Factors associated with vaccination completion and retention among HIV negative female sex workers enrolled in a simulated vaccine efficacy trial in Kampala, Uganda

We enrolled volunteers into a prospective study designed to mimic the rigors of an HIV vaccine trial and administered a licensed recombinant Hepatitis B vaccine (ENGERIX-B™ GlaxoSmitheKline Biologicals Rixensart, Belgium). We used Hepatitis B vaccine because it shares some features with a potential HIV vaccine: the hepatitis B vaccination schedule is similar to that of proposed HIV vaccines and education messages to volunteers would be similar given that Hepatitis B and HIV are both chronic viral infections that share transmission modes. Hepatitis B was also of potential benefit to the volunteers as they are also at high risk for Hepatitis B infection. We enrolled volunteers between August 2014 and May 2016, and each was scheduled to attend 9 visits during a 12 month follow up period. At months 0, 1 and 6 they were scheduled to receive 1 ml of hepatitis B vaccine by intramuscular injection in the deltoid muscle of the non-dominant arm, and have two reactogenicity assessments: at least 30 min and 3 days after each vaccination. Trained study nurses performed baseline clinical assessments (including medical history and physical examination) before giving the first vaccination dose and administered questionnaires to collect study data. The study pharmacist dispensed Hepatitis B vaccine to study nurses who returned all used vials to the pharmacy. Vaccine accountability logs were used to document and monitor vaccine use. The vaccine was stored in a 2–8⁰ C refrigerator and vaccine safety was ensured by monitoring the cold chain daily and assessing volunteers for adverse events. Other study visits were scheduled at months 3, 9 and 12 at which volunteers received HIV testing and counselling (HTC), contraceptive services, syndromic management of sexually transmitted infections (STIs) and free health care for common illnesses. The SiVET visits had scheduled windows as follows: vaccination visits (±3 days), 3-day post vaccination visits (±1 day), all other visits (±7 days) as shown below.

SiVET study visit schedule

We conducted the study at the Good Health for Women Project (GHWP) clinic of MRC/UVRI and LSHTM Uganda Research Unit, located in a peri-urban community in southern Kampala. Field workers conducted mobilization activities with community peer leaders to identify FSWs from commercial hotspots who were then enrolled at the clinic irrespective of HIV status as has been described by Vandepitte et al. [19]. They attended quarterly follow up visits to receive HIV prevention and treatment services which include HIV testing and counselling (HTC); syndromic management for STIs; contraception and, in addition, HIV care and treatment for HIV positive volunteers.

The study field team used phone calls to remind volunteers about follow up visits and for those who needed help to access the clinic, we used a project vehicle to pick them up. Study volunteers received reimbursement of 4USD for time and transport costs when they attended study visits. Free treatment for common illnesses and/ or referral for those who needed specialized health care were provided as needed and the study nurses always informed them about their next scheduled date, which was documented on their appointment card. Field workers kept track of volunteers’ visits using a visit tracker and volunteer visit calendars. When a volunteer missed one visit, the field workers traced her and encouraged her to come to the clinic for the next scheduled visit.

HIV testing was performed on serum using two or more rapid antibody diagnostic tests administered serially as follows: Determine screening test (Alere Medical Co. Ltd., Chuba, Japan), Statpak rapid confirmatory kit (Chembo Diagnostics Systsem Inc., Medford, NY, USA) and Unigold as tie breaker (Trinity Biotech Plc, Wicklow, Ireland). Pregnancy tests were performed on urine using QuickVue tests strips (Quidel Corporation, San Diego CA, USA).

All data were collected using pre-designed interviewer administered questionnaires.

We selected independent variables based on literature review of multi-dose vaccine studies that assessed vaccination completion and/ or retention, and also included socio-demographic variables as potential confounders. The independent variables were as follows: Socio-demographic characteristics; age, education level and marital status.

