To the Editor:
The incorporation of resistance testing into HIV clinical practice has highlighted the importance of developing standardized approaches for test interpretation. Establishment of genotypic resistance scores is based on the genotypic profile at baseline in the context of virologic response. This provides objective information for the guidance of drug selection, in particular for treatment-experienced patients.1-5 Recently, a genotypic resistance score for boosted atazanavir/ritonavir (ATV/RTV) in protease inhibitor (PI)-experienced patients has been proposed using stepwise methodologies, as previously proposed.6-8 Eight mutations (10F/I/V, 16E, 33I/F/V, 46I/L, 60E, 84V, 85V, and 90M) were identified in a resistance score associated with a reduced virologic response to ATV/RTV-containing regimen.9 There was no decrease of ATV/RTV activity in patients harboring variants with less than 3 mutations at baseline, whereas a loss of activity was observed in patients harboring higher numbers of mutations. Thus, the clinical cutoff was set at greater than or equal to 3 mutations (www.hivfrenchresistance.org). This set of mutations has been also introduced within the last International AIDS Society-USA Drug Resistance Mutations list (www.iasusa.org).10 It is now recognized that the validation or measurement of the ability of a rule to predict clinical response is important for establishing clinical utility. Ideally, this is assessed for a model using a completely different data set, independent of the discovery set from which the model was derived. To validate the accuracy of the interpretation rule recently proposed for ATV/RTV, we applied this score in a population of heavily pretreated patients enrolled in a salvage trial with ATV/RTV-containing regimen (randomized ANRS 107 Puzzle 2 trial).
The ANRS 107 Puzzle 2 trial evaluated the efficacy up to 26 weeks of once-daily 300 mg atazanavir plus 100 mg ritonavir combined with tenofovir and recycled nucleoside reverse transcriptase inhibitors (NRTIs) in 53 HIV-infected patients with multiple antiretroviral regimen virologic failures.11 At baseline, median CD4 cell count was 206 cells/mm3, median viral load (VL) 5.0 log10 copies/mL, and median numbers of NRTIs, nonnucleotide reverse transcriptase inhibitors (NNRTIs), and PIs resistance mutations were 7, 1, and 8, respectively, according to the International AIDS Society-USA Drug Resistance Mutations list. At week 26, a mild decrease in median VL from baseline of 0.2 log10 copies/mL was observed, with 16 (31%) and 9 (17%) patients exhibiting a decrease in viral load of at least 0.5 and 1.0 log10 copies/mL, respectively. HIV genotypic resistance mutations patterns were assessed at week 4 using population-based sequencing. Interpretation of genotypic resistance was performed using the 2005 version of the ANRS algorithm of resistance (http://www.hivfrenchresistance.org). In the present analysis, virologic response was defined as a decrease in HIV-1 RNA of more than 0.5 log10 at 26 weeks. Associations between baseline PI mutations and virologic response were analyzed by Fisher exact tests. Mutations detected in more than 8% of patients and with P = 0.2 were considered. The Cochran-Armitage test was used to evaluate the predictability of the resistance genotypic score proposed by Vora et al9 (10F/I/V, 16E, 33I/F/V, 46I/L, 60E, 84V, 85V, and 90M). Robustness of score was investigated by bootstrap resampling method. Significant predictive factors in univariate analysis (P = 0.05) were entered into multivariate analysis.
At baseline, the median number of minor and major PI mutations was 3 (range, 0-5) and 5 (range, 0-11), respectively. According to the 2005 ANRS algorithm, 96% of patients were resistant to at least one NRTI, 85% to NNRTIs, and 94% to at least one PI. Resistance to only 1, 2, and 3 classes of drugs was observed in 2%, 15%, and 81% of patients, respectively.
Considering the 16 patients exhibiting a decrease in VL of at least 0.5 log10 copies/mL, 10 PI mutations, present at baseline in at least 8% of the patients, were associated with a reduced virologic response (<0.5 log10 copies/mL HIV-1 RNA decrease) to ATV/RTV-containing regimen: 10F/H/I/V, 11I/L, T12K/P/Q/S, 33F/I, 54L/M/N/S/T/V, 71I/L/T/V, I72L/T/V, G73S/T, 84V, and 90M. Mutations at positions 10, 33, 84, and 90 are also included in the ATV/RTV score. The following combination of 8 mutations proposed by Vora et al9 (10F/I/V, 16E, 33I/F/V, 46I/L, 60E, 84V, 85V, and 90M) was associated (Cochran-Armitage test, P = 0.003) with a reduced virologic response at 26 weeks.
