Minimizing Resistance Consequences After Virologic Failure on Initial Combination Therapy: A Systematic Overview

Objective: To identify optimal first-line therapies based on the rate of virologic success (VS) and the preservation of future treatment options in antiretroviral therapy (ART)-naive subjects.

Design: Systematic overview of genotypic resistance mutations from clinical trials of combination ART.

Methods: Various sources were searched for studies in ART-naive subjects providing virologic response rates and genotypes from subjects with virologic failure. The International AIDS Society-USA genotypic resistance guidelines were used to calculate regimen resistance cost (RC_reg) and number of active drug (AD) scores for each regimen and to rank the regimens.

Results: Intra- and interstudy comparisons showed higher VS rates for nonnucleoside reverse transcriptase inhibitor (NNRTI) regimens (range: 51%-76%) and boosted protease inhibitor (boosted PI) regimens (range: 55%-79%). Boosted PI failures had the lowest RC_reg (range: 0.12-0.21) and the highest AD (range: 19.80-20.18) scores. NNRTI failures had higher RC_reg (range: 0.00-1.22) and lower AD (range: 16.83-21) scores.

Conclusions: NNRTI and boosted PI regimens provide the highest rates of VS in treatment-naive HIV-infected persons. Treatment option scores were higher in subjects who failed boosted PI- containing regimens versus NNRTI-containing regimens, however.

Treatment of HIV with combination antiretroviral therapy (ART) has resulted in a greatly extended life expectancy for persons infected with HIV. After the introduction of combination ART, AIDS incidence and AIDS-related deaths decreased dramatically in the United States and Europe.^1,2 For many persons infected with HIV and on active ART, HIV has become a chronic condition rather than a fatal disease.

Currently, there are many options for initial combination ART and the sequencing of combination ART in individuals infected with HIV. The US and European guidelines^3,4 differ with regard to their respective criteria for initiation of ART in patients with HIV/AIDS and the optimal order of treatment, but a number of options are provided. Both guidelines state that not all combinations have been compared directly and that regimen choice must consider a complex array of factors. Importantly, preservation of future treatment options through the strategic choice and sequencing of drugs is recognized in both sets of guidelines.

Drug resistance is one of the greatest threats to successful long-term ART. Combination ART delays the onset of drug resistance, but drug resistance can still develop and lead to a reduction in drug efficacy. Once resistance has developed to a drug or drug class, resistant viruses may be archived in lymphoid tissue, and responses to the drug or drug class are compromised indefinitely. Because cross-resistance may also limit the efficacy of unused ART agents, many persons infected with HIV can exhaust effective ART treatment options quickly if careful selection and monitoring of initial ART are not undertaken.

A comprehensive review of the resistance profiles of subjects experiencing virologic failure (VF) after different initial combination ART regimens was performed, emphasizing the implications for subsequent treatment choices. Similar systematic overviews have identified associations between pill counts and the proportion of subjects with plasma HIV RNA levels <50 copies/mL after 48 weeks⁵ and the improving proportion of subjects reaching this threshold in more recent clinical trials.⁶ First-line therapies were ranked by rate of virologic success VS and preservation of future treatment options. This review of the available studies of combination ART and resistance results may help clinicians to make initial treatment decisions.

METHODS

MEDLINE and EMBASE were searched during for the period from January 1997 to February 2004 for published data using the following medical subject headings (MeSHs) and combinations thereof: antiretroviral therapy, highly active antiretroviral therapy (HAART), drug resistance, viral, genotype, phenotype, and epidemiologic study characteristics. The Cochrane Library, the US Food and Drug Administration (FDA) Website, the Websites of various HIV organizations, and reference lists in published articles and abstracts from scientific meetings were manually searched. In addition, we contacted the authors of studies included in the review to provide individual genotyping data for each treatment group in the study.

Studies were abstracted if they met the following inclusion criteria:

* Published in English

* Conducted in clinical settings

* Included treatment-naive adult subjects receiving 3 or more antiretroviral agents

* Presented results for virologic responses and failures (definitions of success and failure were study dependent and differed across studies)

* Reported genotypic resistance patterns in individual subjects (timing of resistance sampling and frequency were study dependent and differed across studies)

Studies were excluded if they did not meet these inclusion criteria and/or:

* Were case reports or case series

* Were preclinical studies

* Did not present complete genotypic testing results for individual subjects experiencing VF (authors of studies that did not report genotypic results for subjects with VF were contacted to request these results)

* Were studies of salvage therapies

During the abstraction process, the following data were entered into evidence tables if available in the publication or study document:

