Objective: To assess phenotype susceptibility testing (PHENO) with standard of care (SOC) to improve antiretroviral therapy.
Design: A prospective, multicenter study of 238 patients taking a stable antiretroviral regimen for > 6 months, with one or two protease inhibitors (PI) and entry HIV RNA > 400 copies/ml.
Method: Patients were randomized to receive or not receive PHENO results for selecting antiretroviral regimens. Primary outcome was HIV RNA measures.
Results: At baseline, median CD4 cell count was 277 × 106 cells/l and HIV RNA was 10 000 copies/ml; 76% had not taken a non-nucleoside reverse transcriptase inhibitor drug (NNRTI). There were significant differences between the groups in selection of baseline nucleoside reverse transcriptase inhibitor (NRTI). At month 6, reduction in HIV RNA was 0.71 and 0.69 log10 copies/ml for PHENO and SOC, respectively; the proportion with < 400 copies/ml (48%) was the same for both groups. No differences were seen at month 12. In a subgroup with resistance to four or more PI, 50% of the PHENO versus 17% of the SOC had HIV RNA < 400 copies/ml at month 6 (P = 0.02). The number of NNRTI and PI, but not NRTI, in the regimen that were active by phenotype at baseline was a strong independent predictor of viral suppression (P < 0.006). Use of alternative NRTI sensitivity cut-offs improved their predictive value.
Conclusions: Although virological outcome was similar in both groups, the potential benefit of PHENO was seen in patients with more resistant virus. Lack of appropriate cut-offs may have partially accounted for the lack of benefit from PHENO and demonstrated the need to identify clinically relevant sensitivity cut-off points.
From the aUniversity of California, San Diego
bCalifornia Collaborative Treatment Group Data and Biostatistical Unit, San Diego
cSanta Clara Valley Medical Center, San Jose and Stanford University School of Medicine, Stanford
dViroLogic Inc., South San Francisco
eDavid Geffen School of Medicine at the University of California, Los Angeles
fHarbor-UCLA Medical Center, Torrance
gUniversity of California, Irvine School of Medicine
hKeck School of Medicine of the University of Southern California, Los Angeles
iVeterans Affairs San Diego Healthcare System, La Jolla, California
jUniversity of Texas, Dallas, Texas, USA.
*Present address: Roche Molecular Systems, Pleasanton, California, USA.
†See Appendix for study members.
Received 27 May, 2004
Revised 30 September, 2004
Accepted 28 October, 2004
Requests for reprints to: Dr R. H. Haubrich, UCSD Antiviral Research Center and the CCTG Data and Biostatistical Unit, 150 West Washington St, Suite 100, San Diego, California 92103, USA. Email: email@example.com
As the number of antiretroviral treatments (ART) approved for the treatment of HIV increases, more options become available for patients who have failed previous regimens. Drug resistance, adherence, toxicity, and pharmacological issues complicate selecting a new regimen. Although there are many potential regimens, the number of those that are likely to achieve virological suppression is greatly limited by cross-resistance to drugs within the classes of agent that the patient has previously received. Drug resistance testing is currently recommended for patients failing one or more prior ART regimens [1–3]. Retrospective analyses have validated the association between drug resistance and diminished response to the next treatment and several prospective studies have shown a benefit when resistance assays are used to aid in the selection of the new drug regimen [4–14]. These studies have generally shown short-term virological benefit, since most studies were for 6 months or less.
Several prospective studies have demonstrated that the addition of a genotype test improves the outcome of antiretroviral therapy [9–11]. Genotype resistance testing has the advantages of lower cost, more rapid test completion and wider availability . The benefit of resistance testing can be augmented by expert interpretation of the test results, especially with genotype testing, which may reveal complex patterns of mutations in patients who have had several prior ART regimens . Two prospective studies demonstrated improved virological responses when expert interpretation was provided in addition to genotyping [9,10]. Phenotype assays have the potential advantage of providing a quantitative assessment of the degree of resistance and are reported in a format that is more familiar to clinicians . Therefore, interpreting a phenotype might be less complicated than interpreting a genotype, especially for patients with high levels of resistance.
