To investigate the effect of the application of therapeutic hypothermia (32–35°C) on survival and major clinical endpoints in critically ill patients.
We searched online database and clinical trial registries dated up to April 30, 2019, and references of relevant studies.
Low risk of bias randomized trials which compared hypothermia applied for at least 24 hours and conventional therapy in critically ill patients were included. We excluded trials investigating therapeutic hypothermia in indications already supported by international guidelines (adult cardiac arrest and hypoxic-ischemic encephalopathy of newborns) or intraoperative hypothermia.
Titles and abstracts were reviewed independently by two authors. If the articles seemed eligible, full-text articles were reviewed, and data were abstracted using a structured template.
Our search retained 14 low risk of bias randomized trials (2,670 patients) performed in three different settings: traumatic brain injury, serious infections, and stroke. Therapeutic hypothermia was associated with an increase in mortality at longest follow-up available (432/1,375 [31%] vs 330/1,295 [25%]; risk ratio, 1.24; 95% CI, 1.10–1.39; p = 0.0004; I2 = 0%). Pooled results showed no difference of good neurologic outcome among survivors between the two treatment arms (493/1,142 [43%] vs 486/1,067 [46%]; risk ratio, 1.04; 95% CI, 0.97–1.12; p = 0.27; I2 = 1%). Arrhythmias were significantly increased among patients undergoing therapeutic hypothermia. We found no difference between groups in pneumonia, serious infections, any infection, hemorrhage, renal failure, deep vein thrombosis, and uncontrollable intracranial hypertension.
High-quality randomized evidence indicates that therapeutic hypothermia is associated with higher mortality and no difference in good neurologic outcome compared with normothermia in critically ill patients. Although there still might be a possibility that therapeutic hypothermia is beneficial in a specific setting, routine application of therapeutic hypothermia would better be avoided outside the settings indicated by international guidelines (adult cardiac arrest and hypoxic-ischemic encephalopathy of newborns).