Emergence of HIV-1 Drug Resistance in Previously Untreated Patients Initiating Combination Antiretroviral Treatment:  A Comparison of Different Regimen Types FREE

Current guidelines recommend the initiation of combination antiretroviral treatment (cART) against human immunodeficiency virus 1 (HIV-1) with 2 nucleoside reverse transcriptase inhibitors (NRTIs) and a nonnucleoside reverse transcriptase inhibitor (NNRTI) or a ritonavir-boosted protease inhibitor (PI/r).¹ These different drug classes vary considerably in the number of mutations required to confer resistance and in the degree of cross-resistance to drugs of the same class. Selection of a single mutation leads to the emergence of high-level resistance against NNRTIs, and resistance usually emerges quickly if treatment failure occurs.^2- 3 In contrast, virological failure with boosted PIs is associated with a reduced risk of selection of drug resistance mutations.⁴

Few data from observational studies exist on how these differences in genetic barriers, defined as the number of amino acid changes required to confer resistance, and the degree of cross-resistance may drive the emergence of drug resistance and hence affect the success of first-line therapy and beyond.^5- 7 Moreover, to our knowledge, only 1 head-to-head comparison of boosted PI (lopinavir) vs NNRTI-based (efavirenz) first-line combination therapy has been conducted to date. The AIDS Clinical Trial Group 5142 presented preliminary data on this randomized, open-label trial and observed no difference in rates of primary study end points, which were virological failure and/or treatment discontinuation due to toxic effects.⁸ However, the authors noted trends for a shorter time to those end points in the lopinavir arm and for more resistance accumulation on virological failure in the efavirenz arm.

Our objectives were to study the frequency and patterns of the emergence of viral drug resistance in patients from a large cohort in whom first-line cART failed in a routine clinical setting and to assess the potential impact on future therapies. Therefore, we analyzed patients from the Swiss HIV Cohort Study (SHCS) who were grouped according to their third drug, defined as the additional drug besides NRTIs or lamivudine or emtricitabine in standard 2-class combination therapy (NNRTI, PI, or PI/r). Further analyses were performed to compare the 3 regimen types with regard to virological failure and treatment discontinuation.

Included in this analysis were the cART-naïve patients from the SHCS^9- 10 who had initiated antiretroviral treatment between January 1, 1999, when genotyping became available in clinical practice in Switzerland, and December 31, 2005, with either 2 NRTIs and 1 NNRTI or 2 NRTIs and 1 PI (PI/r or unboosted). In case patients experienced an early adverse reaction to the initial treatment, the second cART regimen was considered if the initial regimen was followed for less than 30 days and if no treatment interruption occurred between the first and second regimens. As another inclusion criterion, 1 additional on-treatment HIV RNA measurement after the attainment of undetectable viral load levels of less than 50 copies/mL had to be available, or in patients in whom viral suppression was not achieved, 1 viral load after 180 days or more of continuous therapy. This criterion was required to evaluate our study participants with regard to virological failure while receiving therapy, which was defined as follows: (1) viral rebound with 2 consecutive viral loads greater than 500 copies/mL after previous attainment of undetectable HIV RNA levels, (2) 1 value greater than 500 copies/mL followed by a stop or a modification of the current therapy in patients with previous suppression of viral replication to undetectable levels, or (3) 1 viral load greater than 500 copies/mL after 180 days of continuous treatment in patients without previous suppression of viral load to undetectable levels. The date of the first HIV RNA measurement that indicated a virological failure was considered the failure date.