In order to define better the role of phenotype testing, the California Collaborative Treatment Group (CCTG) 575 study was designed to assess the addition of phenotype testing (PHENO) to a structured clinical history [standard of care (SOC)] to guide selection of new and subsequent regimens after failure of at least one previous ART regimen. The primary objective of the study was to determine if the clinical management of ART was improved by PHENO, as assessed by changes in HIV RNA at months 6 and 12.
CCTG 575 was a seven-center, two-arm, randomized clinical strategy study over 12 months to evaluate the virological benefit of using PHENO as an adjuvant to select and adjust ART. Entry criteria included at least 6 months of previous ART, exposure to no more than two prior protease inhibitors (PI), failure of the current regimen (defined by HIV RNA > 400 copies/ml), and a stable regimen (defined as no change for at least 4 weeks prior to screening). Patients with acute infections, active cancers, a resistance assay performed within 6 months, or involvement in another investigational study that dictated the regimen were excluded. Within 30 days of study entry, two measures of HIV RNA [COBAS AMPLICOR HIV-1 MONITOR (standard and ultrasensitive assays); Roche Molecular Systems, Pleasanton, California, USA] and CD4 cell count and a phenotype assay (PhenoSense HIV, Virologic, Inc., South San Francisco, California, USA)  were carried out for all patients. Phenotype assays were provided without cost by ViroLogic. After stratification for prior PI treatment and entry HIV RNA (cut-off point at 100 000 copies/ml), patients were randomized to either the PHENO group, in which the results of the phenotype assay were provided to the clinician selecting the baseline regimen, or to the SOC group, in which the phenotype assay was performed, but not made available.
Randomization was performed centrally at the CCTG data center. At baseline, all patients started a new ART regimen selected by the primary provider based on their independent interpretation of the ART history and, for the PHENO group, phenotypic susceptibility. Accuracy and reproducibility of phenotype interpretation was encouraged by monthly discussions of select cases by CCTG investigators; however, individual expert interpretation of assays was not provided. The phenotype results expressed drug susceptibility as the fold-change (FC) in the drug concentration required to inhibit patient virus replication by 50% (IC50) compared with that of a wild-type reference virus (FC = IC50 patient/IC50 control). The boundary for distinguishing reduced susceptibility was the IC50 value that represented a clear increase compared with wild-type HIV based on the precision (reproducibility) of the assay. During most of the study, a FC value of 2.5 was considered the cut-off point for reduced susceptibility for all drugs . HIV RNA and CD4 cell count measurements were performed at a central laboratory every 2 months. Regimen failure subsequent to baseline assessment was defined by the study site clinician and treatment switches were allowed at any time. Phenotype assays were carried out whenever HIV RNA was > 400 copies/ml, but as at baseline, the results were only reported for those participants in the PHENO group. The study duration was 12 months and all patients signed informed consent documents approved by each study site's Institutional Review Board. The study was approved by the Human Research Protections Program at the University of California, San Diego.
The primary endpoints of the study were the degree of viral suppression as assessed by change in HIV RNA from baseline to month 6 and month 12 using an area-based measure (average area under the curve minus baseline, calculated using the trapezoidal rule)  and the proportion of patients with HIV RNA below detection (< 400 copies/ml) at months 6 and 12. Primary analyses were planned as intent to treat without imputation for missing values. Power calculations were performed using a two-sample t-test approximation and a two-tailed α value of 0.05. The inclusion of 100 patients per arm would provided 80% power to detect a difference of 0.375 in HIV RNA measured as log10 copies/milliliter between the treatment strategies. This number would also provide 81% power to detect at least a 20% difference in the proportion of patients with undetectable HIV RNA at months 6 or 12. An interim analysis was performed when the study completed 75% of targeted accrual and patients had at least 6 months of follow-up. The O’Brien–Fleming boundaries were used to define stopping rules for the study. The interim analysis was reviewed by a data and safety monitoring board that recommended continuing the study as planned. Descriptive statistics were used to examine the baseline patient characteristics, including duration and types of previous ARV therapy. HIV RNA values were log10 transformed for analysis. Changes in HIV RNA at months 6 and 12 were analyzed using two-sample t-tests of an area-based measure, which utilized all available HIV RNA values in the assessed interval and was expressed as the average change. Comparison of proportions used chi-squared tests. Planned secondary analyses included HIV RNA changes and the proportion of patients with HIV RNA < 50 copies/ml. Post-hoc secondary analyses evaluated subgroups based on demographic and prior treatment categorizations. No correction was made for multiple comparisons.
