Antiviral Therapy 7:

Transcription

1 Antiviral Therapy 7: 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 Dale J Kempf*, Jeffrey D Isaacson, Martin S King, Scott C Brun, Jacquelyn Sylte, Bruce Richards, Barry Bernstein, Richard Rode and Eugene Sun Global Pharmaceutical Research and Development, Abbott Laboratories, Abbott Park, Ill., USA *Corresponding author: Tel: ; Fax: ; dale.kempf@abbott.com The virological response of multiple protease inhibitorexperienced, non-nucleoside reverse transcriptase inhibitor-naive, HIV-1-infected subjects was examined with respect to baseline viral phenotype and genotype through 72 weeks of therapy with lopinavir/ritonavir plus efavirenz and nucleoside reverse transcriptase inhibitors (Study M98-957). Using a dropouts as censored analysis, plasma HIV RNA 4 copies/ml was observed in 93% (25/27), 73% (11/15) and 25% (2/8) of subjects with <1-fold, 1- to 4-fold, and >4-fold reduced susceptibility to lopinavir at baseline, respectively. In addition, virological response was observed in 91% (21/23), 71% (15/21) and 33% (2/6) of subjects with baseline lopinavir mutation score of 5, 6 7 and 8, respectively. Through 72 weeks, all subjects experiencing virological failure whose baseline isolates contained six or more protease inhibitor mutations had a common genotypic pattern, with mutations at positions 82, 54 and 1, along with a median of four additional mutations in protease. However, an equal number of subjects with a similar genotypic pattern experienced virological response. Further analysis revealed the baseline phenotypic susceptibility to lopinavir to be an additional covariate predicting response in this subset of subjects. In multivariate analyses, baseline susceptibility to lopinavir was associated with response at each time point examined (weeks 24, 48 and 72). These results provide guidance for clinically relevant interpretation of phenotypic and genotypic resistance tests when applied to lopinavir/ritonavir. Introduction The development of HIV drug resistance during antiretroviral therapy can compromise the efficacy of subsequent regimens following virological failure. Several retrospective studies have shown that changes in baseline viral phenotype [1 6] or genotype [4,6 8], compared to that of wild-type virus, adversely affect virological response of antiretroviral-experienced subjects to therapy with a subsequent regimen. Moreover, prospective studies have verified the utility of baseline genotype [9,1] or phenotype [11] in producing a superior virological response to therapy in treatment-experienced patients. The evaluation of virological response with respect to baseline resistance can aid in defining the genotypic and phenotypic susceptibility ranges over which new antiretroviral agents under clinical investigation display activity. Lopinavir (formerly ABT-378) is a new protease inhibitor (PI) that, when co-administered with lowdose ritonavir (designated lopinavir/r), has shown potent virological activity in both antiretroviral-naive [12] and single PI-experienced, HIV-infected subjects [13]. In the latter group (Study M97-765), virological response after 24 and 48 weeks of therapy was not different between subjects whose baseline viral isolates (at the initiation of lopinavir/r therapy) displayed either >fourfold or <fourfold reduced in vitro susceptibility to lopinavir. This result, coupled with the observation that mean plasma lopinavir trough levels (5.5 µg/ml) exceed the IC 5 of lopinavir for wild-type HIV in the presence of 5% human serum [14] by >75- fold, suggests that a fourfold decrease in phenotypic susceptibility to lopinavir is not a biologically or clinically relevant threshold. Instead, it is likely that decreased response to lopinavir/r requires substantially greater reduced phenotypic susceptibility. In a similar manner, virological response in Study M was independent of the number of mutations in HIV protease within the baseline viral isolates, 22 International Medical Press /2/$

2 DJ Kempf et al. despite the fact that >5% of the baseline isolates contained 3 6 mutations associated with PI resistance [13]. The patterns of mutations selected by lopinavir/r during therapy in treatment-naive subjects have not been determined; however, genetic correlates of reduced in vitro susceptibility to lopinavir in viral isolates from subjects failing therapy with other PIs have been characterised [15]. That analysis revealed mutations at 11 amino acid positions in the HIV protease gene (L1F/I/R/V, K2M/R, L24I, M46I/L, F53L, I54L/T/V, L63P, A71I/L/T/V, V82A/F/T, I84V and L9M) that were likely to contribute to reduced susceptibility to lopinavir. The number of these 11 mutations (referred to as the lopinavir mutation score) provide a potential method by which virological response to lopinavir/r can be evaluated as a function of baseline genotype. Since the baseline HIV isolates from single PI-experienced subjects displayed a level of phenotypic and genotypic resistance that was insufficient to adequately characterise the effect of these parameters on the response to lopinavir/r, we studied the virological response of multiple PI-experienced, non-nucleoside reverse transcriptase inhibitor (NNRTI)-naive subjects to combination therapy with lopinavir/r, efavirenz and nucleoside reverse transcriptase inhibitors (NRTIs). This analysis, by characterizing virological response with respect to baseline phenotype and genotype in a more advanced patient population, provides guidance for the interpretation of HIV resistance testing when applied to this new PI. Methods Analysis of the phenotype and genotype of baseline isolates Plasma samples for baseline viral phenotype and genotype were collected between day 6 and day 1. Viral phenotype was determined by ViroLogic Inc., using a single-cycle recombinant HIV assay [16], and analysed as fold IC 5, compared with the standard laboratory strain. The genotypes of the recombinant baseline isolates