Deliverable 2: The Clinical Problem

Articles

www.thelancet.com Vol 399 May 7, 2022 1799

Aggressive intraoperative warming versus routine thermal management during non-cardiac surgery (PROTECT): a multicentre, parallel group, superiority trial Daniel I Sessler*, Lijian Pei*, Kai Li*, Shusen Cui, Matthew T V Chan, Yuguang Huang, Jingxiang Wu, Xuemei He, Gausan R Bajracharya, Eva Rivas, Carmen K M Lam, on behalf of the PROTECT Investigators†

Summary Background Moderate intraoperative hypothermia promotes myocardial injury, surgical site infections, and blood loss. Whether aggressive warming to a truly normothermic temperature near 37°C improves outcomes remains unknown. We aimed to test the hypothesis that aggressive intraoperative warming reduces major perioperative complications.

Methods In this multicentre, parallel group, superiority trial, patients at 12 sites in China and at the Cleveland Clinic in the USA were randomly assigned (1:1) to receive either aggressive warming to a target core temperature of 37°C (aggressively warmed group) or routine thermal management to a target of 35·5°C (routine thermal management group) during non-cardiac surgery. Randomisation was stratified by site, with computer-generated, randomly sized blocks. Eligible patients (aged ≥45 years) had at least one cardiovascular risk factor, were scheduled for inpatient non- cardiac surgery expected to last 2–6 h with general anaesthesia, and were expected to have at least half of the anterior skin surface available for warming. Patients requiring dialysis and those with a body-mass index exceeding 30 kg/m² were excluded. The primary outcome was a composite of myocardial injury (troponin elevation, apparently of ischaemic origin), non-fatal cardiac arrest, and all-cause mortality within 30 days of surgery, as assessed in the modified intention-to-treat population. This study is registered with ClinicalTrials.gov, NCT03111875.

Findings Between March 27, 2017, and March 16, 2021, 5056 participants were enrolled, of whom 5013 were included in the intention-to-treat population (2507 in the aggressively warmed group and 2506 in the routine thermal management group). Patients assigned to aggressive warming had a mean final intraoperative core temperature of 37·1°C (SD 0·3) whereas the routine thermal management group averaged 35·6°C (SD 0·3). At least one of the primary outcome components (myocardial injury after non-cardiac surgery, cardiac arrest, or mortality) occurred in 246 (9·9%) of 2497 patients in the aggressively warmed group and in 239 (9·6%) of 2490 patients in the routine thermal management group. The common effect relative risk of aggressive versus routine thermal management was an estimated 1·04 (95% CI 0·87–1·24, p=0·69). There were 39 adverse events in patients assigned to aggressive warming (17 of which were serious) and 54 in those assigned to routine thermal management (30 of which were serious). One serious adverse event, in an aggressively warmed patient, was deemed to be possibly related to thermal management.

Interpretation The incidence of a 30-day composite of major cardiovascular outcomes did not differ significantly in patients randomised to 35·5°C and to 37°C. At least over a 1·5°C range from very mild hypothermia to full normothermia, there was no evidence that any substantive outcome varied. Keeping core temperature at least 35·5°C in surgical patients appears sufficient.

Funding 3M and the Health and Medical Research Fund, Food and Health Bureau, Hong Kong.

Copyright © 2022 Published by Elsevier Ltd. All rights reserved.

Introduction Overall, 30-day postoperative mortality is about 2% among inpatients,1,2 corre sponding to the third leading cause of death worldwide.3 About 25% of all 30-day postoperative deaths are cardiovascular or consequent to cardiovascular events, with myocardial injury being by far the most commonly associated complication.4

Fewer than 10% of patients experiencing a perioperative myocardial infarction have chest pain, and 65% are entirely clinically silent and thus go undetected without routine troponin screening.4,5 30-day mortality in patients

with elevated postoperative troponin is nonetheless significantly and substantially increased in patients with and without symptoms.6 Myocardial injury after non-cardiac surgery (MINS) highlights that troponin elevations without a non-ischaemic explanation are clinically important—even in patients who do not have the symptoms and signs required to meet the formal universal definition of myocardial infarction.7,8

Among the factors possibly contributing to MINS is hypothermia. Nearly all unwarmed patients who have surgery become hypothermic. Thermoregulatory

Lancet 2022; 399: 1799–808

Published Online April 4, 2022 https://doi.org/10.1016/ S0140-6736(22)00560-8

See Comment page 1757

For the Chinese translation of the abstract see Online for appendix 1

*Contributed equally

†PROTECT Investigators are listed in appendix 2 (p 2)

Department of Outcomes Research, Anesthesiology Institute, Cleveland Clinic, Cleveland, OH, USA (Prof D I Sessler MD, G R Bajracharya MD, E Rivas MD); Population Health Research Institute, McMaster University, ON, Canada (Prof D I Sessler); Department of Anesthesiology, Peking Union Medical College Hospital, Beijing, China (L Pei MD, Prof Y Huang MD); China-Japan Union Hospital of Jilin University, Jilin, China (K Li MD, Prof S Cui PhD); The Chinese University of Hong Kong, Hong Kong Special Administrative Region, China (Prof M T V Chan PhD); Shanghai Chest Hospital, Shanghai Jiao Tong University, Shanghai, China (Prof J Wu MD); West China Hospital, Sichuan University, Sichuan, China (X He MD); Department of Anesthesia, Hospital Clinic of Barcelona, IDIBAPS, Universidad de Barcelona, Barcelona, Spain (E Rivas); Tuen Mun Hospital, Hong Kong Special Administrative Region, China (C K M Lam MBBS)

Correspondence to: Prof Daniel I Sessler, Department of Outcomes Research, Anesthesiology Institute, Cleveland Clinic, Cleveland, OH 44195, USA [email protected]

or

Articles

1800 www.thelancet.com Vol 399 May 7, 2022

Prof Yuguang Huang, Department of Anesthesiology,

Peking Union Medical College Hospital, Beijing 100730, China

[email protected]

See Online for appendix 2

vasoconstriction, the initial autonomic response to hypothermia, increases systemic vascular resistance and consequently increases blood pressure.9,10 When combined with thermal discomfort and shivering, hypothermia also causes sympathetic activation and increases heart rate10— all of which worsen myocardial supply–demand which is the major cause of post operative myocardial injury.11

Many, but not all, patients having surgery are actively warmed in high-income countries.12 In low-income and middle-income countries, few patients are actively warmed. For example, a 2015 cross-sectional survey of Beijing hospitals found that only 11% of surgical patients were actively warmed, and that more than a third of all patients having surgeries lasting at least 2 h were hypothermic at the end of surgery.13 A more recent survey of six countries in Asia confirms that intra- operative warming is hardly routine.14