Multivariate logistic regression models were used to determine the independent effect of regimen phenotypic susceptibility on the proportion of patients who achieved HIV RNA < 400 copies/ml at month 6 after controlling for baseline viral load and CD4 cell count. Regimen components were separated into classes and various cut-off points were evaluated to assess the impact on prediction of response. Analyses were performed with SAS version 8.1 (SAS Institute, Cary, North Carolina, USA) and SPSS version 10.0.7 statistical software (SPSS, Chicago, Illinois, USA).
Overall, 256 patients from seven university-affiliated clinics were randomized and 238 (119 in each group) completed the baseline evaluation between November 1998 and December 1999. By month 12, 178 patients remained on the study, 95 in the PHENO group and 83 in the SOC group. The time to premature discontinuation tended to be longer in the PHENO group, but the difference was not significant (340 days PHENO and 317 days SOC; P = 0.11 by log rank test). The groups were well balanced for baseline characteristics and there was no significant difference between groups (Table 1). Overall, the patients were 85% male, had a median CD4 cell count of 277 × 106 cells/l, had a median HIV RNA of 4.0 log10 copies/ml, and 58% were non-white. The median duration of nucleoside reverse transcriptase inhibitor (NRTI) use was 36 months and PI had been used for a median of 18 months. Most patients (76%) were naive to non-nucleoside reverse transcriptase inhibitors (NNRTI) and had received one prior PI-containing regimen (55% had taken nelfinavir). At baseline, 22% of the patients switched to a regimen with NRTI alone or NRTI plus NNRTI; 42% switched to regimens of PI plus NRTI, and 36% selected regimens with all three classes (Table 1). Of the PI-containing regimens, half contained a single PI and half two, with the majority (90%) of those taking two PI drugs using ritonavir as one.
Susceptibility of viral isolate to the baseline regimen
The participants’ primary care providers were asked to declare the proposed baseline regimen prior to receipt of the phenotype that was measured at screening (approximately day −30). At baseline, providers for participants in the PHENO group had the results of the phenotype assay to assist with regimen selection and 79% of these providers made changes to the proposed regimen. Notably, 50% of the providers in the SOC arm also made changes to the proposed regimen even though they had no additional information. There was a significant difference (P < 0.001) between the two groups in the proportion of providers who made changes to the proposed baseline regimen. Changes from the proposed to the actual regimen were about twice as frequent on the PHENO arm compared with the SOC arm for all drug classes. Adjustments were made to the PI component in 48% versus 22%, to the NNRTI component in 43% versus 23%, and to the NRTI component in 62% versus 33% in the PHENO group versus the SOC group, respectively.
The use of phenotype results allowed selection of a regimen to which the viral isolate was susceptible in the majority of participants in the PHENO arm of the study. Using an IC50 FC < 2.5 to define susceptibility, 83% of patients in the PHENO group versus 55% of patients in the SOC group had virus that was susceptible to three or more drugs in the baseline regimen (P < 0.001). The proportion of patients selected to receive each PI or NNRTI was not different between the two groups, but there were significant differences in the choice of NRTI used in the baseline regimen. Patients in the PHENO group were more likely to receive didanosine (62% versus 43%; P < 0.01) and stavudine (49% versus 33%; P < 0.01), and less likely to receive abacavir (13% versus 39%; P < 0.01).
Analysis of virological outcomes
The primary analysis compared the change in HIV RNA from baseline to month 6 (n = 211) and month 12 (n = 178) using an area-based measure (Table 2). The mean reduction in of HIV RNA at month 6 was 0.71 log10 copies/ml (SD, 0.7) for the PHENO group versus 0.69 log10 copies/ml (SD, 0.7) for the SOC group. Both groups had mean reductions in log10 HIV RNA of 0.71 log10 copies/ml (SD, 0.7) at month 12. There were no significant group differences. The proportion of patients with HIV RNA < 400 copies/ml at month 6 was the same in both groups (48%) and there was not a significant difference at month 12 (45% in the PHENO group versus 46% in the SOC group).
