To the Editor: In their randomized controlled trial, Dr Cruz and colleagues1 reported that polymyxin B hemoperfusion reduced 28-day mortality in patients with severe sepsis or septic shock. I believe that this conclusion is wrong. The statistically significant difference in mortality rates reported in this study was based on the analysis of a hazard ratio generated by a Cox proportional hazards regression survival model. However, the odds ratio (OR) of the crude 28-day mortality rates did not reach statistical significance (11/34 vs 16/30; OR, 0.42; 95% confidence interval [CI], 0.13-1.29; P = .13). Such a difference between the hazard ratio and the OR can only be explained by a later occurrence of deaths in the treated group, an effect that is suggested in Figure 3.1 This observation is concerning in an unblinded trial, in which investigators could forgo life support later in patients in the treatment group, creating a false appearance of improved survival in that group early in the study. It would therefore be helpful to know the mortality rates at 60 or 90 days; the narrowing difference between mortality rates that seems to be occurring over 28 days in Figure 3 may have decreased further over longer periods.
Vincent J. Polymyxin B Hemoperfusion and Mortality in Abdominal Septic Shock. JAMA. 2009;302(18):1968. doi:10.1001/jama.2009.1606