Tropism Screening by Deep Sequencing Relative to Trofile and Population-Based Sequencing
Overall, genotyping identified 1037 samples (57%) as R5 and 790 (43%) as non-R5 using the PSSMx4/R5 algorithm. PSSM and g2p had ~90% concordance with one another. For ease of presentation, results will be shown for the PSSMx4/R5 algorithm. When screened with Trofile, 1141 samples (62%) were designated as R5 and 686 (38%) were designated as non-R5 (Dual-/Mixed-tropic or X4). Global concordance of genotyping with Trofile-defined tropism was 82%. Detailed analyses for g2p as well as the PSSMSI/NSI algorithm were largely similar to analyses for PSSMx4/R5 and are presented in Supplementary data table 1.
Of the 686 samples identified as non-R5 by Trofile, 575 (84%) were also identified as non-R5 by genotyping. An additional 215 samples that were not identified as non-R5 by Trofile were identified as non-R5 by genotyping. Using Trofile as a reference, sensitivity of genotyping was 84%, and specificity was 81%. Using genotyping as a reference, the sensitivity of the original Trofile assay was 73%, and specificity was 89%. Comparing population-based sequencing against deep sequencing, overall concordance was 80%, with 64% sensitivity and 93% specificity. The sensitivity of both Trofile and population-based sequencing was lower when the proportion of non-R5 variants in the viral population was lower according to deep sequencing ().
Figure 1. Sensitivity of population-based V3 sequencing and original Trofile assay with 454 genotyping as the reference. The ability of the population-based V3 sequencing (black bars) and original Trofile (gray bars) to identify screening samples as having non–CCR5-using (more ...)
Of interest, deep sequencing detected at least some non-R5 HIV in >90% of patients (1700 of 1827 patients), regardless of tropism classification. Samples with R5 HIV by Trofile had a median of 0.1% non-R5 variants at screening according to deep-sequencing results. However, non-R5 levels <2% likely have low reproducibility and should be considered with caution. The Trofile non-R5 group excluding dual/mixed samples (ie, only “pure” X4 by Trofile; n = 39) had a median of 93% non-R5 virus.
Patients with R5 virus by Trofile but non-R5 virus by genotyping had 12% non-R5 HIV present at screening (n
= 167), which was much higher than results obtained when both assays indicated R5: 0.1% (n
= 926). This suggests that the original Trofile assay did not reliably detect patients with low-level non-R5 variants present at screening (), consistent with the finding that 8% of patients had non-R5 results at baseline [27
], and consistent with ESTA results for the Maraviroc versus Efavirenz Regimens as Initial Therapy (MERIT) trial of maraviroc [34
Of the 1827 patients screened, 1093 actually entered the maraviroc (n = 851) or placebo (n = 242) arms of the trials. Baseline characteristics of patient groups screened by both methods are presented in . The R5 groups by either method were similar in terms of baseline pVL and CD4+ cell count, as were the non-R5 groups. Among maraviroc recipients, genotyping identified over twice as many patients as being unlikely to respond to maraviroc, compared with Trofile (n = 240 vs 111).
Baseline Characteristics of Treated Population, Stratified by Tropism Status by Genotype and Trofile
Early Virologic Response to Maraviroc
Screening genotype was a predictor of response to maraviroc-based antiretroviral therapy in treatment-experienced patients. Maraviroc recipients found to have R5 HIV by genotyping had consistently better virologic outcomes than did those found to have non-R5 HIV. Using a number of parameters, genotyping performed similarly to or marginally out-performed the original Trofile assay in predicting virologic response. Virologic performance was slightly better in the maraviroc twice-daily arm, compared with the maraviroc once-daily arm (data not shown), but these arms have been pooled to simplify the presentation of the results.
