Factors influencing pre-exposure prophylaxis adherence among female sex workers in Kinshasa, Democratic Republic of the Congo: a prospective cohort study Authors: Ben Bepouka, William M Miller, Antoinette Tshefu, Pierre Buekens Published: 2026 ABSTRACT AbstractBackgroundFemale sex workers (FSWs) in the Democratic Republic of the Congo (DRC) bear a disproportionately high burden of HIV infection. Daily oral pre-exposure prophylaxis (PrEP) is an effective HIV prevention strategy, but its public health impact depends on sustained adherence. Evidence of PrEP adherence among FSWs in the DRC remains limited. This study assessed PrEP adherence over time and identified factors associated with adherence among FSWs in Kinshasa.MethodsWe conducted a 6-month prospective cohort study among 120 FSWs receiving PrEP at four client-friendly healthcare centers in Kinshasa in 2024, hereafter referred to as Clinic A, Clinic B, Clinic C, and Clinic D. Adherence was assessed using self-reported missed doses over the previous seven days and pill count. Retention in care was documented at months 3 and 6. Multivariable logistic regression guided by a directed acyclic graph (DAG) was used to identify factors associated with adherence at inclusion, month 3, and month 6.ResultsThe median age of participants was 28 years (interquartile range [IQR]: 24–37). Based on pill count, adherence increased from 80.0% at inclusion to 93.6% at month 3 and 95.1% at month 6, whereas self-reported adherence declined over time. In DAG-informed models, clinic site was significantly associated with adherence at baseline. Compared with Clinic A, participants attending Clinic B (adjusted odds ratio [aOR] = 5.7; 95% confidence interval [CI]: 1.4–22.3) and Clinic C (aOR = 11.4; 95% CI: 2.1–61.1) were more likely to be adherent. No independent predictors were identified at later follow-up points. Retention in care was 92% at month 3 and 84% at month 6.ConclusionPrEP adherence among FSWs in Kinshasa was high and improved over time when measured objectively. Baseline site-level differences underscore the importance of contextual and programmatic factors in supporting adherence. These findings provide evidence to inform the optimization and scale-up of PrEP delivery models for key populations in the DRC and similar settings.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-026-13640-9.Keywords: Pre-exposure prophylaxis, Female sex workers, Adherence, Prospective cohort study, Democratic Republic of the Congo Background HIV cases have not declined as expected despite substantial global efforts and remain a major cause of mortality [1]. Worldwide, the most recent UNAIDS report estimates 1.3 million new HIV infections and 630,000 deaths in 2024 [2]. Although sub-Saharan Africa accounts for approximately 11% of the world’s population, it remains the global epicenter of the HIV epidemic, accounting for approximately 50% of all new HIV infections in 2024. An estimated 55% of newly acquired HIV infections among adults occur in key populations, including sex workers [1, 3, 4]. Daily oral pre-exposure prophylaxis (PrEP) effectively reduces the risk of HIV acquisition [5]. In the Democratic Republic of the Congo (DRC), HIV prevalence remains low in the general population [0.8%] but is substantially higher among female sex workers (FSWs) [7.5%], men who have sex with men (MSM) [7.5%], and people who inject drugs (PWID) [3.9%] [6]. Between February and May 2018, the International Center for AIDS Care and Treatment Programs (ICAP) conducted a PrEP demonstration project in seven clinics in the DRC—three in Lubumbashi and four in Kinshasa. Among 469 eligible clients offered daily oral PrEP, 66.3% received at least one adherence assessment, and 39.0% of these reported poor adherence [7]. According to the most recent data from PrEPWatch, approximately 69,838 individuals had initiated oral PrEP in the DRC [8]. PrEP is currently offered to several key populations, including serodiscordant couples, sex workers, men who have sex with men, people who inject drugs, and transgender people [9]. ICAP-supported client-friendly centers in Kinshasa provide HIV prevention and care services to a large number of individuals from key populations, including female sex workers receiving PrEP. After five years of PrEP implementation in the DRC, it is important to assess adherence and retention in PrEP care among FSWs. In this context, sustained engagement refers to continued participation in PrEP care over time, including regular follow-up visits, adherence monitoring, counseling, and retention in care, in line with standard PrEP delivery practices and international recommendations. In ICAP-supported centers, these components are further reinforced through peer educator involvement and client-friendly service