Recommendations to warm surgical patients are based on randomised trials showing that mild hypothermia (body temperature of about 34·5°C) causes complications relative to a temperature of about 36·5°C. But whether smaller temperature differences matter is less obvious. Perioperative normothermia is conventionally defined as a core temperature of at least 36°C. However, even at the circadian nadir (usually about 0300 h), core tempera- ture normally exceeds 36°C; at about 1500 h, core temperature typically exceeds 37°C.15 On average, normal body temperature in humans is about 37°C,16 not the 36°C that is widely accepted as suitable for perioperative patients.17

There is little evidence to support 36°C as the optimal perioperative target temperature. A slightly lower tempera- ture (such as 35·5°C) might be equally safe, or a higher temperature (such as 37°C) might be better. Intraoperative

core temperature using a single intra operative forced-air cover, the conventional approach, averages only 36°C.18 Reliably warming patients to at least 37°C usually requires 30 min of pre-warming, two intraoperative forced-air warming covers, and a fluid warmer. The added time, cost, and difficulty of main taining an intraoperative core temperature of at least 37°C would be a worthwhile investment if it reduced major complications.

Core temperature is normally tightly regulated, suggesting that humans function best at temperatures near 37°C. Consistent with this theory, randomised trials have shown that mild perioperative hypothermia (eg, 34·5°C) causes coagulopathy,19 increases blood trans fusion,19 promotes surgical site infection,20 delays drug metabolism,21 and prolongs recovery time22 and duration of hospital stay.20 Whether aggressive warming to a truly normothermic 37°C improves outcomes remains unknown.

We, therefore, aimed to test the primary hypothesis that aggressive warming to a core temperature target of 37°C, versus routine thermal management (typically 35·5°C), prevents major adverse cardiac events—defined as a composite of 30-day myocardial injury, non-fatal cardiac arrest, and all-cause mortality. We also aimed to test the hypotheses that aggressive warming reduces infection, red blood cell transfusion, duration of hospital stay, and hospital readmission within 30 days of surgery.

Methods Study design and participants PROTECT was a multicentre, parallel group, superiority trial that assessed the effect of aggressive warming on major complications of non-cardiac surgery at 12 sites in China and at the Cleveland Clinic in the USA. Patients

Research in context

Evidence before this study The principal investigator (DIS) supplemented his personal collection of more than 2000 articles about perioperative thermoregulation by searching MEDLINE with the search terms (“temperature” or “thermoregulation”) and (“anaesthesia” or “surgery”) for articles published between Jan 1, 2000 and Dec 31, 2021, with no language restrictions. All articles and relevant references from within citation lists were considered; those with the most robust methodology and largest sample size were given most weight. Small trials, many of which were published two decades ago, suggested that perioperative hypothermia causes myocardial, infectious, and bleeding complications. Influential randomised trials include a 200-patient trial reporting a three-times increase in surgical site infections (mostly superficial) published in 1996; a 300-patient trial reporting that hypothermia more than doubles the risk of morbid myocardial events (mostly ventricular tachycardia) published in 1997; and a meta-analysis of nine small trials reporting a

36% reduction in transfusion requirement that was published in 2020.

Added value of this study Nearly all unwarmed surgical patients become hypothermic, usually to between 34·5°C and 35·5°C. Previous small trials report that hypothermia causes morbid myocardial events, surgical site infections, and increases transfusion requirements. PROTECT is the first robust trial of complications consequent to perioperative hypothermia.

Implications of all the available evidence In PROTECT, incidence of myocardial injury, surgical site infection, and the need for transfusion were similar in patients randomly allocated to intraoperative temperatures of 35·5°C (routine care) and 37·0°C (aggressive warming). At least over a 1·5°C range from very mild hypothermia to full normothermia, there was no evidence that any substantive outcome varied. Keeping core temperature at least 35·5°C in surgical patients appears sufficient to prevent major temperature-related complications.

Articles

www.thelancet.com Vol 399 May 7, 2022 1801

were eligible when they were scheduled for non-cardiac surgery expected to last 2–6 h with general anaesthesia. We enrolled patients who were at least 45 years old, expected to require at least overnight hospitalisation, and expected to have at least half of the anterior skin surface available for warming. Additionally, patients had to have at least one cardiovascular risk factor, such as hypertension, diabetes, previous stroke, or known cardiovascular disease. We excluded patients who required dialysis and those with a body-mass index exceeding 30 kg/m², because patients with a high BMI tend to stay warmer during surgery.23

The Department of Outcomes Research at the Cleveland Clinic (Cleveland, OH, USA) was the trial coordinating centre and was responsible for the design of online randomisation, database maintenance, data verification, and data analyses. The overall trial protocol was approved by the Cleveland Clinic Institutional Review Board and the institutional review boards at each participating hospital. Written consent was obtained from each patient by investigators well before the induction of anaesthesia. The trial was approved by the Chinese Ministry of Science and Technology, including permission to share patient- level data with the trial management centre at the Cleveland Clinic. Data were monitored and audited at trial sites by an independent contract research organisation, and centrally by the trial organisers.

The trial is registered at ClinicalTrials.gov, NCT03111875. Reporting is consistent with CONSORT recommendations. The full protocol, statistical plan, and change log are presented in appendix 2 (pp 8–32).

Randomisation and masking Patients were randomly allocated (1:1) to one of two groups, stratified by site, with computer-generated, randomly sized blocks. To conceal allocation, investigators accessed a web-based site about 1 h before surgery. The treatment groups were (1) routine thermal management with rescue intraoperative forced-air warming to prevent core temperature from decreasing below 35·5°C or (2) aggressive warming to a target intraoperative core temperature of at least 37°C. The lower temperature specified in this trial (35·5°C) reasonably represents routine care in China, where 99% of patients were enrolled.14,24 A more extensive perspective on ethical considerations is presented in appendix 2 (pp 3–4).

It was not possible to mask patients to prewarming or clinicians to intraoperative warming. Clinicians and investigators were thus aware of group assignment. However, all postoperative measurements were made by a separate team of investigators who were unaware of patients’ group assignment and intraoperative management. The trial was thus assessor blinded.

Procedures Patients assigned to routine thermal management were not pre-warmed and ambient intraoperative temperature

was maintained near 20°C per routine. Only transfused blood was warmed. An upper-body or lower-body forced- air cover was positioned over an appropriate non- operative site, but was not activated unless core temperature decreased to 35·5°C.

Patients assigned to aggressive warming were pre-warmed with a full-body forced-air cover for about 30 min before induction of anaesthesia. The warmer was initially set to high mode, which corresponds to about 43°C. It was subsequently adjusted to make patients feel warm, but not uncomfortably so. Pre-warming for 30 min25,26 ameliorates redistribution hypothermia, which is otherwise the major cause of hypothermia during the initial hour of general or neuraxial27 anaesthesia. Patients were aggressively warmed during surgery to a target intraoperative core temperature of at least 37°C, using two forced-air covers when clinically practical. All intravenous fluids were warmed to body temperature. Ambient intraoperative temperature was maintained near 20°C per routine.