Secondary measures of outcome included area-based HIV RNA changes at months 6 and 12 and the proportion of patients with undetectable viral load using the HIV RNA assay with a sensitivity of 50 copies/ml. There were no significant differences between the PHENO and SOC groups at either month 6 or month 12 with these measures (Table 2). Similarly, increases in CD4 cell count cells (21–31 × 106 cells/l at month 6 and 39–41 × 106 cells/l at month 12) did not differ between the groups.
Several sensitivity analyses were carried out to attempt to account for the possible effect of differential drop-out on the study results. More participants from the SOC group dropped out of the study than did from the PHENO group; it is possible that SOC participants, particularly those who were experiencing virological failure, dropped out in order to receive the results of resistance testing. Analyses were done using the last HIV RNA observation carried forward or missing observations imputed as failure. Using these definitions, the proportion of patients with HIV RNA values < 400 copies/ml at months 6 and 12 were not different between the groups (P > 0.5; data not shown).
In order to explore if phenotype assays might be most useful in patients with higher levels of resistance (and those that might require a boosted PI), subgroup analyses were carried out. In a subgroup of participants (post hoc; n = 44) with baseline phenotypic profiles demonstrating resistance to four or more PI (defined as IC50 FC > 2.5), 50% of the PHENO group versus 17% of the SOC patients had HIV RNA < 400 copies/ml at month 6 (P = 0.02). Therefore, phenotype testing appeared to benefit greatly a subgroup of patients with broad cross-resistance to PI on entry to the study. Patients who received dual PI regimens (ritonavir-boosted PI) also tended to respond better in the PHENO arm than in the SOC arm. In this subgroup, the mean fall in HIV RNA at month 6 was 0.9 log10 copies/ml for the PHENO group versus 0.6 log10 copies/ml for the SOC group (P = 0.06; n = 65). Furthermore, among patients with a greater duration of prior treatment (> 60 months of total therapy; n = 46), PHENO patients were more likely to have HIV RNA values < 400 copies/ml at month 6 than SOC patients (69% versus 40%; P = 0.05). Subgroups based on other factors, such as prior use of nelfinavir, naive status with NNRTI, and number of prior PI used (none versus one or two), did not differ in HIV RNA response between the strategy groups.
Phenotypic susceptibility as a predictor of virological responses
Given that more participants in the PHENO group received regimens with three or more drugs to which their viral isolate was susceptible yet there was no difference in HIV RNA changes between the groups, the predictive value of the phenotype assay for virological response was evaluated. The proportion of patients with HIV RNA < 400 copies/ml at month 6 was modeled using multiple logistic regression after accounting for the baseline viral load and CD4 cell count. Each regimen was interrogated as to whether the drugs would have activity against the virus using a FC cut-off point of 2.5. Regimen drugs with a FC value < 2.5 were given a score of 1 and those drugs with a FC > 2.5 were scored 0. The three classes of ARV agent were entered into the model as separate variables in order to assess best their contribution to prediction of virological outcome. The number of regimen NNRTI and PI to which the virus was susceptible was a significant predictor of month 6 virological response even after taking into account baseline covariates, while the number of NRTI was not (model 1 in Table 3). Therefore, even though the phenotype was a potent predictor of virological response and patients in the PHENO group had significantly more drugs in their regimen to which the virus was predicted to be susceptible, the PHENO group failed to have a better outcome.
The discordance was evaluated by comparing which virus-susceptible drugs were present in the regimens. Since the NRTI phenotype was not an independent predictor of response using a cut-off point of 2.5, the interpretation of NRTI susceptibility was suspected to be problematic. In fact, two NRTI, stavudine and didanosine, accounted for the major difference between groups in the number of drugs in the baseline regimen to which HIV appeared to be susceptible. While overall, 83% of the PHENO versus 55% of the SOC participants had virus susceptible to three or more drugs in the regimen (P < 0.001), when stavudine and didanosine were excluded from the count, the proportions were not significantly different (25% PHENO versus 22% SOC; P = 0.4). Since the NRTI phenotype was not an independent predictor of response using a FC cut-off point of 2.5, the use of this cut-off point in this study might have led to excessive use of stavudine and didanosine (when the virus was not susceptible) and may have subverted the value of the phenotype assays.