delivery models. Understanding the determinants of PrEP adherence among FSWs remains suboptimal. Although studies from sub-Saharan Africa suggest considerable variability in adherence across social and cultural contexts [10–16], data from the DRC remain scarce. Understanding context-specific determinants of PrEP adherence is therefore critical to inform targeted interventions and optimize program implementation. Given that PrEP delivery occurs across multiple clinics with potentially varying programmatic characteristics, we hypothesized that site-level factors may influence adherence outcomes among female sex workers. To address this gap, we conducted one of the first quantitative prospective cohort studies in the DRC to measure PrEP adherence among female sex workers. Methods Study design and setting We conducted a 6-month prospective cohort study between June and December 2024. This study is part of a mixed-methods project, with the qualitative component designed to explore contextual and behavioral factors influencing PrEP adherence to be reported separately. This study was designed as an observational cohort focusing on adherence among individuals already receiving PrEP. As such, no comparison group of non-PrEP users was included, as the primary objective was not to evaluate PrEP effectiveness but to identify factors associated with adherence within the cohort. The study was conducted at four healthcare centers in Kinshasa where PrEP was first implemented in 2018 through the ICAP program. For reasons of confidentiality, these sites are hereafter referred to as Clinic A, Clinic B, Clinic C, and Clinic D. Although all sites were located in Kinshasa and supported by the same program, they served different catchment areas with potentially varying socio-economic and behavioral characteristics. Site-level analysis was therefore conducted to account for potential contextual and programmatic differences that could influence adherence outcomes. Specific service delivery characteristics (e.g., staff experience, organization of care, or peer support) were not formally measured in this study. Study population The study population consisted of female sex workers attending the participating study clinics. Gender identity was categorized as cisgender women (assigned female at birth) and transgender women (assigned male at birth). Inclusion criteria were female sex workers who had exchanged sex for money in the previous 30 days, were aged 18 years or older, had been receiving PrEP at a participating clinic for at least three months, and were registered as taking PrEP at their most recent clinic visit. Exclusion criteria comprised a self-reported psychiatric disorder and an intention to leave the country within the subsequent six months. A small number of transgender women were included, as they are part of the key populations receiving PrEP services at the participating clinics and share similar risk profiles and service access as cisgender female sex workers. Sample size The sample size was calculated using a single-proportion formula with a finite population correction, assuming a 50% prevalence of adherence, a 95% confidence level (α = 0.05), and a margin of error of ± 5%. The calculation was performed using the OpenEpi sample size calculator and was based on the estimated number of female sex workers receiving PrEP in participating clinics, informed by prior ICAP program data (approximately 157 individuals). This yielded a minimum required sample size of 112 participants. To account for potential loss to follow-up, 120 participants were ultimately enrolled. Study procedures and data collection All FSWs attending the PrEP clinics were invited to participate in the study and provided written informed consent. Participants were recruited consecutively until the target sample size was reached. The questionnaire was developed in French and administered in French and Lingala, the most commonly spoken languages in Kinshasa. It was translated into Lingala and back-translated into French to ensure accuracy and consistency. Data were collected by a trained study team using mobile phones equipped with Open Data Kit (ODK) applications and uploaded to a secure web server. Data collectors received prior training on the study protocol, standardized interviewing techniques, and the use of the ODK platform. To ensure data quality, electronic forms included built-in validation checks, and data collection activities were supervised regularly. Data were reviewed periodically to identify and resolve inconsistencies. Four researchers from the University of Kinshasa visited the four study sites to recruit participants. Eligible participants were identified from routine clinic records of female sex workers receiving PrEP at participating sites. Health