Baseline demographic and morphometric charac- teristics were recorded, as were cardiovascular risks and medications. Surgery was characterised as ortho- paedic, laparoscopic (including video-assisted thoracic procedures), open abdominal, urological, neurosurgical (including spine), and others. Timing was characterised as elective, urgent, or emergent. To characterise the risk of surgical site infection, we recorded whether the

Figure 1: Trial profile

2527 allocated to aggressive warming

5056 randomly allocated to study group

12 064 patients assessed for eligibility

7008 excluded

7 did not receive allocated treatment

8 did not receive allocated treatment

20 did not receive any treatment

23 did not receive any treatment

39 withdrew from study 16 surgery cancelled

or changed 8 patient withdrew

13 staff decision 2 device malfunction

49 withdrew from study 17 surgery cancelled or

changed 17 patient withdrew 14 staff decision

1 device malfunction

2529 allocated to routine care

2507 included in modified intention-to-treat population

2506 included in modified intention-to-treat population

Articles

1802 www.thelancet.com Vol 399 May 7, 2022

surgery involved colon resection, rectal resection, other abdominal surgery, and whether the wound was contaminated or dirty-infected.

Intraoperative core temperature was measured in the distal oesophagus or nasopharynx 10–20 cm past the nares, and recorded at 15-min intervals throughout surgery. Core temperature was summarised for each patient as time-weighted averages, calculated as the area under the curve for all temperature measurements for a

patient using the trapezoid rule, then dividing by the total measurement time for that patient.

Mean arterial pressure and heart rate were recorded at 15-min intervals throughout surgery and during the initial postoperative hour. Blood for generation 4 cardiac troponin T or troponin I (per clinical routine) was sampled for the first 3 days postoperatively while patients remained in hospital.6 Blood for generation 5 (high-sensitivity) troponin T or I was sampled pre operatively up to 2 weeks before surgery when practical, and on the first 2–3 days postoperatively while patients remained in hospital.5 Troponin samples were also obtained when patients had shortness of breath or experienced chest, neck, or arm pain. The only substantive change to the protocol was to update thresholds for various types and generations of troponin in light of information that was not available when the trial started, as detailed in appendix 2 (pp 5–6).

When blood troponin concentrations exceeded site thresholds, a cardiology consultation was requested along with an electrocardiogram and, when possible, an echocardiogram. Haemoglobin was obtained on the first postoperative morning. Ischaemic symptoms, such as shortness of breath or chest, neck, or arm pain were documented.

The Quality of Recovery-15 score was assessed on the third postoperative day in person, or by telephone for patients already discharged from hospital,28 using a language-specific version.29 Patients were also evaluated at hospital discharge and via telephone 30 days after surgery to evaluate myocardial, infectious, and other complications. Outcomes were censored at 30 days.

Outcomes The primary outcome was a composite of myocardial injury after non-cardiac surgery, non-fatal cardiac arrest, and all-cause mortality within 30 days of surgery. Myocardial injury was diagnosed when available troponin concentrations exceeded generation-specific and type- specific thresholds (appendix 2 pp 5–6) and were apparently of ischaemic origin (ie, no other obvious cause for artifactual elevation). Myocardial infarction diagnosis required both troponin elevation and at least one diagnostic symptom or sign.

Secondary outcomes were (1) deep or organ-space surgical site infection within 30 days of surgery, as defined by US Centers for Disease Control and Prevention criteria (appendix 2 p 7); (2) intraoperative transfusion require- ment, defined as units of red blood cells transfused; (3) duration of hospital stay; and (4) hospital readmission within 30 days of surgery.

Our exploratory outcomes were (1) change in blood haemoglobin (preoperatively vs the first postoperative morning); (2) superficial surgical site infection within 30 days of surgery; (3) Quality of Recovery-15 score on the third postoperative day;28 and (4) myocardial infarction within 30 days of surgery meeting the Third Universal Definition.30 The components of the composite primary

Aggressive warming (n=2507)

Routine care (n=2506)

Absolute standardised difference*

Baseline characteristics

Mean age, years (SD) 67 (8) 67 (8) 0·027

Sex

Female 836 (33%) 801 (32%) 0·029

Male 1668 (67%) 1701 (68%) ··

Mean body-mass index, kg/m² (SD)

23 (3) 23 (3) 0·053

Type of surgery

Orthopaedic 56 (2%) 49 (2%) 0·096

Laparoscopic 1313 (52%) 1301 (52%) ··

Open abdominal 610 (24%) 613 (25%) ··

Neurosurgical 126 (5%) 89 (4%) ··

Urological 138 (6%) 138 (6%) ··

Other 261 (10%) 312 (12%) ··

Type of troponin

Low-sensitivity T 200 (8%) 199 (8%) 0·002

High-sensitivity T 823 (33%) 825 (33%) ··

Low-sensitivity I 1484 (59%) 1482 (59%) ··

Timing of surgery

Elective 2440 (97%) 2432 (97%) 0·015

Urgent or emergent 64 (3%) 70 (3%) ··

Risk of surgical site infection

Any infection risk 1421 (57%) 1414 (57%) 0·005

Colon resection 376 (15%) 392 (16%) 0·018

Rectal resection 378 (15%) 344 (14%) 0·038

Other abdominal surgeries 397 (16%) 407 (16%) 0·011

Contaminated or dirty- infected wound

429 (17%) 438 (18%) 0·01

Cardiac risk

Any cardiac risk 1755 (70%) 1791 (72%) 0·033

Hypertension requiring treatment

1234 (49%) 1201 (48%) 0·026

Diabetes requiring medication (oral or insulin)

534 (21%) 466 (19%) 0·068

End-stage renal failure requiring dialysis

2 (<1%) 2 (<1%) <0·001

Peripheral vascular disease 28 (1%) 23 (1%) 0·02

Previous myocardial infarction

39 (2%) 46 (2%) 0·022

Smoker 558 (22%) 610 (24%) 0·05

Congestive heart failure 14 (1%) 18 (1%) 0·02

Chronic obstructive pulmonary disease

52 (2%) 48 (2%) 0·011

(Table 1 continues on next page)

Articles

www.thelancet.com Vol 399 May 7, 2022 1803

outcome, MINS, myocardial infarction, and surgical site infections, were centrally adjudicated by investigators who were blinded to randomisation and intraoperative core temperatures. Adverse events were tabulated as a measure of safety. We considered unexpected safety events that were not predefined outcomes to be adverse events. Events were designated serious when they prolonged hospitali- sation or were life threatening. Events were reported by site investigators to the central coordination site at the Cleveland Clinic. Each was independently designated as serious or not by two investigators (EYA, MCC, FA-C, and SML) who were blinded to treatment allocation.