The distributions of FC values for stavudine and didanosine in experienced patients were narrower that those of other NRTI, suggesting that lower thresholds for reduced susceptibility might be more accurate. To explore alternative cut-off points for stavudine and didanosine, additional models were constructed using FC susceptibility cut-off points of 2.0 or 1.5 (Table 3). In these models, the odds ratios of achieving HIV RNA < 400 copies/ml improved but were still not significant. One further model used the zidovudine and abacavir susceptibility (FC < 2.5) as surrogates for stavudine and didanosine susceptibility. Using these substitutions, the odds ratio for the NRTI approached significance (1.3; P = 0.18). Consequently, alternative cut-off points for stavudine and didanosine might have better informed the selection of drugs using the phenotype assay.
Based on the results of numerous prospective and retrospective studies demonstrating the predictive ability and the clinical utility of resistance assays [4–14], ART drug resistance testing has become the standard of care to assist in the selection of a new regimen after the failure of one or more regimens [1,2]. Overall, the current study did not find a significant benefit of phenotype testing, but it elucidated important issues in the use of resistance tests. Patients with greater prior treatment experience and a broader degree of PI resistance had better virological response if assigned to phenotypic monitoring. The CERT study  also demonstrated a benefit of phenotype testing only in the subgroup of patients with greater prior treatment. In the CERT study, the risk of virological treatment failure for participants who had received more than four prior ART was 1.7 times greater in the SOC group than in the PHENO group. Patients with prior NNRTI treatment also did better with phenotype testing than with no testing.
The lack of clinical benefit was not a consequence of an inability of the phenotype assay to define HIV resistance accurately. In CCTG 575 and other studies using this phenotype assay, the number of drugs in the baseline regimen to which the patient's HIV isolate was susceptible clearly predicted virological response [6–8,20–22]. The lack of benefit of phenotype testing in this study may have been partially a result of inaccurate clinical cut-offs for some drugs, resulting in suboptimal ART selection.
The cut-off used for this study to define resistance was an IC50 FC of 2.5 for all drugs. This standard was based on the reproducibility of the ViroLogic PhenoSense assay. Subsequent data have suggested that clinical cut-offs, derived from analyses that relate HIV RNA response to FC values, represent better metrics for phenotype interpretation [20–22]. More recent analyses have revealed that didanosine and stavudine have lower FC thresholds (1.7) and abacavir remains active up to a FC of 4.5 [23,24]. Based on the inaccurate 2.5 cut-off, patients in the PHENO group were much more likely to use stavudine and didanosine (49% versus 19% in the SOC group; P < 0.01) and less likely to use abacavir (13% versus 39% in the SOC group; P < 0.01) and these differences in ARV choices may have reduced the benefit of resistance testing in the PHENO group. The improvement in the prediction of virological response with the use of alternative NRTI phenotype cut-off points supports the assumption that greater utility of the phenotype test could be obtained using more refined clinical cut-offs. Therefore, optimal use of the resistance test must be with reference to clinically based cut-off points determined from virological response to a regimen.
Previous studies of both genotype resistance testing and phenotype testing have shown virological benefits, though in some only in subgroups [9–14,25–27]. Two randomized studies compared genotype testing with phenotype testing [12,14]; although overall no difference between assays were noted, one subgroup analysis showed that genotype testing was superior to phenotype  and the other showed the converse . Several previous strategy studies of resistance testing have emphasized the importance of accurate interpretation of the assay results. In the GART study, a group of expert virologists provided genotype interpretation and made regimen recommendations to clinicians . Overall, the study showed a benefit of resistance testing when given with expert advice. Analyses from subgroups suggested that the full benefit of resistance testing was not realized without the aid of expert interpretation. Patients from clinics that had access to genotypes, but did not follow the GART recommendations, did not respond as well as those who followed the recommendations . The Havana study also demonstrated the importance of interpretation of resistance assays . This study used a factorial design to evaluate genotype testing or not, crossed with expert advice or not. In a per-protocol analysis, patients who received expert advice were more likely to have HIV RNA < 400 copies/ml at week 24 than those without advice, even after accounting for genotype testing. The magnitude of the effect of expert advice was similar to genotype testing itself in improving the response to the regimen.