were determined by ViroLogic Inc. using population sequencing, and reported as sequence changes with respect to the sequence of the pnl4-3 laboratory wild-type strain. Baseline genotype was analysed as the lopinavir mutation score (the number of mutations in protease from the following list: L1F/I/R/V, K2M/R, L24I, M46I/L, F53L, I54L/T/V, L63P, A71I/L/T/V, V82A/F/T, I84V and L9M). Analysis of virological response Virological response was defined by plasma HIV RNA 4 copies/ml (Amplicor ) or 5 copies/ml (Ultrasensitive) and assessed using a dropouts as censored definition [6]. In addition, subjects who discontinued therapy prior to week 8 and thus did not undergo therapy for a sufficient length of time to experience complete response, were excluded from analysis. Statistical analyses of virological response (binary) were conducted using univariate and multivariate ( forward-selection stepwise) logistic regression. Stepwise regression models employed a total of 19 baseline parameters, including lopinavir/r dose, baseline resistance parameters (log fold lopinavir IC 5, log fold efavirenz IC 5, NRTI phenotypic susceptibility score [6], lopinavir mutation score, and genotypic susceptibility scores [6] for efavirenz and the NRTIs in the regimen) as well as other data collected per study protocol (age, gender, baseline weight, time since HIV diagnosis, time since the initiation of PI therapy, number of previous antiretroviral agents, number of previous NRTIs, number of previous PIs, baseline CD4, baseline CD8 and baseline plasma HIV RNA). The stepwise regression analyses were performed using entry and exit P-values (significance levels) of.15. The relationship between virological response and individual mutations within the lopinavir mutation score was evaluated using Fisher s exact test. The modified Bonferroni adjustment [17,18] was used to adjust for multiplicity so that a P-value less than.15 (.5 divided by the square root of 11) was considered statistically significant. Results Characterisation of baseline viral isolates In order to assess virological response to lopinavir/r as a function of baseline phenotypic and genotypic susceptibility, we analysed data up to 72 weeks in a Phase I/II clinical study. Study M enrolled 57 NNRTI-naive subjects with plasma HIV RNA >1 copies/ml on either sequential and/or simultaneous therapy with at least two PIs. Subjects entering this study had been treated with a median of three PIs and four NRTIs. Sixteen subjects (29%) had previous treatment with four PIs. Previous experience to indinavir was the highest (86%), followed by ritonavir (77%), saquinavir (71%) and nelfinavir (57%). In addition, the vast majority of subjects had previously been treated with lamivudine (91%), stavudine (93%), zidovudine (93%) and didanosine (8%). All subjects initiated therapy with one of two doses of lopinavir/r (4/1 or 533/133 mg twice-daily), along with efavirenz and two NRTIs. Phenotype and genotype were obtained on 56/57 (98%) baseline samples. The susceptibility of the baseline isolates to the PIs tested is summarised in Table 1. Quantifiably (>2.5-fold) reduced susceptibility to two or more PIs was observed for 44/56 (79%) isolates International Medical Press

3 Lopinavir/r therapy with respect to baseline phenotype and genotype Table 1. Susceptibility of baseline isolates to protease inhibitors Figure 1. Prevalence of baseline mutations associated with protease inhibitor resistance Fold IC 5 No. with 2.5- Compound Median Range fold change Lopinavir % Amprenavir % Indinavir >171 75% Nelfinavir % Ritonavir 28.5 >316 77% Saquinavir >545 63% Percentage of isolates The numbers of isolates with 1- to 2-fold, 2- to 4- fold, and >4-fold reduced susceptibility to four or more of the six PIs tested were 1 (18%), six (11%) and nine (16%), respectively. Approximately twothirds (36/56) of the baseline isolates displayed >2.5-fold reduced susceptibility to lopinavir (median 5.2-fold), ranging up to 96-fold. The numbers of isolates with 1- to 2-fold, 2- to 4-fold, and >4- fold reduced susceptibility to lopinavir were 9 (16%), 7 (13%) and 8/56 (14%), respectively. In general, substantially reduced susceptibility to lopinavir correlated with broad in vitro resistance to the PI class. Thus, all but two of the 24 isolates that displayed >1-fold loss of susceptibility to lopinavir also displayed 1-fold reduced susceptibility to at least three other PIs tested. More than half the isolates displayed >2.5-fold reduced susceptibility to lamivudine, zidovudine and abacavir (data not shown). Only two baseline isolates displayed 2.5-fold reduced susceptibility to efavirenz (4.7- and 2-fold, respectively), consistent with the study requirement for subjects to be NNRTI-naive. Most of the baseline isolates contained multiple mutations associated with PI cross-resistance, reflecting the treatment history (two or more PIs) of the study group (Figure 1). Thus, 44/56 (79%) isolates contained one or more primary mutation [19] at amino acid positions 82, 84 and 9, as well as secondary mutations at positions 1 and/or 71. Concurrently, the occurrence of mutations generally selected by only one of the currently available PIs (for example, positions 3 and 88 from nelfinavir and 48 from saquinavir) was low. The lopinavir mutation score of the baseline isolates ranged from to 1, and 79% of isolates (44/56) had mutation scores of 4 or more. All but one of these 44 isolates contained one or more of the primary mutations at positions 82, 84 or 9. Common mutations associated with NRTI resistance (amino acid positions 41, 67, 69, 7, 184, 21, 215 and 219 in reverse transcriptase) were noted in 16 73% of isolates. One baseline isolate had a multi-nrti-resistant genotype (Q151M plus F116Y) and displayed significantly reduced phenotypic susceptibility to the NRTI class. Amino acid position Univariate analysis of virological response to lopinavir/r with respect to baseline phenotype and genotype A total of 52/56 