Hypothermia has trial evidence, and the trials disagree

Why it matters

Hypothermia is the PACU problem with the best evidence and the most confident teaching, and the evidence is less settled than the teaching. Three randomised trials from the 1990s found that mild hypothermia increased cardiac events, wound infection and blood loss frank-1997 kurz-1996 schmied-1996. One randomised trial from 2022, more than ten times larger than any of them, found that warming patients from 35.6 °C to 37.1 °C changed nothing that was measured sessler-2022. The lower arm of that trial sits at the same temperature as the hypothermic arm of the trial that found the cardiac harm.

The consequence of getting this wrong is either to warm every patient to full normothermia as if the old trials required it, or to conclude that hypothermia does not matter because the new trial found nothing. The evidence supports neither.

Mechanism

Core temperature falls under anaesthesia because vasodilatation redistributes heat from the core to the periphery and because the thermoregulatory thresholds for vasoconstriction and shivering are lowered. Below the body’s normal range, coagulation enzymes and platelets function less well, oxygen consumption rises with shivering, and the sympathetic response to cold raises heart rate and blood pressure. Each of those pathways is a plausible route to bleeding, infection and cardiac events. What the mechanism does not specify is where on the temperature scale each effect becomes clinically important, and that is precisely what the trials disagree about.

Evidence

Frank 1997 randomised 300 patients with documented coronary disease or high cardiac risk having abdominal, thoracic or vascular surgery to routine thermal care or supplemental warming frank-1997. Mean postoperative core temperature was 35.4 °C (SD 0.1) in the hypothermic group and 36.7 °C (SD 0.1) in the normothermic group, P less than .001 frank-1997. Morbid cardiac events, unstable angina or ischaemia, cardiac arrest or myocardial infarction, occurred in 1.4% of normothermic and 6.3% of hypothermic patients, P=.02; hypothermia was an independent predictor, RR 2.2 (95% CI 1.1 to 4.7, P=.04), and postoperative ventricular tachycardia occurred in 2.4% versus 7.9%, P=.04, with outcomes assessed double-blind frank-1997.

Kurz 1996 randomised 200 colorectal surgery patients to routine intraoperative thermal care or additional warming and is the landmark wound-infection and length-of-stay result; the effect sizes have not been read from the primary record, so they are placeholders here: [NUMBER NEEDED: surgical site infection rates and length of stay by arm in Kurz 1996] kurz-1996. Schmied 1996 is the blood-loss limb in total hip arthroplasty schmied-1996.

Rajagopalan 2008 pooled 14 studies for blood loss and 10 for transfusion, with hypothermia defined as 34 to 36 °C and a median temperature difference between arms of 0.85 °C (quartiles 0.60 and 1.1) rajagopalan-2008. The ratio of geometric means for blood loss was 0.84 (0.74 to 0.96), P=0.009, so hypothermia increased blood loss by about 16% (4 to 26%), and the transfusion relative risk was 0.78 (95% CI 0.63 to 0.97), P=0.027, so hypothermia increased the relative risk of transfusion by about 22% (3 to 37%) rajagopalan-2008.

PROTECT randomised 5056 patients, 5013 in the intention-to-treat analysis, at 12 sites: aged 45 or older with at least one cardiovascular risk factor, having inpatient noncardiac surgery of 2 to 6 h under general anaesthesia sessler-2022. Aggressive warming reached a mean final intraoperative core temperature of 37.1 °C (SD 0.3) against 35.6 °C (SD 0.3) with routine care sessler-2022. The 30-day composite of myocardial injury, non-fatal cardiac arrest and all-cause mortality occurred in 246 of 2497 (9.9%) versus 239 of 2490 (9.6%), with no difference in surgical site infection, transfusion, length of stay or readmission; the authors conclude that keeping core temperature at least 35.5 °C appears sufficient sessler-2022.

What this does not show

The reconciliation usually offered, that the old trials and PROTECT cover different temperature ranges, is weaker than it looks. Frank’s hypothermic arm averaged 35.4 °C and PROTECT’s routine-care arm targeted and reached 35.6 °C; those are the same temperature frank-1997 sessler-2022. Frank found a significant cardiac difference at that separation; PROTECT, far larger, found none. Rajagopalan sits at a median separation of only 0.85 °C within a 34 to 36 °C band, so it partly extends below PROTECT’s range and partly does not rajagopalan-2008. This page states plainly that a 1997 trial of 300 patients and a 2022 trial of 5013 reach different conclusions across overlapping temperatures, and that the field has not resolved it. The claim status is uncertain for that reason.

The 1990s trials are era-limited: single surgical populations, pre-modern infection bundles, pre-modern warming equipment kurz-1996. PROTECT excluded patients with a body mass index above 30 and those on dialysis sessler-2022.

Items on the director’s content list for this day that the held sources do not carry are placeholders: [NUMBER NEEDED: effect of hypothermia on drug metabolism and on oxygen consumption with shivering] [NUMBER NEEDED: shivering treatment with meperidine, dexmedetomidine or clonidine, doses and evidence]. Malignant hyperthermia presenting late in the PACU is assigned to this day’s boundary and has no ledger row yet: [NUMBER NEEDED: features distinguishing late malignant hyperthermia from other causes of hyperthermia and tachycardia in the PACU].

At the bedside

Measure core temperature on PACU arrival and treat a reading below 35.5 °C as the level at which the randomised evidence for harm applies sessler-2022 frank-1997. Below that, warm.

Above 35.5 °C, warming is comfort and shivering control, not outcome protection, on the best current evidence; do not tell a resident or a patient that the last degree to 37 °C prevents infarction or infection, because the largest trial found that it did not sessler-2022. [PRACTICE VARIES: the PACU discharge temperature criterion and the warming devices available differ between institutions; follow local protocol.]

When a colleague quotes the 1996 and 1997 trials, ask what temperature their hypothermic arms reached, and compare it with what your patient’s monitor says. The disagreement in the literature is a disagreement about a few tenths of a degree, and the patient in front of you is on one side of it or the other.

Check yourself

Four items. Every option carries an explanation. Progress is not saved.