Other factors may have influenced the results of this study and are potential hazards of open clinical strategy trials. In this study, 68% of the patients enrolled had failed a single PI (55% nelfinavir) and 76% were naive to NNRTI. The resistance profile of patients failing nelfinavir is straightforward, predictable and had been reported by several groups during the course of CCTG 575 [13,29]. As clinical studies and clinical practice utilizing resistance testing became available during the course of this study, clinicians became more aware of patterns of resistance after initial regimens and were better able to select treatment based on treatment history alone . Therefore, an effective regimen could be empirically selected for more than half of the patients in the study based on published resistance profiles of the prior treatment regimen without the need for an individual resistance test. This greatly reduced the power of the study to show a difference between groups, since the intervention could only affect a small fraction of the total patient population (i.e., patients other than NNRTI-naive nelfinavir failures). In participants with higher degrees of PI resistance, the greater reduction in viral load that was seen in the PHENO group compared with the SOC group is consistent with this argument. These patients had a more complex resistance profile that would not be predicted from the treatment history alone.
Further, salvage of nelfinavir failures with a NNRTI and PI regimen had been reported and was successful in up to 60% of patients . The overall response rate in this study was high, with 48% of patients having HIV RNA values < 400 copies/ml at month 12. This also reduced the potential for a strategy to improve the treatment response. As therapy for treatment-experienced patients improved during the conduct of CCTG 575, potential differences between the strategy groups may have been reduced. It is also interesting to note a temporal trend in the success of other clinical trials comparing resistance testing with standard of care. Earlier studies of short duration [9–11] tended to demonstrate greater differences between resistance testing and control groups (0.5 log10 copies/ml), while later studies (as CCTG 575) showed little difference between the strategy groups. It is possible that greater utilization of resistance tests in practice, better description of cross-resistance profiles, and use of resistance tests by investigators in the resistance-testing groups allowed clinicians in later trials to improve their selection of ART regimens empirically, in the SOC group, based on a careful review of past treatment.
This study was limited by its open design and in its power to demonstrate differences between groups. In CCTG 575, as in many clinical strategy studies, randomization at the patient, rather than provider or clinic level, resulted in investigators following patients in both arms. The impact of the resistance information given to the investigators may have informed their decisions for patients in the SOC group. During the conduct of the study, case studies of patients in the PHENO arm were presented to the investigator group in order to encourage appropriate and uniform interpretation of assay results. The clinician–investigators’ expanding knowledge of the cross-resistance profiles could then inform decisions on regimen selection for SOC participants. This would reduce the ability of the study to demonstrate an effect of resistance testing and represents a fundamental difficulty in conducting clinical strategy studies. Alternative study designs, such as cluster randomization (by clinic or provider), may help to reduce dispersion of the intervention effect but have lower power and may be hard to accrue when the intervention is desirable (as was the case in CCTG 575) [32,33].
In summary, this study did not demonstrate an improvement in virological outcome for patients randomized to the PHENO arm compared with the SOC arm. Significant differences were noted in post-hoc analyses of patients with greater degrees of PI resistance. This study emphasizes the need to establish clinical cut-offs for each drug reported by the test and has stimulated efforts to perform analyses aimed at defining and refining such cut-offs [20,34].
Sponsorship: This work was supported by funding from the California University-wide AIDS Research Program (UARP) CC99-SD-003, CC02-SD-003, IS02-SD-701, UCSD Center for AIDS Research (CFAR) 5P30 AI 36214, and ViroLogic Inc., Additional support for D. D. Richman: AI 27670, AI 38858, AI 43638, and AI 29164 from the National Institutes of Health, and the Research Center for AIDS and HIV Infection of the San Diego Veterans Affairs Healthcare System.
1. Hirsch MS, Brun-Vezinet F, Clotet B, Conway B, Kuritzkes DR, D’Aquila RT, et al
. Antiretroviral drug resistance testing in adults infected with human immunodeficiency virus type 1: 2003 recommendations of an International AIDS Society–USA Panel. Clin Infect Dis 2003; 37:113–128.
2. US Department of Health and Human Services Panel on Clinical Practices for Treatment of HIV Infection. Guidelines for the Use of Antiretroviral Agents in HIV-1-Infected Adults and Adolescents
. Washington, DC: US Department of Health and Human Services; November 10, 2003. http://www.aidsinfo.nih.gov
3. No authors listed. European guidelines for the clinical management and treatment of HIV-infected adults in Europe. AIDS
4. Zolopa AR, Shafer RW, Warford A, Montoya JG, Hsu P, Katzenstein D, et al
. HIV-1 genotypic resistance patterns predict response to saquinavir–ritonavir therapy in patients in whom previous protease inhibitor therapy had failed. Ann Intern Med 1999; 131:813–821.