subjects with available baseline phenotype and genotype completed more than 8 weeks of therapy and qualified for the dropouts as censored analysis of virological response [6]. Virological response up to 72 weeks of therapy as a function of baseline phenotypic and genotypic susceptibility to lopinavir is shown in Figures 2 and 3, respectively. The highest response was observed among subjects whose baseline isolates displayed <1-fold reduced susceptibility to lopinavir, compared with wild-type HIV ( 4 copies/ml: 25/27, 93%; 5 copies/ml: 24/27, 89%). The response rate diminished in subjects whose baseline viral susceptibility to lopinavir was between 1- and 4-fold ( 4 copies/ml: 11/15, 73%; 5 copies/ml: 9/15, 6%) and was lowest in subjects with baseline isolates with >4-fold reduced susceptibility ( 4 copies/ml: 2/8, 25%; 5 copies/ml: 2/8, 25%). Two subjects discontinued therapy with viral load <4 copies/ml between weeks 8 and 72, and were thus censored under the dropouts as censored analysis. Using Fisher s exact test (2 3 table), virological response to both 4 and 5 copies/ml at week 72 were found to be statistically significantly associated with baseline phenotype (P<.1 and P=.1, respectively). Similarly, the highest response was observed among subjects whose baseline isolates displayed a lopinavir mutation score of 5 or less ( 4 copies/ml: 21/23, 91%; 5 copies/ml: 2/23, 87%). The response rate diminished in subjects with a lopinavir mutation score of 6 or 7 ( 4 copies/ml: 15/21, 71%; 5 copies/ml: 13/21, 62%), and was lowest in the six subjects with baseline lopinavir mutation score of 8 or more (4 copies/ml: 2/6, 33%; 5 copies/ml: 2/6, 33%). At week 72, baseline genotype using this categorical description was statistically associated with virological response to both 4 and 5 copies/ml (P=.11 and P=.22, respectively). Antiviral Therapy 7:3 167

4 DJ Kempf et al. Figure 2. Virological response with respect to baseline susceptibility to lopinavir Figure 3. Virological response with respect to baseline lopinavir mutation score (a) 1 (a) 1 Percentage 4 copies/ml Percentage 4 copies/ml (b) Weeks (b) Weeks Percentage 5 copies/ml Percentage 5 copies/ml Weeks Weeks (a) 4 copies/ml; (b) 5 copies/ml. Baseline lopinavir phenotype <1-fold (solid circles), 1- to 4-fold (open squares), >4-fold (solid triangles). (a) 4 copies/ml; (b) 5 copies/ml. Baseline lopinavir mutation score 5 (solid circles), 6 7 (open squares), 8 1 (solid triangles). Using logistic regression models that considered the baseline phenotype (log fold lopinavir IC 5 ) and genotype (lopinavir mutation score) as continuous variables, probability curves for virological response to 4 copies/ml, with corresponding 95% confidence intervals, were constructed. The sigmoidal regression curves produced using each variable to predict response are shown in Figure 4, with response as a function of the above categorical descriptors of phenotype (<1-fold, 1- to 4-fold, and >4-fold) or genotype (lopinavir mutation score <5, 6 7 and 8), superimposed. The estimated probability of virological response to 4 copies/ml at week 72 remained above 5% for baseline lopinavir IC 5 values up to 35-fold above wild-type HIV or for baseline lopinavir mutation scores up to 9. Using a more conservative estimate of response (that is, the lower 95% confidence limit), the probability of response was at least 5% for subjects with a baseline lopinavir phenotype up to 17-fold above wild-type HIV or a baseline lopinavir mutation score of 6 or less. Similarly, the estimate of the probability of response to 5 copies/ml at week 72 was at least 5% for subjects with a baseline lopinavir phenotype up to 28-fold above wild-type HIV or a baseline lopinavir mutation score of 8 or less (conservative estimates of 11-fold and 6, respectively, data not shown). Since subjects in Study M were receiving concomitant therapy with efavirenz and NRTIs, the virological response was also analysed with respect to the baseline susceptibility to efavirenz, the baseline NRTI phenotypic susceptibility score (PSS), and the baseline NRTI genotypic susceptibility score (GSS) [6]. All but two isolates displayed <2.5-fold reduced susceptibility to efavirenz, reflecting the NNRTI-naive study population. With the exception of those two subjects, who both also had at least 2-fold reduced susceptibility to lopinavir and who both experienced virological failure, there was no overall relationship of response with baseline susceptibility to efavirenz. In International Medical Press

5 Lopinavir/r therapy with respect to baseline phenotype and genotype Figure 4. Logistic regression models (solid curves) with 95% confidence intervals (dashed curves) for virological response (HIV RNA 4 copies/ml) at week 72 (a) 1. Estimated probability (b) 1. Estimated probability / /23 Baseline lopinavir phenotype 15/21 11/ Baseline lopinavir mutation score Observed categorical response rates are superimposed (black bars). (a) Baseline lopinavir phenotype; (b) Baseline lopinavir mutation score. addition, there was no significant relationship between virological response and baseline parameters describing NRTI phenotypic (PSS) or genotypic (GSS) susceptibility (data not shown). Of the 14 subjects with no active NRTIs in their regimen (as judged by the NRTI PSS), nine (64%) had HIV RNA <5 copies/ml at week 72. Analysis of virological response with respect to specific baseline protease mutations and mutation patterns The relationship of virological response ( 4 and 5 copies/ml) at weeks 24, 48 and 72 to the presence or absence of any single mutation associated with PI resistance was examined using Fisher s exact test. The analysis was not limited to those mutations within the lopinavir mutation score, but also included other amino acid positions previously shown to be associated with PI resistance [19] that appeared in at least four baseline samples. Among all single mutations, only the I54L/T/V mutation was statistically