Drug resistance tests and clinical data collected up until February 1, 2007, were used for this analysis. Resistance tests were included in statistical analyses if the blood sampling was performed on or after the presumed date of virological failure or within 1 month after stopping the initial regimen if the patient continued not to receive treatment. Mean HIV RNA levels at the time of resistance testing were 3.55, 3.33, and 3.23 log copies/mL for the NNRTI group, the PI group, and the PI/r group, respectively (Kruskal-Wallis test, P = .66). Resistance data stem from routine clinical testing (60% of all tests) performed by the 4 laboratories in Switzerland authorized by the Federal Office of Public Health and tests specifically performed for this project from frozen repository samples (40%) if specimens were available. All laboratories performed population-based sequencing of the full protease gene and in minimum codons 28 to 225 of the reverse transcriptase gene using commercial assays (Viroseq vs 1, PE Biosystems, Rotkreuz, Switzerland; Virsoseq vs 2, Abbott AG, Baar, Switzerland; and vircoTYPE HIV-1 Assay, Virco Laboratory, Mechelen, Belgium) and in-house methods.¹¹ Resistance data were collected and stored using SmartGene's Integrated Database Network System (version 3.3.0, SmartGene, Zug, Switzerland). The degree of drug resistance was evaluated using the Stanford algorithm¹² version 4.2.6 and a modified version of the Fall 2006 International AIDS Society–USA drug mutation list,¹³ in which the following major mutations were included: NRTIs: M41L, A62V, K65R, D67N, K70E/R, L74V, V75I, F77L, Y115F, F116Y, Q151M, M184V/I, L210W, T215Y/F, K219Q/E, and insertions at codon 69; NNRTIs: L100I, K103N, V106M/A, V108I, Y181C/I, Y188C/L/H, G190S/A, P225H, and P236L; and PIs: D30N, V32I, L33F, M46I/L, I47V/A, G48V, I50V/L, L76V, V82A/F/T/S/L, I84V, N88S, and L90M. Mutations that confer resistance against entry inhibitors were not considered. The 5 resistance categories from the Stanford algorithm were regrouped for this analysis: viruses with a genotypic sensitivity score less than 15 were considered susceptible, those with a genotypic sensitivity score in the range of 15 to 59 were considered to have intermediate resistance, and those with a genotypic sensitivity score greater than 59 were considered to have high resistance. Throughout this study, lamivudine and emtricitabine were considered separate drug classes. In these drugs, the emergence of resistance on failure follows a pattern that is distinct from other NRTIs because only a single mutation at codon 184 is required to confer high resistance.¹⁴

The SHCS was approved by ethics committees of all participating institutions, and written informed consent was obtained from all participants.

All statistical analyses were stratified by regimen type (PI group [cART that contained 1 unboosted PI], PI/r group [1 PI/r], or NNRTI group [1 NNRTI]). Patterns of drug resistance were assessed using the Kruskal-Wallis test for continuous data and Fisher exact test for categorical data. Failure rates were analyzed using Kaplan-Meier life tables and univariate and multivariate Cox regression. We performed a competing risk analysis to estimate time receiving first-line therapy, which took into account that some patients were stopping cART because of reasons other than virological failure, such as ART-related toxic effects.^15- 16 The primary event of interest was stopping or modification of a regimen because of virological failure, and stopping or modification of a regimen owing to adverse events (adverse effects and laboratory abnormalities) was considered a competing risk event. For patients not reaching these end points, the analysis was censored at the stop date of the initial cART regimen or the last date of follow-up, whichever occurred first. For the estimation of Cox model parameters, the type of end point (virological failure or adverse event) and year of cART initiation were included as strata. Modeling assumptions were verified using the Schoenfeld residuals method.¹⁷ Reasons for stopping cART are systematically assessed in the SHCS using the Data Collection on Adverse Events of Anti-HIV Drugs reporting scheme.¹⁸

Models were adjusted for baseline demographics and the following potential confounders: HIV RNA viral load and CD4 cell count at therapy initiation as well as adherence.¹⁹ Baseline measurements for HIV RNA and CD4 cell counts were missing for approximately one-fifth of all patients (irrespective of treatment group). Hence, those variables were stratified (<1000, 1000-99 999, and ≥100 000 copies/mL for HIV RNA; <200, 200-349, and ≥350 cells/μL for CD4 cell count), and additional strata were included for missing values.

All statistical tests were 2-sided. P<.05 was considered statistically significant. No adjustments for multiple testing were made. Analyses were performed using Stata SE 9.2 (StataCorp, College Station, Texas).

Between January 1, 1999, and December 31, 2005, 2751 study participants in the SHCS initiated therapy with cART, of which 939 patients (34%) were excluded from our primary analyses because of short treatment duration with the same regimen or missing HIV RNA values. Patient characteristics associated with exclusion were following a PI-based therapy, use of injection drugs, female sex, or having higher CD4 cell counts, all characteristics known to be associated with premature therapy discontinuation.^20- 21 The second treatment was considered for 141 patients (44 in the PI group, 48 in the PI/r group, and 49 in the NNRTI group). Of the remaining 1812 patients who contributed 3789 person-years to the competing risk analysis, 489 (27%) initiated cART with unboosted PIs, 518 (29%) with boosted PIs, and 805 (44%) with NNRTIs as the group-specific third drug. A median of 5.7 viral load values per patient per year were obtained.