5. Harrigan PR, Hertogs K, Verbiest W, Pauwels R, Larder B, Kemp S, et al
. Baseline HIV drug resistance profile predicts response to ritonavir–saquinavir protease inhibitor therapy in a community setting. AIDS 1999; 13:1863–1871.
6. Deeks SG, Hellmann NS, Grant RM, Parkin NT, Petropoulos CJ, Becker M, et al
. 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. J Infect Dis 1999; 179:1375–1381.
7. DeGruttola V, Dix L, D’Aquila R, Holder D, Phillips A, Ait-Khaled M, et al
. 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. Antivir Ther 2000; 5:41–48.
8. Call SA, Saag MS, Westfall AO, Raper JL, Pham SV, Tolson JM, et al
. Phenotypic drug susceptibility testing predicts long-term virologic suppression better than treatment history in patients with human immunodeficiency virus infection. J Infect Dis 2001; 183:401–408.
9. Baxter JD, Mayers DL, Wentworth DN, Neaton JD, Hoover ML, Winters MA, et al
. A randomized study of antiretroviral management based on plasma genotypic antiretroviral resistance testing in patients failing therapy. AIDS 2000; 14:F83–F93.
10. Tural C, Ruiz L, Holtzer C, Schapiro J, Viciana P, Gonzalez J, et al
. Clinical utility of HIV-1 genotyping and expert advice: the Havana trial. AIDS 2002; 16:209–218.
11. Durant J, Clevenbergh P, Halfon P, Delgiudice P, Porsin S, Simonet P, et al
. Drug-resistance genotyping in HIV-1 therapy: the VIRADAPT randomised controlled trial. Lancet 1999; 353:2195–2199.
12. Meynard JL, Vray M, Morand-Joubert L, Race E, Descamps D, Peytavin G, et al
. Phenotypic or genotypic resistance testing for choosing antiretroviral therapy after treatment failure: a randomized trial. AIDS 2002; 16:727–736.
13. Cohen CJ, Hunt S, Sension M, Farthing C, Conant M, Jacobson S, et al
. A randomized trial assessing the impact of phenotypic resistance testing on antiretroviral therapy. AIDS 2002; 16:579–588.
14. Wegner SA, Wallace MR, Aronson NE, Tasker SA, Blazes DL, Tamminga C, et al
. Long-term efficacy of routine access to antiretroviral-resistance testing in HIV type 1-infected patients: results of the clinical efficacy of resistance testing trial. Clin Infect Dis 2004; 38:723–730.
15. MacArthur RD. Drug resistance. An updated, user-friendly guide to genotype interpretation. AIDS Read 2000; 10:652–657.
16. Demeter L, Haubrich R. International perspectives on antiretroviral resistance. Phenotypic and genotypic resistance assays: methodology, reliability, and interpretations. J Acquir Immune Defic Syndr 2001; 26(suppl 1):S3–S9.
17. Petropoulos CJ, Parkin NT, Limoli KL, Lie YS, Wrin T, Huang W, et al
. A novel phenotypic drug susceptibility assay for human immunodeficiency virus type 1. Antimicrob Agents Chemother 2000; 44:920–928.
18. Hellmann N, Johnson P, Laboratory, VCR, Petropoulos C, Laboratory VR. Validation of the performance characteristics of a novel, rapid phenotypic drug susceptibility assay, PhenoSense HIV. Antivir Ther 1999; 4(suppl 1):34.
19. Haubrich R, Currier JS, Forthal DN, Beall G, Kemper CA, Johnson D, and the California Collaborative Treatment Group. A randomized study of the utility of human immunodeficiency virus RNA measurement for the management of antiretroviral therapy. Clin Infect Dis 2001; 33:1060–1068.
20. Haubrich R, Jiang H, Swanstrom R, Bates M, Katzenstein D, Fiscus S, et al
. Delavirdine hypersusceptibility (DLV HS): virologic response and phenotypic cut-points-results from ACTG 359. Eleventh Conference on Retroviruses and Opportunistic Infections
. San Francisco, February 2004 [abstract 671].
21. Bosch RJ, Downey GF, Katzenstein DA, Hellmann N, Bacheler L, Albrecht MA. Evaluation of cutpoints for phenotypic hypersusceptibility to efavirenz. AIDS 2003; 17:2395–2396.