associated (P<.1) with diminished virological response to either 4 or 5 copies/ml at all time points examined. However, more than half the subjects whose baseline isolate had a mutation at amino acid position 2/8 2/6 54 experienced virological response. Thus, considered alone, the I54 mutation has limited value in predicting failure. To investigate whether patterns of multiple mutations were associated with lower response to lopinavir/r, the specific genotypes of the subset of 1 subjects with plasma HIV RNA >4 copies/ml at week 72 and whose baseline isolates had six or more mutations generally associated with PI resistance [6], were examined for common mutation patterns. Each baseline isolate from this subset of 1 subjects contained mutations at positions 82, 54 and 1 plus a median of 4 (range 3 8) additional mutations associated with PI resistance (Table 2). Although the presence of this mutation pattern (82/54/1 plus three others) was statistically significantly associated with a lack of virological response to 4 copies/ml (P<.5), an additional 1 subjects who had baseline isolates with a similar genotypic pattern experienced response at week 72 (Table 2). Thus, this pattern also has limited value for predicting virological failure. Results from univariate logistic regression analyses indicated that baseline phenotypic susceptibility to lopinavir was a predictor of response in this group of 2 subjects (P=.14). No other baseline factors were statistically associated with response in this subgroup; however, the product of the fold IC 5 of lopinavir and the fold IC 5 of efavirenz appeared to be a superior predictor (P<.1) of response compared to the susceptibility to lopinavir alone (data not shown). These results suggest that the number of PI mutations (for example, the lopinavir mutation score) more accurately predict response to lopinavir/r rather than the presence or absence of any single mutation or specific patterns of multiple mutations. Multivariate analysis of virological response with respect to baseline parameters In order to evaluate the effect of baseline resistance in the context of other baseline factors that can affect virological response to treatment, as defined by plasma HIV RNA 4 copies/ml, a multivariate (forward) stepwise logistic regression analysis was performed at weeks 24, 48 and 72. This analysis was limited to the 47 subjects for whom response data (success or failure) were available at all three time points. In addition, one subject with substantially reduced baseline susceptibility to efavirenz (2-fold) was also excluded. The analyses included a total of 19 phenotypic, genotypic and other baseline parameters present in the clinical database as well as lopinavir/r dose (see Methods). The final regression models are summarised in Table 3. The log fold lopinavir IC 5 was determined to be significantly associated with virological response at the week 24, 48 and 72 visits. Other Antiviral Therapy 7:3 169

6 DJ Kempf et al. Table 2. Subjects with baseline genotypes containing protease inhibitor mutations at positions 82, 54 and 1 Lopinavir Amino acid position No. Week 72 Fold IC Response 11.3 I I V L/V I A V 2 Response 12.6 I V V A V M 3 Response 3.38 I I I F/L V V T 4 Response 32.9 I R I L V V A M 5 Response 15.7 L/F L/I I V V A M 6 Response 13.9 I L V V G/S A 7 Response 21.6 I F M/L V V S/T I A M 8 Response 18.4 V R I V A M 9 Response 52.7 I M I I L V V V/A I/V M 1 Response 15.5 I I L/V L V V G/S A M 11 Failure 95.8 F/I R I I I L S L V A 12 Failure 64.1 F/V I V V A M 13 Failure 2.3 I V V V I A M 14 Failure 16.9 I I I V V I A 15 Failure 12.1 I V V I T M 16 Failure 62.9 I L V T L I A 17 Failure 44.2 I R I I L/V F/L V V A 18 Failure 34.2 I R I L V V A M 19 Failure 57.3 I/V K/R F I L V V G/S V/A I/V M 2 Failure 66.7 I I V I F M All subjects also had the L63P mutation, which is included in the lopinavir mutation score but not in some lists of protease inhibitor resistance mutations (see [6]). baseline factors associated with virological response at one or more of the three time points included years since HIV-positive diagnosis, baseline subject weight, NRTI PSS (2.5-fold cut-off), the number of new NRTIs, baseline viral load, log fold efavirenz IC 5, and the NRTI GSS. However, none of these parameters were consistently present in all of the models, suggesting that the susceptibility to lopinavir was the most important factor determining response. Interestingly, the odds ratio for the number of years since diagnosis was greater than 1 in the week 24 and week 48 models, indicating that subjects with a longer history of HIV infection experienced a superior virological response. However, the parameter years since first PI therapy initiated was not associated with response. Since the phenotypic susceptibility to lopinavir and the lopinavir mutation score are highly correlated [15], the two parameters could not co-exist within the stepwise logistic regression model. Therefore, to assess the effect of baseline protease genotype in the context of other parameters, a second set of stepwise logistic regression analyses, excluding the baseline lopinavir fold IC 5, were performed. The results, shown in Table 4, were similar to the first model. Thus, the lopinavir mutation score was present in the model as a significant predictor of response to all time points. Other parameters found in one or more models included the number of years since HIV-positive diagnosis, subject weight, NRTI PSS, the number of new NRTIs and baseline viral load. Discussion In this study, we examined the effect of baseline genotype and phenotype on the virological response to therapy with lopinavir/r plus efavirenz and NRTIs in multiple PI-experienced, NNRTI-naive subjects. In a previous study of the activity of lopinavir/r plus nevirapine and NRTIs in single PI-experienced subjects (Study M97-765), we did not observe a significant relationship between virological response and baseline resistance parameters, presumably because of the small number of isolates that displayed levels of reduced phenotypic or genotypic susceptibility (that is, >1- fold change in lopinavir IC 5 or lopinavir mutation score of >5) needed to impact response [13]. In the present study, the ranges of baseline phenotypic and genotypic susceptibilities were substantially greater than those found in single PI-experienced subjects and we observed a strong association between virological response and both baseline phenotypic susceptibility to lopinavir and baseline lopinavir mutation score, beginning at week 24 and continuing through to week 72. In univariate analyses, virological response was associated with baseline lopinavir phenotype either as a categorical or as a continuous variable. When used as a categorical variable, a statistically significant associ International Medical Press