professionals at the participating clinics identified eligible FSWs during routine PrEP follow-up visits and referred them to the study team for enrollment. Meetings were also organized with peer educators to introduce the study and facilitate engagement. Study staff then screened participants for eligibility, obtained written informed consent, and conducted data collection. Each participant was assigned a unique alphanumeric study identification number. Data were collected using an inclusion questionnaire specifically developed for this study and adapted from Franks et al. [7]. The questionnaire captured individual-level factors (sociodemographic characteristics, sex work as the sole source of income, consistent condom use, stable sexual partnerships, and forced sex in the previous month), structural factors (access to transportation and staff attitude), PrEP duration, daily income, sexually transmitted infection (STI) status, alcohol use, and drug use. Measurement of PrEP adherence PrEP adherence was assessed using both subjective and objective measures to improve the robustness of adherence assessment. Self-reported adherence was assessed by asking participants about missed doses during the seven days preceding the interview; missing at least two doses was classified as non-adherence. Objective adherence was assessed using pill counts conducted at each follow-up visit, based on returned unused medication tablets. Low adherence was defined as a pill count coverage of less than 80%. This 80% threshold is commonly used in PrEP implementation studies to classify adherence as high (≥ 80%) or low (< 80%) [17]. Similar operational thresholds have been applied in PrEP studies conducted in sub-Saharan Africa [18, 19]. Pill count adherence was calculated as follows [20]: Retention in care Retention in care was defined as continued engagement in PrEP follow-up visits at the study clinics at each scheduled time point (month 3 and month 6). Participants were considered retained if they attended their scheduled follow-up visit within the predefined visit window. Participants who missed a scheduled visit and did not return to care within the follow-up period were classified as not retained. Retention in care differs from adherence, as retention reflects continued participation in care, whereas adherence refers to the extent to which participants correctly followed the prescribed PrEP regimen. Thus, a participant could be retained in care but have low adherence, or conversely, not retained but previously adherent. Bias Several strategies were implemented to minimize potential sources of bias. Consecutive recruitment at all participating sites was used to reduce selection bias. Standardized questionnaires, trained interviewers, and electronic data collection with built-in validation checks were used to minimize information bias. Because adherence was assessed using both self-report and pill count, the use of complementary adherence measures was intended to improve measurement validity. Nevertheless, residual confounding and measurement error cannot be completely excluded. Data analysis Data was extracted from ODK and analyzed using Stata version 18 (StataCorp LLC, College Station, TX, USA). Participants’ baseline characteristics were summarized using frequencies and percentages for categorical variables, and medians with interquartile ranges for non-normally distributed continuous variables. Differences in adherence at inclusion were assessed using the chi-square test or Fisher’s exact test, as appropriate. Adherence proportions were calculated at inclusion, month 3, and month 6 for both adherence measurement methods. Concordance between self-reported adherence and pill count was assessed using the kappa coefficient. Retention in care was calculated at months 3 and 6. No imputation procedures were performed for missing data. Analyses at each follow-up time point were restricted to participants with available adherence data. Adherence was treated as a binary outcome (adherent vs. non-adherent). Multivariable logistic regression models were used to assess factors associated with adherence at inclusion, month 3, and month 6. Covariate selection was guided by a Directed Acyclic Graph (DAG), developed using DAGitty, to ensure appropriate control of confounding while avoiding adjustment for mediators or colliders. The DAG is presented in Supplementary Figure S1. Clinic site was considered the primary exposure of interest, reflecting contextual and programmatic differences across facilities. Age and duration of sex work were identified as confounders due to their potential influence on both clinic attendance and PrEP adherence. Consistent condom use was included as a behavioral covariate related to risk perception and prevention experience. Variables