22. Katzenstein DA, Bosch RJ, Hellmann N, Wang N, Bacheler L, Albrecht MA. Phenotypic susceptibility and virological outcome in nucleoside-experienced patients receiving three or four antiretroviral drugs. AIDS 2003; 17:821–830.
23. Lanier E, Hellmann N, Scott J, Ait-Khaled M, Melby T, Paxinos E, et al
. Determination of a clinically relevant phenotypic resistance ’cutoff’ for abacavir using the phenosense assay. Eighth Conference on Retroviruses and Opportunistic Infections
. Chicago, February 2001 [abstract 254].
24. Kempf D, Brun S, Rode R, Isaacson J, King M, Xu Y, et al
. Identification of clinically relevant phenotypic and genotypic breakpoints for ABT-378/r in multiple PI-experienced, NNRTI-naive patients. Antivir Ther 2000; 5(suppl 3):70.
25. Torre D, Tambini R. Antiretroviral drug resistance testing in patients with HIV-1 infection: a meta-analysis study. HIV Clin Trials 2002; 3:1–8.
26. Perez-Elias MJ, Garcia-Arota I, Munoz V, Santos I, Sanz J, Abraira V, et al
. Phenotype or virtual phenotype for choosing antiretroviral therapy after failure: a prospective, randomized study. Antivir Ther 2003; 8:577–584.
27. Mazzotta F, Lo Caputo S, Torti C, Tinelli C, Pierotti P, Castelli F, et al
. Real versus virtual phenotype to guide treatment in heavily pretreated patients: 48-week follow-up of the Genotipo–Fenotipo di Resistenza (GenPheRex) trial. J Acquir Immune Defic Syndr 2003; 32:268–280.
28. Badri SM, Adeyemi OM, Max BE, Zagorski BM, Barker DE. How does expert advice impact genotypic resistance testing in clinical practice? Clin Infect Dis 2003; 37:708–713.
29. Kemper CA, Witt MD, Keiser PH, Dube MP, Forthal DN, Leibowitz M, et al
. Sequencing of protease inhibitor therapy: insights from an analysis of HIV phenotypic resistance in patients failing protease inhibitors. AIDS 2001; 15:609–615.
30. Hirsch MS, Brun-Vezinet F, D’Aquila RT, Hammer SM, Johnson VA, Kuritzkes DR, et al
. Antiretroviral drug resistance testing in adult HIV-1 infection: recommendations of an International AIDS Society–USA Panel. JAMA 2000; 283:2417–2426.
31. Zolopa A, Tebas P, Gallant J, Keiser P, Sension M, Smith P, et al
. The efficacy of ritonavir (RTV)/saquinavir (SQV) antiretroviral therapy (ART) in patients who failed nelfinavir (NFV): a multi-center clinical cohort study.39th Interscience Conference on Antimicrobial Agents and Chemotherapy
. San Francisco, September 1999 [abstract 2065].
32. Donner A. Some aspects of the design and analysis of cluster randomization trials. Appl Stat 1998; 47:95–113.
33. Eccles M, Grimshaw J, Steen N, Parkin D, Purves I, McColl E, et al
. The design and analysis of a randomized controlled trial to evaluate computerized decision support in primary care: the COGENT study. Fam Pract 2000; 17:180–186.
34. Phillips AN, Costagliola D, DeGruttola V. for the Forum for Collaborative HIV Research. Standardization and Clinical Relevance of HIV Drug Resistance Testing: Refinement of Viral Genotype Interpretation Systems for ddI and Abacavir, Based on Virologic Response, Data Analysis Plan
. Washington, DC: School of Public Health and Health Services; August, 2003. http://www.hivforum.org
The CCTG 575 Study Team: Edward L. Seefried, Andrew Rigby, Craig Ballard, Kari Abulhosn, Chris Miller and Graziella Bruni (University of California, San Diego Antiviral Research Center and the California Collaborative Treatment Group Data and Biostatistical Unit, San Diego); Bobi Keenan (University of California, Irvine School of Medicine); Judith Currier and Maria Palmer (David Geffen School of Medicine at the University of California, Los Angeles); Shay M. Martinez (Santa Clara Valley Medical Center, San Jose).