| |
An updated systematic overview of triple combination therapy in ART-naive HIV+ adults
|
| |
| |
AIDS: Volume 20(16) 24 October 2006 p 2051-2064
Bartlett, John Aa; Fath, Michael Jb; DeMasi, Ralphc; Hermes, Ashwaqb; Quinn, Josephd; Mondou, Elsad; Rousseau, Franckd
From the aDuke University Medical Center, Durham, North Carolina, USA
bAbbott Laboratories, Abbott Park, Illinois, USA
cTrimeris Inc, Morrisville, North Carolina, USA
dGilead Sciences, Durham, North Carolina, USA.
This systematic overview was performed to re-examine the activity of initial three-drug combinations in antiretroviral-naive persons with contemporary agents, the predictors of activity, and the temporal changes in activity over time.
Over the period of observation, the weighted mean responses for the percentage of patients with plasma HIV RNA levels < 50 copies/ml at week 48 steadily improved, (prior to 1998, 41%; 1999-2000, 50%; 2001-2002, 56%; 2003-2004, 64%), (Fig. 1). During the same periods of time the weighted mean responses for increases in CD4+ cell counts at week 48 were similar, (prior to 1998, +173 cells/ml; 1999-2000, +174 cells/ml; 2001-2002, +180 cells/ml; 2003-2004, +169 cells/ml).
The authors said: The results of this expanded overview offer important new observations on the success of contemporary antiretroviral treatment regimens in trials studying ART-naive populations. The percentages of patients achieving plasma HIV RNA levels < 50 copies/ml at 48 weeks is increasing over time, and in reports from most recent trials, more than 80% of patients may reach this threshold [121-123]. The improving virologic responses were seen despite the enrollment of patient populations with lower CD4+ cell counts at entry, and surprisingly, lower entry CD4+ cell counts were associated with improved responses in some analyses. Drug class was the most significant predictor of regimen potency, with the best responses achieved by NNRTI and BPI-containing regimens. Pill count was not consistently identified as a predictor of virologic response, in contrast to our previously published overview.
Abstract
Objective: To compare the effectiveness of three drug combination antiretroviral therapy (ART) in treatment-naive HIV-infected persons, and identify the predictors of responses.
Design and methods: Overview of trials identified by searching public domain publications and conference presentations. The three-drug combination therapy was defined as two nucleoside reverse transcriptase inhibitors (NRTI) or nucleotide and NRTI, and either: (1) a protease inhibitor (PI); (2) a non-nucleoside RTI (NNRTI); (3) a third NRTI; or (4) a ritonavir-boosted PI (BPI). Week 24 and 48 results for the proportions of patients with plasma HIV RNA levels < 400 and < 50 copies/ml, and change in CD4+ cell counts were recorded.
Results:
Fifty-three trials met the entry criteria, and enrolled 14 264 patients into 90 treatment arms.
Overall 55% of patients had plasma HIV RNA levels < 50 copies/ml at week 48 and this percentage increased with later publication dates.
In unadjusted pairwise comparisons at week 48, significantly greater percentages of patients receiving NNRTI (64%) and BPI (64%) had RNA < 50 copies/ml than NRTI (54%) or PI (43%), and CD4+ cell count increases were significantly greater in the BPI group (+200 cells/_l) than the PI (+179), NNRTI (+173), or NRTI (+161) groups.
Pill count and percentage of patients with week 48 plasma HIV RNA levels < 50 copies/ml were correlated in the univariate analysis (P = 0.0053; r = -0.323), but pill count was not a significant predictor in the multivariate analyses.
Drug class and baseline CD4+ cell counts were significant predictors, but explained only a modest amount of the treatment effect, (R 2 = 0.355).
Conclusions: NNRTI and BPI-containing regimens offer superior virologic suppression over 48 weeks, supporting existing guidelines for the choice of initial ART. Pill count was not a consistent predictor of virologic suppression.
