Factors associated with pre-treatment HIV RNA: Application for the use of abacavir and rilpivirine as the first-line regimen for HIV-infected patients in resource-limited settings

Abstract

© 2017 The Author(s). Background: Abacavir and rilpivirine are alternative antiretroviral drugs for treatment-naïve HIV-infected patients. However, both drugs are only recommended for the patients who have pre-treatment HIV RNA <100,000 copies/mL. In resource-limited settings, pre-treatment HIV RNA is not routinely performed and not widely available. The aims of this study are to determine factors associated with pre-treatment HIV RNA <100,000 copies/mL and to construct a model to predict this outcome. Methods: HIV-infected adults enrolled in the TREAT Asia HIV Observational Database were eligible if they had an HIV RNA measurement documented at the time of ART initiation. The dataset was randomly split into a derivation data set (75% of patients) and a validation data set (25%). Factors associated with pre-treatment HIV RNA <100,000 copies/mL were evaluated by logistic regression adjusted for study site. A prediction model and prediction scores were created. Results: A total of 2592 patients were enrolled for the analysis. Median [interquartile range (IQR)] age was 35.8 (29.9-42.5) years; CD4 count was 147 (50-248) cells/mm3; and pre-treatment HIV RNA was 100,000 (34,045-301,075) copies/mL. Factors associated with pre-treatment HIV RNA <100,000 copies/mL were age <30 years [OR 1.40 vs. 41-50 years; 95% confidence interval (CI) 1.10-1.80, p = 0.01], body mass index >30 kg/m2 (OR 2.4 vs. <18.5 kg/m2; 95% CI 1.1-5.1, p = 0.02), anemia (OR 1.70; 95% CI 1.40-2.10, p < 0.01), CD4 count >350 cells/mm3 (OR 3.9 vs. <100 cells/mm3; 95% CI 2.0-4.1, p < 0.01), total lymphocyte count >2000 cells/mm3 (OR 1.7 vs. <1000 cells/mm3; 95% CI 1.3-2.3, p < 0.01), and no prior AIDS-defining illness (OR 1.8; 95% CI 1.5-2.3, p < 0.01). Receiver-operator characteristic (ROC) analysis yielded area under the curve of 0.70 (95% CI 0.67-0.72) among derivation patients and 0.69 (95% CI 0.65-0.74) among validation patients. A cut off score >25 yielded the sensitivity of 46.7%, specificity of 79.1%, positive predictive value of 67.7%, and negative predictive value of 61.2% for prediction of pre-treatment HIV RNA <100,000 copies/mL among derivation patients. Conclusion: A model prediction for pre-treatment HIV RNA <100,000 copies/mL produced an area under the ROC curve of 0.70. A larger sample size for prediction model development as well as for model validation is warranted.