* Study design and subject characteristics

* First-line drug combinations

* Definition of VF

* Intention-to-treat (ITT) results, including the following:

* Percentage of patients with plasma HIV RNA levels below protocol-specified thresholds at protocol-specified time points, percentage of subjects experiencing VF, time to last virologic response, and CD4 cell count at protocol-specified time points

* Resistance mutations in individual subjects and percentage of patients developing those mutations

* Time between meeting the threshold for VF and collection of samples for genotypic resistance testing

The definition of VF differed among individual studies. Given the heterogeneity of VF definitions, our analyses focused on VS, defined as the proportion of study subjects with plasma HIV RNA levels <50 copies/mL at week 48. Genotypic resistance results were analyzed systematically, and the results were used for inter- and intratrial comparisons. Genotypic results from individual subjects were provided by study investigators and GlaxoSmithKline. Abbott Laboratories provided the regimen resistance cost (RC_reg) and mean number of individual drugs with number of active drug (AD) calculations. Individual combination ART regimens were ranked by their rates of VERSUS, RC_reg scores, and AD scores.

Published scoring methods to quantify resistance were applied to studies for which detailed resistance data with mutations could be obtained.⁷ The International AIDS Society-USA (IAS-USA) genotypic resistance guidelines were used to ascertain resistance and cross-resistance for the different drug classes and the individual drugs within each class.^8,9

The mean AD and RC_reg scores were calculated based on the future drug options (FDO) calculation, which assesses the number of remaining active drugs and classes based on genotypes.⁷ The mean AD score can range from 21 (all drugs remain) to 0 (no drugs remain), whereas RC_reg scores can range between 0 (no resistance) and 5.5 (resistant to all 21 drugs on the US market).

Ranked results are presented in tabular (Table 1) and graphic formats (Figs. 1, 2). Because definitions of VF varied across studies, interstudy statistical comparisons of VS, RC_reg scores, and AD scores would be inappropriate and therefore were not conducted. Intrastudy statistical comparisons were made for VS and AD results from selected studies using the T statistic and are presented in Table 2.

Figure 1	Figure 2	Table 1
Table 2

Initially, the analysis planned to focus on the class of the third drug in a HAART regimen (protease inhibitor [PI] vs. nonnucleoside reverse transcriptase inhibitor [NNRTI] vs. nucleoside/nucleotide reverse transcriptase inhibitor [NRTI]) taking into account the NRTI backbone. Relatively few trials with different NRTI backbones and the same third agent were available, however. Therefore, all presented analyses address only the third drug component of combination ART (PI vs. boosted PI vs. NNRTI vs. NRTI).

RESULTS

The literature search yielded 176 citations of studies on combination ART. Studies fulfilling inclusion/exclusion criteria of 3-drug combination ART with data on VF were abstracted. Studies were most commonly excluded because they included treatment-experienced subjects, did not present VF data, or did not report individual genotype results. In addition, many studies not appearing in the National Library of Medicine Gateway search were identified through screening of reference lists in publications and searches of scientific meeting abstracts. Ultimately, only 15 studies met the inclusion criteria.

Table 3 provides an overview of the studies included in the review.^10-26 All the studies are randomized clinical trials, and the number of subjects in each study arm ranged from 19 to 327. Most studies provided virologic results at 48 weeks, although the total duration ranged from 12 to 204 weeks. There were a total of 30 study arms and 18 unique triple-drug combinations or dosing schedules in the 15 studies. The summarized studies investigated a total of 15 PI-containing regimens, 7 NNRTI-containing regimens, 6 boosted PI-containing regimens, and 3 triple-NRTI-containing regimens (1 study arm alternated between PI-containing regimens and NNRTI-containing regimens). Twenty of the 30 study arms used a thymidine analogue (zidovudine [ZDV] or stavudine [d4T]) plus lamivudine (3TC), 7 arms used abacavir (ABC) plus 3TC without a thymidine analogue, 3 arms used d4T plus didanosine (ddI), and 1 arm used ZDV plus ddI (1 study arm alternated between ZDV plus 3TC and d4T plus ddI). VS was commonly defined as the percentage of subjects with plasma HIV RNA levels <50 copies/mL, and VF was defined as failure to achieve RNA levels <400 copies/mL or rebound above this threshold. The definition of VF was variable among trials and ranged from failure to achieve RNA <50 to 500 copies/mL and/or rebound from above 50 to 1265 copies/mL. When available, the ITT missing equals failure (ITT: M = F) results were used.

