Postoperative hypothermia and outcomes after cardiac surgery

Abstract

We read with interest the article by Karalapillai et al. [1]. As the authors pointed out, the study’s strength lies in its large size and integration of data from multiple centres. Despite being retrospective, the use of a simple outcome measure (mortality) from a high quality database enhances the data’s reliability. However, we would question the group allocation. During the 24-hour measurement period, normothermia was defined as a temperature persistently > 36 °C, transient hypothermia as a temperature at any point > 36 °C, and persistent hypothermia as temperatures remaining < 36 °C. Table 1 showed the normothermic group with a mean (SD) lowest temperature of 36 (0.3) °C; if the data are normally distributed, 50% of the normothermic patients will have had a temperature < 36 °C, meaning they were not normothermic, but instead transiently hypothermic, according to the authors’ stated definitions. Similarly, the mean (SD) highest temperature of the persistently hypothermic patients was 35.5 (0.5) °C, again indicating that, if normally distributed, approximately 16% of patients would have had a temperature ≥ 36 °C; according to a normal distribution, 34% of patients will lie one SD above and below the mean. Table 2 compared the mean (SD) lowest and highest temperatures of survivors and non-survivors (respectively 35.6 (0.4) °C and 35.5 (0.4) °C, and 37.2 (0.6) °C and 37.0 (0.8) °C). Given the stated SDs and the resulting confidence intervals, if the data are normally distributed, we are surprised that the authors found a statistically significant difference (p < 0.001) between the lowest and highest temperatures of the survivors and non-survivors. It is possible that the data are not normally distributed but instead skewed. If this explains the patients’ group allocation and the statistically significant differences in the temperatures of survivors vs non-survivors, it would have been preferable to present the data as median (IQR). Regardless, a mean lowest temperature of 36 °C in the normothermic group indicates that some patients must have had a temperature < 36 °C, thereby defining themselves as being transiently hypothermic. Arguably, the key finding of the study was the high risk of death within the persistently hypothermic patients (18/111). The authors acknowledged that it is not possible to elucidate whether persistent hypothermia is a marker of illness severity or that failing to correct hypothermia is detrimental. With regard to the former explanation, in Table 1, 39 patients in the persistent hypothermia group had neither coronary artery surgery, valve surgery, nor combined coronary and valve surgery. These patients represent a sizeable proportion of the persistent hypothermia group (35%), compared with only 7% of the overall study population undergoing such surgery. Whilst the precise nature of such surgery is not clearly defined by the authors, it is possible that it was complicated high-risk surgery, with increased risks of significant blood transfusion, requirement for renal replacement therapy, or return to the operating theatre. In such circumstances, persistent hypothermia is not an independent predictor of death but merely a surrogate indicator of surgery associated with both persistent hypothermia and an increased risk of death. Given our concerns about the statistics, the data presentation and the potentially confounding factor of the nature of surgery in the persistent hypothermia group, we interpret the results with caution. However, if the results are borne out when clarified then they are worthy of further study in a prospective setting.