Study design
The study was a 12-year open retrospective cohort of cART recipients. The outcome event was DRC. This was recorded as diagnosed by the patient’s physician and listed in the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) [5]. Exposures of interest were cART regimens and patients’ underlying biomedical and demographic characteristics. Survival was calculated as the time elapsed from the date of cART initiation until the date of DRC development. The study was conducted in Gaborone, Botswana’s capital city [15]. HIV period prevalence (2008–2013) was estimated at 19% among persons aged 18 months and above [16].Two clinics, Princess Marina Hospital (PMH) HIV clinic and Bontleng HIV clinic, were chosen as study sites out of 14 sites. The two clinics were selected because of their high capacity to cater for HIV patients and were the best at keeping patient records.
Data collection
Data were collected from patient medical records. The records included patient files, referral and discharge booklets. Only data from patients who met the inclusion criteria were considered. A cohort that included: (i) patients who were receiving the first-line cART, (ii) patients who were receiving the second-line cART and (iii) those on the third-line cART was identified and followed between 2002 and 2015. The follow up endpoint was when DRC occurred or the end of the study. We excluded patients who had DRCs upon entry into the treatment program, pregnant women, recipients who were initiated on cART after 2012 (allowing for at least 3 years of follow-up), recipients aged less than 18 years and recipients with discrepant data from different records within the same clinic to assure data accuracies.
From each patient, we collected data on the date of enrolment into the program, the date of cART initiation, weight in kilograms (weight-1) and height in centimetres when entering the program, weight at the time of data collection (weight-2), CD4 cell count at cART initiation (CD4–1), CD4 cell count at the time of data collection (CD4–2), whether or not the treatment regimen was the same for a patient since the enrolment in the program, cART treatment regimen received before if there was a switch and reason for the switch. Only switches between different cART regimens were considered. Data on whether or not adherence to the treatment had been continued or discontinued and information on DRCs as well as the date of the diagnosis were also collected.
The sample size was estimated as by Hennekens and Buring [17], using a sampling error of 0.05 and a beta level of 0.20. The proportion of baseline DRC among HIV patients recipients of cART was 17.6% [18] and the expected magnitude of association between cART and DRC was set at 1.9 odds ratio. This led to an estimated sample size of 483. However, we included 540 participants in the study.
Data analysis
Data were entered into a computer using Microsoft Excel (Redmond, WA) before being imported into IBM SPSS version 21 (Chicago, IL). Proportions of participants on the first-line cART, the second-line, the third-line and on second-, third-line cART regimens were estimated. Potential associations between the different cART regimens and DRCs were investigated by computing unadjusted hazard ratios (UHR) of each regimen and their 95% confidence intervals (CIs) using univariate Cox regression analysis.
Similarly, proportions of participants with underlying biomedical and demographic characteristics known as potential DRC risk factors were assessed, namely CD4 cell count as a continuous variable, CD4 cell count categories, gender, age as a continuous time dependent variable, age groups (< 35 years old/≥35 years old), body mass index (BMI) categories, whether patients adhered to treatment or not and attended PMH clinic or Bontleng clinic, and whether they switched cART regimens or not. Using DRCs as status /event/dependent variable, time in years, potential associations with these factors were assessed by computing UHR and their 95% CIs in univariate Cox regression analysis. Continuous variables namely CD4 cell count at cART initiation, CD4 cell count after cART initiation and age were not included in this analysis unless they were statistically significant in the two independent samples t-test analysis. Covariates that achieved p ≤ 0.9 in the univariate analysis were re-examined in a multivariate Cox regression model to identify those that significantly (p < 0.05) affected the time to event. Time spent in the study was expressed in years. Survival was estimated as the time elapsed from the date of initiation to cART until the date of the first DRC development or the last attendance. Recipients lost to follow-up, those referred to other facilities were left censored at the last date they were seen. Those who got to the end of the study without developing any type of DRC were right censored at the date of data collection. Adjusted hazard ratios (AHRs) and their 95% CIs were computed. The survival functions were plotted and the probability of survival in years from the date of initiation on: first-line cART (cART line1), second-line cART (cART line2), third-line cART (cART line3) or second-, third-line cART (cART line2/3) to the development of DRCs were compared using Log rank X2. The Cox regression presented herein is the best model where only covariates with P < 0.05 are kept in. The overall statistical significance of the model or how well the model fits the data was tested by computing the likelihood chi-square statistic.
Ethical approval and consent to participate
Ethical approval to collect data from HIV clinics was sought and obtained from the University of Botswana Review Board and the ethics committee of the Ministry of Health and Wellness, Botswana. Permission to consult clinic record books and systems was also sought and obtained from the clinic management. As this was a record based study, no consent to participate was required.