Using this genotypic score, a virologic response was observed in 100% of patients with 0 mutations and in 75%, 40%, 23%, 22%, 29%, and 0% of patients with 1, 2, 3, 4, 5, and 6 mutations, respectively (Fig. 1). In the 100 bootstrap samples, P values of the Cochran-Armitage test were less than 0.0001 and less than 0.005 in 13% and 86% of cases, respectively. In univariate analysis, the genotypic score was associated to the virologic response at 26 weeks: the median number of mutations among the set of 8 ATV/RTV mutations was 4 in nonresponders and 3 in responders (P = 0.03). When entered in multivariate analysis, ATV/RTV genotypic mutation score remained predictive of virologic response with odds ratio of 11.2 (95% CI, 1.8-70.9).
When tested with the same methodology on this data set, a previous algorithm based on phenotype/genotype correlations was not relevant because the distribution of the virologic response according to this score was dichotomous (data not shown).12
This study confirms the predictive accuracy of a genotypic score comprising 8 mutations (10F/I/V, 16E, 33I/F/V, 46I/L, 60E, 84V, 85V, and 90M) for the virologic response to an ATV/RTV-containing regimen in a different set of patients, independent of the discovery set from which the score was derived. It is worth noting that the patients' characteristics of the establishment and the validation data sets were completely different. Patients enrolled in the data set used for the establishment of the score harbored in a relatively small number of cases high number of PI-resistant mutations at baseline, whereas the patients enrolled in the ANRS 107 trial (validation data set) were heavily pretreated patients with a higher number of PI-resistant mutations. This study is limited by the small sample size of the validation data set and the poor virologic response observed in the ANRS 107 trial. However, this score was relevant when tested on this data set because the distribution of the virologic response according to this score was monotonic, and this finding validated by the Cochran-Armitage test.
In conclusion, the genotypic resistance score recently proposed by Vora et al9 is a useful tool for the prediction of virologic response to ATV/RTV in PI-experienced patients.
Anne-Geneviève Marcelin, PharmD, PhD
Corine Chazallon, PhD
Laurence Gérard, PhD
Yacine Saïdi, PhD
Jean-Pierre Aboulker, PhD
Pierre-Marie Girard, MD, PhD
Vincent Calvez, MD, PhD
Christophe Piketty, MD, PhD
*Department of Virology
Hôpital Pitié, Salpêtrière
Assistance Publique-Hôpitaux de Paris
Université Pierre et Marie Curie-Paris 6
†Inserm SC10, Villejuif
‡Department of Infectious Diseases
Hôpital Saint Antoine
and §Department of Immunology
Hôpital Georges Pompidou
Assistance Publique-Hôpitaux de Paris
1. Kempf DJ, Isaacson JD, King MS, et al. Analysis of the virological response with respect to baseline viral phenotype and genotype in protease inhibitor-experienced HIV-1-infected patients receiving lopinavir/ritonavir therapy. Antivir Ther
2. Marcelin AG, Dalban C, Peytavin G, et al. Clinically relevant interpretation of genotype and relationship to plasma drug concentrations for resistance to saquinavir-ritonavir in human immunodeficiency virus type 1 protease inhibitor-experienced patients. Antimicrob Agents Chemother
3. Marcelin AG, Lamotte C, Delaugerre C, et al. Genotypic inhibitory quotient as predictor of virological response to ritonavir-amprenavir in human immunodeficiency virus type 1 protease inhibitor-experienced patients. Antimicrob Agents Chemother
4. Valer L, de Mendoza C, Soriano V. Predictive value of drug levels, HIV genotyping, and the genotypic inhibitory quotient (GIQ) on response to saquinavir/ritonavir in antiretroviral-experienced HIV-infected patients. J Med Virol
5. Valer L, Gonzalez de Requena D, de Mendoza C, et al. Impact of drug levels and baseline genotype and phenotype on the virologic response to amprenavir/ritonavir-based salvage regimens. AIDS Patient Care STDS
6. Brun-Vezinet F, Costagliola D, Khaled MA, et al. Clinically validated genotype analysis: guiding principles and statistical concerns. Antivir Ther
7. Brun-Vezinet F, Descamps D, Ruffault A, et al. Clinically relevant interpretation of genotype for resistance to abacavir. AIDS
8. Flandre P, Marcelin AG, Pavie J, et al. Comparison of tests and procedures to build clinically relevant genotypic scores: application to the Jaguar study. Antivir Ther
9. Vora S, Marcelin AG, Gunthard HF, et al. Clinical validation of atazanavir/ritonavir genotypic resistance score in protease inhibitor-experienced patients. AIDS
10. Johnson VA, Brun-Vezinet F, Clotet B, et al. Update of the drug resistance mutations in HIV-1: fall 2005. Top HIV Med
11. Piketty C, Gérard L, Chazallon C, et al. Salvage therapy with atazanavir/ritonavir combined to tenofovir in HIV-infected patients with multiple treatment failures: randomized ANRS 107 trial. Antivir Ther
12. Colonno RJ, Thiry A, Limoli K, et al. Activities of atazanavir (BMS-232632) against a large panel of human immunodeficiency virus type 1 clinical isolates resistant to one or more approved protease inhibitors. Antimicrob Agents Chemother