such as staff attitude, STI status, and income were not adjusted for, as the DAG indicated that they functioned as mediators or colliders. Ethical considerations The study was approved by the Ethics Committee of the University of Kinshasa School of Public Health (ESP/CE/036/2023) and Tulane University (ref. 2023 − 1828). Written informed consent was obtained from all participants. To preserve participant anonymity, all informants are referred to using pseudonyms. All study procedures were conducted in accordance with the ethical standards of the institutional research committee and the principles of the Declaration of Helsinki. Results General characteristics of FSWs at inclusion The study was conducted at four sites (Clinic A–D) where PrEP was first implemented in 2018 through the ICAP project. Among 140 FSWs who attended the sites for follow-up visits during the study period, 120 FSWs receiving PrEP were included in the analysis (Supplementary Figure S2). Thirty participants were enrolled at each site. The median age of participants was 28 years (IQR: 24–37), and 83 (69%) were aged 25 years or older. Overall, 80% of participants were classified as adherent at inclusion. The proportion of adherent participants was significantly higher at Clinic C compared with the other sites (p = 0.035). Participants with more than 10 years of sex work experience represented the largest subgroup (38%).In bivariate analyses, marital status was statistically associated with adherence (p = 0.043). However, when categories were grouped, no clear association was observed (OR = 1.20; 95% CI: 0.39–3.66), and no consistent pattern across categories emerged. The small number of participants in some subgroups further limited the interpretability of this finding. Adherence did not differ significantly according to gender identity (cisgender vs. transgender women) (p = 0.253), level of education (p = 0.820), age group (p = 0.542), duration of sex work (p = 0.729), or whether sex work was the sole source of income (p = 0.453). Similarly, no significant differences in adherence were observed according to history of forced sex (p = 0.116), STI status (p = 0.158), access to transportation (p = 0.707), income level (p = 0.401), tobacco use (p = 0.155), alcohol use (p = 0.261), having a stable sexual partner (p = 0.315), difficulty taking PrEP (p = 0.380), or staff attitude (p = 0.374). In contrast, participants who reported consistent condom use were less likely to be adherent compared with those who did not report consistent condom use (p = 0.043) (Table 1). Table 1General characteristics of FSW at inclusionVariablesAll n Adherentn (%)Non-adherentn (%)p-value Sites Clinic AClinic BClinic CClinic D3030303018(60.0)25(83.3)28(93.3)25(83.3)12(40)5(16.7)2(6.7)5(16.7)0.010 Comparison of sites Clinic COther sites (Clinic A, Clinic B, Clinic D)Clinic BOther sites (Clinic A, Clinic C, Clinic D)3090309028(93.3)68(75.6)25(83.3)71(78.9)2(6.7)22(24.4)5(16.7)19(21.1)0.0350.44 AgeAge Median (IQR)Extremes< 20 years20–24>= 25 years28(24–37)18–52433833(75.0)29(87.8)66(79.5)1(25.0)4(12.1)17(20.5)0.542 Gender Cisgender Female Transgender female115591(79.1)5(100.0)24(20.9)0(0.0)0.253 Civil status SingleDivorcedMarriedCohabitingWidowed901718472(80.0)17(100.0)0(0.0)4(50.0)3(75.0)18(20.0)0(0.0)1(100.0)4(50.0)1(25.0)0.043 Level of education NonePrimarySecondary931807(77.8)26(83.9)63(78.8)2(22.2)5(16.1)17(21.3)0.820 Sex as the only source of income YesNo1011982(81.2)14(73.7)19(18.8)5(26.3)0.453 Consistent condom use YesNo685250(73.5)46(88.5)18(26.5)6(11.5)0.043 Having a stable sexual partner YesNo358530(85.7)66(77.6)5(14.3)19(22.4)0.315 Forced to have sex last month YesNo1710316(94.1)80(77.7)1(5.9)23(22.3)0.116 STI test NegativePositiveNot done21603915(71.4)46(76.7)35(89.7)6(28.6)14(23.3)4(10.3)0.158Access to transportYesNo744660(81.1)36(78.3)14(18.9)10(21.7)0.707 Daily income < 10 dollars10–49 dollars50–100 dollars3183623(74.2)69(83.1)4(66.7)8(25.8)14(16.9)2(33.3)0.401 Drug use* YesNo279319(70.4)77(82.8)8(29.6)16(17.2)0.155 Alcohol use* YesNo952578(82.1)18(72.0)17(17.9)7(28.0)0.261 Difficulty taking PrEP YesNo813963(77.8)33(84.6)18(22.2)6(15.4)0.380 Staff attitude Fairly goodgoodvery good181381(100.0)62(76.5)33(86.8)0(0.0)19(23.5)5(13.2)0.374 Sex work duration 6–11 month12–23 month2–4 years5–10 years> 10 years163041421(100.0)5(83.3)24(80.0)35(85.4)31(73.8)0(0.0)1(16.7)6(20.0)6(14.6)11(26.2)0.729*Self-reported use of illegal psychoactive substances (such as cocaine, amphetamines, and cannabis) at least once during the previous three months is referred to as “drug use,” and alcohol consumption during the same time frame is referred to as “alcohol use” General characteristics of FSW at inclusion Clinic A Clinic B Clinic C Clinic D 30 30 30 30 18(60.0) 25(83.3) 28(93.3) 25(83.3) 