Predictors of adherence among antiretroviral therapy naive patients on first-line regimen at Themba Lethu Clinic inJohannesburg: results from a prospective cohort study
Mbengue, Mouhamed Abdou Salam
Introduction Viral load is the most reliable indicator of poor adherence to anti-retroviral therapy (ART). However, this assay is difficult to implement in resource-limited settings due to financial and technical constraints. Laboratory markers, combined with the patient’s demographic and clinical details, have been described as better proxies of adherence than the current self-reported adherence measures. However, the real diagnostic value of these biomarkers remains unknown. Therefore, the aim of this study was to assess the usefulness of a composite marker to identify poor adherence to ART defined as a detectable plasma viral load in HIV-positive patients on first-line regimen at Themba Lethu Clinic (TLC) in Johannesburg, South Africa Materials and Methods: This study was retrospective cohort analysis of data collected on HIV-positive ART naïve adults initiating first line antiretroviral regimen at TLC following the 2010 South African antiretroviral treatment guidelines. The data collection was carried out as part of the low-cost monitoring (LCM) study at Themba Lethu Clinic in Johannesburg from February 2012 to 2014. The LCM cohort which aims to look at low cost monitoring of HIV treatment in resource limited settings was initiated in 2009 in Johannesburg, South Africa. The study or treatment outcome was failure to suppress viral load (VL ≥ 400 copies/ml) at 6 and at 12 months. Adherence to antiretroviral treatment was assessed using four (4) self-reported adherence (SRA) measures namely: a self-reporting questionnaire, a Visual Analogue Scale (VAS), a pill identification test (PIT) and the Simplified Medication Adherence Questionnaire (SMAQ). The result of each self-reported measure was classified as either positive or negative given a conventional threshold. In our study three (3) self-reported adherence (SRA) measures were combined into a multi-method approach tool which included self-reports combined with VAS and the pill identification test (PIT). Continuous variables were summarized by median with interquartile range. Categorical variables were summarized by giving their frequencies. To compare continuous variables, we used an unpaired t-test if the variable was normally distributed. When continuous variables were compared from baseline to the previous 6 months, a paired t-test was done. In the case of skewed distribution, we used a non-parametric variant of the t-test such as the Mann-Whitney U-test. To compare categorical variables, we used cross-tables with corresponding chi-square test or Fisher exact test. A Modified Poisson Generalized Linear Model (GLM) with robust variance was used to estimate adjusted relative risks (aRR) of failing to suppress viral load at 6 and at 12 months adjusting for age age, gender, self-reported adherence measures, changes in laboratory markers and missed appointments at 6 and 12 months after ART initiation. As there was missing values in the covariatess and the outcome, we performed a multiple imputation technique under missing at random (MAR) assumption in order to compare the robustness of the estimations between the complete case analysis and the imputation model under MAR after imputing missing values. with the imputed dataset. Additionally, we calculated the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for each self-reported adherence measure using viral load as the reference standard. Thus, we derived two diagnostic risk scores from rounding and adding together the adjusted regression coefficients used to estimate adjusted relative risk and following the Spiegelhalter and Knill-Jones approach, at 6 and at 12 months. The Receiver Operating characteristic (ROC) curves were computed to see the overall discriminative value of each continuous risk score. To assess the clinical usefulness of the continuous riskscores we dichotomized them from 2 ≥ vs < 1 to 5 ≥ vs < 5 and calculated the sensitivity (Se), specificity (Sp), positive predictive value (PPV) and negative predictive value (NPV) at each cut-off, taking detectable viral load as a gold standard. Results: There were 353 HIV-positive patients initiated on first line ART at TLC for the LCM cohort study. Of these, 80.7% did not suppress viral load after 6 months while 30.1% did not suppress viral load at 12 months. The proportion of patients classified as being highly adherent was 86.7% but this proportion decreased to 60% at 12 months. By 6 months, after adjusting for gender and age, the variables that were significantly associated with detectable viral load included: having missed at least two ARV visits by ≥ 7 days (aRR: 2.35 95% CI: 1.08 -5.11); platelet count < 150 cells/mm3 (aRR: 2.73 95% CI: 1.04 -7.18) and VAS ≤ 95% (aRR: 1.65. 95% CI: 1.01-2.71). At 12 months, the estimates showed a positive relationship only with age group and unemployment. There were no similarities in the results found using complete case analysis and analysis with imputed datasets. However, the largest standard errors were obtained from the complete case analysis. At 6 months, the AUC ROC curve was calculated as 0.63 (95% CI, 0.53 - 0.72) while, for the visual analogue scale, the AUC decreased to 0.55 (95% CI, 0.49 - 0.62); for the Simplified Medication Adherence Questionnaire (SMAQ), the AUC decreased to 0.52 (95%CI, 0.45 - 0.60), while for the multi-method approach, it decreased to 0.53 (95% CI, 0.46 - 0.58). The optimal diagnostic accuracy was obtained with the score 5 (≥5 vs <5 Se: 64% and a Sp: 50.0%) followed by a risk score of 4 (Se of 76.0%, Sp of 34.7%). At 12 months, the AUC of the diagnostic risk score was calculated as 0.44 (95%CI, 0.40 - 0.60) while for the three self-reported adherence methods, it decreased to 0.48 (95% CI, 0.40 - 0.60), 0.51 (95%CI, 0.40 - 0.60) and 0.50 (95%CI, 0.41 - 0.59) respectively for the visual analogue scale, the SMAQ and the multi-method approach method respectively. Conclusion. This study shows that after ART initiation, the 6-month’s adherence can be better diagnosed using laboratory markers combined with patient’s information and traditional self-reported adherence measures at Themba Lethu Clinic. The advantage of this proposed method is that it is based on routine and accessible informations collected during HIV-positive patient visits, thus incurring no additional cost for its implementation. An external validation of this diagnostic risk score is needed for its translation into clinical practice in resource-limited settings.
A dissertation submitted to the Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, in partial fulfilment of the requirements for the Degree of Master of Science in Epidemiology and Biostatistics. Johannesburg, November 2017.