Power in the Study of Mortality and Necrotizing Soft Tissue Infections

To the Editor:

We read with interest the study reported by George et al., “Hyperbaric Oxygen Does Not Improve Outcome in Patients with Necrotizing Soft Tissue Infection,” published in the February 2009 issue of your journal [1]. The authors' retrospective study of mortality associated with necrotizing soft tissue infections (NSTIs) compared 30 NSTI patients at the University of Minnesota Medical Center (UMMC) who received surgery, antibiotics, and supportive care, with 48 NSTI patients at the Hennepin County Medical Center (HCMC) who received surgery, antibiotics, supportive care, and adjuvant hyperbaric oxygen (HBO) treatment. The study showed a 5% difference in mortality in favor of HBO, but this result did not reach statistical significance. The authors concluded that HBO does not improve outcomes in patients with NSTIs. We are glad to see the authors' attention to this important topic. However, we take issue with several aspects of the study design, the authors' interpretation of prior literature, and the conclusion the authors draw from their data.

The authors mention that the patients in the two groups they compare are similar. Yet their tabulated results show significant differences between the groups in body mass, blood sugars, white blood cell counts, and number of organ transplants, as well as differences in age and immunosuppression that approach statistical significance. The authors note that there may also have been potentially biasing differences in wound care, general health, socioeconomic status, and culture. These potentially biasing differences are pertinent because the patients in one group were admitted to a tertiary care university hospital with a large multi-organ transplant service (UMMC), whereas patients in the other group were admitted to a county hospital, level one trauma center in the same city (HCMC). Without comment, the authors omitted patients at HCMC with NSTIs who did not receive HBO. Further, the authors did not ascertain whether all patients included in the HBO group actually had the prescribed three HBO treatments of 90 min at 3.0 atm absolute within the first 24 h. Given these shortcomings, it is our opinion that the authors did not demonstrate the comparability of their two study groups or establish with certainty the key treatment variable in the HBO group.

The authors' statement that prior studies have not consistently demonstrated the benefit of HBO for NSTIs misrepresents the literature. They cite seven prior studies of HBO for NSTIs. Four of these studies have demonstrated mortality benefits [2–5]. Two others showed notable trends toward increased survival in patients with NSTIs who received adjuvant HBO treatment [6, 7]. Only one of the studies cited by the authors had a negative result [8]. An additional meta-analysis, not cited by the authors, has demonstrated a statistically significant survival benefit from HBO therapy for patients with NSTIs [9]. Whereas the majority of these studies were retrospective, viewed in aggregate their results favor the use of HBO for NSTIs. A simple odds ratio, derived from the results of all the studies cited by the authors, is a statistically significant 2.27 (95% confidence interval 1.39, 3.70) in favor of increased survival using adjuvant HBO.

The authors of the current study overstate the certainty of their findings when they conclude that HBO does not improve outcome in patients with NSTIs. This conclusion risks a type 2 error, or the failure to find a difference where one exists, a point of some importance in a study where a key outcome variable is mortality. Their study found a 5% difference in mortality, but this difference did not achieve statistical significance due to the sample size. It is our opinion that a 5% difference in mortality would be meaningful to most clinicians. Yet the study's power to find that the 5% difference is statistically significant, with an alpha of 0.05, is at best 0.18 (study size 2.0), leaving around an 80% chance of a type 2 error. The conclusion that HBO does not improve outcomes in patients with NSTIs therefore overstates this study's power to find a significant difference.

A prospective, multi-center study would be useful to define the most appropriate role of HBO in NSTIs. In lieu of such information, it is important for clinicians to understand that the bulk of available literature, including the study mentioned here, favors the use of adjuvant HBO therapy to decrease mortality in patients with NSTIs.