7 Lopinavir/r therapy with respect to baseline phenotype and genotype Table 3. Stepwise logistic regression models of virological response* Week 24 Week 48 Week 72 Variable Odds ratio [95% CI] P value Odds ratio [95% CI] P value Odds ratio [95% CI] P value Log fold lopinavir IC 5.13 [.13.69].37.1 [.12.44].1.62 [.6.32].4 Years since HIV-positive 2.3 [ ] [ ].32 diagnosis Weight (kg).83 [.67.96].32.9 [.8 1.].69 NRTI PSS [ ].51 Number of new NRTIs 14.2 [ ].69 Log baseline viral load.32 [ ] [.4.64].18 Log fold efavirenz IC 5.17 [<.1.47].48 NRTI GSS 3.77 [.93 22].87 *Odds ratios reflect the change in the odds of experiencing virological response per 1 unit increase of the parameter. If the odds ratio is not presented, the indicated parameter was not found to be significantly associated with virological response at that time point (P>.15). PSS, phenotypic susceptibility score; GSS, genotypic susceptibility score (see [6]); NRTI, nucleoside reverse transcriptase inhibitor. ation was observed irrespective of the use of one or both of two apparent lopinavir/r susceptibility breakpoints (that is, 1- and 4-fold change in baseline IC 5 ). Indeed, even though the number of subjects whose baseline isolates displayed >4-fold reduced susceptibility was relatively small, the statistical correlation of lowered response with a >4-fold change in susceptibility appeared equal to or somewhat superior to the correlation of lowered response with a >1-fold change in baseline susceptibility. When phenotype and genotype were analysed as continuous variables in univariate logistic regression models, at least 5% of the subjects were predicted to maintain response for up to 72 weeks with up to a 35-fold change in lopinavir IC 5 or nine baseline lopinavir-associated mutations. Taken together, these results suggest that the activity of lopinavir/r, as used in this study, is likely to be highest in those subjects with <1-fold reduced baseline susceptibility to lopinavir, intermediate in those subjects whose baseline isolates display between 1- and 4-fold reduced susceptibility, and low in subjects whose baseline isolates display >4-fold reduced susceptibility. Similarly, activity appears to be high, intermediate and low in subjects with baseline lopinavir mutation scores of 5, 6 7 and 8 or more, respectively. The observed activity in this study undoubtedly reflects a partial contribution of efavirenz, as only 2/56 baseline isolates had in vitro susceptibility to efavirenz that was decreased relative to that of wild-type virus. Thus, the susceptibility range observed in this population might be different than those found with more highly experienced patients. However, the robust association between virological response and baseline lopinavir susceptibility establishes that lopinavir/r contributes substantially to the observed response in these subjects. In the univariate and multivariate analyses of response, both baseline susceptibility to lopinavir and the baseline lopinavir mutation score were significantly associated with the virological response at weeks 24, 48 and 72. As univariate predictors of response, baseline susceptibility to efavirenz and the baseline NRTI PSS were not statistically associated with virological response at week 24. However, in stepwise logistic regression analyses, both of these parameters appeared to add to the predictive power of the baseline lopinavir susceptibility at one or more time points. This observation suggests that all three anti- Table 4. Stepwise logistic regression models of virological response (excluding baseline lopinavir phenotype)* Week 24 Week 48 Week 72 Variable Odds ratio [95% CI] P value Odds ratio [95% CI] P value Odds ratio [95% CI] P value Lopinavir mutation score.57 [.32.88] [ ] [ ].9 Years since HIV-positive diagnosis 1.36 [ ] [ ].11 Weight (kg).89 [.78.98].34 NRTI PSS 6.19 [ ].84 Number of new NRTIs 12.7 [ ].35 Log baseline viral load.26 [.7.73].21 *Odds ratios reflect the change in the odds of experiencing virological response per 1 unit increase of the parameter. If the odds ratio is not presented, the indicated parameter was not found to be significantly associated with virological response at that time point (P>.15). PSS, phenotypic susceptibility score [6]; NRTI, nucleoside reverse transcriptase inhibitor. Antiviral Therapy 7:3 171

