amazonian region, with access only by boat or airplane (Figure S1). Both municipalities have a high representation of indigenous groups (38.3%) and rural populations (45.9%) [11].

Choco is a department located in northeast Colombia. Istmina is near the department's capital city with an accessible road and a low proportion of rural population. In contrast, Nuqui is accessible only by boat or airplane, and the rural population is 93.2%. The Choco municipalities are characterized by large Afro‐Colombian populations (73.7%) [11].

2.2. Study Design

A two‐arm community trial with a crossover design was carried out from September 2024 to June 2025 in four rural municipalities (two departments) in Colombia (Figure 1). Municipalities were randomly allocated to begin with a self‐collection or clinician‐collection campaign and then move to a 3‐week routine screening with the alternative (opposite) collection method for HPV testing. Participating women were offered cervical cancer screening accordingly.

FIGURE 1.

Study design. Municipality Access—Town A: Urban towns—Sibundoy Valley (Putumayo), Istmina (Choco); Town B: Distant towns—Puerto Leguizamo (Putumayo), Nuqui (Choco). The sample size in arm 1 was rounded to 720 women to comply with the size of the batch in the BD Onclarity technology.

Women aged 30–50, with no record of cervical cancer screening in the previous 30 months or unknown results of screening tests performed in the previous 24 months, were eligible. Potential participants with cognitive impairment, known pregnancy, history of hysterectomy, or cervical cancer were excluded.

A team from Hospital Universitario San Ignacio in Bogotá trained the healthcare staff in local hospitals in all components of the strategy, assisted with the campaigns, and provided the treatment and gynecological assessment of positive‐screened women.

Due to the lack of population‐based screening, the invitation to participate was extended through local channels. Local hospitals developed an outreach strategy that combined individual calls to women eligible for cervical cancer screening (when valid contact information was available) with broad invitations in local media promoting the campaigns. Screening campaigns were carried out during 3 days at each municipality. During those days there was a greater availability of staff at the local hospitals to support the sample collection, the logistics, and the guidance of women attending the campaigns. Immediately after the campaign, a 3‐week routine screening period took place, with the opposite collection method (Figure 1); the availability of the HPV‐based screening was informed to the community by regular communication strategies. Recruitment was sequential, based on arrival (Figure 2).

FIGURE 2.

Study recruitment flowchart.

For the women assigned to self‐collection, instructions were provided by the healthcare staff, and an infographic brochure was provided with the test kit. All women were tested for the presence of HPV deoxyribonucleic acid (DNA) using a clinically validated HPV assay [12]. The sample was collected by the patient (dry swab) or by a trained lab professional according to assignment to a study arm. Samples were collected at the health centers and sent to Hospital Universitario San Ignacio for processing according to the manufacturer's instructions. Each specimen collected was transferred to an HPV liquid‐based cytology diluent tube. The specimens were then loaded onto a Viper system that performed all steps for the extraction and amplification of target DNA. Samples were run in batches of 30, each with an individual control. The results were individually reported and returned to the health centers in the participating municipalities to inform the participating women.

HPV‐positive women were scheduled for a gynecological evaluation and were managed according to a treatment algorithm based on HPV genotyping (Figure S2). Treatment was provided to all women with a positive triage test with ablative treatment (thermal ablation or cryotherapy) upon availability of treatment devices. Women not eligible for immediate treatment underwent colposcopy/biopsy.

All HPV‐positive women were reassessed at a 6‐month follow‐up visit. During the follow‐up visit, visual inspection with acetic acid (VIA) and Lugol's iodine (VIA‐VILI) was performed for all women [13], and ablative treatment was provided to all women with positive results [14].

2.3. Educational Activities

An educational session was conducted during the screening visit (regardless of delivery strategy) to educate women about cervical cancer prevention and screening. At the second visit, an additional session was conducted to educate HPV‐positive women in detail about treatment options for precancerous lesions.

2.4. Sample Size and Statistical Analysis

We expected a 10% increase in participation with self‐collection (reported cytology coverage was 27.5% for Choco and 57.6% for Putumayo) [3]. With 95% confidence, 90% power, and design effect 2, we expected 1368 participants (684 per arm); thus, we anticipated 1400 women invited with a stratified sample of 700 women per district (350 by municipality). We planned a subgroup analysis by district (Choco and Putumayo) and delivery strategy (campaign and routine screening).