12(40) 5(16.7) 2(6.7) 5(16.7) Clinic C Other sites (Clinic A, Clinic B, Clinic D) Clinic B Other sites (Clinic A, Clinic C, Clinic D) 30 90 30 90 28(93.3) 68(75.6) 25(83.3) 71(78.9) 2(6.7) 22(24.4) 5(16.7) 19(21.1) 0.035 0.44 Median (IQR) Extremes < 20 years 20–24 >= 25 years 28(24–37) 18–52 4 33 83 3(75.0) 29(87.8) 66(79.5) 1(25.0) 4(12.1) 17(20.5) Gender Cisgender Female Transgender female 115 5 91(79.1) 5(100.0) 24(20.9) 0(0.0) Single Divorced Married Cohabiting Widowed 90 17 1 8 4 72(80.0) 17(100.0) 0(0.0) 4(50.0) 3(75.0) 18(20.0) 0(0.0) 1(100.0) 4(50.0) 1(25.0) None Primary Secondary 9 31 80 7(77.8) 26(83.9) 63(78.8) 2(22.2) 5(16.1) 17(21.3) Yes No 101 19 82(81.2) 14(73.7) 19(18.8) 5(26.3) Yes No 68 52 50(73.5) 46(88.5) 18(26.5) 6(11.5) Yes No 35 85 30(85.7) 66(77.6) 5(14.3) 19(22.4) Yes No 17 103 16(94.1) 80(77.7) 1(5.9) 23(22.3) Negative Positive Not done 21 60 39 15(71.4) 46(76.7) 35(89.7) 6(28.6) 14(23.3) 4(10.3) Access to transport Yes No 74 46 60(81.1) 36(78.3) 14(18.9) 10(21.7) Daily income < 10 dollars 10–49 dollars 50–100 dollars 31 83 6 23(74.2) 69(83.1) 4(66.7) 8(25.8) 14(16.9) 2(33.3) Drug use* Yes No 27 93 19(70.4) 77(82.8) 8(29.6) 16(17.2) Alcohol use* Yes No 95 25 78(82.1) 18(72.0) 17(17.9) 7(28.0) Difficulty taking PrEP Yes No 81 39 63(77.8) 33(84.6) 18(22.2) 6(15.4) Fairly good good very good 1 81 38 1(100.0) 62(76.5) 33(86.8) 0(0.0) 19(23.5) 5(13.2) 6–11 month 12–23 month 2–4 years 5–10 years > 10 years 1 6 30 41 42 1(100.0) 5(83.3) 24(80.0) 35(85.4) 31(73.8) 0(0.0) 1(16.7) 6(20.0) 6(14.6) 11(26.2) *Self-reported use of illegal psychoactive substances (such as cocaine, amphetamines, and cannabis) at least once during the previous three months is referred to as “drug use,” and alcohol consumption during the same time frame is referred to as “alcohol use” Adherence during follow-up, by measurement method Based on self-reported measures, adherence declined over follow-up, from 81.7% at inclusion to 76.2% at month 3 and 67.3% at month 6. In contrast, adherence measured using pill count increased over time, from 80.0% at inclusion to 93.6% at month 3 and 95.1% at month 6 (Table 2). Table 2Adherence during follow up by measurement methodAdherence measurement methodsAt inclusionn (%)Month 3n (%)Month 6n (%) Self-report AdherentNon adherent98 (81.7)22 (18.3)83(76.2)26(23.9)68 (67.3)33 (32.7) Pill count AdherentNon adherent96 (80.0)24 (20.0)102 (93.6)7 (6.4)96 (95.1)5 (4.9) Adherence during follow up by measurement method Adherent Non adherent 98 (81.7) 22 (18.3) 83(76.2) 26(23.9) 68 (67.3) 33 (32.7) Adherent Non adherent 96 (80.0) 24 (20.0) 102 (93.6) 7 (6.4) 96 (95.1) 5 (4.9) Concordance between adherence measures Agreement between self-reported adherence and pill count was poor (κ = 0.14) and not statistically significant (observed agreement: 73.3%; expected agreement: 69.0%; standard error: 0.09; z = 1.53; p = 0.063). Given the low concordance between the two measures, pill count was used as the reference method for assessing adherence in subsequent analyses. Retention in care Retention in PrEP care was 92% at month 3 and 84% at month 6. Factors influencing PrEP adherence among FSWs In the final multivariable logistic regression model, clinic site remained significantly associated with PrEP adherence at baseline. Compared with Clinic A, participants attending Clinic B (aOR = 5.7; 95% CI: 1.4–22.3; p = 0.012) and Clinic C (aOR = 11.4; 95% CI: 2.1–61.1; p = 0.004) were more likely to be adherent, whereas no statistically significant difference was observed for Clinic D. At month 3, no statistically significant association between clinic site and PrEP adherence was observed. At month 6, adherence measured by pill count reached 95.1%, with only five participants classified as non-adherent. This near-uniform distribution of the outcome reflects a ceiling effect, resulting in limited variability and sparse outcome data. As a consequence, multivariable logistic regression models at month 6 were unstable, with non-convergence and/or quasi-complete separation, leading to infinite or non-estimable odds ratios and very wide confidence intervals, and therefore did not produce reliable estimates. Adjusted estimates are not presented. All models were adjusted for age, consistent condom use, and duration of sex work as confounders identified through a Directed Acyclic Graph. Adjusted estimates for these variables are not presented, as they represent conditional direct effects and are not interpreted as primary results (Table 3). Table 3Final multivariable logistic regression model based on the DAG framework for predictors of PrEP adherence among female sex workers at inclusion, month 3, and month 6VariableInclusionMonth 3Month 6aOR95% CIP- valueaOR95% CIp-valueaOR95% CIp-valueSiteClinic A11Clinic B5.71.4–22.30.0120.40.03–5.40.525--Clinic