8 DJ Kempf et al. retroviral drug classes are, as anticipated, contributing to the overall virological response; however, in this study population, susceptibility to lopinavir is a strong determinant of successful viral suppression. Other factors found to be associated with response in the multivariate analyses included baseline viral load, the number of new NRTIs, baseline subject weight, and the number of years since HIV-positive diagnosis. The association of response with the time since HIV-positive diagnosis was positive (that is, subjects with longer intervals since diagnosis were more likely to have a successful response). This correlation is consistent with a survivorship effect, whereby individuals with favourable disease or host factors represent an increasing proportion of patients with longer disease duration. For example, patients with more aggressive viruses could experience more rapid treatment failure, resulting in earlier eligibility for this study than patients with more indolent viruses. Finally, those subjects with low body weight at baseline were more likely to experience virological response at weeks 24 and 48. This observation is presumably related to the inverse correlation of body weight and plasma lopinavir levels observed upon dosing lopinavir/r [2]. The observation of activity of lopinavir/r against viruses with substantially reduced phenotypic susceptibility to other PIs is consistent with its pharmacokinetic profile, in which mean trough levels are sustained far above (>75-fold when dosed at 4/1 mg twice-daily) the human serum-adjusted IC 5 [14] for wild-type virus. There is mounting evidence that trough levels are the best pharmacokinetic predictor of the activity of PIs in vivo [21 23]. The ratio of C trough /IC 5 (inhibitory quotient, IQ) calculated for non-enhanced PIs is 5 or less. Co-incidentally, a fourfold change in baseline susceptibility has been retrospectively associated with diminished response to most current PIs [1,2,4]. Indeed, in several studies, a fourfold change in susceptibility has been more strongly associated with diminished response than a 1-fold change [6], suggesting that relevant drug concentrations at the site of action (that is, within nascent viral particles) are fourfold or less above those needed to adequately suppress replication. The estimated average IQ for lopinavir/r for a virus with up to 4-fold reduced susceptibility is 2 or greater, consistent with the evidence of significant in vivo activity against viruses with 4-fold reduced susceptibility. The high rate of virological response in subjects with up to five mutations associated with reduced phenotypic susceptibility to lopinavir compares favourably with previous studies with single or dual PI regimens [4,7]. For example, the presence of one or two baseline mutations was associated with diminished response of PI-experienced patients to therapy with ritonavir/saquinavir, and none of the subjects with three or more major protease mutations had a complete virological response [7]. Within the study M baseline viruses, mutations that produce changes in phenotypic susceptibility to ritonavir and indinavir produce a similar change in phenotype to lopinavir [15]. Therefore, the increased activity of lopinavir/r in subjects with mutant virus is likely to be a consequence of the high plasma trough levels of lopinavir, in contrast to the mechanism of action observed with PIs such as tipranavir, in which activity against mutant viruses results from a low incidence of in vitro phenotypic cross-resistance [24]. As such, the lopinavir mutation score should predict response to the extent that this measure correlates with reduced phenotypic susceptibility to lopinavir. This correlation has been shown in a previous analysis of 112 viruses from PI-experienced patients, in which the median (mean) IC 5 of lopinavir against viruses with five and seven lopinavir mutations was 4.2-fold (6.8-fold) and 16.-fold (2.6-fold), compared to wild type HIV [15]. Indeed, in this study, the response rates at week 72 in subjects with 5 and 6 7 baseline mutations (91 and 71%, respectively) were similar to the response rates in subjects whose baseline isolates displayed up to 1-fold and 1- to 4-fold reduced susceptibility to lopinavir (93 and 73%, respectively). The correlation between these response rates suggests that the lopinavir mutation score provides a reasonable surrogate, or virtual phenotype for lopinavir, at least for patients with experience with those prior PIs encountered in this study (indinavir, nelfinavir, saquinavir and ritonavir). The results of these analyses are highly relevant for the interpretation of HIV resistance testing with regard to lopinavir/r. With the increasing availability of both genotypic and phenotypic assays, the proper interpretation of the results of these in vitro tests is critical for maximizing the care of patients who have already failed one or more therapeutic regimens and have fewer treatment options available. Although access to either genotypic or phenotypic information has been shown in prospective studies to improve the response to salvage therapy [9,1], there is little information currently available on the appropriate interpretation of such tests. Thus, definition of the clinically relevant phenotypic and genotypic susceptibility ranges for new antiretrovirals, as well as combinations of existing antiretrovirals, is needed. This is particularly important for combinations in which the plasma levels of one or more drugs in the combination are substantially altered through metabolic interactions. In this study we observed a uniformly high response in association with 1-fold phenotypic changes and 5 mutations in the lopinavir mutation score, and generally poor response with >4-fold phenotypic changes and >7 lopinavir International Medical Press

9 Lopinavir/r therapy with respect to baseline phenotype and genotype mutations; intermediate responses were observed between these ranges. These values may provide guidance for interpretation of the results of both phenotypic and genotypic resistance testing. However, ranges of susceptibility to lopinavir/r might differ from those above in more highly experienced patient populations where a second new drug is not available, as it was in this study. The virological response in this study with respect to the baseline lopinavir mutation score also provides insight into the genetic barrier to in vivo resistance to lopinavir/r in treatment-naive subjects. The average change in lopinavir IC 5 per mutation contained in the lopinavir mutation score was <twofold [15], although some mutations not included in the lopinavir mutation score might also produce significant changes in susceptibility [25]. The modest phenotypic change per mutation for lopinavir is relatively typical of the PI class; however, the high plasma drug levels of lopinavir are likely to mandate a greater number of mutations required to overcome drug suppression than required for non-enhanced PI regimens. In contrast to most PIs, NNRTIs may sustain plasma concentrations that are well above their respective IC 5 throughout the dosing interval. However, the change in susceptibility per mutation is often large, thus few mutations are required to overcome even high plasma levels of drug. The results of the present analysis suggest that the potent activity of lopinavir/r, in combination with stavudine and lamivudine, observed in antiretroviralnaïve subjects [12,26] is, in part, a consequence of the high genetic barrier to resistance to lopinavir/r in vivo. The existence of a high genetic barrier is also supported by the lack of development of resistance to lopinavir in antiretroviral-naive subjects with plasma HIV RNA >5 copies/ml [27]. Although the lopinavir mutation score was a good predictor of virological response in this study, the investigation of specific mutation patterns associated with diminished response to lopinavir/r is of interest. The pattern of mutations common in those subjects with six or more baseline mutations who failed therapy (positions 82, 54 and 1 plus a median of four additional PI mutations) suggests one pathway to resistance to lopinavir/r in subjects who have failed prior therapy with resistance to other PIs. The implications of this pattern on the selection of de novo resistance to lopinavir/r in PI-naive subjects are unknown. Further, an equal number of subjects with very similar mutation patterns experienced virological response ( 4 copies/ml) through to week 72. Within this group of 2 subjects with similar baseline genotype, the baseline phenotypic susceptibility to lopinavir, as well as the combined susceptibility to both lopinavir and efavirenz, appeared to influence virological response. These results suggest that for complex genotypic patterns involving multiple protease mutations, viral phenotype may provide additional predictive power over genotype. In conclusion, the analyses performed in this study illustrate two critical considerations for interpreting HIV resistance data. First, clinically meaningful drug resistance is defined not only by virological characteristics, but also by the in vivo pharmacokinetic properties of individual drugs. Thus, high drug levels may overcome resistance mutations, much like the activity of antibiotics against resistant bacteria at a site of infection where greatly elevated local drug concentrations are possible (for example, the urinary tract). Second, resistance to protease inhibitors should be viewed as a continuum, with genotypic and phenotypic ranges that correspond to a range of virological responses. The correlation of virological response with baseline lopinavir phenotype and mutation score through 72 weeks observed in this study suggests that HIV resistance testing may be beneficial in optimising the use of lopinavir/r in PI-experienced patients. The data included in this report provide the basis for clinical interpretation of such assays. Acknowledgements The investigators and study site co-ordinators for Study M are gratefully acknowledged: Stephen Becker (Pacific Horizon Medical Group), Frank Bergmann (Charite-Humboldt University Berlin), Giampiero Carosi (University di Brescia), Nathan Clumeck (Hospital Universitaire St Pierre Brussels), Sven Danner (Academic Medical Center Amsterdam), Pierre-Marie Girard (Hopital Rothschild), David Ho (Aaron Diamond AIDS Research Center), Adriano Lazzarin (Ospedale San Raffaele ), Guiseppe Pantaleo (Hopital de Beaumont), Jurgen Rockstroh (Medizinische Klinik und Poliklinik der Uni Bonn) and Roland Tubiana (Hospital Pitie-Salpetriere). References 1. Deeks SG, Hellmann NS, Grant RM, Parkin NT, Petropoulos CJ, Becker M, Symonds W, Chesney M & Volberding PA. Novel four-drug salvage treatment regimens after failure of a human immunodeficiency virus type 1 protease inhibitor-containing regimen: antiviral activity and correlation of baseline phenotypic drug susceptibility with virologic outcome. Journal of Infectious Diseases 1999; 179: Harrigan PR, Hertogs K, Verbiest W, Pauwels R, Larder B, Kemp S, Bloor S, Yip B, Hogg R, Alexander C & Montaner JS. Baseline HIV drug resistance profile predicts response to ritonavir-saquinavir protease inhibitor therapy in a community setting. AIDS 1999; 13: Piketty C, Race E, Castiel P, Belec L, Peytavin G, Si-Mohamed A, Gonzalez-Canali G, Weiss L, Clavel F & Kazatchkine MD. Efficacy of a five-drug combination including ritonavir, saquinavir and efavirenz in patients Antiviral Therapy 7:3 173

10 DJ Kempf et al. who failed on a conventional triple-drug regimen: phenotypic resistance to protease inhibitors predicts outcome of therapy. AIDS 1999; 13:F71 F Para MF, Glidden DV, Coombs RW, Collier AC, Condra JH, Craig C, Bassett R, Leavitt R, Snyder S, McAuliffe V & Boucher C. Baseline human immunodeficiency virus type 1 phenotype, genotype, and RNA response after switching from long-term hard-capsule saquinavir to indinavir or soft-gel-capsule saquinavir in AIDS Clinical Trials Group protocol 333. Journal of Infectious Diseases 2; 182: Miller V, Cozzi-Lepri A, Hertogs K, Gute P, Larder B, Bloor S, Klauke S, Rabenau H, Phillips A & Staszewski S. HIV drug susceptibility and treatment response to mega- HAART regimen in patients from the Frankfurt HIV cohort. Antiviral Therapy 2; 5: DeGruttola V, Dix L, D Aquila R, Holder D, Phillips A, Ait-Khaled M, Baxter J, Clevenbergh P, Hammer S, Harrigan R, Katzenstein D, Lanier R, Miller M, Para M, Yerly S, Zolopa A, Murray J, Patick A, Miller V, Castillo S, Pedneault L & Mellors J. The relation between baseline HIV drug resistance and response to antiretroviral therapy: re-analysis of retrospective and prospective studies using a