Baseline characteristics at the time of cART initiation for the 1812 patients are given in Table 1. Of note, patients who initiated cART with boosted PIs in general had a lower CD4 cell count and a higher HIV viral load at baseline compared with the PI group and the NNRTI group. Most patients received lamivudine as part of their treatment, ranging from 88% in the unboosted PI group to 95% in the 2 other groups. The most common third drug in the 3 groups was unboosted nelfinavir mesylate (88%), boosted lopinavir (76%), and efavirenz (91%).

In total, 177 patients experienced a virological failure (108 in the unboosted PI group, 24 in the boosted PI/r group, and 45 in the NNRTI group). For those patients, 112 baseline resistance test results (64 [59%], 14 [58%], and 34 [76%] of patients in the PI, PI/r, and NNRTI groups, respectively; Table 2) were available, and preexisting resistance was detected in 6 samples (5%). Resistance test results at the time of virological failure could be obtained for 142 patients (84 [78%], 20 [83%], and 38 [84%] in the PI, PI/r, and NNRTI groups, respectively) Table 3. At least 1 relevant resistance mutation was detected in 71 of the 84 patients (84%; 95% confidence interval [CI], 75%-92%) in the unboosted PI group, 6 of 20 patients (30%; 95% CI, 12%-54%) in the boosted PI/r group, and 25 of 38 (66%; 95% CI, 49%-80%) in the NNRTI group (Fisher exact test, P < .001; Figure 1). Patients in the unboosted PI and NNRTI groups acquired a median of 2 mutations during treatment (interquartile range [IQR], 1-2.5 and 0-3, respectively), whereas the median number of mutations in the boosted PI group was 0 (IQR, 0-1) (Kruskal-Wallis test, P < .001; Figure 1).

A similar pattern was observed when resistance against the 4 drug classes (NRTI, lamivudine or emtricitabine, NNRTI, and PI) was compared using the Stanford algorithm, whereby intermediate or high resistance against 1 or more drugs in a class signified class resistance. We observed that more classes were affected in unboosted PI-based (median, 2; IQR, 1-2) and NNRTI-based (median, 2; IQR, 0-2) regimens compared with boosted PI-based regimens (median, 0; IQR, 0-1; Kruskal-Wallis test, P < .001; Figure 1). This difference remained statistically significant when only high-level resistance was considered (data not shown). High-level resistance against lamivudine or emtricitabine was observed in 73% (95% CI, 62%-82%) of the PI group, 25% (95% CI, 9%-49%) of the PI/r group, and 58% (95% CI, 41%-74%) of the NNRTI group (Fisher exact test; P < .001). High-level resistance against the respective third drug was most prevalent in the NNRTI group at 55% (95% CI, 38%-71%), followed by the unboosted PI group at 20% (95% CI, 12%-30%) and the boosted PI group at 5% (95% CI, 0%-25%) (Fisher exact test, P < .001). Multidrug resistance predominantly occurred as resistance against the third drug and against lamivudine or emtricitabine and was also more frequent in the NNRTI group (50%; 95% CI, 33%-67%) than in the PI group (17%; 95% CI, 9%-26%) or PI/r group (10%; 95% CI, 0.1%-32%, respectively) (Fisher exact test, P < .001).

We further assessed rates of virological failure and adverse events in the different treatment groups (Figure 2). Analyzing the 1812 included patients with respect to virological failure events, we found that crude rates (per 100 person-years) were highest in the unboosted PI group (10.3; 95% CI, 8.5-12.4) and comparable between the boosted PI group (2.7; 95% CI, 1.8-4.0) and the NNRTI group (2.4; 95% CI, 1.8-3.3; log-rank P < .001; Table 4). A multivariate Cox regression adjusted for potential confounders confirmed this finding. Using the NNRTI group as a reference, hazard ratios were 1.19 (95% CI, 0.71-2.0) for the boosted PI group and 3.83 (95% CI, 2.62-5.59) for the unboosted PI group.