Behavioral characteristics: alcohol use in the past month (none, once a week or less and daily); being drunk before sex in the past month (none, sometimes/ most times and always); drug use in the past month (Yes/ No); new sexual partner(s) in the past 3 months (Yes /No); condom use with new sexual partner(s) (Yes/ No); receiving payment for sex in the past 3 months (Yes/ No); giving payment for sex in the past 3 months (Yes/ No). Payment for sex was in form of money, gifts or favors.

Clinical characteristics: reported vaginal discharge syndrome (VDS) in the past 3 months (Yes/ No), reported genital ulcer disease in the past 3 months (Yes/ No).

The study data was double entered in OpenClinica (version 3.1, USA) and analyzed in STATA 14 (StataCorp, College Station, TX, USA). We summarized volunteer socio demographic and behavioral characteristics at baseline by counts and percentages. We estimated the proportion of volunteers that completed the entire vaccination regimen (i.e. all three vaccination visits) as number completing all vaccinations divided by the total number of volunteers enrolled. We further estimated the proportion of volunteers that were retained in the study as the number of volunteers that completed all three vaccinations and also attended month 12 visit divided by the total number that were expected to attend (those who received all 3 vaccination doses and were not censored between month 6 and month 12. The two proportions were stratified by the volunteer baseline characteristics. Data for each volunteer was divided into observation time points corresponding to scheduled clinic visits. For each outcome i.e. vaccination completion and retention, we fitted bivariable Poisson regression models with time-varying covariates to allow for intra-individual correlation (because individuals had multiple records) by using robust standard errors. After bivariable analyses, two multivariable models were fitted (i) for retention and (ii) for vaccination completion. Only factors for which the association attained a statistical significance at the 20% level using a likelihood ratio test in a bivariate analysis were considered for the multivariable model. In the multivariable model, factors were removed from the model using a backward elimination algorithm if removing the term did not make the fit of the model significantly worse at the 5% level on a likelihood ratio test.

We screened 381 volunteers for the study of whom 304 (79.8%) were eligible and 290 (76.1%) enrolled. The common reasons for not enrolling if eligible were returning late for the enrolment visit (8) and volunteers declining enrolment because they thought blood draws were too frequent (4). The median age of study volunteers was 27 years (range 18–56 years), 94.5% had attained primary level education or higher and 70.3% were either separated or widowed. New sexual partners in the past 3 months were reported by 94.5% of volunteers of whom 79.9% reported always using condoms with new partners. All except one reported receiving money, gifts or favors in exchange for sex while 16.2% also reported giving money, gifts or favors in exchange for sex in the past 3 months. Vaginal discharge syndrome (VDS) and genital ulcer disease (GUD) were reported by 29.7 and 18.6% respectively. Table 1 shows volunteer baseline characteristics stratified by study outcomes.

Of the 290 enrolled women, 239 (82.4%) completed their 3-dose vaccine schedule, and 212 (73.1%) of these also attended the 12-month study visit. Nine women who did not complete their vaccination schedule also attended their 12-month study visit; these were not included in our final estimate of retention. In total, 230 reached a study end-point as follows: 7 became HIV infected, 11 became pregnant and 212 completed both the 3-dose vaccination schedule and the last visit at month 12 giving a retention of 77.9% (212/272).

At adjusted analysis, volunteers who dropped out by 12 months were more likely to report GUD symptoms in the past 3 months (IRR 1.90; 95% CI 1.09–3.32) and be < 35 years (IRR 6.59; 95% CI 2.11–20.57).

At adjusted analysis, missing any vaccination visit was associated with being < 35 years (IRR 13.10; 95% CI 1.89–90.92), reporting GUD symptoms in the past 3 months (IRR 3.02; 95% CI 1.71–5.33) and reporting consistent condom use with new sexual partners (IRR 2.57; 95% CI 1.10–6.07). We did not find education level, marital status and substance use (alcohol and illicit drugs) to be associated with either vaccination completion or retention. Details are in Table 2.