AUTHOR DISCUSSION
The results of this expanded overview offer important new observations on the success of contemporary antiretroviral treatment regimens in trials studying ART-naive populations. The percentages of patients achieving plasma HIV RNA levels < 50 copies/ml at 48 weeks is increasing over time, and in reports from most recent trials, more than 80% of patients may reach this threshold [121-123]. The improving virologic responses were seen despite the enrollment of patient populations with lower CD4+ cell counts at entry, and surprisingly, lower entry CD4+ cell counts were associated with improved responses in some analyses. Drug class was the most significant predictor of regimen potency, with the best responses achieved by NNRTI and BPI-containing regimens. Pill count was not consistently identified as a predictor of virologic response, in contrast to our previously published overview.
The improving response rates over time are encouraging evidence for the identification of new treatment regimens and their strategic use in ART-naive patients. Similar observations have been reported in numerous observational cohorts [124-127] and a portion of the declining risk of virologic failure over time could be attributed to the choice of starting regimen [127]. The multivariable linear regression in our study offered the opportunity to assess the predictive value of publication date for plasma HIV RNA levels < 50 copies/ml in comparison with other factors such as drug class. The results demonstrate that improving virologic success is due to the use of more potent drug classes, and not simply a function of publication date. The superiority of NNRTI and BPI-containing regimens over PI and NRTI-containing regimens in suppressing plasma HIV RNA levels below 50 copies/ml observed in this overview is supported by the results of smaller randomized clinical trials [128-131] and provides an additional evidence base consistent with current therapeutic guidelines. Greater CD4+cell count increases were seen in patients receiving BPI-containing than NNRTI, PI and NRTI-containing regimens after 48 weeks. Importantly, the superior CD4+ cell count increases were seen in both the unadjusted analyses and were associated with drug class in the multivariable linear regression analyses. The potential mechanisms for the greater CD4+cell count increases on BPI-containing regimens are uncertain, and must be confirmed in additional randomized studies, especially those directly comparing BPI and NNRTI-containing regimens given the apparent lack of differences between them in observed antiretroviral potency.
Additional factors such as drug-related toxicities and the resistance consequences of first regimen failure may also influence the choice of first line treatment regimens. This analysis considered drug-related toxicities only to the extent that toxicities resulted in treatment discontinuation, and therefore affected the ITT:M = F results. Certainly, many of the toxicities of NNRTI and BPI-containing regimens [132-134] may be non-overlapping, thus allowing for individualization of treatment choices. Nonetheless, the most robust recommendations on drug-related toxicities should come from directly comparative, randomized clinical trials. Drug resistance appears to be less commonly reported from randomized clinical trials among patients receiving BPI-containing regimens, interestingly in both the protease and reverse transcriptase genes. A recent systematic overview identified a lower resistance cost associated with the use of BPI versus NNRTI-containing regimens [135]. However, once again these analyses cannot substitute for resistance results generated from randomized clinical trials comparing the two approaches.
In contrast to our previously published overview, pill count was not identified as a significant predictor of plasma HIV RNA levels below 50 copies/ml at week 48 in the multivariable analysis. However, pill count should not be forgotten as a potential factor influencing the outcome of treatment interventions; pill count was predictive in the current univariable and dichotomous (≦ 10 versus > 10) analyses. Why did the current analysis not duplicate the results from the previous overview? A number of factors must be considered in assessing the contrasting results. The current overview is much larger with more participants, more trials, more unique treatment regimens, and a broader range of pill counts, and as a result, findings from the current overview are more robust. The current overview included a greatly expanded number of patients receiving NNRTI-containing regimens and BPI-containing regimens, identified as the most potent drug classes and therefore diminishing the impact of pill counts.
There are a number of important limitations to the current overview. Overviews cannot replace randomized clinical trials, which provide the most robust evidence to guide clinical practice. The data used for this overview were generated from population-based results and not individual patients, and therefore the overview may lack the ability to discern some predictors of treatment response, especially baseline characteristics. Analyses focused on the contribution of a third agent to regimen success, and it is possible that the nucleoside or nucleotide reverse transcriptase inhibitor component may affect these results. A recently completed trial highlighted the potential importance of these agents [123]. In the multivariable logistic regression analysis, the R 2 value for the model predicting plasma HIV RNA levels < 50 copies/ml at week 48 was 0.355, reflecting the importance of other unmeasured factors in contributing to treatment responses. In addition, some trials which prematurely closed such as ESS 30009 are not analyzed in this overview [136]. Finally, the time span of included trials now exceeds 10 years, and during this period substantial improvements in care have been implemented beyond the choice of treatment regimen such as improved care access and delivery, monitoring, adherence education, and support for co-morbidities.