Crude rates for toxic effects related to cART were higher in the PI-based regimens (unboosted PIs: 8.9; 95% CI, 7.3-10.9; boosted PIs: 10.9; 95% CI, 9.0-13.3) compared with 7.8 (95% CI, 6.6-9.2) in the group taking NNRTIs (log-rank test; P = .03). When considering both types of events (virological failure and toxic effects) simultaneously, patients taking unboosted PIs had the highest rate at 19.2 (95% CI, 16.7-22.1) followed by boosted PIs (13.6; 95% CI, 11.4-16.3) and NNRTIs (10.3; 95% CI, 8.9-11.8; log-rank test; P < .001), which was consistent with results from the multivariate competing risk Cox model. Compared with the NNRTI group, relative hazards for experiencing either a virological failure or a cART-related toxic effect were 1.36 (95% CI, 1.07-1.73) and 1.75 (95% CI, 1.41-2.18) for patients taking boosted PIs and unboosted PIs, respectively. When repeating the competing risk analysis including all patients who had started cART (intent to treat, n = 2751), we obtained a lower hazard ratio for the unboosted PI group (1.24; 95% CI, 1.05-1.46) but similar estimates for the boosted PI group (1.28; 95% CI, 1.07-1.52; Table 4). Reasons for premature therapy discontinuation of the 939 patients additionally included in the intent-to-treat analysis are shown in Table 5.

We also looked at subsequent virological failures in 1001 of 1812 patients (55%; including 141 of 177 patients in whom first-line therapy failed) who initiated a new treatment after stopping first-line therapy and who were not lost to follow-up (107; 6%). We observed that patients in whom first-line therapy had already failed virologically were more likely to experience second-line therapy failure (18 cases; incidence rate, 10.4 per 100 person-years; 95% CI, 6.6-16.5) compared with patients without previous failure (27 patients; incidence rate, 2.4 per 100 person-years; 95% CI, 1.7-3.6; log-rank test; P < .001). Because of the small number of cases, this analysis could not be stratified further.

In this study, we observed clinically relevant differences in the emergence of drug resistance for the 3 types of initial regimens. Although cART-containing boosted PIs and NNRTIs appeared to be equivalent with regard to the virological failure rate (ie, 2.7 and 2.4 per 100 person-years, respectively), significantly more resistance emerged in the NNRTI group at virological failure. Virus strains resistant against lamivudine or emtricitabine were present in 58% of patients with virological failure in the NNRTI group, and 55% of all viral samples in this group exhibited resistance against NNRTI. Even more worrisome, multidrug resistance against NNRTIs and at least 1 component of NRTIs and lamivudine or emtricitabine was observed in half of all patients with virological failure in this group (Table 2). Nonetheless, NNRTI-containing combinations showed better tolerability than PI-based regimens. Thus, our analysis demonstrates that despite similar potency, boosted PIs are superior to more tolerable NNRTI regimens when comparing the emergence of drug resistance in patients in whom therapy is failing.

The low genetic barrier of NNRTIs and the high potential for cross-resistance in this drug class have long been known from in vitro experiments and clinical trials,^2,14,22 but the impact of those aspects on the emergence of drug resistance in routine clinical practice and the consequences for future therapies have rarely been investigated. In a retrospective cohort analysis, Phillips et al⁶ reported a significantly higher risk of acquiring third drug–specific resistance and multidrug resistance for patients initiating cART with NNRTI compared with boosted PI. Preliminary results from AIDS Clinical Trial Group 5142 also indicated more drug resistance accumulation in NNRTI-based regimens compared with boosted PI regimens.⁸ However, no differences in virological failure and/or stopping because of toxic effects between NNRTI-based and boosted PI-based regimen types were observed.

Our results confirm and extend these previous observations. We found that virological failure of NNRTI-containing regimens led to the emergence of multidrug resistance in half of all cases, 5 times more often than in the boosted PI group. In addition, we observed that almost all cases of multiclass resistance (with 1 exception in the PI group and 1 in the NNRTI group; Table 2) involved the reverse transcriptase mutation M184V/I, whereas high-level resistance against NRTIs other than lamivudine or emtricitabine remained rare. Hence, our data demonstrate that NNRTI-based regimens carry a markedly higher risk of losing multiple drug classes in case of failure.