The characteristics of the study population are described. The primary intention‐to‐screen (ITS) population included all patients according to the initial arm assignment and delivery strategy. Screening participation was defined as the proportion of invited women who underwent the screening test. Follow‐up compliance was defined as the proportion of HPV‐positive women who attended the triage/treatment and 6‐month follow‐up visits. A comparison between self‐ and clinician collection was done as a difference in proportions.

The average participation and compliance rates by community (municipality) are described, as well as the effect in the crossover design. The difference in proportions was estimated with a X ² test (95% CI reported) and the corresponding p‐values.

A Mantel–Haenszel stratified analysis was conducted to evaluate the effect of the strata on differences in participation and compliance rates between the study arms. The analysis was conducted across three categorical confounders: department (territory), delivery strategy, and municipality access (distant vs. urban towns). This approach allowed for the estimation of a weighted pooled odds ratio, accounting for the potential effect of the strata on the primary outcomes per study arm.

All analyses were performed using STATA version 19.5 and the R statistical software.

3. Results

A total of 532 women attended the screening visit and underwent the screening test, 380 in the self‐collected arm and 152 in the clinician‐collected arm. Moreover, 385 women attended the screening visit during the campaign and 147 during the routine screening period (Figure 2). The positivity rate of HPV testing was 25.2% overall, with 28.4% in the self‐collected arm and 17.1% in the clinician‐collected arm.

Given the challenges of providing individual invitations and the absence of population‐based screening, data on nonattending women were not available or were inaccurate; therefore, they are not reported. The mean age for women attending screening was 38.9 (SD: 5.9), 77.6% lived in rural areas, 35.0% were Afro‐Colombian, and 22.0% were indigenous. Most participants belonged to the subsidized health insurance regimen for low‐income populations (67.1%). In total, 6.6% of participants reported cytology screening within the previous 24 months (with unknown results), 67.1% between 25 and 36 months, and 26.3% over 36 months since the last screening test. The study arms differed by ethnic identity and health insurance regimen (Table 1). Characteristics of participants by municipality are described in detail in Table S1.

TABLE 1.

Baseline demographic and clinical characteristics of participant women.

Baseline characteristics	HPV self‐collected n = 380 (100%)		HPV clinician‐collected n = 152 (100%)		Total n = 532 (100%)		p
Age
Mean (SD)	39.1 (6.0)		38.5 (5.7)		38.9 (5.9)		0.412
Area of residence
Urban	300	78.9	113	74.3	413	77.6	0.129
Rural	71	18.7	38	25.0	109	20.5
NA	9	2.4	1	0.7	10	1.9
District
Putumayo	231	60.8	116	76.3	347	65.2	0.001
Choco	149	39.2	36	23.7	185	34.8
Ethnic group
Indigenous	86	22.6	31	20.4	117	22.0	0.001
Afro‐Colombian	149	39.2	37	24.3	186	35.0
Other	145	38.2	84	55.3	229	43.1
Health insurance
Payroll (contributory)	139	36.6	32	21.1	171	32.1	0.002
Subsidized	239	62.9	118	77.6	357	67.1
No insurance	2	0.5	2	1.3	4	0.8
HPV vaccination
One dose or more	71	18.7	24	15.8	95	17.9	0.431
Not vaccination history	309	81.3	128	84.2	437	82.1
Time since last screening
≤ 24 months	24	6.3	11	7.2	35	6.6	0.916
25–36 months	255	67.1	102	67.1	357	67.1
> 36 months	101	26.6	39	25.7	140	26.3
Parity
Mean (SD)	2.71 (1.9)		2.54 (1.5)		2.66 (1.8)		0.662

Note: Payroll corresponds to the insurance regimen for population with payment capacity and the subsidized for low‐income populations without payment capacity.

Abbreviations: HPV, Human papillomavirus; NA, Not available; SD, standard deviation.