Statistical analysis We assumed that the incidence of MINS, non-fatal cardiac arrest, and total mortality would be 10%, 0·1% and 1%, respectively, based on data from the Perioperative Ischemic Evaluation-2 trial.31,32 4000 patients would provide a power of 0·9 at the 0·05 significance level to detect a 30% or more relative reduction in the primary outcomes using the common effect generalised estimating equation model.33 After accounting for interim analyses and allowing for 5% of patients dropping out, we planned to enrol 5050 patients.

The statistical analysis plan was included in the original protocol approved by the institutional review boards. Our primary outcome was analysed in the modified intention- to-treat population; that is, we included all patients who had surgery and received thermal management, whether or not target temperatures were reached. Any imbalanced baseline characteristics with absolute standardised differences larger than 0·10 between randomised groups would be adjusted for in all analyses.

Primarily, we assessed the effect of thermal manage- ment on the binary-event composite vector outcome of MINS, non-fatal cardiac arrest, and all-cause mortality within 30 days of surgery using a multivariate (ie, multiple components measured on each patient) generalised estimating equation model, in which we estimated the common (or global) effect relative risk across the three outcome components.34,35 We used the log link (ie, a log-binomial model) to estimate relative risk instead of odds ratio. We also reported the relative risk across components using the average relative effect method via a generalised estimating equation distinct effects model, in which we first estimated the treatment effect for each component of the composite and then averaged them. This method has the advantage of not being driven by components with the highest frequency, as occurs with more standard methods.34,35 Treatment effects for each component were also reported. We also conducted post- hoc treatment effect heterogeneity tests for age, sex, and surgery type, and further estimated the effect within each subgroup.

For secondary outcomes, we used log-binomial models to estimate the treatment effect for all binary outcomes, Cox proportional hazards regression model for time to

discharge alive, and linear regression models for all continuous outcomes.

The primary analysis we describe above did not include the imputation of missing outcome data. As a sensitivity analysis, missing outcomes were imputed after examining the missing mechanism and dropout rates. If fewer than 5% of patients had missing data for the primary outcome, we planned to assign worst outcome to the active treatment patients and best outcome to the control patients. We planned to apply the last observation carried forward method for patients who were missing 30-day follow-up data but had discharge assessment data. If 5–20% of patients had missing primary outcome data, we planned for multiple imputation.

Interim monitoring for efficacy and futility was conducted at each 25% of the maximum planned enrolment, following a group sequential design with gamma spending function parameters of –4 for efficacy and 0 for futility. Boundaries for efficacy (futility in

Aggressive warming (n=2507)

Routine care (n=2506)

Absolute standardised difference*

(Continued from previous page)

Medication use

Any use of the following 606 (24%) 599 (24%) 0·006

Beta blockers 204 (8%) 199 (8%) 0·007

Angiotension converting enzyme inhibitors

143 (6%) 128 (5%) 0·026

Angiotension receptor blockers

211 (8%) 197 (8%) 0·02

Statin 294 (12%) 313 (13%) 0·024

Intraoperative management

Mean final core temperature, °C (SD)

37·1 (0·3) 35·6 (0·3) ··

Mean TWA core temperature, °C (SD)

36·8 (0·3) 35·8 (0·3) ··

Mean MAP, mm Hg (SD) 84 (10) 85 (10) ··

Median MAP area under <65 mm Hg, mm Hg per min (IQR)

0 (0–6) 0 (0–6) ··

Mean heart rate, beats per min (SD)

70 (10) 67 (10) ··

Mean surgery duration, h (SD)

4·4 (1·8) 4·3 (1·6) ··

Mean crystalloids, L (SD) 1·47 (0·91) 1·41 (0·82) ··

Mean colloids, mL (SD) 627 (514) 622 (508) ··

Median estimated blood loss, mL (IQR)

200 (50–300) 200 (50–300) ··

Transfusion 254 (10%) 236 (9%) ··

Use of cell saver 14 (1%) 18 (1%) ··

Summary statistics are presented as n (%) for categorical variables, and as mean (SD) or medians (IQR) for continuous variables. Baseline data were missing in three patients in the aggressive warming group and four patients in the routine thermal management group. Intraoperative data were missing in 11 patients who received aggressive warming and 18 patients who received routine care. MAP=mean arterial pressure. TWA=time-weighted average. *Presented for all baseline characteristics. >0·10 was considered imbalanced. Since no imbalance was found, none of these were adjusted for in our analyses.

Table 1: Baseline characteristics and intraoperative management

Articles

1804 www.thelancet.com Vol 399 May 7, 2022

parentheses) at each interim and final analysis were p≤0·001 (p>0·880), p≤0·005 (p>0·525), p≤0·014 (p>0·169), and p≤0·044 (p>0·044).

With an overall alpha of 0·05, the significance criterion was 0·044 for analysis of the primary outcome after accounting for alpha spent in three interim analyses (with Bonferroni correction for the three components), and 0·0125 for each secondary outcome (ie, 0·05 / 4, Bonferroni). Since only the primary outcome was subject to repeated testing and decision making in the interim analyses, the interim-adjusted alpha level of 0·044 was not applied to the secondary outcomes. Interactions were considered significant when the p values were less than 0·10.

We used SAS software (version 9.4) for all analyses.

Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or the writing of the report.

Results Between March 27, 2017, and March 16, 2021, 5056 participants were enrolled from 14 sites and randomly allocated to treatment; 5013 of whom were included in the intention-to-treat population (2507 in the aggressively warmed group and 2506 in the routine thermal management group; figure 1). Follow-up for the primary outcome was achieved in 99% of participants.

All baseline characteristics were well balanced, and therefore none were adjusted for in our analyses. The mean age was 67 years (SD 8), 99% of participants were Chinese (appendix 2 p 33), and 33% were female. Intraoperative fluid management, blood loss, and transfusion were similar between the two groups (table 1). Four patients on dialysis were incorrectly enrolled, but were retained under our intention-to-treat definition.

Patients assigned to aggressive warming had a mean final core temperature of 37·1°C (SD 0·3), and a mean time-weighted average core temperature of 36·8°C ( 0·3). Patients assigned to routine care had a mean final core temperature of 35·6°C (0·3), and a mean time-weighted average core temperature of 35·8°C (0·3; table 1, figure 2). There was excellent temperature separation between the groups.