C11.42.1–61.10.0040.10.01–1.30.078--Clinic D2.90.8–11.40.105----aOR: adjusted odds ratio CI: confidence interval-: Not estimable due to sparse outcome data and quasi-complete separation (ceiling effect)Models were adjusted for age, consistent condom use, and duration of sex work, identified as confounders using a Directed Acyclic Graph (DAG). Adjusted estimates for these variables are not shown, as they represent conditional direct effects rather than total causal effects and are therefore not interpreted as primary results Final multivariable logistic regression model based on the DAG framework for predictors of PrEP adherence among female sex workers at inclusion, month 3, and month 6 aOR: adjusted odds ratio CI: confidence interval -: Not estimable due to sparse outcome data and quasi-complete separation (ceiling effect) Models were adjusted for age, consistent condom use, and duration of sex work, identified as confounders using a Directed Acyclic Graph (DAG). Adjusted estimates for these variables are not shown, as they represent conditional direct effects rather than total causal effects and are therefore not interpreted as primary results Discussion This study aimed to examine factors influencing adherence to pre-exposure prophylaxis (PrEP) among female sex workers (FSWs) in Kinshasa, Democratic Republic of the Congo (DRC). Adherence decreased over time according to self-reported measures but increased when assessed using pill counts. This divergence may reflect improved counseling and accountability during follow-up visits, as participants were required to return unused medication. While self-reported adherence is subject to recall and social desirability biases, it typically tended to overestimate adherence, especially when participants wished to present themselves in a positive light. The observed decrease in self-reported adherence, in contrast to the increase in pill count adherence, raises questions about the factors influencing these discrepancies. The discrepancy observed between self-reported adherence and pill count is more likely explained by recall bias and reporting inaccuracies inherent to self-reported measures. Participants may have had difficulty accurately recalling missed doses over the previous seven days, particularly in contexts where daily routines are unstable. In addition, substance use, including alcohol and other psychoactive drugs reported in this population, may impair memory and further reduce the reliability of self-reported adherence. These findings are consistent with prior studies showing that self-reported adherence often diverges from more objective measures due to recall limitations and social desirability bias. In contrast, the pill count measure, which is a more objective method, likely better reflects actual adherence behavior, although it may still be influenced by factors such as pill dumping or sharing. However, in the absence of biological measures of drug exposure, these interpretations should be considered hypothesis-generating rather than definitive. Similar findings have been reported in studies conducted in Uganda and Kenya, which documented low concordance (κ < 0.20) between subjective and objective adherence measures. These observations suggest that combining complementary methods such as pill count, self-report, and biological assays may provide a more accurate assessment of PrEP adherence among sex workers [21–28]. Recent evidence suggests that oral PrEP may remain highly effective for cisgender women even with less-than-daily dosing, such as four doses per week. This challenges the traditional definition of perfect adherence and highlights the need to consider adherence thresholds that more closely reflect biological protection rather than strict daily use [29]. Although our study was not designed to assess adherence at this level of granularity, future research should explore the proportion of individuals achieving protective levels of PrEP coverage, rather than relying solely on daily adherence metrics. Retention in PrEP care was high, with 92% of participants retained at month 3 and 84% at month 6. These findings are comparable to those reported in a recent systematic review and meta-analysis evaluating PrEP retention across facility-based and community-based models in sub-Saharan Africa. In that review, six-month retention was estimated at 83% for facility-based models and 76% for community-based models, with retention generally exceeding 80% at earlier follow-up points before declining over time [30]. The retention rates observed in our study therefore align with regional evidence and suggest good continuity of care within the study sites. In bivariate analyses, consistent condom use was associated with lower PrEP adherence, although this