Table 3

Table 3

Table 1 summarizes the measurements of resistance in the different trials. The treatment arms are listed by the lowest RC_reg scores. All 6 boosted PI-containing regimens had RC_reg scores among the lowest 10 regimen scores. In contrast, 6 of the 7 NNRTI-containing regimens had RC_reg scores in the higher half of treatment regimens (the lowest score was measured in the seventh NNRTI-containing regimen, where it was alternated with a PI-containing regimen). RC_reg scores for unboosted PI- and NRTI-containing regimens did not show consistent trends.

Figure 1 displays a bubble plot of RC_reg plotted on the y-axis and VS rate plotted on the x-axis. Figure 2 is a bubble plot of mean AD and VS. In both figures, treatment regimens are color coded by the class of the third drug, and the size of the bubble is related to the number of subjects receiving this regimen. Overall, boosted PI-containing regimens were associated with high VS rates (range: 55%-79%), the lowest RC_reg score (range: 0.12-0.21), and the highest mean AD score (range: 19.80-20.18). As both figures demonstrate, NNRTI-containing regimens were associated with high VS rates (range: 51%-76%), more variable RC_reg scores (range: 0.00-1.22), and varying mean AD scores (range: 16.83-21.00).

The unboosted PI regimens generally had lower VS rates (range: 40%-58%) and varying mean AD scores (range: 16.60-20.00). The 3 triple-NRTI regimens had VS rates ranging from 40% to 62%, with mean AD score ranging from 17.86 to 18.44.

We further contrasted the results of this systematic overview against the results of smaller randomized trials. When comparing regimens in randomized subjects, similar results in VS rates and mean AD scores were observed (see Table 2). These observations support the analyses of the systematic overview by controlling for differing definitions of VS and VF, baseline differences in study populations, and other potential study differences.

DISCUSSION

The optimal initial treatment of persons infected with HIV should be chosen based on therapeutic efficacy, minimal drug-related toxicities, and potential future drug options (FDOs). The results of this systematic overview suggest overlapping virologic success rates of NNRTI-containing regimens and boosted PI-containing regimens; however, the resistance consequences of initial treatment regimens favored boosted PI-containing regimens over NNRTI-containing regimens. The observation of delayed resistance with boosted PI-containing regimens has been made in several smaller studies, but this systematic overview has extended these observations to larger numbers of subjects and to boosted regimens containing lopinavir (LPV), amprenavir, or fosamprenavir.

The results of this systematic overview are also consistent with data from available randomized comparative trials. Two studies have compared PI-containing regimens versus boosted PI-containing regimens, and their results demonstrate better preservation of future treatment options associated with boosted PIs.^12,25 Only 1 study has compared NNRTI-containing regimens with boosted PI-containing regimens, and among subjects with VF in the boosted PI arm, relatively few resistance mutations were observed.²⁶ Several studies have compared unboosted PI-containing regimens versus NNRTI-containing regimens or NRTI-containing regimens, with results suggesting equal or improved virologic suppression for NNRTI-containing regimens^9,24,27-29 and no difference in FDOs.^9,24 Important new results are likely to be generated from ongoing clinical trials directly comparing NNRTI-containing regimens with boosted PI-containing regimens such as Adult Clinical Trials Group (ACTG) 5142.

The mechanism for the delayed appearance of resistance mutations with boosted PI-containing regimens remains uncertain but is likely to include a combination of antiretroviral efficacy, broad cell/compartment penetration, and a high genetic barrier to resistance.^30,31 The increase in PI levels seen with ritonavir (RTV) enhancement seems to represent the crucial process. RTV enhancement raises PI levels through inhibition of the CYP3A isozymes of the cytochrome P450 system. The plasma area under the curve concentration of LPV is increased more than 80-fold by RTV coadministration.³² For amprenavir and fosamprenavir, the plasma area under the curve concentration is also increased more than 2-fold by RTV enhancement.³³ The decrease in resistance mutations observed at VF when PI levels are boosted with RTV suggests that drug concentrations promoting viral evolution (ie, concentrations under which any preexisting variants with reduced drug susceptibility have a growth advantage over the predominant wild-type strain) occur less commonly with RTV boosting than during use of the same PIs unboosted. This difference may reflect an increased time with boosted regimens for plasma levels to decline to a point at which significant viral replication can begin (most likely during periods of intermittent adherence) and/or atypical terminal pharmacokinetics wherein the clearance of the boosted PI increases over time because of the decline in RTV CYP3A inhibition.^31,32 For example, the plasma half-life of LPV is 8.5 hours between 6 and 12 hours after dosing but only 2.2 hours between 24 and 36 hours after dosing.³² Such nonlinear pharmacokinetics, only observed with boosted PIs, contrast greatly with agents such as NNRTIs with low innate clearance, which provides extended active drug concentrations but also extended times of selective pressure for viral evolution.