At least one of the primary outcome components (MINS, cardiac arrest, or mortality) occurred in 246 (9·9%) of 2497 aggressively warmed patients and in 239 (9·6%) of 2490 patients assigned to routine thermal management. The common effect relative risk of aggressive versus routine thermal management was an estimated 1·04 (95% CI 0·87–1·24, p=0·69). The result after imputation of missing outcomes (ten aggressively warmed patients and 16 control patients) was consistent, with a common effect relative risk across the components of the composite outcome of 1·10 (0·92–1·32, p=0·27). No boundary for efficacy or futility was crossed during the interim monitoring; consequently, the trial continued to completion.

The relative risk across the individual components estimated using the average relative risk method was 0·68 (0·43–1·08, p=0·103). The observed association varied between the individual components of the composite primary outcome, but the treatment-by-component interaction was not significant (p=0·13); however, the interaction test was underpowered. The incidence and the estimated treatment effect for each individual component of the composite are presented in figure 3 and table 2 . No effect of aggressive warming was detected on any of the components of the primary outcome.

The effect of aggressive warming on the primary outcome did not differ across age and sex (both pinteraction>0·20) or by surgery type (pinteraction=0·37; figure 4). However, these interaction tests are underpowered and do not preclude potentially clinically important interactions for type of surgery.

Myocardial infarctions occurred in 51 (2·1%) of 2469 patients assigned to aggressive warming and

Figure 2: Core temperature over time after induction as averages Error bars are standard deviations. About 50% of patients remained in surgery at 4·5 h, and 25% remained in surgery at 5 h. Data after 5 hours are not shown.

0 1 2 3 4 5

35·5

36·0

36·5

37·0

37·5 Co

re te

m pe

ra tu

re (°

C)

Time since induction (hours)

Aggressive warming Routine care

Figure 3: Forest plot of the primary and secondary outcomes The circles represent the relative risk of aggressive warming versus routine care. Error bars show 95% CIs for primary outcomes or Bonferroni-adjusted CIs (95·6% CI for individual components of primary outcome, 97·5% CI for secondary outcomes). MINS=myocardial injury after non-cardiac surgery.

Aggressive warming, n/N

Routine care, n/N

Relative risk

Primary outcomes

Common effect

Average relative effect

MINS

Non-fatal cardiac arrest

Mortality

Secondary outcomes

Surgical site infection

Transfusion requirement

Hospital readmission

246/2497

233/2470

6/2493

13/2489

178/2487

254/2494

161/2462

239/2490

223/2468

15/2486

17/2482

157/2479

236/2486

135/2447

Favours aggressive warming Favours routine care

0·60·4 1·40·20·1 1·0 2·0

Articles

www.thelancet.com Vol 399 May 7, 2022 1805

45 (1·8%) of 2468 patients assigned to routine management (p=0·54). The 30-day incidence of deep or organ-space surgical site infection, red blood cell transfusions, and 30-day readmissions did not differ between the two groups. The median length of hospital stay was 8 days (IQR 6–12) in both groups and did not differ significantly by thermal management. Exploratory outcomes are reported descriptively (table 2, figure 3).

There were 17 serious adverse events in patients assigned to aggressive warming and 30 in patients assigned to routine thermal management. Only one serious adverse event—fever with negative blood cultures and no clinical evidence of infection, in an aggressively warmed patient—was deemed to be possibly related to thermal management. There were 22 other adverse events in patients in the aggressively warmed group versus 24 in the routine thermal management group (table 3). All adverse events are listed in appendix 2 (pp 34–44). Death was a defined outcome and, therefore, was not reported as an adverse event. There

Aggressive warming (n= 2507)

Routine care (n=2506) Treatment effect p value

Nmiss Summary Nmiss Summary

Primary outcomes

Overall*

Common effect 10 246 (9·9%) 16 239 (9·6%) 1·04 (0·87–1·24) 0·69

Distinct†

Average relative effect ·· ·· ·· ·· 0·68 (0·43–1·08) 0·10

MINS 37 233 (9·4%) 38 223 (9·0%) 1·05 (0·85–1·30) 0·57

Non-fatal cardiac arrest 14 6 (0·2%) 20 15 (0·6%) 0·40 (0·13–1·27) 0·057

All-cause mortality 18 13 (0·5%) 24 17 (0·7%) 0·76 (0·32–1·84) 0·46

Secondary outcomes

Surgical site infection‡ 20 178 (7·2%) 27 157 (6·3%) 1·13 (0·87–1·47) 0·25

Transfusion requirement‡ 13 254 (10%) 20 236 (9·5%) 1·07 (0·87–1·33) 0·41

Readmission‡ 45 161 (6·5%) 59 135 (5·5%) 1·19 (0·89–1·57) 0·13

Median length of hospital stay, days (IQR) 35 8 (6–12) 41 8 (6–12) 0·98 (0·91–1·05)§ 0·46

Exploratory outcomes

Mean reduction in hemoglobin from preoperative to POD1

161 1·2 (1·5) 148 1·1 (1·5) ·· ··

Mean Quality of Recovery-15 score on POD3 27 110 (17) 33 110 (17) ·· ··

30-day superficial surgical site infection 18 109 (4%) 25 97 (4%) ·· ··

30-day myocardial infarction 38 51 (2%) 38 45 (2%) ·· ··

Summary statistics are presented as mean (SD) for continuous variables and n (%) for categorical variables. For primary outcomes, the treatment effect is presented as relative risk (95·6% CI). For secondary outcomes, the treatment effect is presented as relative risk (98·75% CI). p values of 0·0125 (ie, 0·05 / 4) were considered significant for individual secondary outcomes. Treatment effects and p values are not reported for exploratory outcomes. 10 aggressively warmed patients and 16 patients given routine care were missing data for all three components of the composite outcome. MINS=myocardial injury after non-cardiac surgery. Nmiss=number of patients with missing data. POD=postoperative day. GEE=generalised estimating equation. *The overall relative risk (RR) was estimated using a common effect GEE model, which estimates a single treatment effect across the components while adjusting for within-patient correlation. The incidence is presented here as the collapsed composite (ie, any occurrence of the three components) for simplicity. A p value <0·044 was considered to be significant for the overall common effect across all the components. CIs of the common effect reflect the correction for interim analyses in order to maintain overall type I error rate at 5%. †As a sensitivity analysis, the average relative effect and individual effects were estimated using a GEE distinct effects model. The treatment-by-outcome interaction p value was 0·13, meaning that there was not sufficient evidence to conclude that the effects differed across components. A p value of <0·05 was considered to be significant for the average relative effect; 95% CI was presented. p values of 0·017 (ie, 0·05 / 3) were considered significant for individual components and 98·3% CI were presented with their RRs. ‡The relative risk was estimated using a GEE model to adjust for within-patient correlation across components and log link (to estimate relative risk instead of odds ratio). Summary statistics were presented as the incidence of each outcome by group. §The hazard ratio of being discharged alive was estimated from a Cox proportional hazard model.