association was not retained in multivariable models. This finding may reflect differences in perceived HIV risk or prevention strategies and underscores the importance of emphasizing the complementary roles of PrEP and condoms in prevention counseling. Multivariable analyses indicated higher PrEP adherence among participants attending Clinic B and Clinic C. Similar site-level differences have been reported in studies of PrEP implementation across sub-Saharan Africa, highlighting the influence of contextual and programmatic factors on adherence [31–33]. Facilities with active peer educator involvement, well-trained staff, and extensive experience in supporting sex workers on PrEP have consistently demonstrated better adherence outcomes. Evidence from Kenya and South Africa similarly shows that clinics offering peer support and individualized adherence counseling achieve higher continuation rates than standard service delivery models [34–35]. The higher adherence observed in Clinics B and C may reflect underlying programmatic differences, such as greater peer educator involvement, more structured adherence counseling, or differences in staff experience and service organization. However, these factors were not formally measured in the present study and should be interpreted cautiously. These aspects are being explored in the qualitative component of this mixed-methods study, which will provide further insight into the mechanisms underlying site-level differences. The variability observed across sites in our study likely reflects differences in service organization, quality of adherence counseling, client engagement, and peer educator involvement, rather than intrinsic participant characteristics. The very high level of adherence observed at month 6 resulted in a ceiling effect, with very few non-adherence events. This limited variability in the outcome reduced statistical power and contributed to model instability and quasi-complete separation, preventing reliable estimation in multivariable analyses. Similar methodological challenges have been reported in studies with highly adherent populations, where sparse outcome data limit the performance and interpretability of conventional regression approaches. The emergence of long-acting PrEP formulations, including injectable agents administered every six months, represents a promising strategy to address adherence challenges associated with daily oral regimens [36]. Although such options are not yet widely available in the DRC, they may be particularly beneficial for key populations facing structural and behavioral barriers to daily adherence. Future implementation efforts should consider how long-acting PrEP could complement existing service delivery models. From a causal inference perspective, the Directed Acyclic Graph (DAG) developed for this analysis conceptualized clinic site as the primary exposure influencing adherence through mechanisms such as counseling intensity and peer support quality. By adjusting only for key confounders—age, consistent condom use, and duration of sex work—this approach strengthened causal interpretability while avoiding over-adjustment for mediators. Within this framework, attendance at Clinic B and Clinic C remained associated with higher adherence. Programmatic implications The high adherence and retention rates observed suggest that core components of the ICAP-supported model of PrEP delivery, including dedicated adherence counseling, structured pill tracking, and continuous peer support, may contribute to improving adherence in similar settings. The variability in adherence observed across sites further suggests that programmatic factors such as peer educator engagement, quality of adherence counseling, and staff experience play a critical role in supporting PrEP adherence. Strengthening these components through standardized training, structured adherence support, and peer-led interventions may improve adherence outcomes. High-performing sites could serve as models for replication in other regions. Replicating and scaling up these programmatic elements within national HIV prevention programs could enhance the effectiveness of PrEP delivery among key populations in the DRC and similar settings across sub-Saharan Africa. Strengths and limitations This study has several limitations. Participants were not randomly selected, which may limit the generalizability of the findings to all FSWs in the DRC. In addition, eligibility criteria requiring at least three months of PrEP use introduced a prevalent user (survivor) bias, likely overestimating adherence and retention and limiting generalizability to newly initiated users. Adherence measures were subject to measurement error, and biological assays of antiretroviral drug levels in blood or urine, the gold