In this study, several large cohort studies combined their data and tried to answer a few questions. For several reasons, I am not a fan of this type of study. However, the number of patients included in such a study is often phenomenal, the P values have an enormous amount of zeros and, on occasion, you can get some interesting information. Unfortunately, you are frequently left with an uneasy feeling that your results will get screwed up. Because of this, I do not put a lot of faith in the results of this kind of study.

The authors of this study examined the mortality of HIV-infected individuals based on their initial antiretroviral regimen. The 12 cohorts (from 3 different continents) looked at patients who had started highly active antiretroviral therapy (HAART) and examined their risk of death according to 8 different "third" drugs: efavirenz (EFV, Sustiva, Stocrin), nevirapine (NVP, Viramune), nelfinavir (NFV, Viracept), indinavir (IDV, Crixivan), ritonavir (RTV, Norvir), a ritonavir-boosted protease inhibitor (amprenavir [APV, Agenerase], lopinavir [LPV], saquinavir [SQV, Fortovase, Invirase]), saquinavir soft and hard gel, and abacavir (ABC, Ziagen). They also compared the nucleoside pairs zidovudine (AZT, Retrovir) + lamivudine (3TC, Epivir) and stavudine (d4T, Zerit) + lamivudine.

The numbers were extraordinary: 17,666 patients with more than 55,000 patient-years of follow-up, and a total of 895 deaths.

The bottom line is that there were not really any significant mortality differences based on the drug that was initiated. It looked like nevirapine-based therapy had a higher hazard ratio (1.65, 1.16-2.36) than the rest of the third drugs and that stavudine + lamivudine had a higher hazard ratio than zidovudine + lamivudine (1.35, 1.14-1.59).

The problem with studies like this is that it is enormously difficult to control for all the confounding variables. For example, it is likely that patients with certain characteristics might be given some specific medication. Nevirapine, for instance, may be given to intravenous drug users. If those patients have an increased mortality, it could look like nevirapine is associated with the negative outcome, when in fact it has nothing to do with it.

Although an attempt was made to control for variables like CD4+ cell count and viral load, age or gender at the time of treatment initiation, it is simply impossible to control for all the variables. Even the authors recognized this problem and in their abstract suggested that the variability in prognosis was due to "unmeasured confounding by indication" rather than difference between the regimens.

In conclusion, although this study showed that there are no significant differences in mortality depending on baseline regimen selected, this does not mean that all regimens are equally effective and good. Which, in the end, leaves us in much the same situation as before the presentation of this poster and with no new information.