Of the expected number of invited women, screening participation was 38.0%, with 49.6% (95% CI: 45.9–53.3) in the HPV self‐collected arm and 25.7% (95% CI: 22.5–29.2) in the HPV clinician‐collected arm (p < 0.001). Participation rates were significantly higher during the campaigns (55.0%; 95% CI: 51.2–58.7) in comparison to routine screening (21.0%; 95% CI: 18.0–24.2). We also observed significant differences in participation by district (Table 2).

TABLE 2.

Screening participation rates.

Screening participation	HPV self‐collected		HPV clinician‐collected		p
	(n) %	95% CI	(n) %	95% CI
Total	(357) 49.6	45.9–53.3	(175) 25.7	22.5–29.2	< 0.001
District
Putumayo	(235) 65.3	60.1–70.2	(112) 32.9	28.0–38.2	< 0.001
Choco	(122) 33.9	29.0–39.0	(63) 18.5	14.5–23.1	< 0.001
Delivery strategy
Campaign	(239) 66.4	61.3–71.3	(146) 42.9	37.6–48.4	< 0.001
Routine screening	(118) 32.8	27.9–37.9	(29) 8.5	5.8–12.0	< 0.001

Abbreviations: CI, confidence interval; HPV, Human papillomavirus.

Compliance with the triage/treatment visit among HPV‐positive women was 89.6%, with 86.7% (95% CI: 78.4–92.7) in the self‐collected arm and 97.2% (95% CI: 85.5–99.9) in the clinician‐collected arm. There were no significant differences among the study arms or subgroups of analysis (Table 3a). The compliance at the 6‐month visit was 70.9% overall, without significant differences between study arms. No differences were observed between the predefined study subgroups (Table 3b). No differences in compliance at 6 months were observed between women with positive triage (immediate treatment) and women with negative triage results (follow‐up without treatment). Sensitivity analysis using a per‐protocol approach confirmed the ITS findings, with no significant changes in the outcomes (Tables S2 and S3).

TABLE 3.

Follow‐up compliance.

Study arm/Subgroup	HPV self‐collected		HPV clinician‐collected		p
	(n) %	95% CI	(n) %	95% CI
A. Compliance at the treatment visit
Total	(85) 86.7	78.4–92.7	(35) 97.2	85.5–99.9	0.079
District
Putumayo	(54) 87.1	76.1–94.3	(22) 100.0	—	0.077
Choco	(31) 86.1	70.5–95.3	(13) 92.9	66.1–99.8	0.510
Delivery strategy
Campaign	(53) 84.1	72.7–92.1	(34) 97.1	85.1–99.9	0.050
Routine screening	(32) 91.4	76.9–98.2	(1) 100.0	—	0.760
B. Compliance at the 6‐months visit
Total	(70) 71.4	61.4–80.1	(25) 69.4	51.9–83.7	0.823
District
Putumayo	(42) 67.7	54.7–79.1	(17) 77.3	54.6–92.2	0.401
Choco	(28) 77.8	60.8–89.9	(8) 57.1	28.9–82.3	0.054
Delivery strategy
Campaign	(48) 76.2	63.8–86.0	(24) 68.6	50.7–83.1	0.413
Routine screening	(22) 62.9	44.9–78.5	(1) 100.0	—	0.446

Abbreviations: CI, confidence interval; HPV, Human papillomavirus.

After adjusting for district, delivery strategy, and municipality access with the Mantel–Haenszel test, no interaction was observed across the strata over participation rates, triage/treatment visit compliance, and 6‐month follow‐up compliance (OR 1.9, 2.1, and 0.5, respectively, with respective p‐values of 0.513, 0.437, and 0.281) (Table S4).

4. Discussion

Our study found higher participation with HPV self‐collection but also a significant effect of delivery strategies with higher screening uptake among women participating in the campaigns in comparison to routine screening offered by local hospitals. To our knowledge, this is the first study analyzing delivery strategies for HPV self‐collection in Latin America beyond its comparison with clinician collection.

HPV self‐sampling has consistently shown higher women's participation than clinician‐sampling in clinical trials [7, 8]; however, participation rates vary greatly between delivery strategies [15, 16]. In general, opt‐out strategies outperform opt‐in strategies; the most frequently studied opt‐out strategies are door‐to‐door and direct mailing for the delivery of sampling devices [16].

