Genetic association studies related to septic shock have traditionally evaluated 2 end points: susceptibility to development of septic shock and mortality due to septic shock. Recently, Lorenz et al1 reported an association between toll-like receptor 4 (TLR4) mutations and susceptibility to gram-negative septic shock in the ARCHIVES. The C3H/HeJ mice, which are lipopolysaccharide hyporesponsive due to mutations within the Toll gene, were more susceptible to dissemination of gram-negative bacteria.2 In in vitro studies, when cells with the Asp299Gly mutation were exposed to lipopolysaccharide, they released less tumor necrosis factor α and interleukin 1, which are important mediators of septic shock.3 Although it is difficult to extrapolate these results to humans, one may hypothesize that patients who develop gram-negative infections and have the Asp299Gly mutation are more susceptible to bacteremia. However, whether these patients would be more or less susceptible to develop septic shock, which is mediated by proinflammatory cytokines, remains unknown. Therefore, the use of a 1-sided Fisher exact test cannot be justified.4 Furthermore, the same authors have demonstrated that patients with the Asp299Gly mutation have a trend for lower graft-vs-host reaction after stem cell transplantation.5 Although one may argue that the pathophysiologic mechanisms are different, the use of 1-sided statistical tests should be limited to conditions in which overwhelming evidence suggests that the effect will occur only in 1 direction. The authors report that 5 of 91 patients with septic shock had the mutation compared with 0 of 73 patients in the control group (P = .07 for a 2-sided Fisher test rather than P = .05 for a 1-sided test). The prevalence of the Asp299Gly mutation in control population is approximately 5%.6 The rationale of choosing only 73 patients in the control group remains unclear, especially when none of these patients had the Asp299Gly mutation. Although the authors report an important association, the validity of the study remains questionable because of these issues. I am also curious whether the authors evaluated other end points for increased risk of dissemination of gram-negative bacteria, such as the number of patients with positive blood cultures.