Drug-resistant genotyping in HIV-1 therapy

Abstract

Authors' replySir—Jean-Pierre Aboulker suggested that the p values are not consistent with the corresponding figures for the proportion of patients with plasma HIV-1 RNA load below the detection limit of 200 copies/mL. We agree with his remarks. However, the legend for table 3 was incorrect: the legend representing the two groups was reversed (see Department of error, p 1128)Aboulker also states that at month 3, the mean reduction in HIV-1 RNA load was not significantly greater in the genotypic than in the control group. He retrospectively reanalysed the results without the benefit of the study protocol and compared two groups with the Fisher's exact test which should usually be used to compare two groups. We compared three groups of patients: complete responders (viral load < 200 copies/mL), partial responders (viral load decrease >0·5 log10 copies/mL but still >200 copies/mL), and non-responders (viral load decrease <0·5 log10 copies/mL), which is more logical and in accordance with the protocol design. This description was withdrawn at the request of one of the reviewers. To do the statistical analysis according to the actual protocol design, we used the corrected x2 test, which is the appropriate test for our study design: therefore our conclusions are correct.At month 3, mean change in HIV-1 RNA was significant. Effectively this value is a mean log10 change, and therefore the 95% CI must exclude 0 and not 1. The 95% CI does exclude 0, and because all our analyses and the figure are log10 values of HIV-1 RNA load, our results are correct. We confirm our statistical analysis and our conclusions and respectfully submit that Aboulker's concerns were based on incomplete information about the protocol design and data analysis plan for the VIRADAPT study. Authors' reply Sir—Jean-Pierre Aboulker suggested that the p values are not consistent with the corresponding figures for the proportion of patients with plasma HIV-1 RNA load below the detection limit of 200 copies/mL. We agree with his remarks. However, the legend for table 3 was incorrect: the legend representing the two groups was reversed (see Department of error, p 1128) Aboulker also states that at month 3, the mean reduction in HIV-1 RNA load was not significantly greater in the genotypic than in the control group. He retrospectively reanalysed the results without the benefit of the study protocol and compared two groups with the Fisher's exact test which should usually be used to compare two groups. We compared three groups of patients: complete responders (viral load < 200 copies/mL), partial responders (viral load decrease >0·5 log10 copies/mL but still >200 copies/mL), and non-responders (viral load decrease <0·5 log10 copies/mL), which is more logical and in accordance with the protocol design. This description was withdrawn at the request of one of the reviewers. To do the statistical analysis according to the actual protocol design, we used the corrected x2 test, which is the appropriate test for our study design: therefore our conclusions are correct. At month 3, mean change in HIV-1 RNA was significant. Effectively this value is a mean log10 change, and therefore the 95% CI must exclude 0 and not 1. The 95% CI does exclude 0, and because all our analyses and the figure are log10 values of HIV-1 RNA load, our results are correct. We confirm our statistical analysis and our conclusions and respectfully submit that Aboulker's concerns were based on incomplete information about the protocol design and data analysis plan for the VIRADAPT study. Drug-resistant genotyping in HIV-1 therapyThe results of the VIRADAPT trial (June 26, p 2195)1 are puzzling from a statistical point of view. J Durant and colleagues1 report that the percentage of patients with HIV-1 RNA lower than the detection level was better in the genotypic than in the control group. However, the p values repeatedly shown in the results section and in the summary are not consistent with the corresponding figures of the proportions of patients with plasma HIV-1 RNA below the detection limit of 200 copies/mL. With Fisher's exact two-sided test, the comparison of 19 of 65 patients versus six of 43 patients being below the limit at 3 months yields a p value of 0·102 and not 0·017. Full-Text PDF