Opt‐in strategies are the most frequent strategies implemented in high‐income countries, supported by organized screening programs, robust infrastructure, and availability of resources [17, 18], however, in some countries opt‐in strategies are being used for nonattenders or underscreened populations [19, 20]. Yet, several studies have shown no effect of opt‐in strategies on screening uptake when compared with regular care, and the most frequent opt‐in strategies among hard‐to‐reach women in clinical trials are different invitation means of collecting the sampling devices or directly collecting the samples in health centers (regular care) [8, 15, 16].

Despite the advantage of opt‐out strategies on participation rates, implementation of door‐to‐door strategies and the affordability of direct mailing strategies are challenging for most LMICs lacking organized screening, and data from controlled trials comparing different opt‐in strategies are scarce. In Latin America, we found three controlled trials reporting participation rates with HPV self‐collection, but all of them used door‐to‐door delivery and compared HPV self‐collection to cytology testing or HPV clinician‐collection in health centers [21, 22, 23]; thus, the comparison of collection methods is biased by delivery strategy.

We used two opt‐in strategies in our study and compared self‐ and clinician‐collection methods independently in every strategy, observing higher participation rates in campaigns vs. routine screening for both collection methods (Table 2). Our results are consistent with data from a previous controlled trial (the only one found) comparing community health campaigns and facility‐based HPV self‐collection, where participation rates were 60% and 37%, respectively [24]. Although the observed participation rates in campaigns are lower than those reported for opt‐out strategies in clinical trials, we deem our results promising since the reported coverage for both territories in our study is even lower, as described in the methods section. Furthermore, door‐to‐door or mail‐in strategies are not feasible in Colombia, as in many regions in LMICs.

There is no extended experience with campaigns aimed at increasing screening uptake. However, lessons learned from immunization programs show that campaigns should not be considered the primary strategy for improving screening coverage; instead, they should be used as a complementary “periodic intensification of routine screening” [25]. Complementary mass screening campaigns could be suitable for delivering HPV self‐collection tests to hard‐to‐reach women and underserved populations, and consequently could have a positive impact on equitable access to screening [26]. This is particularly relevant during the implementation of HPV self‐collection, given the low literacy level concerning this alternative among the general population but its high acceptability once the women learn of it [27, 28].

Our findings confirm that the use of HPV self‐collection as part of a screen‐and‐treat real‐world delivery model is an effective strategy that enables higher participation in mass screening and allows adequate follow‐up of HPV‐positive women in a short timeframe [29]. As previous studies have reported, the flexibility of implementing HPV self‐collection in different settings helps reduce logistical barriers in remote areas [30]. In addition, the high acceptability in terms of ease of use and reduction of individual barriers, enhancing physical privacy and comfort, and reducing providers‐associated anxiety results in higher uptake in special populations (indigenous/rural women) [16, 31, 32].

Acceptability measured via questionnaires or as acceptance/willingness to use the test among women who received the kit is in all cases restricted to participating women; thus, acceptability is not equivalent to participation, and it is not a surrogate measure of screening coverage. Consequently, the high acceptability of HPV self‐collection reported in clinical trials and cross‐sectional studies is not necessarily translated into participation rates and coverage, as demonstrated by the difference in participation rates between the two territories in our study (Table 2). Although the invitation and recruitment strategies were the same in Putumayo and Choco, the former had active participation of health centers and local Non‐Governmental Organizations (NGOs), whereas the latter had poor participation of health centers and no NGOs involved. This indicates the relevance of health system support and contextual barriers to implementing any alternative technology or programmatic approach.

We also found a difference in the HPV positivity rate among study arms, which is consistent with previous studies that have shown higher positivity rates with the self‐collection sampling method [33, 34]. A detailed analysis of this finding is out of the scope of this manuscript and deserves an independent publication.

