Low bone mass in people living with HIV on long-term anti-retroviral therapy in Uganda

This was data collected for a cross-sectional study titled Low bone mass in people living with HIV (PLWH) on long-term anti-retroviral therapy in Uganda that study set out to determine the prevalence of low bone mass following long-term exposure to antiretroviral therapy in Ugandan people living with HIV.This was a cross-sectional study conducted over an 18-month period in a cohort of people living with HIV (PLWH) receiving treatment who had been on antiretroviral therapy (ART) for more than 10 years. One hundred ninety nine adult participants were randomly selected. The sample size was calculated using the www.openepi.com online proportions sample size calculator for: an 80.4% prevalence of low BMD among HIV positive persons on ART population size of 1000 HIV positive clients on ART, confidence limits of 5% and design effect of 1.3 arising from the multiple medications for each client at any one time to give a calculated sample size of 190 respondents for a power of 90. To this was included a 5% allowance for loss and omissions, binging the final sample size to 200 respondents.All participants were subjected to lumbar spine, left total hip and left femoral neck body scans, to assess body mineral content and bone mineral density (BMD), using dual energy X-ray absorptiometry (DXA) (Hologic Discovery Wi Apex 3.1, Hologic Bedford Inc., Bedford, MA, USA) using standard protocols. Bone mineral density was expressed in grams of mineral per square centimetre. The reference population for this scanning was age, sex, race and BMI matched from the National Health and Nutrition Examination Survey (NHANES) cohort. Bone mass was categorized using the official position of the International Society of Clinical Densitometry (ISCD) since most participants were below 50 years of age. The ISCD official position recommends the use of the Z-score for pre-menopausal females, males younger than 50 years and children (1). A Z-score of -2.0 or lower was defined as having low bone mass for age and a Z-score above -2.0 was defined as bone mass within the expected range for age. The World Health Organization (WHO) was used for post-menopausal females and males above 50 years of age. A T-score that was within one standard deviation (SD) of the reference BMD was classified as normal, 1 to 2.5 SD below the reference BMD as osteopenia, and greater than 2.5 SD below the reference BMD as osteoporosis. For this study all participants with low bone mass or osteopenia/osteoporosis were categorized as having “low bone mass” while the rest were categorized as having normal bone mass (BM). This was done for ease of data analysis. The following data was also collected at enrolment: age, sex, height and weight for use in calculating Body Mass Index (BMI), the type of ART (categorized as Nucleoside reverse transcriptase inhibitors, Non-nucleoside reverse transcriptase inhibitors, protease inhibitors, and Integrase inhibitors), duration on specific antiretroviral agents, behavioural risk information including alcohol and tobacco use, and history of previous fractures of the wrist, hip or spine, using a study questionnaire and data abstraction tool.Data were collected initially collected using case report forms and then entered into DataFax forms that were specially designed for this study. This data management system is designated to manage paper data forms. The forms are faxed to the DataFax server where they are read using intelligent characters-recognition and then added to the study database. The latest viral load results and nadir CD4 count were retrieved from the Integrated Clinic Enterprise Application (ICEA) database, an in-house software that provides good quality data collection, with minimal missing or incorrect information (2). These validated records were then exported as excel files for transformation so that each row represented an individual drug exposure prior to further analysis in the R-statistical computing environment (3). The final long dataset comprised of 1384 anti-retroviral drug exposures. The analysis was informed by the study hypothesis that: <i>exposure (time on treatment and type of drugs) to ART was associated with differences in BM among respondents after controlling for age, alcohol consumption, sex, smoking, most recent viral load, CD4 cell count, line of treatment (switch) and body mass index.</i> During analysis descriptive statistics were generated and summarised as frequencies and correlations in tables. Inferential statistics were generated with multilevel binomial longitudinal Markov chain Monte Carlo (MCMC) mixed multivariate regression modelling using the <i>rstanarm</i> package (4-6). The diagnostics for each model, one for the spine and other for hip BM, were used to identify any poorly performing values and assess the sampling quality. The absence of a value of “zero” in the 95% confidence interval for the coefficients was used to identify statistically significant output values. The mean coefficients were summarised in tables but exponentiated to obtain odds ratios. All records with missing data were removed from the database prior to the inferential MCMC part of the analysis. Ethics approval was obtained from a Research Ethics Committee and the Uganda National Council of Science and Technology. All participants provided informed consent to participate in the study.