Reports

This report is based on medical evidence presented at sanctioned medical congress, from peer reviewed literature or opinion provided by a qualified healthcare practitioner. The consumption of the information contained within this report is intended for qualified Canadian healthcare practitioners only.

17th Annual Canadian Conference on HIV/AIDS Research

Montreal, Quebec / April 24-27, 2008

As the HIV population matures, more attention is being paid to health risks that may not be directly related to the effects of infection. Of these, the cardiovascular (CV) system attracts a substantial proportion of attention for many reasons, including the importance of CV disease as a cause of death in individuals over the age of 50, the high prevalence of CV risk factors in an HIV population, and the accelerated CV risk associated with antiretrovirals that alter lipid metabolism. In a D:A:D (Data Collection on Adverse Events of Anti-HIV Drugs) analysis presented several months ago, launched initially to evaluate the contribution of lipid-altering nucleoside reverse transcriptase inhibitors (NRTIs) to CV risk, there was a surprise finding: the NRTIs that did not affect lipids rather than those that did appeared to increase CV risk.

However, subsequent efforts to reproduce this risk have failed. Unlike the association of COX-2 inhibitors with CV risk, which was seen in both observational and randomized trials, “we are not seeing [an association between lipid-neutral NRTIs and CV risk] in the randomized trials so far,” observed CAHR chair Dr. Marina B. Klein, Assistant Professor of Medicine, McGill University, Montreal.

D:A:D Analysis Results

In the most recent D:A:D analysis, the pooled data from 33,347 patients in the 11 cohorts were tested for the hypothesis that the NRTIs zidovudine (AZT) and stavudine (d4T) were contributing to the increased risk of myocardial infarction (MI) that had previously been associated with HAART containing a protease inhibitor (PI). In the analysis, abacavir (ABC) and didanosine (ddI) were employed as controls because of their lipid neutrality. However, the results were the opposite. D:A:D associated both ABC and ddI with an increased risk but did not see such an association with AZT or d4T.

The Search for Corroborative Findings

This observation first reported at the 2008 Conference on Retroviruses and Opportunistic Infections (CROI) immediately launched an effort to corroborate the findings. The focus has largely been on ABC because it has been listed as a preferred agent in several treatment guidelines. Not surprisingly, the first step was to go back to the randomized data. There have been 54 controlled trials with ABC randomizing approximately 14,600 patients. When any HAART regimen, including those with a PI or a non-NRTI (NNRTI) that contained ABC in these trials was compared to any HAART regimen without ABC, no trend or indication for a difference in CV risk could be detected. Indeed, the absolute number of MIs or other ischemic events was greater among those not receiving ABC.

When an independent cohort analysis was launched from a Veterans Affairs (VA) dataset of 41,213 patients, a group with an older average age than in the D:A:D cohort, the same result was produced. No signal of a difference in MI, ischemic events, or all-cause mortality was observed among those who did or did not receive ABC. In this population, which had a higher proportion of African Americans than D:A:D, 26.7% had a major modifiable risk factor, such as hypertension or diabetes. According to Dr. Jaime Hernández, Triangle Park, North Carolina, the analyses of the randomized data and the VA dataset are being prepared for publication and presentation at the 2008 International AIDS Conference in Mexico City. The complete absence of any corroboration with D:A:D is striking.

Confounding and Biological Credibility

According to Dr. Klein, the risk of confounding in observational trials is why randomized trials are necessary. As an example, Dr. Klein described a phenomenon known as channelling in which one compound may be preferentially used in a specific group of patients, thereby creating a false association. In this case, the perception of greater CV safety for a lipid-neutral NRTI could have channelled ABC to those patients at greatest CV risk. It is not clear whether this phenomenon explains the recent results of D:A:D, but Dr. Klein noted that several of the criteria that strengthen observational associations were not met in the D:A:D association of ABC with CV risk. This includes a biological credibility for the effect observed and consistency of data across other datasets.

The absence of biological credibility prior to the D:A:D observation has been reinforced in a series of subsequent studies conducted to re-examine activities that might increase CV risk. None were found. In one of the studies, lipid activity was examined in five treatment arms from three prospective studies in antiretroviral treatment-naïve subjects (Hernández et al. Can J Infect Dis 2008;19(suppl A):62A, Abstract P-169). The studies included several different NRTI combinations with the NNRTI efavirenz or the PI nelfinavir. The authors found that independent of the NNRTI or PI, the combinations that included ABC had the most favourable lipid profile, whereas a combination of d4T and lamivudine (3TC) produced the largest increases in total cholesterol, LDL-C levels and triglycerides.

In another study, the relative effects of combination ABC/AZT/3TC plus nelfinavir were compared to d4T/3TC and nelfinavir over 96 weeks in 261 non-diabetic subjects (Hernández et al. Can J Infect Dis 2008;19(suppl A):63A, Abstract P-171). Consistent with previous studies, there was very little change in plasma glucose, fasting plasma glucose, or plasma insulin levels over the course of follow-up in either arm, and there was no difference between arms.

In a much smaller study conducted over a shorter time frame, the goal was to evaluate changes in tissue plasminogen activator (tPA) and plasminogen activator inhibitor- 1 (PAI-1), which are both important determinants of endogenous anti-thrombotic effect (Hernández et al. Can J Infect Dis 2008; 19 (suppl A):63A, Abstract P-172). In this study, changes in tPA and PAI-1 antigens were measured over eight weeks in 14 non-diabetic subjects given ABC/3TC plus the PI amprenavir. Over this period, there was a marked and significant reduction (P=0.02) in tPA antigen, a predictor of greater thrombolytic activity, and no change in PAI-1 antigen. The data suggest that improvement in a marker of impaired thrombolysis can occur with antiretroviral therapy that contains ABC. The authors speculated that the benefit may be due to a reduction in HIV-related inflammation. Observational studies have played an important role in identifying clinical effects not seen in clinical trials, but there are many well-known examples when an action could not be subsequently corroborated. Dr. Klein cited the widespread use of estrogen replacement therapy to achieve the CV protection identified in observational studies but not confirmed when a prospective randomized study was finally conducted. She cautioned that causality is often elusive. For example, she noted that alcohol might easily be mistaken as a cause of lung cancer because of the frequency with which smokers drink.

Dr. Klein implied that the recent D:A:D results were likely a product of the type of confounding to which observational studies are vulnerable. She emphasized that the “confounding which occurs in all observational studies can only be eliminated in randomized trials.” She suggested that in addition to a biological credibility for the effect observed, consistency across multiple studies, the strength of the statistical associations observed, and a dose response are among factors that need to be considered when assessing an observational result. Based on the lack of consistency of evidence, there has been no change in the recommendations of ABC as a preferred NRTI in current HAART regimens.

Summary

Intensive efforts to corroborate an association between ABC and increased CV risk, initially reported from an evaluation of D:A:D observational data, have so far been unproductive. These included an analysis of a separate patient cohort with more than 41,000 patients and a re-analysis of all prospective randomized trials conducted with ABC. The failure to identify any signal of increased CV risk in the controlled trials is particularly significant due to the importance of blinded randomized studies for objective conclusions.