Another relevant finding in our study was the high compliance observed with the second visit for triage and treatment versus the reduced compliance with the follow‐up visit at 6 months (Table 3). Low follow‐up compliance has been consistently reported as a drawback for cervical cancer screening based on HPV self‐collection [35]. Among controlled trials in Latin America, only the study by Arrossi et al. analyzed compliance with triage visits after positive HPV findings. That study reported 78% follow‐up compliance for diagnostic workout at 12 months with additional loss to follow‐up among patients with positive findings in the colposcopy/biopsy [23]. In addition, the extended implementation of HPV self‐collection in the Tucuman region of Argentina resulted in 42.9% compliance with the triage visit to perform cytology among HPV‐positive women [36].

The available information and the results in our study clearly indicate that, regardless of the implementation strategy, introducing HPV self‐collection in hard‐to‐reach populations should be complemented with proper programmatic approaches, such as screen‐and‐treat or screen‐triage‐and‐treat in one or two visits [37].

Our study has some limitations. The design as a community trial does not ensure proper comparison between study arms at the individual level; however, we consider this a convenient approach to evaluate the implementation of HPV self‐collection close to real world conditions. In addition, we used a stratified sample with proper adjustment, showing no interaction across the strata. Moreover, we found no significant differences in sociodemographic variables, except for the participation of ethnic communities belonging to every territory and affiliation to health insurance plans. The Colombian health system is based on individual insurance, and it has universal insurance coverage with no differences in benefit packages between insurance plans (contributory or subsidized); however, a higher percentage of women affiliated with the contributory insurance plan might be indicative of better socioeconomic conditions. Again, we observed that differences in health insurance plans are essentially driven by territory (76.8% and 62.0% subsidized regimens in Choco and Putumayo, respectively), and we found higher participation rates with HPV self‐collection and campaigns within every territory independently.

The differences in participation rates by delivery strategy should be carefully interpreted since the comparison could be limited by the sequential implementation of routine screening after the campaigns. Thus, the campaigns could have initially reached women who were more inclined to undergo screening. However, we consider this condition does not invalidate our findings of intensive campaigns on participation rates, considering that routine screening was already available before the campaigns took place and it is unlikely that all motivated women attended screening exclusively during the campaigns.

Despite the limitations, we believe our data to be of greater relevance since they represent the introduction of new technologies and programmatic approaches in real‐world settings, and they are the result of a trial stratified by delivery strategy and local context (territory). Accordingly, we expect the knowledge provided to inform policy making and screening program design and organization in LMICs.

Author Contributions

Natascha Ortíz: writing – review and editing, investigation, supervision, methodology. Carolina Morante‐Caicedo: investigation, writing – review and editing, methodology. Ginna Fernández‐Deaza: formal analysis, investigation, project administration, writing – original draft, validation, software, data curation, conceptualization, methodology, writing – review and editing. Alexander Salazar: investigation, writing – review and editing, methodology. Raúl Murillo: conceptualization, methodology, investigation, formal analysis, supervision, writing – original draft, writing – review and editing, funding acquisition, resources, project administration. Mario Rodríguez: investigation, writing – review and editing, methodology. Ana Karina Ceballos: data curation, validation, writing – review and editing. Juliana Rodríguez: investigation, writing – review and editing, methodology. Angélica Parra: writing – review and editing, investigation, methodology, supervision. Maria Alejandra Arria: data curation, writing – review and editing, validation. Diana Acosta: methodology, writing – review and editing, investigation, supervision. Juan David Toca: data curation, writing – review and editing, validation.

Funding

This study was funded by a grant from Pfizer Inc. through a call by the Asociación Colombiana de Hematología y Oncología (ACHO).

Ethics Statement

The study was approved by the Institutional Research and Ethics Committee of Pontificia Universidad Javeriana and Hospital Universitario San Ignacio in Bogota, Colombia, with approval number: FM‐CIE‐0951‐24. Since it was a community‐based trial, authorization by local health authorities in every municipality was obtained, and all participating women signed informed consent.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Geographical location, study sites.

Figure S2: Screen, triage and treat approach.

Table S1: Baseline demographic and clinical characteristics of participant women by municipality.

Table S2: Screening participation rates (Per‐protocol analysis).

Table S3: Follow‐up compliance (Per‐protocol analysis).

Table S4: Probability of compliance to the screening and follow‐up.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.