The inclusion of competing risks in our time-to-event analysis has allowed us to gain a more accurate picture of the success of initial cART. The differences in hazard ratios between the competing risk approach and the conventional approach of estimating Cox models imply that the benefits of higher potency of newer regimens compared with unboosted PIs are partially outweighed by the higher frequency of adverse events, especially in boosted PI–containing regimens. This loss of benefit was even more pronounced in the intent-to-treat analysis, in which event rates and relative hazards were no longer different between boosted and unboosted PIs. Further analyses showed that patients following boosted PI regimens had more adverse events related to cART in the first 180 days of therapy (39.4 per 100 person-years; 95% CI, 33.1-46.9) than any of the other groups, with 33.3 (95% CI, 28.7-38.7) and 25.0 (95% CI, 20.3-30.7) in patients taking NNRTIs and PIs, respectively. Overall, NNRTIs were generally better tolerated, and patients following those regimens continued with their initial therapy for a longer period in both the on-treatment analysis and the intent-to-treat analysis.

This study has some limitations. Potentially because of the somewhat shorter follow up-time, relatively few virological failures occurred in the group starting cART with newer regimen types and especially boosted PIs, but our results are largely consistent with other reports on resistance development.⁴ Like all observational studies, this analysis is prone to confounding. The time point of resistance testing might have affected our results because it has been shown that the longer maintenance of failing regimens leads to more resistance.²³ However, sensitivity analyses neither revealed a systematic confounding of our results by the time (per month) between presumed date of virological failure and the sampling date for resistance testing (odds ratio, 1.01; 95% CI, 0.97-1.06) nor indicated any influence of viral load levels (per log increase) at the time of testing (OR, 0.94; 95% CI, 0.69-1.29).

Furthermore, studies²⁴ have demonstrated detrimental effects of transmitted resistance mutations on treatment success. In the present analysis, pre-cART resistance tests were available for 63% of all patients with a virological failure, and we therefore can not fully exclude potential confounding of our results by baseline resistance.

Another limitation of this study compared with clinical trials is the high variation of drug combinations within the 3 treatment groups. To verify our findings, we limited the time-to-event analyses to the most frequent drugs in the boosted PI group and the NNRTI group (lopinavir [76%] and efavirenz [91%]), which yielded similar results for the comparison of virological failure rates. For the competing risk analysis, however, the previously observed difference in the hazard ratios between boosted PIs and NNRTIs disappeared. A secondary analysis showed that this difference was mainly because of the exclusion of boosted indinavir sulfate, which exhibited a far higher rate of treatment discontinuation caused by toxic effects (29.5 per 100 person-years; 95% CI, 20.4-42.7) compared with the remaining third drugs in the boosted PI group (8.7; 95% CI, 6.9-11.1). This finding implies that the apparent tolerability disadvantage of boosted PI compared with the NNRTI regimen might not be as pronounced or even absent with newer, more tolerable combinations that contain boosted atazanavir sulfate or lopinavir.

Previous studies²⁵ have demonstrated associations of adherence and pharmacokinetic drug levels with virological failure and the emergence of drug resistance. Systematic collection of adherence information in the SHCS started in the middle of 2003.¹⁹ Since we do not have complete assessments of adherence for all patients, we performed a sensitivity analysis by confining our time-to-event analyses to patients with at least 1 adherence measure while receiving first-line cART (n = 927, data not shown). Of note, our findings were not altered by this, and we are therefore confident that the observed differences in virological failure rates truly reflect drug potency. Therapeutic drug monitoring has been introduced only recently in the SHCS, so systematic analyses of therapeutic drug monitoring are not possible at this time.

In summary, our findings have potentially important clinical implications. Given the risk of severely compromising future therapy options in case of virological failure while taking NNRTIs and the equivalent potency of boosted PI regimens, physicians should critically assess a patient's ability to adhere to NNRTI-based regimens and to cope with the potential toxic effects, such as adverse effects on the central nervous system. If problems are expected, starting or changing to PI/r therapy most likely represents a safer choice, since a PI/r tends to accumulate less resistance in case of virological failure. Should our findings be confirmed, they should be implemented in treatment guidelines and also be considered for future treatment strategies in developing countries.

Correspondence: Huldrych F. Günthard, MD, Division of Infectious Diseases and Hospital Epidemiology, University Hospital Zurich, Rämistr 100, CH-8091 Zurich, Switzerland (huldrych.guenthard@usz.ch).

Accepted for Publication: April 17, 2007; final revision received April 24, 2007; accepted May 7, 2007.