The median week 8 pVL change from baseline was examined to minimize the number of patients who had discontinued the study because of reasons such as treatment failure or loss-to-follow-up but who were receiving the drug for a sufficient time to measure the efficacy of maraviroc. Maraviroc recipients who were found to have R5 virus by genotyping had a combined median week 8 decrease in pVL from baseline of 2.4 log10 copies/mL (interquartile range [IQR], 1.7–2.9 log10 copies/mL; n = 611), which was approximately twice as large as the 1.4 log10 copies/mL decrease (IQR, 0.2–2.7 log10 copies/mL; n = 240) for patients classified as having non-R5 virus by genotyping. Results in which missing patients were censored or where data were restricted to only those screened for MOTIVATE-1 were largely similar (data not shown).
Using Trofile, the corresponding results were similar: 2.4 log10 copies/mL (IQR, 1.3–2.8 log10 copies/mL; n = 740) for patients with R5 virus versus 1.3 log10 copies/mL (IQR, 0.3–2.7 log10 copies/mL; n = 111) for patients with non-R5 virus. For placebo recipients, the week 8 pVL decreases were modest (0.5–.8 log10 copies/mL) and similar regardless of genotypic tropism. Median pVL responses for patients who received maraviroc and those who received placebo over the course of the studies are shown in and , respectively.
Figure 2. A, Median change in plasma viral load (pVL) from baseline in the maraviroc arms. Patients identified at screening as having CCR5-using (R5) virus by either genotyping or Trofile had much larger median decreases in pVL from baseline, relative to patients (more ...)
Longer-Term Virologic Efficacy
The efficacy of maraviroc in patients identified by genotyping as having R5 virus was sustained to week 48 (). These patients were more likely to achieve a pVL <50 copies/mL at week 48, compared with the non-R5 group. In total, 49% (301 of 611) of the R5 group and 26% (62 of 240) of the non-R5 group had virologic suppression at week 48 when screened by genotyping. By Trofile, these were 46% (337 of 740) and 23% (26 of 111), respectively.
Figure 3. Percentage of patients with plasma viral load (pVL) <50 human immunodeficiency virus (HIV) RNA copies/mL in maraviroc arms. A higher proportion of patients identified at screening by either method as having CCR5-using (R5) virus had a pVL <50 (more ...)
The genotypic non-R5 group could be divided roughly in half, with 127 patients having low-prevalence (2%–20%) non-R5 virus and 113 patients having >20% non-R5 virus. The group of patients with 2%–20% non-R5 virus according to deep sequencing had minority non-R5 variants that were not reliably detected by standard population-based sequencing methods (). Importantly, this group of patients had poor response to maraviroc, with 27% (34 of 127) of the patients achieving virologic suppression at week 48, similar to the non-R5 group as a whole (26%) and to patients with >20% non-R5 virus (25%; 28 of 113). The rate of virologic response among placebo recipients was similar to that among maraviroc recipients identified as having non-R5 HIV, ranging from 17% to 23% depending on tropism or assay.
Interestingly, the virologic outcomes for maraviroc recipients showed a general inverse relationship with the percentage of non-R5 virus present at screening according to genotyping. Patients with 0% non-R5 virus had the greatest success, showing a median week 8 pVL decrease of 2.6 log10 copies/mL, with 65% (58 of 89) of the patients having week 48 virologic suppression. Patients with 0%–1% non-R5 virus had slightly poorer outcomes, with a median week 8 pVL decrease of 2.4 log10 copies/mL and a 48% rate of virologic suppression (234 of 491 patients). This decreased again in patients with 1%–2% non-R5 virus, who had a median decrease of 2.1 log10 copies/mL and a 29% rate of virologic suppression (9 of 31 patients). Patients with >2% non-R5 virus (ie, the group identified as having non-R5 virus by genotyping) all showed similar low virologic responses, as detailed above.
Changes in Viral Tropism
As a separate endpoint, patients were analyzed according to whether they experienced a change in their Trofile result from R5 to non-R5 (a tropism “switch”) over the course of the studies (). This parameter is both clinically relevant for maraviroc-based therapy and functioned as a measure separate from changes in pVL. Among those patients originally identified as having R5 virus by Trofile screening, those identified as having non-R5 virus by genotyping were almost twice as likely to have non-R5 HIV emerge by week 24, compared with patients identified as having R5 virus by both methods. A total of 40% (72 of 180) of the maraviroc recipients who switched tropism were identified by genotyping as having ≥2% non-R5 virus at screening. Tropism switches occurred in 18% (111 of 612) of the patients identified as having R5 virus by genotyping, which was lower than the rate among those identified as having R5 virus by Trofile alone (25%; 180 of 724 patients).
