Lamivudine-based maintenance antiretroviral therapies in patients living with HIV-1 with suppressed HIV RNA: derivation of a predictive score for virological failure

Objectives: Two-drug antiretroviral regimens based on lamivudine (3TC) plus either a protease inhibitor (PI) or dolutegravir (DTG) are becoming increasingly popular in switch strategies. Our goal was to derive a predictive score for virological failure (VF). Methods: We retrospectively analysed data for a cohort of 587 virologically suppressed (HIV RNA < 37 HIV-1 RNA copies/mL), adult (≥ 18 years old) patients starting lamivudine plus either a boosted PI or dolutegravir. Predictors of VF (defined as a single HIV RNA measurement ≥ 1000 copies/mL or two consecutive HIV RNA measurements ≥ 50 copies/mL) were identified using a multivariate Cox regression model. A ‘weighted’ score was assigned to each variable associated with VF; the discriminative power of the score obtained was expressed as the area under the receiver-operator characteristic curve (ROC-AUC). Results: During a median 2 years of follow-up time, 35 VFs occurred; predictors of VF were baseline residual HIV RNA between 20 and 36 copies/mL, African ethnicity, ≥ 10 therapeutic lines, the presence of at least one resistance-associated mutation (RAM) for resistance to current drugs (excluding M184V), a non-B viral subtype and a baseline CD4 count < 200 cells/μL. A score of 2 was assigned to non-B viral subtype, 3 to residual viraemia ≥ 20 copies/mL, ≥ 10 previous therapeutic lines and African ethnicity, 4 to baseline CD4 count < 200 cells/μL, and 7 to the presence of at least one RAM (excluding M184V). The ROC-AUC was 0.67 (95% confidence interval 0.57–0.77). Conclusions: The presence of at least one RAM, higher residual viraemia and African ethnicity were among the major predictors of VF in our cohort. Studies with larger sample sizes are warranted to improve the predictive value of the derived score.