 |
|
|
| |
Faster Death May Be the Price Paid for Delayed Switching After Failure
|
| |
| |
15th Conference on Retroviruses and Opportunistic Infections
February 3-6, 2008
Boston
Mark Mascolini
Delaying a drug switch after failure of a nonnucleoside (NNRTI) regimen raised the risk of death almost 25% for every extra 3 months of lag time, according to analysis of two large prospective US cohorts [1]. Delayed switching after failure of a PI seemed to lower the risk of death, at least in the short term.
But this study has one clear limitation--most first failures analyzed happened in the late 1990s and early 2000s, when switch options remained much narrower than they are today.
Guidelines recommend starting a new regimen quickly after virologic failure. But Maya Petersen (University of California, Berkeley) and colleagues at other sites noted that switching often gets delayed because of sporadic follow-up, slow detection of failure, and lack of options for new regimens. Petersen used a statistical method called marginal structural modeling to correct for a likely bias in drug switching: Physicians tend to swap regimens quickly for people with a poor prognosis but may dally more when people have a good prognosis. Marginal structural models mimic randomized controlled trials in which patients remain on a failing regimen for a random duration before switching to a new combination.
To get a handle on clinical consequences of tardy switching after failure, the investigators analyzed 982 people after their first confirmed virologic failure at the University of North Carolina (UNC), Chapel Hill, and Johns Hopkins Hospital in Baltimore. They defined failure as consecutive viral loads above 1000 copies 12 to 24 weeks after starting therapy or above 500 copies after week 24. A regimen switch meant starting a new antiretroviral class or at least two drugs not in the failing regimen.
In 76% of the combined cohorts, a first-line PI regimen failed, and 32% of those failures involved a ritonavir-boosted PI. Among people taking a non-PI regimen, almost all took an NNRTI. About 80% of cohort members were not white, and about three quarters were men. Median dates of first HAART failure were September 1998 for Hopkins patients taking a PI, January 2000 for UNC patients taking a PI, April 2001 for Hopkins patients taking a non-PI regimen, and August 2002 for UNC patients taking a non-PI regimen.
After failure of a non-PI regimen, every additional 3 months of delay in starting a new combination raised the risk of death 23% (hazard ratio [HR] 1.23, 95% confidence interval [CI] 1.08 to 1.40, P = 0.002) while raising the risk of immunologic failure or death 21% (RH 1.21, 95% CI 1.07 to 1.36, P = 0.002). In contrast, putting off a switch after PI failure appeared to lower the risk of death 7% for every 3 months (RH 0.93, 95% CI 0.87 to 0.99, P = 0.03) and to have little effect on immunologic failure or death. Failure of a second non-PI or PI regimen did not have a significant impact on risk of death or risk of immunologic failure or death. The impact of delayed switching was similar in the two cohorts.
Next Petersen used a method called weighted data-adaptive regression to gauge interactions between switch delays and elapsed time since failure. This analysis showed a sharp increase in death risk when people stayed on a failing non-PI regimen. That climbing death risk hit a plateau when a backup regimen began but remained elevated despite the change. After failure of a PI regimen this method saw little difference in short-term mortality risk between people who modified therapy immediately and those who delayed--at least over the short term. But overall mortality risk rose over time after PI failure.
Petersen and coworkers posited two main conclusions and one warning:
· Delayed modification of a failing NNRTI regimen boosts the risk of progression and death.
· Delayed modification of a failing PI regimen has a slower impact than delayed switching from an NNRTI regimen.
· Most people in low-income countries--where viral load monitoring may be unavailable--start NNRTI regimens.
What explains the clinical progression difference after failure of an NNRTI versus a PI? Petersen and colleagues (who included Steven Deeks at the University of California, San Francisco, Richard Moore at Johns Hopkins, and Joseph Eron at UNC) did not venture answers to that question. But in introducing their study, they noted three forgiving features of PI failure:
First, mutations making virus resistant to PIs--especially ritonavir-boosted PIs--emerge more slowly than mutations to NNRTIs and stir up less cross-resistance. Second, PI mutations that do evolve may make HIV less fit than mutations to other classes. And third, CD4 counts often remain stable after failure of a PI regimen.
Reference
1. Petersen M, van der Laan M, Napravnik S, et al. Long-term consequences of the delay between virologic failure of HAART and regimen modification: a prospective cohort study. 15th Conference on Retroviruses and Opportunistic Infections. February 3-6, 2008. Boston. Abstract 798.
|
| |
|
|
|
|
|