standard for adherence assessment, were not available. In addition, the relatively small sample size may have limited the ability to detect statistically significant associations, particularly in multivariable analyses. Furthermore, a ceiling effect at month 6, with adherence exceeding 95%, resulted in very few non-adherence events, limiting statistical power and preventing reliable multivariable analysis at this time point. Residual confounding from unmeasured variables, such as social support or mental health status, cannot be excluded. Although no comparison group was included, this was consistent with the study objective, which focused on adherence among PrEP users rather than evaluating intervention effectiveness. The findings may not fully reflect the diversity of social and cultural contexts of female sex workers across different regions of the DRC, particularly outside urban settings such as Kinshasa. Nevertheless, the longitudinal design represents a key strength, and this study constitutes one of the first prospective cohort analyses of PrEP adherence among FSWs in the DRC. While this study provides quantitative estimates of adherence and retention, further qualitative research is needed to better understand the underlying reasons for non-adherence and attrition over time. Future research should focus on integrating qualitative and mixed-methods approaches to better understand the contextual, behavioral, and structural determinants of PrEP adherence among female sex workers. Longitudinal studies incorporating biological measures of drug exposure are needed to improve adherence assessment. In addition, implementation research evaluating peer-led interventions, enhanced adherence counseling, and differentiated service delivery models across diverse settings would help refine strategies to improve PrEP adherence and retention. Conclusion PrEP adherence among FSWs in Kinshasa was high and improved over time when measured objectively. Differences observed across clinic sites suggest that contextual and programmatic factors may influence adherence outcomes. However, as specific service delivery characteristics were not formally measured, these findings should be interpreted with caution. These results highlight the potential importance of site-level implementation factors in shaping adherence, while underscoring the need for further research to better understand the mechanisms underlying these differences. Despite these limitations, the high adherence levels observed in this ICAP-supported program are encouraging and comparable to findings from other sub-Saharan African settings. These results provide valuable insights for the optimization and scale-up of PrEP delivery among key populations in the DRC. Future interventions should focus on strengthening adherence support strategies, including peer engagement and structured follow-up, while ensuring that programmatic adaptations are informed by context-specific evidence. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgements The authors wish to thank Matthieu Sikando, Axel Okolo, and Armel Demule for their contribution to data collection, as well as all participants in this study. Abbreviations Democratic Republic of the Congo Pre-exposure prophylaxis Female sex workers Directed acyclic graph Men who have sex with men People who inject drugs International center for AIDS care and treatment programs Open data kit Author contributions BB developed the study proposal and was involved in data collection, data entry, data analysis, and manuscript writing. PB and AT were involved in proposal development and reviewed all versions of the manuscript. WM was involved in manuscript preparation and editing of all versions of the manuscript. All authors read and approved the final manuscript. Funding This work was supported by the Fogarty International Center of the National Institutes of Health under award number NIH/FIC D43TW009340. The funding did not include specific provisions for article processing charges. The content is solely the responsibility of the authors. Data availability The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate The study was approved by the Ethics Committee of the University of Kinshasa School of Public Health (ESP/CE/036/2023) and Tulane University (ref. 2023 − 1828). Written informed consent was obtained from all participants. Following consent, selected participants were invited to take part in an in-depth individual interview using a semi-structured interview guide. To preserve participant anonymity, all informants are referred to using pseudonyms. All study procedures were conducted in accordance with the ethical standards of the institutional research committee and the principles of the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.