Author Contributions: Viktor von Wyl and Huldrych Günthard had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Günthard. Acquisition of the data: Böni, Burgisser, Klimkait, Yerly, Battegay, Furrer, Telenti, Hirschel, Vernazza, Bernasconi, Rickenbach, and Perrin. Analysis and interpretation of the data: von Wyl, Ledergerber, and Günthard. Drafting of the manuscript: von Wyl, Ledergerber, and Günthard. Critical revision of the manuscript for important intellectual content: Böni, Burgisser, Klimkait, Yerly, Battegay, Furrer, Telenti, Hirschel, Vernazza, Bernasconi, Rickenbach, and Perrin. Statistical analysis: von Wyl and Ledergerber. Obtained funding: Günthard, Yerly, and Ledergerber. Study supervision: Günthard.

Swiss HIV Cohort Study Members: Heiner Bucher, MD; Sandro Cattacin, PhD; Matthias Cavassini, MD; Rolf Dubs, PhD; Matthias Egger, MD; Lucia Elzi, MD; Peter Erb, PhD; Marek Fischer, PhD; Markus Flepp, MD; Adriano Fontana, MD; Patrick Francioli, MD (president of the SHCS, Centre Hospitalier Universitaire Vaudois, Lausanne); Meri Gorgievski, PhD; Hans Hirsch, MD; Irene Hösli, MD; Christian Kahlert, MD; Laurent Kaiser, MD; Urs Karrer, MD, PhD; Olivia Keiser, MSc; Christian Kind, MD; Gina Martinetti, MD; Begogna Martinez, MD; Nicolas Müller, MD; David Nadal, MD; Milos Opravil, MD; Fred Paccaud, MD; Giuseppe. Pantaleo, MD; Christoph Rudin, MD (chairman of the Mother & Child Substudy); Patrick Schmid, MD; Detlev Schultze, PhD; Jörg Schüpbach, MD; Roberto Speck, MD; Patrick Taffé, PhD; Philippe Tarr, MD; Alexandra Trkola, PhD; and Rainer Weber, MD.

Financial Disclosures: Dr Klimkait has served on the advisory boards of Abbott, Bayer, Bristol-Myers Squibb, and Roche. Dr Battegay is a consultant for Roche Pharma Switzerland and Boehringer Ingelheim Switzerland. Drs Furrer and Vernazza have served on the advisory boards of Abbott, GlaxoSmithKline, Bristol-Myers Squibb, Roche, Gilead, Merck Sharp & Dohme, Boehringer-Ingelheim, and Tibotec (Dr Vernazza). The institution of Dr Furrer has received unrestricted educational grants from Abbott, GlaxoSmithKline, Bristol-Myers Squibb, Roche, Gilead, Merck Sharp & Dohme, and Boehringer-Ingelheim. Dr Bernasconi has received travel grants and honoraria from Gilead, Roche, GlaxoSmithKline, Pfizer, Boehringer Ingelheim, and Tibotec. Dr Ledergerber has received travel grants, grants, or honoraria from Abbott, Aventis, Bristol-Myers Squibb, Gilead, GlaxoSmithKline, Merck Sharp & Dohme, Roche, and Tibotec. Dr Günthard has been an adviser and/or consultant for GlaxoSmithKline, Abbott, Novartis, Boehringer Ingelheim, Roche, and Bristol-Myers Squibb and has received unrestricted research and educational grants from Roche, Abbott, Bristol-Myers Squibb, GlaxoSmithKline, and Merck Sharp & Dohme.

Funding/Support: This study has been financed in the framework of the Swiss HIV Cohort Study, supported by the Swiss National Science Foundation (SNF) (grant 3345-062041). Further support was provided by SNF grant 3247B0-112594/1 (Drs Günthard and Ledergerber and Ms Yerly) and SHCS project 470.

Role of the Sponsors: The funding sources had no role in study design; in collection, analysis, or interpretation of data; or in the writing of the report.

Previous Presentation: Results of this study were presented in part at the 15th International Drug Resistance Workshop; June 15, 2006; Sitges, Spain (abstract 72); and the 14th Conference on Retroviruses and Opportunistic Infections; February 26, 2007; Los Angeles, California (abstract M-102).

Additional Information: All sequences have been submitted to GenBank (accession Nos. EF449774 to EF450027).

Additional Contributions: Chantal Gaille (Laboratory of Virology and AIDS Center, Geneva University Hospital) provided the retrospective sequencing, SmartGene provided technical support, and Joseph Wong, MD, performed critical review of the manuscript. We thank the patients for participation in the SHCS, the physicians and study nurses for excellent patient care, and the laboratory technicians of the Swiss resistance laboratories for the quality of the data.