Figure 4. Time to change in tropism from CCR5-using (R5) to dual/mixed (DM) or X4 virus in maraviroc arms. The change in tropism in the maraviroc arms, where all patients had R5 virus identified at screening by Trofile and switched tropism to DM or X4 over the (more ...)
Among patients who switched tropism, maraviroc recipients classified as having R5 virus by Trofile but as having non-R5 virus by genotyping were treated for a mean of 4.6 weeks before Trofile gave a non-R5 result, which was more than twice as quickly as for patients for whom both tests indicated R5 virus (9.7 weeks).
Response Stratified by Background Drug Activity
Patients were also classified according to a weighted optimized background therapy susceptibility score (wOBTss)—in general, the number of active drugs in the patient's background regimen at baseline, with nucleoside reverse-transcriptase inhibitors scoring 0.5 [35
]. Genotyping was predictive of virologic success for patients who received maraviroc-based therapy, regardless of wOBTss ().
Figure 5. Median change in plasma viral load (pVL) from baseline in patients with CCR5-using (R5) virus stratified by weighted optimized background therapy susceptibility score (wOBTss). Maraviroc-treated patients identified as having R5 virus at screening by genotyping (more ...)
Maraviroc was successful in either of the R5 groups where the wOBTss was between 1 and 2. The proportions of these patients with a week 48 pVL <50 copies/mL were 58% (179 of 311) and 53% (205 of 389) when screened by genotyping and Trofile, respectively. The predictive ability of genotyping was more pronounced at more-compromised background regimens (). The proportions of patients with undetectable viral loads were 33% (81 of 232) and 29% (78 of 271) of R5-classified patients with wOBTSS <1 by genotyping or by Trofile, respectively.
Where screening assay results differed, virologic outcomes for patients who received maraviroc slightly favored genotyping. Among the discordant patients, in cases in which Trofile indicated R5 but genotyping identified 2% non-R5 virus (n = 135), median log10 copies/mL decreases in pVL at week 8 were lower, at 1.8 log10 copies/mL, as was the decrease in the concordant non-R5 group (1.2 log10 copies/mL; n = 105). In cases in which genotyping screening identified patients as having R5 virus but Trofile screening identified them as having non-R5 virus, the median week 8 pVL decrease was 2.6 log10 copies/mL (n = 6), which was similar to the decrease in the concordant R5 group (2.4 log10 copies/mL; n = 605).
Either or both assays indicating non-R5 virus was a poor prognostic indicator of longer-term maraviroc response. At week 48, the proportions of patients with suppressed pVLs were 27% (36 of 135) for the Trofile R5 and genotyping non-R5 group and 0% (0 of 6) for the Trofile non-R5 and genotyping R5 group. The rate of viral suppression at week 48 was twice as high in the concordant R5 group than in the concordant non-R5 group: 50% (301 of 605) versus 25% (26 of 105).
In the Trofile R5 and genotyping non-R5 group, 55% (74 of 135) of the maraviroc recipients changed tropism, which was a much higher rate than that in the concordant R5 group (18%; 111 of 605). Most patients identified by Trofile as having non-R5 virus continued to have non-R5 virus over the course of the study period, regardless of concordance with genotyping.
Finally, an analysis was performed to approximate ESTA screening data by combining the screening and baseline Trofile results. If either result indicated non-R5 virus, the patient was classified as being in the Trofile non-R5 group and likely had non-R5 HIV fluctuating near the limits of detection of Trofile. In this analysis, the combined Trofile assays performed similarly to deep sequencing at screening alone, with similar virologic responses among the discordant groups (data not shown).