Efavirenz was studied in the greatest number of trials, in the largest number of patients, and in more diverse regimens than any other single agent. Efavirenz-containing regimens have provided impressive results in recent trials with suppression of plasma HIV RNA levels below 50 copies/ml in approximately 80% of patients in ITT:M = F analysis, [121-123]. However, NNRTI-containing regimens have been associated with higher resistance costs when virologic failure occurs [135], which may compromise response to subsequent treatment regimens. Important results will be generated from randomized trials which are currently underway comparing directly NNRTI-containing versus BPI-containing regimens (ACTG 5142, 5202).
In summary, this overview has identified improving temporal trends in virologic responses, superior virologic responses for NNRTI and BPI-containing regimens, and an inconsistent and less pivotal role for lowered pill counts.
Results
Descriptive characteristics of included trials
Fifty-three trials met the inclusion criteria and are listed in Table 1, (representing the addition of 30 trials since the previously published overview). These 53 trials enrolled 14 264 patients into 90 independent treatment arms. The median baseline plasma HIV RNA level was 4.75 log10 copies/ml and the median baseline CD4+ cell count was 315/_l. Seventy-six percent of patients were male.
Design characteristics and regimens studied
Thirty-four trials were randomized and 19 included only one treatment arm. Thirty-two trials were open-label, 19 were double-blinded, and two were partially blinded. A total of 90 treatment arms were included in these 53 trials with 40 unique treatment regimens. NNRTI-containing regimens were most commonly studied (n = 38), followed by PI-containing (n = 32), NRTI-containing (n = 12) and BPI-containing regimens (n = 8). Efavirenz-containing regimens were most commonly studied; they were included in 20 different trials and 29 treatment regimens in 5798 individuals. In addition, efavirenz-containing regimens were combined with more NRTI combinations than any other drug, (nine stavudine and lamivudine, eight zidovudine and lamivudine, four abacavir and lamivudine, three stavudine and didanosine, two didanosine and lamivudine, two didanosine and emtricitabine, one tenofovir and lamivudine). In comparison with the previous overview, there was a slight decrease in the representation of PI-containing regimens [15/31 regimens in 2001 (48%) versus 32/90 in 2005 (36%)], no difference in the representation of NNRTI-containing regimens [13/31 regimens in 2001 (41%) versus 38/90 in 2005 (42%)], a substantial increase in the number of efavirenz-containing regimens (five in 2001 versus 29 in 2005), and the addition of BPI-containing regimens. The median pill count was 10/day, with a range of 3-22/day.
Temporal trends in baseline characteristics and responses by publication date
In the previous overview, the median baseline plasma HIV RNA levels and CD4+ cell counts were 4.69 log10 copies/ml and 375/_l. In the current update, the values were 4.75 log10 copies/ml (P = 0.10) and 315/_l (P = 0.001). Over the period of observation, the weighted mean responses for the percentage of patients with plasma HIV RNA levels < 50 copies/ml at week 48 steadily improved, (prior to 1998, 41%; 1999-2000, 50%; 2001-2002, 56%; 2003-2004, 64%), (Fig. 1). During the same periods of time the weighted mean responses for increases in CD4+ cell counts at week 48 were similar, (prior to 1998, +173 cells/_l; 1999-2000, +174 cells/_l; 2001-2002, +180 cells/_l; 2003-2004, +169 cells/_l) (Fig. 1).
