Background

Both population and ultra-deep sequencing (UDS) of the HIV-1 V3 loop are useful in selecting candidates for maraviroc (MVC) therapy. We used mathematical modeling to determine that patients whose non-R5 HIV comprises <2% of the viral population by UDS are likely to respond to a MVC-containing regimen. However, the predictive value of absolute amount of non-R5 HIV is unknown.

Objective

To determine whether non-R5 viral load contributes to predicting response to a MVC-containing regimen.

Methods

Patients enrolled in the MERIT study (MVC or efavirenz plus zidovudine/lamivudine in treatment-naïve patients) with R5 virus at screening (by original Trofile assay) and randomized to the twice-daily MVC arm were included. UDS was performed with a 454/Roche GS-FLX instrument. Tropism was predicted using the "geno2pheno" co-receptor algorithm (g2p). A sample was considered R5 if <2% of variants had a score below 3.5 FPR. MVC responses at Week 48 were predicted by descriptive statistics and mathematical modeling.

Results

Samples for 343 patients (308 R5, 35 non-R5) were available. Baseline median CD4 and mean viral load (VL) were 247 and 232 cells/µL and 4.9 and 4.6 log10 c/mL in patients with R5 and non-R5 virus. No CXCR4-using viruses were detected in 249/343 (73%) patients. Among the 94 patients with detectable CXCR4-use, median (q25, q75) percent and absolute levels of CXCR4-using viruses were 0.8% (0.4-8.1) and 2.9 (2.3-3.5) log10 c/mL, respectively. Week 48 virologic responses are shown in Table 1.

Table 1
Full size table

In univariate models, baseline CD4 and percent of CXCR4-using virus were not significant predictors of week 48 response (p=0.12; p=0.26); VL and absolute amount of CXCR4-using virus were significant (p=0.02; p=0.03) and were included in the multivariate model (p=0.02 for both in final model).

Conclusion

In MVC-treated patients in the MERIT study, baseline VL and absolute amount of CXCR4-using virus were predictive of Week 48 response. It is possible that total burden of CXCR4-using virus in drug-naive individuals may play a greater role than the percentage of such virus in predicting response to regimens containing a CCR5 antagonist.