In analyses from this survey comparing resistance consequences of NNRTI-containing regimens versus NRTI-containing regimens, the mean AD scores were similar but the RC_reg scores were lower with the NRTI-containing regimens. Previous studies have demonstrated the appearance of NNRTI resistance mutations at low plasma HIV RNA levels, with resulting NNRTI cross-resistance and higher RC_reg scores. The accumulation of resistance mutations on NRTI-containing regimens may be a function of time with detectable viremia, resulting in the early detection of wild-type virus or M184V mutations and followed by the appearance of thymidine analogue mutations after 24 weeks. Ideally, this overview would have addressed the resistance consequences of the NRTI components of combination ART. Unfortunately, there was a preponderance of thymidine analogue plus 3TC-containing combinations. The resistance consequences of thymidine analogue versus not, 3TC-containing regimens or emtricitabine (FTC)-containing regimens versus not, and tenofovir-containing regimens versus not could not be elucidated in this overview and are best addressed in randomized clinical trials.

There are important limitations for the observations of this systematic overview. Baseline resistance testing was not always available on subjects; thus, preexisting resistance leading to VF cannot be excluded. When ART-resistant virus is transmitted, it may affect virologic response and clinical prognosis, including survival.^34-36 Most studies to date suggest a low prevalence of drug resistance among newly infected and previously untreated persons, however.^37,38

Many clinical trials could not be included in this systematic overview because resistance results were not presented or were presented in summary only and did not include resistance results from individual subjects. One hundred seventy-six studies were identified for potential inclusion, but only 15 met inclusion criteria, introducing the potential for selection bias. The results from these 15 studies are consistent with results from prospective randomized trials, however. Great effort was undertaken to obtain genotypic resistance results from studies before their exclusion, but additional information was unavailable or was not shared by investigators and/or study sponsors. Heterogeneity among subject populations from different studies may also affect resistance results. In addition, these analyses have only considered the role of a third agent in the appearance of resistance, and it is possible that the choice of dual NRTIs may have an impact on the emergence of resistance mutations. Resistance results were also not available in treatment-naive subjects for several boosted PI-containing combinations, including saquinavir, indinavir, and atazanavir. Definitive conclusions can only be reached from directly comparative randomized clinical trials.

Importantly, many clinical trials used differing definitions of VF, and there was no standardized approach to obtaining samples for resistance assays. It is plausible that the time with detectable viremia on a given regimen, absolute plasma HIV RNA level, CD4 cell count, and other clinical factors may have an impact on the appearance of resistance mutations. Finally, genotyping methodologies were not standardized, and our understanding of resistance patterns is evolving, although recent studies suggest a high level of genotype concordance and we used the most current consensus genotype definitions.

These limitations illustrate the need for a standardized approach to the presentation of resistance results. Clinical trials should present a disposition for all randomized subjects in the ITT population. The reasons for discontinuation of an initial ART regimen, including VF, drug-related toxicities, voluntary withdrawal, and loss to follow-up, should be listed. It is unlikely that consensus can be reached on an absolute plasma HIV RNA level as the definition for VF, but a clear and transparent definition must be presented by each study. Definitions of VF in treatment-naive populations may require a lower HIV RNA threshold because of the potential accumulation of drug resistance mutations, which could compromise second-line treatment options. The time from reaching VF to the obtaining of samples for genotyping must also be given, recognizing that the duration of continuing treatment in a patient with detectable viremia on the same regimen may be an important determinant of additional drug resistance mutations. All genotypic results should be listed, including results from individual subjects and samples with unsuccessful genotyping.

In summary, this systematic overview has extended the results of individual clinical trials that identified the potent activity of NNRTI-containing regimens and boosted PI-containing combinations as well as the delayed appearance of resistance mutations in subjects receiving boosted PI-containing ART. Clinicians should seriously consider this factor in combination with regimen efficacy and toxicity when choosing an initial regimen.

ACKNOWLEDGMENT

The authors gratefully acknowledge the assistance and provision of summary data from Abbott Laboratories, Inc., and the provision of individual genotyping data from the SWATCH Study Team, the COMBINE Study Team, and the START 1 and START 2 Study Teams.