 |
|
|
| |
Abacavir (But Not Efavirenz) Weaker in Cohort Study Than in Clinical Trial
|
| |
| |
15th Conference on Retroviruses and Opportunistic Infections
February 3-6, 2008
Boston
Mark Mascolini
Virologic response to abacavir by previously untreated people proved significantly worse in the "real-life" conditions of a large cohort study than in an AIDS Clinical Trials Group (ACTG) randomized trial [1]. But treatment-naive people responded as well to efavirenz in the cohort study as they did in the ACTG study.
ACTG A5095 and ART Cohort Collaboration (ART-CC) investigators undertook the comparison because antiretroviral "efficacy" in clinical trials often differs from "effectiveness" of the same drug in practice. To explore 24-week virologic responses and 48-week clinical responses to abacavir and efavirenz in people starting their first regimen, Michael Mugavero (University of Alabama, Birmingham) and colleagues compared results in 377 people starting abacavir and 376 starting efavirenz in A5095 versus 2009 patients beginning abacavir and 2426 beginning efavirenz in one of 15 North American or European cohorts.
ACTG investigators closed the abacavir-containing triple-nucleoside arm of A5095 early because people responded more poorly to that regimen than to efavirenz-containing combinations [2]. The new analysis did not detail which drugs ART-CC patients started with abacavir.
Possible prescribing biases in the ART-CC data include heavier use of efavirenz than abacavir for men (80% versus 72%) and higher median CD4 counts in people starting abacavir than efavirenz (251 versus 209). But the ACTG and cohort groups matched closely in age and pretreatment viral load (around 60,000 copies) regardless of which drug they used. Starting median CD4 counts measured 197 in the ACTG abacavir group, 251 in the ART-CC abacavir group, 201 in the ACTG efavirenz group, and 209 in the ART-CC efavirenz group.
Defining virologic failure as a week 24 viral load above 400 copies, Mugavero found significantly fewer overall failures in ACTG A5095 than in the cohorts (13.8% versus 18.6%, P < 0.01). At the 24-week point efavirenz virologic failure rates were equivalent in ACTG A5095 and in the ART-CC (odds ratio [OR] 0.78, 95% confidence interval [CI] 0.54 to 1.13).
But abacavir failed significantly less often in the ACTG study than in the ART-CC. People in the "real-life" cohorts had a 55% higher risk of abacavir failure than people in the randomized trial (OR 0.45, 95% CI 0.32 to 0.64). The differing antiviral "efficacy" in A5095 and "effectiveness" in the ART-CC for abacavir versus efavirenz reached statistical significance (OR 0.58, 95% CI 0.35 to 0.97).
Significantly more people in the ART-CC cohorts than in the ACTG trial progressed to AIDS or death by week 48 regardless of whether they took abacavir or efavirenz (6.9% versus 3.1%, P < 0.01). So abacavir and efavirenz did not differ in clinical outcomes over the short term of 48 weeks.
A study like this cannot extract the reasons for abacavir's apparently better performance in a randomized trial than in clinic patients enrolled in big cohorts. But Mugavero and colleagues offered some speculation:
· The statistical models may have overlooked factors that could sway results one way or the other. Such factors may include reasons why physicians prescribe abacavir or efavirenz in practice.
· Antiretrovirals may have intrinsic "efficacy" and "effectiveness" differences that could reflect differing adherence and resistance relationships.
The higher progression rate in the ART-CC versus ACTG A5095--regardless of treatment with abacavir or efavirenz--could reflect biases related to which people volunteer for trials, which antiretrovirals physicians prescribe for whom in practice, and closer follow-up in trials than in the clinic.
References
1. Mugavero MJ, May M, Ribaudo HJ, et al. Outcomes of abacavir and efavirenz-based HAART: comparison of ACTG 5095 trial results with observational cohort studies. 15th Conference on Retroviruses and Opportunistic Infections. February 3-6, 2008. Boston. Abstract 783.
2. Gulick RM, Ribaudo HJ, Shikuma CM, et al. Triple-nucleoside regimens versus efavirenz-containing regimens for the initial treatment of HIV-1 infection. N Engl J Med. 2004;350:1850-1861.
ABSTRACT
Outcomes of Abacavir- and Efavirenz-based HAART: Comparison of ACTG 5095 Trial Results with Observational Cohort Studies
Michael Mugavero*1, M May2, H Ribaudo3, J Sterne2, S Napravnik4, M Egger5, M Saag1, R Gulick6, and ACTG and ART-CC
1Univ of Alabama at Birmingham, US; 2Univ of Bristol, UK; 3Harvard Sch of Publ Hlth, Boston, MA, US; 4Univ of North Carolina at Chapel Hill, US; 5Univ of Berne, Switzerland; and 6Weill Med Coll of Cornell Univ, New York, NY, US
Background: Randomized controlled trials are the study design of choice for evaluating the effects of ART, but trials often focus on virologic suppression, rather than clinical endpoints. Large observational cohort studies have the power to examine clinical outcomes, but results may be affected by confounding by indication. We compared the ACTG 5095 trial (A5095) with data from the ART Cohort Collaboration (ART-CC), 15 cohort studies in Europe and North America.
Methods: We compared efavirenz (EFV) vs abacavir (ABC) (combined with zidovudine/lamivudine [ZDV/3TC] or stavudine [d4T]/3TC) stratifying by study design, and A5095 vs ART-CC stratifying by third drug in ARV-naive patients starting ART. A5095 data and ART-CC data were analyzed using identical logistic regression models to evaluate 24-week virologic failure (>400 copies/mL), and Cox models to compare 48-week progression to AIDS or death, adjusting for age, sex, baseline CD4 count and viral load.
Results: We identified 5796 patients who met inclusion criteria: 753 from A5095 (ABC = 377, EFV = 376) and 5043 from ART-CC (ABC = 2009, EFV = 3034). Overall, virologic failure was observed in 13.8% of A5095 patients (89 of 643) and 18.6% of ART-CC patients (813 of 4368) with available 24-week viral load measures (p <0.01) and 3.1% (23 of 753), and 6.9% (350 of 5043) progressed to AIDS or death (p <0.01). Prognostic factors were well balanced in A5095. In ART-CC, patients starting with ABC had higher median CD4 counts than patients starting with EFV (251 vs 209 cells).
Conclusions: Patients on EFV were less likely to experience virologic failure in both A5095 and ART-CC. Of note, EFV-based regimens performed similarly in ART-CC and A5095, while virologic responses to ABC-regimens were considerably better in A5095 compared to ART-CC. There was no clear evidence of differences in clinical progression between EFV and ABC, but the attenuation of the hazard ratio for AIDS/death after adjustment for covariates in the ART-CC suggests residual confounding by indication in the observational data. Clinical progression rates were higher in ART-CC than A5095, independently of the regimen used.
|
| |
|
|
|
|
|
