2504. Laboratory Evaluation of HIV-1 Viral Load Pooling with Marker-Assisted Deconvolution

Abstract Background Cost still limits HIV-1 viral load (VL) routine monitoring in resource limited settings (RLS), preventing early detection of virologic failure (VF). Pooled VL testing reduces cost over individual testing (IND). We previously showed in simulation, that additional cost benefits over previously-used pooling deconvolution algorithms can be achieved by using low-cost, routinely-collected clinical markers to determine the order for VL testing in deconvolution (termed marker-assisted minipool plus algorithm; mMPA). This algorithm has not been assessed in-vitro. Methods 150 samples from 99 Ghanaian adults with HIV on first-line therapy (VF 17%; CD4-VL correlation −0.35) were used to construct 30, 5-sample pools: n = 10 with 0, n = 5 with 1, and n = 15 with 2 individuals with VF. VL testing was with Abbott M2000. Accuracy, number of tests and rounds of testing to deconvolute pools were estimated using four strategies: (1) IND; (2) Minipooling (MP); (3) Minipooling with algorithm (MPA); and (4) mMPA. Results Compared with IND, MP and MPA, mMPA reduced total number of tests per pool needed to ascertain VF: MP average 4.3 (95% confidence interval (CI) 3.5–5.2, p> 0.05), MPA 3.0 (95% CI 2.4–3.5, P < 0.001), and mMPA 2.5 (CI 2.0–3.0, P < 0.001). Compared with MP and MPA, mMPA further reduced VL tests by 42% (1.9 tests/pool, CI 1.3–2.4, P < 0.001) and 17% (0.5, CI 0.2–0.8, p = 0.004); and required fewer testing rounds than MPA by 17% (P < 0.01), thus producing results quicker. IND and MP had 100% sensitivity and specificity. MPA and mMPA had similar sensitivity of 96.1% (MPA CI 90.7–100%; mMPA CI 88.0–100.0%) and specificity of 99.5% and 99.2% (98.5–100.0% for MPA and 97.5–100.0% for mMPA). Specifically, 3/150 samples were misclassified with MPA and mMPA: one suppression as VF, and two VF as suppressed. Conclusion Laboratory evaluation confirms that deconvolution using mMPA with CD4 or other routinely-collected clinical information as low-cost biomarkers reduces the number of VL assays required to identify VF in a setting with a low prevalence of VF. Implementation of pooled VL testing using mMPA for deconvolution may increase the availability of VL monitoring in RLS. Work is ongoing to reduce complexity and misclassification, required prior to widespread implementation. Disclosures All authors: No reported disclosures.