Clinical Updates

An analysis of the relative contributions of specific drugs to virological response

Virco presented research at the 2009 ICAAC meeting about the relative contributions of specific drugs to virological response. The statistical evaluation allowed construction of a weighted susceptibility score based on drug class and, for protease inhibitors, on specific drug potency that showed improvements over current methods in predicting response to treatment.

Resistance testing has increasingly become standard-of-care in the clinical management of patients with HIV; at the same time, systems for interpretation of raw data from HIV resistance tests have become more reliable. However, despite improvements in assessing viral resistance, the need to construct potent treatment regimens may still involve value judgments concerning the different intrinsic potency of various antiretroviral drugs. Bart Winters and colleagues from Virco presented research at the 2009 ICAAC meeting about the relative contributions of specific drugs to virological response. The statistical evaluation allowed construction of a weighted susceptibility score based on drug class and, for protease inhibitors, on specific drug potency that showed improvements over current methods in predicting response to treatment.

The analysis started with phenotypic resistance calculations made using the virco®TYPE HIV-1 system. Statistical techniques, including double robust estimation technique were then used to model the contribution of individual agents to virological response across a dataset comprising over 7000 episodes of changes in treatment where baseline resistance and virological response data at 8 and, in many cases, at 24 weeks were available. The result was a hybrid of drug class weights- for NNRTIs, 1.0; for NRTIs, 0.3 for one and 1.1 for two or more fully active agents - and a combined class and individual drug weight for PIs of 1.1 multiplied by a drug-specific factor set at 1 for lopinavir and adjusted up or down based on the relative impact of the particular PI on virological response in the clinical dataset. Thus PIs had drug-specific factors ranging from 0.54 for nelfinavir to 1.57 for boosted darunavir at 600 mg twice-daily. The sensitivity score for any individual drug in a regimen was then calculated by multiplying its fixed drug weight by the amount of residual activity predicted by the virco®TYPE HIV-1 analysis against a particular virus.

Winters et al compared results based on the new method (wPSS) to the standard virco®TYPE report (unweighted) PSS, the ANRS GSS, HIVdb GSS and Rega GSS (weighted GSS) scores. Diagnostic accuracy (area under the ROC curve) on unseen validation data for all regimens ranged from 76-78% for other methods and was 82% for wPSS. For salvage regimens, accuracy declined for other methods to 62% to 72%; however accuracy was maintained at 79% for wPSS. Superior accuracy was also maintained across analyses of week-8 or week-24 virological response and in various subgroup analyses.

If you are interested in viewing the complete poster, please click here. 

This exploratory analysis demonstrated the potential for increased predictive accuracy for sensitivity scores incorporating both accurate assessments of resistance, such as those provided by virco®TYPE HIV-1, and weighting based on data driven analyses of intrinsic drug potency. Additional studies will no doubt be needed to assess the potential to incorporate weighted scores into future clinical use.