Inhibitory quotient as a prognostic factor of response to a salvage antiretroviral therapy containing ritonavir-boosted saquinavir. The CIVSA Study.

BACKGROUND

The addition of a low dose of ritonavir to protease inhibitors (PIs) has become a widespread strategy to improve PI pharmacokinetics. As resistance is a major barrier to long-term suppression, in salvage therapy genotype and/or phenotype scoring is currently used to predict the response. We evaluated the relationship between the saquinavir (SQV) inhibitory quotient (IQ) (virtual and genotypic) and virological response.

METHODS

Eligible patients were on a PI-containing highly active antiretroviral therapy (HAART) regimen excluding SQV and had a viral load >5000 HIV-1 RNA copies/mL. The PI was switched to SQV/ritonavir (RTV) 1000/100 mg twice a day (bid) and the same two backbone nucleoside reverse transcriptase inhibitors (NRTIs) were maintained at least until week 4, when the resistance test results became available. Genotype and virtual phenotype were determined at baseline, while the SQV trough plasma concentration was determined at week 4.

RESULTS

Fifty-three patients were included in the study. Mean baseline viral load and CD4 count were 137,693 copies/mL and 263 cells/microL, respectively, the mean number of previous PIs was 2.3 and the mean number of protease gene mutations (PGMs) was 4.1. Using an on-treatment analysis, at week 16 the mean increase in CD4 count was 70.9 cells/microL, viral load was <200 copies/mL in 17 out of 37 patients (45.9%), and 30 out of 45 patients (66.7%) were considered virological responders (VRs) (viral load <200 copies/mL or viral load declined > or =1 log(10) at week 16). Median virtual phenotype was 1.3 (0.6-6.9). Baseline differences were detected between VR and non-VR populations: the mean numbers of PGMs were 3.2 and 5.8 (P<0.05), the mean numbers of SQV-associated mutations were 2 and 3.8 (P<0.05), and the mean CD4 counts were 365.9 and 184.3 cells/microL (P<0.05), respectively. Mean SQV trough concentrations at week 4 were 1.1 and 1.0 microg/mL (not significant), and mean virtual IQs were 0.7 and 0.1 (P<0.01), respectively. Multivariate analysis showed that baseline PGMs >5 or SQV-associated mutations>5, virtual phenotype, baseline viral load >50,000 copies/mL, and virtual IQ <0.5, but not genotypic IQ, were the variables independently associated with non-VR.

CONCLUSION

In heavily pretreated patients, the use of SQV virtual IQ or alternatively virtual phenotype, as well as PGMs, is a useful tool for the prediction of virological response.