 |
|
|
| |
Optimization of Clinically Relevant Cutpoints for the Determination of HIV Co-Receptor Usage to Predict Maraviroc Responses in Treatment Experienced (TE) Patients Using Population V3 Genotyping
|
| |
| |
Reported by Jules Levin
EACS Nov 13 2009 Cologne Germany
Rachel A. McGovern1, Winnie Dong1, Theresa Mo1, Conan Woods1, Xiaoyin Zhong1, Alexander Thielen2, Mark Jensen3, Jayvant Heera4, Suzanne Ellery4, Marilyn Lewis5, Ian James5, Pinaki Biswas6, Doug Chapman6, Hernan Valdez,6 and Richard Harrigan1
1BC Centre for Excellence in HIV/AIDS, Vancouver, Canada; 2Max Planck Institute for Bioinformatics, Saarbrucken, Germany; 3Fortinbras Consulting, Buford, GA, USA; 4Pfizer Global Research and Development, New London, CT, USA, 5Pfizer Global Research and Development, Sandwich UK, 6Pfizer, Inc., New York, NY, US
AUTHOR CONCLUSION
Population V3 genotyping has utility for clinical prediction of initial MVC responses in TE patients: using non-optimized cut-points, viral load decreases after MVC in TE patients with R5 are comparable to those observed with original Trofile (IAS 2009, Cape Town, Abstract WELBA101)
Alternatively, the use of two cutpoints can define different levels of MVC activity as done for other antiretroviral agents. We identify two useful cut-points at 2% and 5.75% false positive rate with g2P which define responses associated with very poor and slightly diminished response to MVC, respectively
BACKGROUND
The advent of CCR5 antagonists, such as Maraviroc (MVC), has introduced the need to identify HIV co-receptor usage (tropism) prior to initiation of therapy to ensure an adequate contribution to clinical response
Currently, the most widely used tropism assay is the recombinant phenotypic Enhanced Sensitivity Trofile Assay (ESTA) developed by Monogram Biosciences. ESTA data is not available for the MOTIVATE studies
A genotypic tropism assay comprising population sequencing of the V3-loop of the HIV-1 envelope has the potential for use in routine clinical practice. This method has recently been shown to be comparable to the original Trofile assay in predicting clinical outcome in the maraviroc trials (IAS 2009, Cape Town, Abstract WELBA101)
These trials compared maraviroc (MVC) + optimized background therapy (OBT) vs placebo + OBT in treatment-experienced patients (TE) with CCR5-using virus (R5). Patients with non-R5 virus were enrolled in a sister safety study of the same format, A4001029. Tropism for these study groups was determined using the original Trofile assay
Here we optimize the performance of population-based genotyping at the V3 loop of the HIV-1 envelope by searching for better cut-points in the interpretative algorithms, using the MOTIVATE-1, -2 and A4001029 maraviroc studies
METHODS
Triplicate V3 amplicons were prepared by RT-PCR from stored plasma screening samples and sequenced using standard ABI methods by operators blinded to outcome (Figure 1)
Sequence calls were performed automatically without manual review using the custom software ReCall
Therapeutic response to MVC was predicted using the g2P algorithm (cutpoints 2 and 5.75). G2P<2 predicted a poor response, whereas g2P >5.75 predicated a successful response. A score of 2< g2P<5.75 was considered intermediate (Figure 2)
Different cutpoints were used to optimize the algorithm for this use. Cutpoints were chosen to reflect an initial, MVC-specific decrease in viral load or complete viral suppression
A random 75% of available patient data was used as the training set; 25% was reserved for validation. Bootstrapping was also used to verify the cutpoints, a random 75% of the available data was re-sampled with replacement 1000 times
Primary outcome was change in viral load (VL) at week 8 in the combined maraviroc arms (Figure 3a). Week 8 VL changes were modeled with the inclusion of terms for baseline plasma VL and CD4 cell count, as well as OBT sensitivity score (wSS)
Other endpoints examined included percent undetectable (Figure 3b), time to change in tropism (Figure 4a) and time to study discontinuation (Figure 4b), as well as change in CD4 count and time to loss of virological response (data not shown)
Endpoint results for g2P were compared to those of the original Trofile assay
In the case where patients did not complete the 48 study weeks, LOCF (last observation carried forward) and M=F (missing equals failure) were applied to both Trofile and g2P results
Figure 1. Study design for the re-analysis of Motivate-1, 2 and A4001029
RESULTS
Figure 2. Frequency distribution of patient samples as determined by the assigned g2P score in accordance with therapeutic response categories. The majority of patients cluster as either poor or good responders. Note two scales along the x-axis, <10 is incremented by units of 0.5 and >10 is incremented by units of 10. Red indicates patients likely to respond poorly to MVC, yellow predicts compromised MVC efficacy, green indicates patients likely to respond well to MVC.
Figs 3a and 3b. Population sequencing with optimized g2P cutoff points can equally predict the therapeutic outcome of maraviroc compared to the original Trofile as demonstrated with a) viral load change and b) percent suppressed below 50 HIV RNA copies/mL. Outcomes for the combined maraviroc arms (75%), placebo arm, validation set (25%) and all samples from the dataset. Coloured lines represent g2P scores (<2, 2-5.75, >5.75); black lines represent R5 (solid) or non-R5 (dotted) as determined by the original Monogram Trofile assay. A g2P score of <2 inferred a poor response, a score of >5.75 inferred therapeutic success. 2< g2P<5.75, marks the point at which MVC efficacy becomes compromised. Population sequencing with the g2P algorithm is shown to be equally able to predict therapeutic outcome following the initiation of MVC when compared to Trofile. Results suggest g2P may be better able to predict a poor therapeutic response to MVC.
Figures 4a and 4b. Population sequencing with optimized g2P cutoff points can effectively predict a) the probability of a tropism change and b) time to the withdrawal from the study. Labeled as above. The time in which tropism may shift from R5 to non-R5 following initiation of MVC. Those predicted to exhibit a poor response to MVC by g2P were far more likely to switch tropism over the course of the study than those predicted to respond well and less likely to remain in the study for the 48 week duration following the initiation of MVC.
|
| |
|
|
|
|
|