Virologic and immunologic responses
The virologic and immunologic responses at weeks 24 and 48 are shown in Table 2 with the number of treatment arms and patients, the weighted means, and the 95% CI. These responses are unadjusted for baseline characteristics, and are presented as the percentages with plasma HIV RNA levels below 400 and 50 copies/ml and the change from baseline in CD4+ cell counts by drug class. Overall, 55% of patients had plasma HIV RNA levels below 50 copies/ml at week 48 with a CD4+ cell count change from baseline of +176 cells/_l. In unadjusted pairwise comparisons of the proportions of patients with plasma HIV RNA < 50 copies/ml at week 48, the BPI and NNRTI groups were each superior to the PI and NRTI groups (BPI versus PI, P < 0.001; BPI versus NRTI, P = 0.017; NNRTI versus PI, P < 0.001; NNRTI versus NRTI, P = 0.001), but were not significantly different from each other, (BPI versus NNRTI, P = 0.945). The NRTI group was also superior to the PI group in pairwise comparisons (P = 0.001). In pairwise comparisons of the unadjusted changes from baseline in CD4+ cell counts, the BPI group was superior to NNRTI, PI and NRTI groups (BPI versus NNRTI, P < 0.001; BPI versus PI, P = 0.004; BPI versus NRTI, P < 0.001). The PI and NNRTI groups did not have significantly greater increases in CD4+ cell counts than the NRTI group, and were not significantly different from each other.
Figure 2 displays a bubble plot of the relationship between pill count and plasma HIV RNA levels < 50 copies/ml at week 48. In a univariate analysis this relationship was significant (P = 0.0053, r = -0.323). However, when evaluated using multivariable linear regression analysis, this relationship was no longer significant. Additional analyses were performed to adjust for baseline differences and identify predictors of virologic and immunologic responses (Table 3). Lower baseline CD4+ cell counts were associated with increased percentages of plasma HIV RNA < 50 copies/ml at weeks 24 and 48, and greater CD4+ cell count increases at week 48. Baseline plasma HIV RNA levels were not consistently associated with plasma HIV RNA responses or CD4+ cell count increases. Drug class was significantly associated with plasma HIV RNA levels < 400 copies/ml at week 24, plasma HIV RNA levels < 50 copies/ml at weeks 24 and 48, and with increases in CD4+ cell counts at week 48. Lower pill count was associated with higher percentages of plasma HIV RNA levels < 50 copies/ml at week 24 but not with plasma HIV RNA < 50 copies/ml at week 48. Higher pill counts were associated with a greater percentage of patients having plasma HIV RNA < 400 copies/ml at week 48. Publication interval was only predictive of plasma HIV RNA levels < 400 copies/ml at week 24. The modeling of all factors had limited predictive value with a maximum R 2 value of 0.436 for plasma HIV RNA levels < 400 copies/ml at week 48, and an R 2 value of 0.355 for plasma HIV RNA levels < 50 copies/ml at week 48.
Further analyses were undertaken to explore the relationship between pill count and the weighted mean response of week 48 plasma HIV RNA levels < 50 copies/ml. When pill count was analyzed as a continuous variable in all patients with week 48 results (79 treatment arms and 13 558 participants), no significant overall relationship was identified. Pill counts were next analyzed by drug class dichotomously (≦ 10 pills/day and > 10 pills/day), continuously within these two categories ≦ 10 or > 10 pills/day), when one study with disproportionately low response rates (AI424-034) was excluded in a sensitivity analysis, and as a continuous variable within each drug class. In six of nine secondary analyses no significant association was identified; in the analyses examining dichotomous categories (≦ 10 versus > 10) and within the NRTI class results favored lower pill counts, and in the PI class the results favored higher pill counts.
