 |
|
|
| |
Best Genotypic Resistance Tests To Evaluate Sensitivity To Kaletra
|
| |
| |
"Genotypic Algorithms Specific to Lopinavir/ritonavir Outperform Nonspecific Lists of PI Mutations in Predicting Virologic Response to Lopinavir/ritonavir in PI-Experienced Patients"
M Norton, D Kempf, S Brun, J Omachi, P Cernohous, M King; Abbott Laboratories, Abbott Park, IL, USA
XIII International HIV Drug Resistance Conference,Tenerife, Canary Islands, Spain, June 8--10, 2004
Mike Norton reported this poster at Tenerife.
Lopinavir (LPV) is a human immunodeficiency virus (HIV) protease inhibitor (PI) that is co-formulated with ritonavir (r), which functions as a pharmacokinetic enhancer. The approved adult dose of LPV/r is 400/100 mg twice daily (BID). Potent antiviral activity of LPV/r has been demonstrated in antiretroviral (ARV)-naïve and PI-experienced patients. In study 888, LPV/r was compared to a regimen of investigator-selected protease inhibitors ISPI(s) in single PI-experienced, NNRTI-naive patients. Through 48 weeks, LPV/r had significantly better efficacy per the FDA time to loss of virologic response analysis (p<0.001).
A number of algorithms have been developed to identify mutations associated with greater phenotypic or clinical resistance to lopinavir. Ultimately, the validity of these algorithms should be determined by assessing their performance in predicting response to lopinavir-based regimens in large study populations.
The study enrolled 288 single PI-experienced, NNRTI-naive patients experiencing virologic failure on their current regimen: Kaletra (400/100)+nevirapine+2 NRTIs (n=148).
BACKGROUND
Virologic response was based on dropouts-as-censored (DAC) analysis. Patients discontinuing prior to Week 8 were excluded from the analysis. Patients with HIV RNA <400 copies/mL at Week 40-48 were considered responders. Patients who discontinued during weeks 8-40 were considered nonresponders if the final HIV RNA value was above 400 copies/mL and were censored if it was <400 copies/mL.
Six genotypic algorithms were employed. These included algorithms based on genotype-phenotype relationships (LPV mutation score, ViroLogic, Virco), genotype-virologic response relationships (ATU mutation set, Clinic-Based Investigator Group [CBIG1/CBIG2]), or a general list of PI mutations (Data Analysis Plan of the Resistance Collaborative Group [DAP]). The ViroLogic and CBIG2 algorithms include weights for certain mutations. Mutations at positions 10, 20, 24, 46, 54, 71, 82, 84, and 90 were most commonly included in these algorithms, as each appeared in at least 5 of 6 algorithms. Logistic regression was used to assess virologic response based on genotype scores for each algorithm.
RESULTS
Patients were primarily male (86%) and white (67%), with mean baseline HIV RNA and CD4 count of 4.1 log10 copies/mL and 322 cells/mm3, respectively. The most common prior PIs were nelfinavir (43%) and indinavir (42%). The most common control arm regimens used were RTV/SQV (44%), RTV/IDV (21%), and NFV (21%).
The distribution of baseline scores for each genotypic algorithm is shown below.
| LPV | | | | | | | Mutation Score | Virologic | Virco | ATU | CBIG1 | DAP | | Median | 2 | 3 | 2 | 1 | 1 | 3 | | Range | 0-8 | 0-12 | 0-7 | 0-6 | 0-4 | 0-9 |
The most common mutations that patients had: 10, 36, 46, 54, 63, 71, 72, 77, 82, 90.
The ViroLogic (p=0.008) algorithm and the ATU mutation set (p=0.028) were significantly associated with virologic response, and the number of Virco mutations was marginally associated (p=0.053). The CBIG2 algorithm was not associated with response.
Genotypic algorithms and baseline characteristics (age, gender, baseline HIV RNA) were entered into a multivariable logistic regression model using forward stepwise selection. The final model indicated that the ViroLogic algorithm remained significant after adjusting for baseline HIV RNA and gender.
The advantage of the ViroLogic algorithm appears to be due in part to the 3-fold higher weights assigned to mutations at positions 54 and 82 (certain mutations at positions 47 and 50 were also assigned higher weights in the ViroLogic algorithm but were not observed in this study). If weights of 3 are assigned to positions 54 and 82 in the other algorithms, each demonstrates an improved model fit. Similarly, if an unweighted version of the ViroLogic algorithm is used, the logistic regression model is no longer statistically significant.
To assess the potential impact of other weighting schemes on the performance of the ViroLogic and ATU mutation set algorithms, they reported assigning weights of 2, 3, and 4 to each mutation that occurred in at least 5% of patients and compared results of the best logistic regression model based on these new weights to the results from the unweighted model. In contrast to the results from mutations at positions 54 and 82 shown above, weights for other mutations did not demonstrate appreciable improvement of algorithm performance. Results were similar if the mutations were assessed in the context of the 3-fold weights for positions 54 and 82.
Norton summarized:
In our assessment of a variety of genotypic algorithms for predicting the likelihood of response to LPV/r-based regimens, we found that algorithms based on larger numbers of patients and those that are specific to lopinavir performed better.
Results suggest that in our study population of single PI-experienced patients, in each algorithm, mutations at positions 54 and 82 were of greater importance than other mutations in determining virologic response. However, mutations at positions 54 and 82 did not completely prevent the possibility of response, as 20 of 39 (51%) of LPV/r-treated patients with these mutations had HIV RNA <400 copies/mL at Week 48.
These findings are consistent with previously reported results for multiple PI-experienced patients treated with an LPV/r-based regimen.8 In that analysis, 20 patients had baseline mutations at positions 10, 54, and 82, plus at least 3 other mutations from the LPV mutation score. 10/20 patients (50%) demonstrated virologic response through 48 weeks. While this was lower than the response rate in patients without this mutation pattern (28/30, 93%), the results demonstrated that mutations at positions 54 and 82 do not completely preclude the possibility of virologic response.
Among the 20 patients in the prior study, nonresponders had greater reductions in baseline LPV phenotypic susceptibility (median 50.5-fold) compared to responders (16-fold), suggesting that for complex genotypes, viral phenotype may provide additional predictive power over genotype. In the current study, 23 of the 39 patients with mutations at positions 54 or 82 had baseline phenotypic data available, and a similar trend was observed, although the difference was not statistically significant: the median baseline reduction in LPV susceptibility was 14-fold for 17 nonresponders and 5.4-fold for 6 responders (p=0.18).
Limitations of this analysis include the fact that the patients in this study had relatively modest amounts of pre-treatment, as they had been exposed to only one prior protease inhibitor and were NNRTI-naive. Thus, certain mutations that are likely to have an impact on phenotype and clinical resistance to LPV (e.g., I50V) could not be assessed since they were not present in this population.
Norton concluded:
Among the various algorithms that may be used to predict virologic response to a LPV/r-containing regimen, those that are specific to LPV and based on a large number of samples tended to perform better.
As observed in prior trials, mutations at positions 54 and 82 appeared to be of greater importance than other protease mutations; however, the presence of mutations at positions 54 and 82 did not preclude virologic response.
With more complex genotypes, viral phenotypes may provide additional predictive power. In some algorithms, lower response was observed in patients with no baseline mutations, suggesting that some of these patients may have experienced failure on both their prior and current regimens due to poor adherence.]
|
| |
|
|
|
|
|