Table 2: Effects of thermal management on postoperative outcomes

Figure 4: Subgroup analysis of the primary outcome Circles represent the relative risk of aggressive warming versus routine care, with 95% CI error bars.

p valueRelative risk (95% CI)Aggressive warming, n/N

Routine care, n/N

Age, years

<65

≥65

Sex

Male

Female

Surgery type

Orthopaedic

Laparoscopic

Open abdominal

Neurosurgical

Urological

Other

0·81

0·29

0·37

48/753

198/1754

169/1668

77/836

6/56

110/1313

75/610

7/126

14/138

34/261

51/765

188/1741

156/1701

83/801

6/49

121/1301

71/613

3/89

9/138

29/312

Favours aggressive warming Favours routine care

1·00·5 1·50·1 2·0 3·5

Articles

1806 www.thelancet.com Vol 399 May 7, 2022

were 13 deaths in aggressively warmed patients and 17 deaths in patients assigned to routine thermal management.

Discussion PROTECT enrolled 5056 patients, making it more than five times larger than any previous perioperative thermal management trial to date.36 Temperature control was excellent, with time-weighted average intraoperative core temperatures of 36·8°C in patients allocated to aggressive warming and 35·8°C in those assigned to routine thermal management. The corresponding temperatures at the end of surgery were 37·1°C and 35·6°C, respectively. There was, nonetheless, no significant or clinically meaningful difference in the primary composite outcome of myocardial injury, non- fatal cardiac arrest, and 30-day mortality. The incidence of myocardial infarction was also similar in each group. Furthermore, the 95% CIs around the primary outcome were relatively small (0·87–1·25), making a type II error relatively unlikely.

Our results contrast with those reported by Frank and colleagues,37 who conducted the only previous substantive trial of thermal management and perioperative cardio- vascular outcomes, in which 300 patients having vascular surgery were randomly allocated to routine thermal management (which, in 1997, was no active warming) or to forced-air warming. Final intra operative core temperatures were 35·4°C and 36·7°C, respectively, which are similar to those seen in PROTECT patients. A cardiovascular composite outcome was observed in 6·3% of the hypothermic patients versus 1·4% of those kept normothermic (p=0·02), with the composite being largely driven by electrocardiographic evidence of myocardial ischemia and ventricular tachycardia.37 Because the trial was conducted before the troponin biomarker became available, diagnoses were primarily based on Holter electrocardiogram findings. The reported incidence of myocardial infarction was less than 1%, whereas the true incidence is at least 14% in patients who have vascular surgery.38 In another trial of 100 patients having abdominal aortic

surgery, again based on Holter monitoring, there was no difference in postoperative myocardial ischemia in hypothermic (35·6°C) and normothermic (36·4°C) patients.39 Among PROTECT and relevant trials,37,40,41 nearly all outcome events are from the current trial and clearly indicate that cardiovascular outcomes are similar at 35·5°C and 37°C.

Hypothermia potentially promotes surgical site infections via three mechanisms. First, hypothermia triggers postoperative thermoregulatory arterio-venous shunt constriction,42 which potentially reduces delivery of immune cells to injured tissue. Second, hypothermia decreases tissue oxygenation,43 which is necessary for oxidative killing—the major defence against bacterial infection.44 Third, mild core hypothermia directly impairs immune function (including chemotaxis and phagocytosis of granulocytes), reduces macrophage motility, and decreases antibody production. Nonetheless, the incidence of deep and organ-space infection was similar between the two groups in PROTECT: 6·3% in patients assigned to routine care and 7·2% in those assigned to aggressive warming. Similarly, superficial infections (which are less serious) did not differ between the two groups.

The similar incidence of serious wound infections between the two groups in PROTECT contrasts with observational analyses17,45 and the only previous substantive trial of perioperative thermal management and surgical site infections.20 Kurz and colleagues20 randomly allocated 200 patients, who were having colorectal resections, to receive either passive insulation or forced-air warming, with final intraoperative core temperatures being 34·7°C and 36·6°C, respectively. This 1996 trial differs from PROTECT in two important ways. First, patients assigned to routine thermal management for open abdominal surgery had final intraoperative temperatures that were about 1°C lower than patients in our current trial. And second, the primary outcome included superficial infections, which are far more common (and less serious) than the deep and organ- space infections that defined the main infection outcome in the current trial. It remains plausible that mild hypothermia to core temperatures near 34·5°C promotes surgical site infections, although our current results indicate that infection risk is similar at 35·5°C and 37·0°C. It is also important to recognise that the previous trial was conducted 25 years ago, and that perioperative management has changed much since then, including the routine use of laparoscopic surgical approaches. Furthermore, the previous trial was small, the findings had wide CIs, and the reported three-times reduction in infection incidence seems biologically implausible.20 It is clear from the results of PROTECT that very mild hypothermia to 35·5°C does not promote serious wound infections, although lower temperatures might.

Hypothermia decreases platelet aggregation46 and impairs enzymes of the coagulation cascade.47,48 It is,

Aggressive warming (n=39) Routine care* (n=54)

All participants (n=93)

Number of events

Related to treatment

Definitely not or probably not related

Probably related

Serious adverse events 17 16 1 30 47

Other adverse events 22 18 4 24 46

Data are number of events. *All adverse events in the routine thermal management group were definitely not or probably not related to treatment.

Table 3: Adverse events

Articles

www.thelancet.com Vol 399 May 7, 2022 1807

thus, unsurprising that the best-documented compli- cation of perioperative hypothermia is coagulopathy and increased transfusion requirement. Meta-analyses in 2008 and 2020 both reported that active warming reduced the odds of transfusion by at least 22%.19,36 Although patients in the current trial had major surgery lasting an average of more than 4 h, only about 10% required red blood cell transfusions. There was no difference in the transfusion requirements of patients between the two groups. Furthermore, the change in haemoglobin from before surgery to the first postoperative morning—a rough measure of blood loss—was similar between groups. As might be expected from an absence of cardiovascular, infectious, or coagulation effects, very mild hypothermia did not prolong hospitalisation or increase the risk of readmission within 30 days. Quality of Recovery-15 scores, assessed on postoperative day 3, were also similar between the two thermal management strategies.

There were fewer serious adverse events in patients assigned to aggressive warming than to routine care (17 vs 30 respectively), and only one (fever without infection) was deemed to be possibly related to warming. The difference seems more likely to be spurious than a serious safety signal.

PROTECT has several limitations. It was impossible to ensure that the preoperative and intraoperative teams were unaware of the thermal management plan for each patient. Similarly, patients knew that they were pre-warmed, although they were not told that pre-warming was restricted to the aggressive warming group. The primary outcome (a composite of myocardial injury, non-fatal cardiac arrest, and all-cause mortality) is objective and seems unlikely to have been influenced by patient perception of warming. Because non-fatal arrests and death were relatively uncommon in both groups, the composite primary endpoint essentially represents myocardial injury. The secondary outcomes were also objective or unlikely to be influenced by the absence of blinding to allocation. There was no protocol for transfusions, which decreases the precision for detecting temperature-related effects on transfusion.