METHODS
Study selection
A search of public domain and recent conference presentations was conducted for the period from 1994 to July 2004. A MEDLINE search was undertaken with the following keywords: clinical trial, plasma HIV-1 RNA, HAART, antiretroviral therapy, and naive. In addition, the package inserts for US Food and Drug Administration (FDA)-approved antiretroviral drugs were examined and listed trials were reviewed. Proceedings from the following conferences were reviewed: Conference on Retroviruses and Opportunistic Infections (CROI), the World AIDS Conference, the International AIDS Society (IAS) Conference on HIV Pathogenesis and Treatment, the European Conference on Clinical Aspects and Treatment of HIV Infection, the Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), the International Conference on Drug Therapy in HIV Infection (ICDT), and the Annual Meeting of the Infectious Diseases Society of America (IDSA). Triple combination therapy was defined as either two nucleoside or a nucleoside and a nucleotide reverse transcriptase inhibitor and either: (1) a protease inhibitor (PI); (2) a non-nucleoside reverse transcriptase inhibitor (NNRTI); (3) a third nucleoside reverse transcriptase inhibitor (NRTI); or (4) a ritonavir-boosted protease inhibitor (BPI). Regimens were assigned to these four categories based upon the class of the third agent in the regimen, (PI, NNRTI, NRTI and BPI; the inclusion of BPI-containing regimens represents an addition since the previous overview). Trials were included if they were at least 24 weeks in duration, had at least 30 patients per treatment arm, included chronically HIV-infected patients who were treatment-naive or had very limited prior exposure to antiretroviral therapy (< 2 weeks prior NRTI exposure), and reported the percentages of patients with undetectable plasma HIV RNA levels (< 400 and < 50 copies/ml) using an intent-to-treat missing equals failure analysis, (ITT:M = F). Trials were excluded if they studied induction/maintenance strategies, acutely or recently infected patient populations (< 6 months since seroconversion), or experimental treatments which did not successfully receive US FDA approval.
Data collection
Following inclusion of a trial in the overview, all published or presented source documentation was reviewed for each trial, (journal citations, conference abstract books, posters, slides, package inserts, etc.). Information on each trial was abstracted on the following trial characteristics and results: (1) trial design: treatment regimens, open-label or blinded, randomization if used, control groups, and average daily pill count (defined as the number of pills or tablets/day including placebo) in the treatment regimen; (2) baseline characteristics: number of individuals enrolled, percentage male, race, age, log10 plasma HIV RNA, and CD4+ cell counts; and (3) response rates: percentages of patients with plasma HIV RNA < 400 and < 50 copies/ml at 24 and 48 weeks in the ITT M = F populations, and change in CD4+ cell counts at 24 and 48 weeks in the as-treated populations. The date of a study's first presentation or publication was defined as the 'publication date'. Publication dates were divided into the following intervals; prior to 1998, 1999-2000, 2001-2002 and 2003-2004. Abstractions were performed by one reviewer and were confirmed by a second; any discrepancies were reconciled by conference with the study team. If median values were unavailable then the mean was used. All trials reported plasma HIV RNA results with the Roche Amplicor assay (< 400 copies/ml) and the Roche Ultrasensitive assay (< 50 copies/ml) (Roche Diagnostics, Indianapolis, Indiana, USA) with the exception of the START I and II trials and the VIRGO I/II trials, where the results were reported as the percentage < 500 copies/ml with the Chiron bDNA assay (Chiron Corporation, Emeryville, California, USA).
Statistical methods
For plasma HIV RNA responses, the primary analysis utilized reported results for ITT:M = F population, and for changes in CD4+ cell counts, the primary analysis used reported results for the as-treated population. When ITT:M = F results were unavailable, the ITT:M = F results were calculated using the number of patients with plasma HIV RNA undetectable at weeks 24 or 48 divided by the number of enrolled patients. Response rates for each of the four drug classes (PI, NNRTI, NRTI and BPI) were estimated with the number of treatment groups and patients contributing to the estimate. These resulting estimates were then compared by constructing confidence intervals (CI) for the differences in response rates between drug classes. Correlation analyses and weighted least squares multivariate linear regression analyses (MLR) using backwards stepwise selection procedures were used to assess the variability in treatment group response rates as a function of the following factors; baseline log10 plasma HIV RNA level and CD4+ cell count, drug class, and average daily pill count. For the MLR analysis of categorical virologic endpoints, the natural log of the response rate was used. Statistical significance was determined using alpha = 0.05. Bubble plots were constructed, and the results for each treatment arm were weighted in direct proportion to the numbers of patients enrolled. Sensitivity analyses were conducted to evaluate different factors including dichotomous pill counts (≦ 10/day versus > 10/day), analyses of continuous pill counts over these ranges (3-10/day and 11-22/day), and pill counts as predictors of response within each regimen category.
|
|
| |
| |
|
|
|
