Abstract Body

Background: Although virological failure is common in adults on second-line ART in resource limited settings, major protease inhibitor (PI) resistance mutations occur infrequently. Therefore, empiric switches to salvage therapy would waste resources. Resistance testing identifies patients needing salvage therapy, but is expensive and access is limited. Therefore, a clinical prediction rule (CPR) to rationalise access to resistance testing would be a valuable tool.

Methods: We identified predictors of major PI resistance mutations in a large South African cohort. We included adults with virological failure and ≥ 4 months exposure to lopinavir/atazanavir-based ART. PI resistance was defined as ≥1 major resistance mutations to current PI. We constructed a multivariate logistic regression model including age, sex, PI duration, adherence by pharmacy claims, concomitant CYP3A4-inducing drugs and viral load (VL) at time of genotyping as candidate predictors based on an a priori decision. We selected variables included in the CPR using a stepwise approach. We internally validated by bootstrapping. We categorised included variables, and assigned points to each level based on adjusted beta coefficients.

Results: There was PI resistance in 146/339 (43%) patients. Median age was 42 years (IQR 36 to 47), 211 (62%) were female, and 309 (91%) and 30 (9%) were on ritonavir-boosted lopinavir and atazanavir respectively. Median PI duration was 2.6 years (IQR 1.6 to 4.7). Median adherence was 97% (IQR 73 to 100) and 9% took concomitant inducing drugs. Median log VL was 4.9 (IQR 4.3 to 5.4).

Presence of major PI resistance mutations were associated with age (adjusted odds ratio (aOR) for 10 year increase 1.9 (95% CI 1.4 to 2.6)); PI duration (aOR per year 1.1 (95% CI 1.0 to 1.3)); and adherence (aOR per 10% increase 1.2 (95% CI 1.1 to 1.3)). There was no association with sex, inducer exposure and VL. We included age, PI duration and adherence in the CPR.

Area under the ROC curve was 0.736 (95% CI 0.683 to 0.789) before and 0.739 (95% CI 0.736 to 0.742) after bootstrapping. The model had acceptable calibration. The optimal cut point corresponded to a score of 8/15 (75% sensitivity and 67% specificity).

Conclusions: Older patients with high adherence and prolonged PI exposure are most likely to benefit from resistance testing. The CPR may be a useful tool to rationalise patient selection for resistance testing in resource limited settings, but requires external validation

viewer?TAG_ACTION=SECURE_FILE_DOWNLOAD&V
Table: Clinical prediction rule for major PI resistance mutations