Background: Extensive NRTI resistance, common at the time of switch to second-line therapy in ART programme settings, is expected to affect outcomes. We examine the impact of baseline NRTI resistance on responses to PI+NRTIs in the EARNEST trial.
Methods: 1277 patients aged ≥12y who met WHO-defined treatment failure criteria after >12 months on NNRTI-based first-line ART in African rollout programmes were randomised to receive bPI (standardised to lopinavir/ritonavir 400mg/100mg bd) with either 2/3 NRTIs selected by clinician based on algorithms without resistance testing (PI/NRTI); with raltegravir (RAL 400mg bd)(PI/RAL), or as monotherapy (+RAL induction for first 12 weeks)(PI-mono). PI-mono was stopped after 96 weeks; other groups continued randomised treatment to week 144. Drug resistance at baseline (done in 391/426 in PI/NRTI) and VL during study were tested retrospectively on stored samples (results blinded during trial).
Results: Patients had advanced treatment failure (42% VL≥100,000 c/ml, 62% CD4<100 cells/mm3) at baseline. In PI/NRTI, 80% received TDF+3TC/FTC (±ZDV). Based on resistance testing, the PI/NRTI regimen contained 0 predicted active NRTIs (at most low-level resistance, Stanford criteria) in 230 (59%, PI/NRTI(0)), 1 active NRTI in 128 (33%, PI/NRTI(1)) and ≥2 active NRTIs in 33 (8%, PI/NRTI(2)). VL suppression in PI/NRTI(0) was markedly superior to PI-mono (76% vs 44% respectively <50 c/ml at week 96; P<0.001), and similar to PI/RAL (76% vs 72% <50 c/ml for PI/NRTI(0) and PI/RAL respectively at week 144, P=0.28, Figure). Response in PI/NRTI(1) was identical to PI/NRTI(0) (both 76% < 50 c/ml at week 144; P=0.92) but slightly lower in PI/NRTI(2) (62% < 50 c/ml; P=0.12 vs PI/NRTI(1)).
Conclusions: Even when there is little or no predicted activity due to resistance, NRTIs make a major contribution to efficacy of PI/NRTI second-line therapy with clear added activity over the PI alone, equivalent to adding a drug from a new class. No difference between 0 and 1 active NRTIs suggests this contribution is not due to direct drug activity (possibly represents a viral fitness effect). The paradoxical trend to worse outcome with 2 active NRTIs may reflect a small group of patients with very poor adherence on first-line, continued during second-line. Algorithmic NRTI drug selection and attention to adherence are likely to achieve optimal outcomes in standardised PI/NRTI second-line therapy in resource-limited settings with resistance testing to select NRTIs of little added value.
Figure: VL suppression in PI/NRTI arm by genotype-predicted NRTI activity, compared to supression in PI/RAL and PI-mono arms