An additional limitation is that there are thermo- regulatory responses we did not evaluate, including thermal comfort and shivering, which might have been more common at 35·5°C than at 37°C. However, both responses are minor and transient. Although none of the outcomes we considered differed between patients assigned to routine care (35·5°C) or aggressive warming (37°C), available evidence suggests that tempera- tures near 34·5°C cause complications, including surgical site infection,20 coagulopathy,19 prolonged recovery,22 shivering,42 and thermal discomfort.49 Patients having surgery should thus be warmed as necessary to keep their core temperature at 35·5°C or above. As with any trial, our conclusions directly apply to the patients we enrolled, and can reasonably be extrapolated to similar patients. Nearly

all enrolled patients were Chinese, few had orthopaedic or vascular surgery, and all had general anaesthesia. As such, results might differ in other populations, including patients with obesity, those having emergency surgery, or those at greater cardiovascular risk than patients enrolled in PROTECT.

In summary, the incidence of a 30-day composite of major cardiovascular outcomes was similar in patients allocated to routine care (35·5°C) or aggressive warming (37°C). Serious wound infections and transfusions were also similar between groups, as were the duration of hospital stay and hospital readmissions within 30 days. At least over a 1·5°C range from very mild hypothermia to full normothermia, there was no evidence that any substantive outcome varied. Keeping core temperature at least 35·5°C in patients having surgery appears sufficient. Contributors DIS, EJM, LP, MTVC, AK, and PJD designed the trial. JY, XH, RW, HC, YZ, ZS, WH, CP, XL, XZ, SW, YK ZM, XG, CM, ER, GRB, MB, AT, KR, KM, MR, SE, CKML, BCPC, and RPLY acquired data. EYA, MMC, FAC, SML, CH, DC, MT, and MW were responsible for auditing and adjudication. GM, YH, and EJM analysed data. GM, EJM, and DIS had full access to and verified the underlying data. DIS drafted the manuscript. All authors reviewed and approved the manuscript for publication.

Declaration of interests DIS occasionally consults for various temperature-related companies; all fees are donated to charity. Unrelated to this trial, DIS is a consultant for Pacira Biosciences (Parsippany, NJ, USA). DIS also reports advisory board participation and equity interests in Calorint (Philadelphia, PA, USA), TransQtronics (Philadelphia, PA, USA), the Health Data Analytics Institute (Boston, MA, USA), Medasense (Tel Aviv, Israel), Serenno (Tel Aviv, Israel), Sensifree (Cupertino, CA, USA), Perceptive Medical (Newport Beach, CA, USA), and Neuroindex (Tel Aviv, Israel). All other authors declare no competing interests.

Data sharing The informed consent and statistical plan will be freely available from the corresponding author. Complete deidentified individual-patient data, including a data dictionary and supporting documentation, will be available for collaborative analyses 1 year after publication. Interested investigators should submit proposed protocols to the corresponding author at [email protected]. Approval by the trial steering committee and a data use agreement will be required.

Acknowledgments This investigator-initiated trial was initially supported by 3M (St Paul, MN, USA). The company withdrew funding after 50% of the planned patients were enrolled, a decision they made without access to any outcome data. The trial was also supported by the Health and Medical Research Fund, Food and Health Bureau, Hong Kong Special Administrative Region (HMRF 06170306). This work was presented at the American College of Cardiology Annual Meeting in April, 2022.

References 1 Henderson WG, Khuri SF, Mosca C, Fink AS, Hutter MM,

Neumayer LA. Comparison of risk-adjusted 30-day postoperative mortality and morbidity in Department of Veterans Affairs hospitals and selected university medical centers: general surgical operations in men. J Am Coll Surg 2007; 204: 1103–14.

2 Semel ME, Lipsitz SR, Funk LM, Bader AM, Weiser TG, Gawande AA. Rates and patterns of death after surgery in the United States, 1996 and 2006. Surgery 2012; 151: 171–82.

3 Nepogodiev D, Martin J, Biccard B, et al. Global burden of postoperative death. Lancet 2019; 393: 401.

4 Devereaux PJ, Sessler DI. Cardiac complications in patients undergoing major noncardiac surgery. N Engl J Med 2015; 373: 2258–69.

Articles

1808 www.thelancet.com Vol 399 May 7, 2022

5 Devereaux PJ, Biccard BM, Sigamani A, et al. Association of postoperative high-sensitivity troponin levels with myocardial injury and 30-day mortality among patients undergoing noncardiac surgery. JAMA 2017; 317: 1642–51.

6 Devereaux PJ, Chan MT, Alonso-Coello P, et al. Association between postoperative troponin levels and 30-day mortality among patients undergoing noncardiac surgery. JAMA 2012; 307: 2295–304.

7 The Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Investigators. Myocardial injury after noncardiac surgery: a large, international, prospective cohort study establishing diagnostic criteria, characteristics, predictors, and 30-day outcomes. Anesthesiology 2014; 120: 564–78.

8 Ruetzler K, Smilowitz NR, Berger JS, et al. Diagnosis and management of patients with myocardial injury after noncardiac surgery: a scientific statement from the American Heart Association. Circulation 2021; 144: e287–305.

9 Greif R, Laciny S, Rajek A, Doufas AG, Sessler DI. Blood pressure response to thermoregulatory vasoconstriction during isoflurane and desflurane anesthesia. Acta Anaesthesiol Scand 2003; 47: 847–52.

10 Frank SM, Higgins MS, Breslow MJ, et al. The catecholamine, cortisol, and hemodynamic responses to mild perioperative hypothermia. A randomized clinical trial. Anesthesiology 1995; 82: 83–93.

11 Helwani MA, Amin A, Lavigne P, et al. Etiology of acute coronary syndrome after noncardiac surgery. Anesthesiology 2018; 128: 1084–91.

12 Weirich TL. Hypothermia/warming protocols: why are they not widely used in the OR? AORN J 2008; 87: 333–44.

13 Yi J, Xiang Z, Deng X, et al. Incidence of inadvertent intraoperative hypothermia and its risk factors in patients undergoing general anesthesia in Beijing: a prospective regional survey. PLoS One 2015; 10: e0136136.

14 Koh W, Chakravarthy M, Simon E, et al. Perioperative temperature management: a survey of 6 Asia-Pacific countries. BMC Anesthesiol 2021; 21: 205.

15 Sessler DI, Lee KA, McGuire J. Isoflurane anesthesia and circadian temperature cycles in humans. Anesthesiology 1991; 75: 985–89.