standardized data analysis plan. Antiviral Therapy 2; 5: Zolopa AR, Shafer RW, Warford A, Montoya JG, Hsu P, Katzenstein D, Merigan TC & Efron B. HIV-1 genotypic resistance patterns predict response to saquinavir-ritonavir therapy in patients in whom previous protease inhibitor therapy had failed. Annals of Internal Medicine 1999; 131: Condra J, Holder JJ, Schleif WA, Bakshi K, Danovich RM, Graham DJ, Shivaprakash M, Holmes K, Saah AJ, Leavitt RY, Chodakewitz JA & Emini EA. Genetic correlates of virological response to an indinavir-containing salvage regimen in patients with nelfinavir failure. Antiviral Therapy 1999; 4: Baxter JD, Mayers DL, Wentworth DN, Neaton JD, Hoover ML, Winters MA, Mannheimer SB, Thompson MA, Abrams DI, Brizz BJ, Ioannidis JPA & Merigan TC. A randomized study of antiretroviral management based on plasma genotypic antiretroviral resistance testing in patients failing therapy. AIDS 2; 14:F83 F Clevenbergh P, Durant J, Halfon P, del Giudice P, Mondain V, Montagne N, Schapiro JM, Boucher CAB & Dellamonica P. Persisting long-term benefit of genotypeguided treatment for HIV-infected patients failing HAART. The Viradapt study: week 48 follow-up. Antiviral Therapy 2; 5: Cohen C, Hunt S, Sension M, Farthing C, Conant M, Jacobson S, Nadler J, Verbiest W, Hertogs K, Ames M, Rinehart A & Graham N. Phenotypic resistance testing significantly improves response to therapy: a randomized trial (VIRA 31). 7th Conference on Retroviruses and Opportunistic Infections. San Francisco, California, 3 Jan 2 Feb 2, Abstract Murphy RL, Brun S, Hicks C, Eron JT, Gulick R, King M, White AC, Benson C, Thompson M, Kessler HA, Hammer S, Bertz R, Hsu A, Japour A & Sun E. ABT- 378/ritonavir plus stavudine and lamivudine for the treatment of antiretroviral-naive adults with HIV-1 infection: 48-week results. AIDS 21; 15:F1 F Benson CA, Deeks S, Brun SC, Gulick R, Eron JJ, Kessler HA, Murphy RL, Hicks C, King M, Wheeler D, Feinberg J, Stryker R, Sax P, Riddler S, Thompson M, Real K, Hsu A, Kempf D, Japour A & Sun E. 48 Week safety and antiviral activity of lopinavir/ritonavir plus nevirapine and two nucleoside analogs in human immunodeficiency virus protease inhibitor-experienced patients. Journal of Infectious Diseases 22; 185: Molla A, Vasavanonda S, Kumar G, Sham HL, Johnson M, Grabowski B, Denissen JF, Kohlbrenner W, Plattner JJ, Leonard JM, Norbeck DW & Kempf DJ. Human serum attenuates the activity of protease inhibitors toward wildtype and mutant human immunodeficiency virus. Virology 1998; 25: Kempf DJ, Isaacson JD, King MS, Brun SC, Xu Y, Real K, Bernstein BM, Japour AJ, Sun E & Rode RA. Identification of genotypic changes in human immunodeficiency virus protease that correlate with reduced susceptibility to the protease inhibitor lopinavir among viral isolates from protease inhibitor-experienced patients. Journal of Virology 21; 75: Petropoulos CJ, Parkin NT, Limoli KL, Lie YS, Wrin T, Huang W, Tian H, Smith D, Winslow GA, Capon DJ & Whitcomb JM. A novel phenotypic drug susceptibility assay for human immunodeficiency virus type 1. Antimicrobial Agents and Chemotherapy 2; 44: Mantel N. Assessing laboratory evidence for neoplastic activity. Biometrics 198; 36: Tukey JW, Ciminera JL & Heyse JF. Testing the statistical certainty of a response to increasing doses of a drug. Biometrics 1985; 41: Hirsch MS, Brun-Vezinet F, D Aquila RT, Hammer SM, Johnson VA, Kuritzkes DR, Loveday C, Mellors JW, Clotet B, Conway B, Demeter LM, Vella S, Jacobsen DM & Richman DD. Antiretroviral drug resistance testing in adult HIV-1 infection - Recommendations of an International AIDS Society-USA panel. Journal of the American Medical Association 2; 283: Bertz R, Lam W, Hsu A, Granneman GR & Sun E. Effects of gender, race, age and weight on the pharmacokinetics of lopinavir after single-dose Kaletra. 2nd International Workshop on Clinical Pharmacology of HIV Therapy. Noordwijk, the Netherlands, 2 4 April 21, Abstract 3, Kakuda TN, Page LM, Anderson PL, Henry K, Schacker TW, Rhame FS, Acosta EP, Brundage RC & Fletcher CV. Pharmacological basis for concentrationcontrolled therapy with zidovudine, lamivudine, and indinavir. Antimicrobial Agents and Chemotherapy 21; 45: Durant J, Clevenbergh P, Garraffo R, Halfon P, Icard S, Del Giudice P, Montagne N, Schapiro JM & Dellamonica P. Importance of protease inhibitor plasma levels in HIV-infected patients treated with genotypicguided therapy: pharmacological data from the Viradapt Study. AIDS 1333; 14: Kempf D, Hsu A, Jiang P, Rode R, Hertogs K, Larder B, Zolopa A, Shulman N, Havlir D, Gallant J, Race E, Boller S, Swerdlow J, Jasinsky O, Renz C & Sun E. Response to ritonavir (RTV) intensification in indinavir (IDV) recipients is highly correlated with virtual inhibitory quotient. 8th Conference on Retroviruses and Opportunistic Infections. Chicago, Illinois, 4 8 February, 21, Abstract Larder BA, Hertogs K, Bloor S, van den Eynde C, DeCian W, Wang YY, Freimuth WW & Tarpley G. Tipranavir inhibits broadly protease inhibitor-resistant HIV-1 clinical samples. AIDS 2; 14: Parkin NT, Chappey C, Maranta M, Whitehurst N & Petropoulos CJ. Genotypic and phenotypic analysis of a large database of patient samples reveals distinct patterns of protease inhibitor cross-resistance. Antiviral Therapy 21; 6: Ruane P, Mendonca J, Timerman A, Cernohous P, Bauer E, Bernstein B & Sun E. Kaletra vs. nelfinavir in antiretroviral-naive subjects: week 6 comparison in a phase III, blinded, randomized clinical trial. 1st IAS Conference on HIV Pathogenesis and Treatment. Buenos Aires, Argentina, 8 11 July 21, Abstract Kempf D, Bernstein B, King M, Cernohous P, Moseley J, Gu K, Bauer E & Sun E. Comparison of the emergence of genotypic resistance over 6 weeks of therapy with lopinavir/ritonavir (Kaletra) or nelfinavir plus d4t/3tc. 1st IAS Conference on HIV Pathogenesis and Treatment. Buenos Aires, Argentina, 8 11 July 21, Abstract 129. Received 11 December21; accepted 27 March International Medical Press