Thesis (M.Sc. in Epidemiology)--University of the Witwatersrand, Faculty of Health Sciences,
2014. / Introduction: The prevalence of the Human Immunodeficiency Virus (HIV) in South Africa was 17.8% among 15 to 49 year olds in 2010. Antiretroviral therapy (ART) has thus played a crucial role in mitigating the impact of the HIV epidemic. Themba Lethu Clinic is one of the largest single clinics providing ART in South Africa. One of the challenges of ART provision is ensuring adherence to taking the medication. To date there has been no clear consensus on the ideal way to measure adherence in resource limited settings (RLS). Viral load is perhaps the best and most reliable indicator of poor adherence but is expensive and not easily accessible or available in many RLS. Surrogate markers such as mean cell volume (MCV), CD4 cell count, self-reported adherence and missed visits have been shown to be useful to measure adherence but their reliability remains unclear. The aim of the study was to identify other markers that can be used to measure adherence using viral load as the gold standard.
Materials and methods: The study was a retrospective analysis of HIV-positive ART-naïve
adults (≥ 18 years) initiating standard first-line ART at the Themba Lethu Clinic in Johannesburg, South Africa between April 2004 and January 2012. The association between the last self-reported adherence, change in MCV calculated from baseline to 6 months, change in CD4 count calculated from baseline to 6 months (≥ or < the expected increase of 50 cells/mm3 at 6 months) and missed visits (defined as a scheduled appointment that had been missed by ≥ 7 days but not by more than 3 months) and poor adherence (defined as a viral load ≥ 400copies/ml after 6 months on ART) was tested using Poisson regression models with robust error variance to estimate incidence rate ratio (IRR) and 95% confidence interval (CI). The IRR was used to approximate the relative risk (RR) of poor adherence. Interacting variables were stratified by each other, to create a new variable. The diagnostic accuracy of each identified marker of adherence was also tested using sensitivity, specificity, positive predictive values and negative predictive values.
Results: 7160 patients were eligible for the study and of these 63.2% were female. The median age was 36.7 years. The median CD4 count was 101 cells/mm3 at baseline and 18.9% of the patients had poor adherence at 6 months. Variables associated with poor adherence at 6 months were change in CD4 count stratified by change in MCV at 6 months (change in CD4 count ≥ expected and change in MCV ≥ 14.5fL; adjusted relative risk (aRR) 1, change in CD4 count ≥ expected and change in MCV < 14.5fL; aRR 3.11 95% CI 2.41 – 4.02, change in CD4 < expected and change in MCV ≥ 14.5fL; aRR 1.23 95% CI 0.76 – 2.00 and change
in CD4 count < expected and change in MCV < 14.5fL; aRR 6.98 95% CI 5.35 – 9.09), CD4 count at baseline (> 200 cells/mm3; aRR 1, 101 – 200 cells/mm3; aRR 1.05 95% CI 0.80 – 1.38, 51 – 100 cells/mm3; aRR 1.08 95% CI 0.80 – 1.47 and ≤ 50cells/mm3; aRR 1.34 95% CI 1.02 – 1.76) , WHO stage at baseline (stage I; aRR 1, stage II; aRR 1.16 95% CI 0.90 – 1.48, stage III; aRR 1.27 95% CI 1.04 – 1.55 and stage IV; aRR 1.44 95% CI 1.12 – 1.84) and MCV at baseline (< 80fL; aRR 1, 80 – 100fL; aRR 1.33 95% CI 1.01 – 1.75 and > 100fL aRR 0.98 95% CI 0.62 – 1.55). Sensitivity and specificity of the change in CD4 stratified by change in MCV at 6 months to predict poor adherence were 86.5% and 37.3% respectively for all eligible patients. For patients on AZT-based regimens the variables associated with poor adherence at 6 months were change in CD4 count at 6 months (≥ expected; aRR 1 and < expected; aRR 7.66 95% CI 0.98 – 59.91) and pregnancy during the first 6 months on ART (Never pregnant; aRR 1 and pregnant during follow up; aRR 9.11 95% CI 2.17 – 38.25). Sensitivity and specificity of the change in CD4 count at 6 months to predict poor adherence were 64.7% and 75.2% respectively for all eligible patients on AZT-based regimens. Sensitivity and specificity of pregnancy during the first 6 months on ART to predict poor adherence were 20% and 97.6% respectively for all eligible patients on AZT-based regimens.
Discussion: Change in CD4 count stratified by change in MCV at 6 months was an expected marker of adherence as CD4 count is expected to rise in adherent patients on ART and since most patients (62.9%) were on d4T or AZT-based regimens. Pregnancy during the first 6 months on ART appeared as a marker of adherence for patients on AZT-based regimens before multiple imputation possibly due to missing data hence results for this variable should be interpreted with caution. Contrary to previous studies, self-reported adherence was not associated with poor adherence at 6 months before multiple imputation. This could have been due to the fact that that > 50% of patients had missing data for this variable. The variable is also vulnerable to recall and reporting bias so even after multiple imputation, the area under the receiver operating characteristic (ROC) curve remained < 0.55. The number of missed medical visits and regimen change were also markers of adherence in a few of the models after multiple imputation and require further investigation. In conclusion, the markers of adherence to ART are change in CD4 count stratified by change in MCV at 6 months and pregnancy during the first 6 months on ART for patients on AZT-based regimens. These could help health workers identify poor adherence in the absence of viral load testing and target patients for adherence interventions to prevent virological failure.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:wits/oai:wiredspace.wits.ac.za:10539/17621 |
| Date | 05 May 2015 |
| Creators | Nnambalirwa, Maria Tegulifa |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
Page generated in 0.0022 seconds