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 randomized trials from the 1990s found that mild hypothermia increased cardiac events, wound infection and blood loss frank-1997 kurz-1996 schmied-1996. One randomized 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 anesthesia 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 randomized 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 ischemia, 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 randomized 200 colorectal surgery patients to routine intraoperative thermal care or additional warming: final core temperature was 34.7 °C (SD 0.6) against 36.6 °C (SD 0.5); surgical wound infection occurred in 18 of 96 (19%) hypothermic and 6 of 104 (6%) normothermic patients, P=0.009; sutures were removed one day later in the hypothermic group, P=0.002; and hospitalization was prolonged by 2.6 days, about 20%, P=0.01 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 randomized 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 anesthesia 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.

The other consequences the day’s content list expects have small, old sources. Leslie 1995 was read in full on 2026-09-06: six volunteers were studied on two randomly assigned days at 34 °C and at 37 °C, receiving atracurium 0.5 mg/kg after two hours of a propofol infusion leslie-1995. Mild hypothermia raised propofol concentrations by about 28%, P less than 0.05, and prolonged the time to recovery of the first twitch of the train-of-four to 10% of control by about 60%, from 44 plus or minus 4 minutes to 68 plus or minus 7 minutes, while recovery from 25% to 75% was unchanged leslie-1995. Oxygen consumption with shivering was measured by indirect calorimetry in 24 patients: shivering occurred in 8 of 12 after isoflurane, with a maximum oxygen consumption of 639 mL/min against 121 to 225 mL/min during steady-state anesthesia, and meperidine 25 mg intravenously reduced it significantly hausmann-1991.

Shivering treatment has a quantitative review: meperidine 25 mg had a number needed to treat of 1.3 at 5 min (relative risk 9.6, 95% CI 5.7 to 16), clonidine 150 µg 1.3 at 5 min (RR 6.8, 3.3 to 14.2), doxapram 100 mg 1.7 and ketanserin 10 mg 2.3, with adverse-effect reporting sparse and the search closed in 2000, before dexmedetomidine kranke-2002. A later meta-analysis of six trials after spinal anesthesia found dexmedetomidine superior to clonidine for effective rate, time to cessation and recurrence, with less hypotension and more sedation, in a setting that is not the post-general-anesthesia PACU and with effect sizes not captured wang-2020-dex-shivering. Dexmedetomidine’s label carries no indication for shivering; its indications are ICU sedation of intubated adults and procedural sedation, with hypotension and bradycardia, including fatalities, as its principal warnings fda-dexmedetomidine-label-2026.

Malignant hyperthermia presenting late is this day’s boundary. In the North American registry, postoperative malignant hyperthermia was confirmed in 10 of 528 suspected cases (1.9%); all had received a volatile agent and five also succinylcholine; all showed generalized rigidity, hypercapnia or tachypnea, tachycardia and hyperthermia; no subject had hyperthermia as the presenting sign; and the latency from the end of anesthesia to the first sign was 0 to 40 min litman-2008. Across 286 registry episodes from 1987 to 2006, the frequent initial signs were hypercarbia, sinus tachycardia and masseter spasm; a temperature abnormality, median maximum 39.1 °C, was among the first three signs in 63.5%; 78.6% had muscular abnormality with respiratory acidosis but only 26.0% metabolic acidosis; the median dantrolene dose was 5.9 mg/kg; complications occurred in 63 of 181 (34.8%), with the likelihood rising 2.9-fold per 2 °C rise in maximum temperature and 1.6-fold per 30-min delay in dantrolene larach-2010. The bedside distinction is therefore not the temperature, which comes late, but the rising carbon dioxide, the rigidity and the tachycardia in a patient who has just left a volatile anesthetic.

At the bedside

Measure core temperature on PACU arrival and treat a reading below 35.5 °C as the level at which the randomized 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. Local practice differs. 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.

General recommendations

These are the numbers residents are commonly taught and expected to know. This project has not yet read them from a primary source, so each is labeled and none carries a ledger key. Where one conflicts with your institution’s protocol, the protocol wins.

Local practice differs. the core temperature below which a unit will not transfer a patient to the ward, and how often a temperature is taken in recovery, are set locally. A threshold of 36.0 °C is in wide use. This project cites no source for that figure: the national warming guidance usually credited with it is not in this ledger, and the ASPAN normothermia guideline often named beside it is in the ledger with its target unread, because the full text is paywalled [[aspan-normothermia-2010]. Note also that 36.0 °C sits half a degree above the level at which the randomized evidence of harm on this page applies. The two are different kinds of number.]

For malignant hyperthermia: stop the volatile agent, hyperventilate with 100% oxygen, give dantrolene, cool the patient and treat hyperkalemia. The dose is where two authorities disagree, and the disagreement is the thing to learn rather than smooth over. Every FDA dantrolene label says the same: begin at a minimum of 1 mg/kg by rapid intravenous push and continue until the symptoms subside or a maximum cumulative dose of 10 mg/kg has been reached, repeating the regimen if the abnormalities reappear fda-dantrolene-label-2024 fda-ryanodex-label-2024 fda-revonto-label-2024. The Malignant Hyperthermia Association of the United States says something different at both ends: 2.5 mg/kg rapidly through a large-bore line, repeated as frequently as needed until the patient responds, and it states that large doses above 10 mg/kg may be required in patients with persistent contractures or rigidity mhaus-treatment-2026. The label’s maximum is the society’s starting point for a difficult case. Follow your institution’s protocol and its cart, and know which of the two it is written from. The MHAUS page carries no revision date, which is a currency weakness worth stating, and the hotline is 1-800-644-9737 mhaus-treatment-2026.

Check yourself

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