Study design
This was a secondary analysis of two longitudinal cohorts of individuals on long-term ART: the ART cohort of The Infectious Diseases Institute (IDI), College of Health Sciences at Makerere University in Kampala, Uganda [22] and the ART cohort of The AIDS Support Organization, (TASO) in Jinja, Uganda [23]. The two cohorts had a 2-year difference in ART initiation timelines and this limited direct comparison.
Study setting
The IDI urban ART cohort in Kampala
The adult clinic of the Infectious Diseases Institute (IDI) is a Centre of excellence for HIV care and treatment located in the Mulago Teaching Hospital in Kampala-Uganda. IDI is a large outpatient clinic with over 30,000 patients ever enrolled and cares for people living in five Kampala municipalities. ART is provided free of charge, initially through the Global Fund and later mainly through the US President’s Emergency Plan for AIDS Relief (PEPFAR).
Following written informed consent, adults (age ≥ 18 years) starting ART between April 2004 and April 2005 were enrolled and followed up (hereafter referred to as the IDI urban cohort). ART was started in patients with WHO Stage 4 or CD4 count ≤ 200 cells/ml, (according to national treatment guidelines implemented at that time), and consisted of stavudine (weight-adjusted and its use discontinued in 2013 in Uganda), lamivudine and nevirapine (fixed-dose combination) or zidovudine, lamivudine (fixed-dose combination) and efavirenz. A study doctor and adherence counsellor evaluated patients at study enrolment and every 6 months thereafter, while they attended the general clinic for monthly ART medication refills [22]. At each study visit, a physical examination was performed and information was recorded about HIV status of the partner(s), social support, and sexual history in the past month including promotion of condom use, adherence using visual analogue scale, and reasons for non-adherence. ART was monitored every 6 months through viral load (VL) testing. Participants with a VL ≥ 1000 copies/ml were offered enhanced adherence counselling and support including a face-to-face session to discuss the implications of unsuppressed viral load (VL). Adherence strategies used by the patient were reviewed, and an adherence action plan developed. Patients with two consecutive viral loads ≥ 1000 copies/ml were considered eligible for a second-line ART regimen.
The TASO rural ART cohort in Jinja
The AIDS Support Organization is one of the first and largest non-governmental organizations in SSA, providing treatment to over 98,000 patients in Uganda. Since 2004, TASO Jinja, one of 11 TASO care facilities in Uganda, has cared for people living with HIV in the Jinja district within a radius of 75 km and currently provides ART to over 5000 people.
In June 2012, TASO Jinja initiated a prospective longitudinal cohort (Long-term Outcome on ART study) among patients who initiated ART between 2004 and 2007 using the WHO Stage 4 or CD4 count ≤ 200 cells/ml, (according to national treatment guidelines implemented at that time), and consisted of stavudine (weight-adjusted and its use discontinued in 2013 in Uganda). The Long-Term Outcomes on ART study recruited patients receiving first-line ART for a minimum of 4 years (hereafter referred to as the TASO rural cohort) [23, 24]. Between July 1, 2012, and December 31, 2013, all individuals receiving first-line ART for at least 4 years were eligible to participate in the cohort study and were included after written informed consent. Participants continued to receive adherence counselling and condom use from TASO counsellors and peer educators mostly in groups as part of routine care. From August 2014 until January 2015, enrollment was restricted to patients already on ART for at least 4 years with CD4 cell count ≤ 450 cells/mm3 in order to over-sample for participants who may have virologic failure. All participants continued to receive routine comprehensive HIV care from the TASO service providers. Routine VL testing was not available at TASO Jinja at the time of the study but all participants received a VL test at enrollment. Participants with a VL ≥ 1000 copies/ml were offered enhanced adherence counselling and support including a face-to-face session to discuss the implications of unsuppressed VL, assessment of the adherence strategies used by the patient, and development of an adherence action plan. VL was re-assessed after 3 months.
Patients were followed up for additional 3.5 years. Every 6 months, cohort participants completed an interviewer-administered standardized questionnaire (adapted from the HAARP study) [24] to collect behavioral and treatment outcomes. Laboratory monitoring included CD4 counts and VL every 6 months. CD4 counts were performed at the Jinja referral hospital regional laboratory. Plasma samples were stored in liquid nitrogen and shipped to the MRC/UVRI laboratory in Entebbe, Uganda for VL testing and resistance testing.
Data collection and management
The IDI urban cohort data were directly collected into an electronic behavioral medical record by study counsellors who provided risk reduction counselling including condom use promotion and periodically validated by a senior data manager. The TASO rural cohort, data were collected by the research assistants (counsellors) using an interview administered behavioral questionnaire but they did not provide counselling on condom use promotion. Data were double entered using Epi-Info and imported into Access for data management and storage. Laboratory test results were transmitted electronically from the MRC laboratory to the data Centre in TASO Jinja, manually entered in the study database.
For this study, data extracted from TASO and IDI database included socio-demographic information (age, gender, educational level, employment, and marital status), ART start date and regimen, clinical data collected at enrollment, including WHO clinical stage, CD4 cell counts, viral load and sexual behavior (number of partners in past 6 months and consistency of condom use) and adherence data at enrolment and subsequent clinic visits. Risky sexual behavior was defined as sexual intercourse with ≥ 2 partners or sexual intercourse with 1 partner and inconsistent condom use.
Statistical analysis
For this analysis, baseline was defined as the time of enrollment into the cohort which occurred after ≥ 4 year on ART (rural cohort) or at the 4-year on ART visit (urban cohort). For all patients included in the cohort, data was extracted for an additional 3.5 years after this baseline visit. All participants were followed until death, loss to follow up, transfer out, withdrawal of consent, or 3.5 years after the “baseline” visit.
We described participants characteristics at baseline overall, by gender and cohort (urban/IDI and rural/TASO) using means and frequencies. Continuous variables were compared using an unpaired t-test or Wilcoxon rank sum test if not normally distributed. Pearson Chi-square was used to compare categorical variables.
The main exposure of interest was time on ART (coded as 6-month periods). Other covariates of interest were the site, age, CD4 cell count, viral load, gender, education level, marital status, occupation, and ART regimen at baseline and calendar year of ART initiation. In order to examine the effect of long-term ART on sexual behaviors, separate analyses were performed. At baseline, we performed multivariable logistic regression and included variables in the initial model based on prior knowledge or association of the selected covariates with risky sexual behavior in bivariable analysis. Using a backwards elimination procedure, we started with all covariates in the model and then stepwise removed the covariate with the largest p-value until all the remaining covariates had a p-value less than 0.10 for the association with the outcome of interest.
To model the association between the binary outcome (risky sexual behavior) and the exposure of time on ART, we used a Generalized Estimating Equations (GEE) logistic regression model, with a logit link function to model the association between the binary outcome (risky sexual behavior) and the exposure of time on ART. Using a GEE logistic regression, odds ratios (ORs) were derived that took into account the repeated measures in individual participants and missing data in response variables, using robust standard errors to account for within-subject correlation. Co-variates with a p-value less than 0.10 for the association with the outcome of interest was included in the final model. Data were analyzed using STATA version 15 (StataCorp, College Station, TX).
Effect modifiers exploratory analysis
We assessed for effect modification by the following variables: gender and sexual partners, marital status and sexual partners, current age group and sexual partners. Interaction terms with a p-value < 0.05 were considered statistically significant.