Studies of hepatitis B vaccination among FSWs and minority populations in South America have found that less than one third of volunteers complete the vaccination schedule [25, 26]. In our study women aged < 35 years were less likely to complete the vaccination schedule or be retained with the risk of non-retention being higher among the youngest age group (18–24 years). Similarly, in a study done among at-risk youth who were randomized to a standard 0, 1 and 6 months’ hepatitis B vaccination schedule versus an accelerated 0, 1 and 2 months’ schedule, younger volunteers were less likely to complete a vaccination schedule [27]. These findings are corroborated by work done in the HIV vaccine field including: HIV vaccine preparedness cohorts, and clinical trials of both vaginal microbicides and candidate HIV vaccines [28‐30] and yet younger age has been associated with increased risk of HIV acquisition [31, 32]. In the broader context of HIV, studies of the “test and treat” intervention among HIV positive women from the same cohort as the SiVET study group [33] and elsewhere [34, 35] have found that younger people are less likely to initiate prompt anti-retroviral therapy (ART). Younger women from high risk communities are vulnerable and experience a greater degree of unequal power relations, gender inequities, exploitation, limited social support and intimate partner violence when compared to older women [36, 37] who also make the working environment uncomfortable for the younger ones by patronizing, bullying and exploiting them [36, 38]. This may affect their ability to engage with health services [39]. FSWs also work through pimps and middlemen, a group of stakeholders who usually control and make decisions for them, limit their independence [40, 41] and subsequently freedom to make the required study visits. The stakeholders are usually influential community members or those who hold power and vary by community. They include but are not limited to older FSWs, middlemen and community leaders, and have an important role in empowering younger women. Empowering young FSWs to participate more in HIV prevention programs as peer leaders and decision makers has been identified as a need [42] and may improve retention and vaccination completion in future HIV vaccine efficacy trials.

Volunteers who reported GUD in the past 3 months were less likely to complete the vaccination schedule or be retained. Studies among high risk women attending primary health care facilities have not found an association between STIs and vaccination completion [43], however elsewhere, STIs have been associated with Hepatitis B vaccination completion among MSM [44, 45]. Risk behavior such as injection drug use and multiple sexual partners have been associated with low completion of multi-dose vaccine schedules [25, 45]. GUD is a surrogate marker of risky sexual behavior yet low risk perception may lead to complacency with disease prevention and health promotion interventions; GUD has long been understood as a risk factor for HIV acquisition as well [46, 47]. Population based surveys done in different settings and age groups give mixed results on the association between risky sexual behavior and uptake of health interventions [48, 49]. Volunteers who report genital symptoms such as GUD need close follow up to ensure high retention in future trials.

Our finding that women who used condoms consistently with new sexual partners were less likely to complete the vaccination schedule, though unexpected, is similar to findings from two FSW cohort studies in Kenya and China that assessed attrition among HIV negative volunteers [50, 51]. On the contrary, Shokoohi et al. have reported inconsistent condom use among HIV negative women to be associated with a lower likelihood of recent HIV testing, a finding that is consistent with high risk behavior [52]. We believe that women may have felt protected by consistent condom use with new partners and were therefore less keen to attend study vaccination visits. Hepatitis B vaccine being used to simulate an HIV vaccine product may not have been perceived beneficial for HIV prevention like condoms which actually protect against HIV infection if used consistently. A survey of acceptability of future HIV vaccines has shown that acceptability increases if perceived benefits of the vaccine are higher [53]. Volunteers who report consistent condom use may miss vaccination visits in future HIV vaccine efficacy trials but also those who adhere to vaccination visits may engage in risk behavior such as inconsistent condom use or multiple sexual partnerships (Risk Compensation) as has been reported elsewhere [54, 55]. Volunteers who report consistent condom use need on-going education and counselling to enable vaccination completion in future HIV vaccine trials.