Background: The baseline lopinavir mutation score (number of mutations at positions 10, 20, 24, 46, 53, 54, 63, 71, 82, 84, and 90 in HIV protease) is useful for predicting the virologic response to lopinavir/ritonavir (LPV/r) therapy in ARV treatment-experienced patients. However, initial definition of the LPV mutation score was based on a limited number (112) of isolates and reflected the effect of viral genotype on phenotype rather than on response. Consequently, the effect of mutations that occur at relatively low incidence could not be evaluated. Furthermore, the quantitative contribution of individual mutations within the LPV mutation score to lowered virologic response has not been examined.

Methods: An observational cohort of 792 heavily ARV-experienced patients with available protease genotype received LPV/r as part of their ARV-treatment regimen under the Kaletra Autorisation Temporaire d’Utilisation (ATU) program. Virologic response was defined as achieving a viral load of <400 copies/mL within 12 months of initiating LPV/r. Response as a function of baseline LPV mutation score was analyzed using logistic regression. Response models including and excluding each of the 11 mutations constituting the mutation score were compared.  The effect of PI mutations outside of the mutation score that were present at >1% prevalence was also assessed in a similar manner.

Results: Virologic response was observed in 372 (47%) patients. Baseline LPV mutation score was highly predictive of response (odds ratio [OR] 0.824, p<0.0001). Mutations at positions 10, 20, 54, and 82 were significantly (p<0.05) predictive of lower response within the context of other mutations in the mutation score (OR 0.48-0.68). Although not statistically significant, the OR for mutations at positions 24 and 84 were <0.8, suggesting that these mutations may also contribute to lowered response. Furthermore, mutations at positions 33, 36, 47, and 48, which are outside the LPV mutation score, also had OR <0.8 in the context of the LPV mutation score response model. Parameter estimates obtained from a logistic regression model formed the basis of an exploratory algorithm that assigns weighted values to the above mutations.

Conclusions: In this large observational cohort, virologic response with respect to individual PI mutations in the context of complex genotypes provides insight into the development of quantitative algorithms for predicting response to LPV/r in ARV-experienced patients.