16 Protsiv M, Ley C, Lankester J, Hastie T, Parsonnet J. Decreasing human body temperature in the United States since the industrial revolution. eLife 2020; 9: e49555.

17 Scott AV, Stonemetz JL, Wasey JO, et al. Compliance with Surgical Care Improvement Project for body temperature management (SCIP Inf-10) is associated with improved clinical outcomes. Anesthesiology 2015; 123: 116–25.

18 Sun Z, Honar H, Sessler DI, et al. Intraoperative core temperature patterns, transfusion requirement, and hospital duration in patients warmed with forced air. Anesthesiology 2015; 122: 276–85.

19 Rajagopalan S, Mascha E, Na J, Sessler DI. The effects of mild perioperative hypothermia on blood loss and transfusion requirement. Anesthesiology 2008; 108: 71–77.

20 Kurz A, Sessler DI, Lenhardt R. Perioperative normothermia to reduce the incidence of surgical-wound infection and shorten hospitalization. N Engl J Med 1996; 334: 1209–15.

21 Heier T, Clough D, Wright PM, Sharma ML, Sessler DI, Caldwell JE. The influence of mild hypothermia on the pharmacokinetics and time course of action of neostigmine in anesthetized volunteers. Anesthesiology 2002; 97: 90–95.

22 Lenhardt R, Marker E, Goll V, et al. Mild intraoperative hypothermia prolongs postanesthetic recovery. Anesthesiology 1997; 87: 1318–23.

23 Kurz A, Sessler DI, Narzt E, Lenhardt R, Lackner F. Morphometric influences on intraoperative core temperature changes. Anesth Analg 1995; 80: 562–67.

24 Torossian A. Survey on intraoperative temperature management in Europe. Eur J Anaesthesiol 2007; 24: 668–75.

25 Sessler DI, Schroeder M, Merrifield B, Matsukawa T, Cheng C. Optimal duration and temperature of prewarming. Anesthesiology 1995; 82: 674–81.

26 Lau A, Lowlaavar N, Cooke EM, et al. Effect of preoperative warming on intraoperative hypothermia: a randomized-controlled trial. Can J Anaesth 2018; 65: 1029–40.

27 Matsukawa T, Sessler DI, Christensen R, Ozaki M, Schroeder M. Heat flow and distribution during epidural anesthesia. Anesthesiology 1995; 83: 961–67.

28 Stark PA, Myles PS, Burke JA. Development and psychometric evaluation of a postoperative quality of recovery score: the QoR-15. Anesthesiology 2013; 118: 1332–40.

29 Chan MT, Lo CC, Lok CK, Chan TW, Choi KC, Gin T. Psychometric testing of the Chinese quality of recovery score. Anesth Analg 2008; 107: 1189–95.

30 Thygesen K, Alpert JS, Jaffe AS, Simoons ML, Chaitman BR, White HD. Third universal definition of myocardial infarction. Circulation 2012; 126: 2020–35.

31 Devereaux PJ, Mrkobrada M, Sessler DI, et al. Aspirin in patients undergoing noncardiac surgery. N Engl J Med 2014; 370: 1494–503.

32 Devereaux PJ, Sessler DI, Leslie K, et al. Clonidine in patients undergoing noncardiac surgery. N Engl J Med 2014; 370: 1504–13.

33 Mascha EJ. Power calculations for tests on a vector of binary outcomes (MULTBINPOW). Cleveland Clinic Statistical Software Series. http://www.lerner.ccf.org/qhs/software/multbinpow.php (accessed March 29, 2022).

34 Mascha EJ, Imrey PB. Factors affecting power of tests for multiple binary outcomes. Stat Med 2010; 29: 2890–904.

35 Mascha EJ, Sessler DI. Statistical grand rounds: design and analysis of studies with binary- event composite endpoints: guidelines for anesthesia research. Anesth Analg 2011; 112: 1461–71.

36 Balki I, Khan JS, Staibano P, et al. Effect of perioperative active body surface warming systems on analgesic and clinical outcomes: a systematic review and meta-analysis of randomized controlled trials. Anesth Analg 2020; 131: 1430–43.

37 Frank SM, Fleisher LA, Breslow MJ, et al. Perioperative maintenance of normothermia reduces the incidence of morbid cardiac events. A randomized clinical trial. JAMA 1997; 277: 1127–34.

38 Biccard BM, Sigamani A, Chan MTV, et al. Effect of aspirin in vascular surgery in patients from a randomized clinical trial (POISE-2). Br J Surg 2018; 105: 1591–97.

39 Elmore JR, Franklin DP, Youkey JR, Oren JW, Frey CM: Normothermia is protective during infrarenal aortic surgery. J Vasc Surg 1998; 28: 984–92.

40 Krenzischek DA, Frank SM, Kelly S: Forced-air warming versus routine thermal care and core temperature measurement sites. J Perianesth Nurs 1995; 10: 69–78.

41 Yi J, Liang H, Song R, Xia H, Huang Y: Maintaining intraoperative normothermia reduces blood loss in patients undergoing major operations: a pilot randomized controlled clinical trial. BMC Anesthesiol 2018; 18: 126.

42 Sessler DI, Rubinstein EH, Moayeri A. Physiologic responses to mild perianesthetic hypothermia in humans. Anesthesiology 1991; 75: 594–610.

43 Sheffield CW, Sessler DI, Hopf HW, et al. Centrally and locally mediated thermoregulatory responses alter subcutaneous oxygen tension. Wound Repair Regen 1996; 4: 339–45.

44 Allen DB, Maguire JJ, Mahdavian M, et al. Wound hypoxia and acidosis limit neutrophil bacterial killing mechanisms. Arch Surg 1997; 132: 991–96.

45 Seamon MJ, Wobb J, Gaughan JP, Kulp H, Kamel I, Dempsey DT. The effects of intraoperative hypothermia on surgical site infection: an analysis of 524 trauma laparotomies. Ann Surg 2012; 255: 789–95.

46 Michelson AD, MacGregor H, Barnard MR, Kestin AS, Rohrer MJ, Valeri CR. Reversible inhibition of human platelet activation by hypothermia in vivo and in vitro. Thromb Haemost 1994; 71: 633–40.

47 Felfernig M, Blaicher A, Kettner SC, Felfernig D, Acimovic S, Kozek-Langenecker SA. Effects of temperature on partial thromboplastin time in heparinized plasma in vitro. Eur J Anaesthesiol 2001; 18: 467–70.

48 Rohrer MJ, Natale AM. Effect of hypothermia on the coagulation cascade. Crit Care Med 1992; 20: 1402–05.

49 Kurz A, Sessler DI, Narzt E, et al. Postoperative hemodynamic and thermoregulatory consequences of intraoperative core hypothermia. J Clin Anesth 1995; 7: 359–66.