A Clinical Prediction Rule for Protease Inhibitor Resistance in Patients Failing Second-Line Antiretroviral Therapy

BACKGROUND: Most adults with virological failure on second-line antiretroviral therapy (ART) in resource-limited settings have no major protease inhibitor (PI) resistance mutations. Therefore, empiric switches to third-line ART would waste resources. Genotypic antiretroviral resistance testing (GART) is expensive and has limited availability. A clinical prediction rule (CPR) for PI resistance could rationalize access to GART. SETTING: A private sector ART cohort, South Africa. METHODS: We identified adults with virologic failure on ritonavirboosted lopinavir/atazanavir-based ART and GART. We constructed a multivariate logistic regression model including age, sex, PI duration, short-term adherence (using pharmacy claims), concomitant CYP3A4-inducing drugs, and viral load at time of GART. We selected variables for the CPR using a stepwise approach and internally validated the model by bootstrapping. RESULTS: 148/339 (44%) patients had PI resistance (defined as ≥ 1 major resistance mutation to current PI). The median age was 42 years (interquartile range 36–48), 212 (63%) were females, 308 (91%) were on lopinavir/ritonavir, and median PI duration was 2.6 years (interquartile range 1.6–4.7). Variables associated with PI resistance and included in the CPR were age {adjusted odds ratio (aOR) 1.96 (95% confidence interval [CI]: 1.42 to 2.70) for 10-year increase}, PI duration (aOR 1.14 [95% CI: 1.03 to 1.26] per year), and adherence (aOR 1.22 [95% CI: 1.12 to 1.33] per 10% increase). The CPR model had a c-statistic of 0.738 (95% CI: 0.686 to 0.791). CONCLUSIONS: Older patients with high adherence and prolonged PI exposure are most likely to benefit from GART to guide selection of a third-line ART regimen. Our CPR to select patients for GART requires external validation before implementation.