In 2002, two trials reported that cooling comatose patients after resuscitation from out-of-hospital cardiac arrest (OHCA) to 32–34°C for 12–24 hours improved survival (1) and neurologic function (1 , 2). Mild induced hypothermia was subsequently recommended in international guidelines (3). However, the overall evidence supporting the intervention was of low quality (4), and it was unclear whether the benefits seen might not simply result from avoiding fever (5). The “target temperature management (TTM) 33°C versus 36°C after OHCA trial” (the TTM trial) compared two temperature control regimens and reported that comatose patients cooled to 36°C with strict prevention of fever for the first 72 hours after resuscitation from OHCA had similar survival and neurologic outcomes to patients cooled to 33°C (6). This finding led to changes in international cardiac arrest guidelines which now recommend cooling postcardiac arrest patients to 32–36°C (7). The Australian and New Zealand (ANZ) Intensive Care Society Adult Patient Database (8) offers the opportunity to evaluate whether practice changed after the TTM trial and whether such translation of new knowledge into practice was in adherence with the TTM protocol and associated with changes in patient outcomes.
Our primary hypothesis was that there has been widespread adoption of TTM at 36°C in ANZ ICUs since the publication of the TTM trial (6). As data from a number of countries have demonstrated that risk-adjusted mortality following OHCA is reducing over time (9–12), we hypothesized that similar temporal trends would be present in ANZ ICUs. Although recent data evaluating whether implementation of TTM at 36°C improves patient outcomes compared with cooling to 32–34°C are contradictory (13 , 14), we reasoned that it might logically have resulted in an increased focus on postcardiac arrest care generally. Accordingly, we further hypothesized that temporal mortality reductions may have accelerated after publication of the TTM trial.
Study Design and Oversight
We conducted a retrospective cohort study using data from the ANZ Intensive Care Society Centre for Outcome and Resource Evaluation Adult Patient Database (8), which contains greater than 2 million admission episodes from 186 ANZ ICUs.
The study protocol and statistical analysis plan were developed by the study management committee and were posted online on October 3, 2017 (supplementary materials, Supplemental Digital Content 1, http://links.lww.com/CCM/D847) (15). The critical care community was invited to provide feedback on the analysis plan by October 30, 2017, via an e-mail sent to the 8,272 people on the http://www.criticalcarereviews.com mailing list on October 9, 2017. No changes to the statistical analysis plan were made in response to feedback.
The study was approved by the New Zealand Central Health and Disability Ethic Committee (16-CEN-197). As the study involved retrospective review of deidentified data, requirements for informed consent were waived.
Patients were eligible for inclusion if they were mechanically ventilated adults (≥ 18 yr old) admitted to ICU from an emergency department within 24 hours of hospital admission following OHCA. Patients admitted following OHCA were defined as those with an Acute Physiology and Chronic Health Evaluation admission diagnostic code of “cardiac arrest” or with a documented cardiac arrest in the 24 hours prior to ICU admission. Patients who were admitted to ICU for palliative care or with limitations of treatment in place were excluded.
We divided our study population into three cohorts based on ICU admission date. The pre-TTM trial cohort included patients admitted in the 3 years prior to publication of the TTM trial (6) (January 2011 until December 2013 inclusive). The post-TTM trial cohort included patients admitted in the 3 years after publication of the TTM trial (6) (January 2014 until December 2016 inclusive). The extended pre-TTM trial cohort included patients admitted to ICU between January 2005 and December 2013.
Exposures and Outcomes
The primary exposure of interest was admission before versus after December 2013 (the TTM trial was published on December 5, 2013). The primary outcome variable to evaluate temperature management was the lowest temperature in the first 24 hours in ICU. To further describe temperature management, we reported the highest temperature in the first 24 hours in ICU and the difference between the highest and lowest temperature in the first 24 hours. Additional temperature variables were reported on a “post hoc” basis as outlined in the supplementary methods (supplementary materials, Supplemental Digital Content 1, http://links.lww.com/CCM/D847). The primary clinical outcome variable of interest was inhospital mortality. Other clinical outcome variables of interest were 1) the proportion of patients with a body temperature of greater than 38°C in the first 24 hours in ICU, 2) the ICU and hospital length of stay, 3) the proportion of patients discharged home, and 4) the proportion of survivors discharged home and discharged to a rehabilitation facility.
Age, gender, chronic comorbidities, location of ICU (Australia vs New Zealand), and type of ICU (tertiary/metropolitan/rural/private) were recorded. See supplementary materials (Supplemental Digital Content 1, http://links.lww.com/CCM/D847) for additional details of physiologic data collected. ANZ Risk of Death (ANZROD), which is based on comorbidities and physiologic data from the first 24 hours in ICU, was calculated (16 , 17). Because mild induced hypothermia directly alters body temperature, ANZROD was also calculated with the contribution from the temperature component removed. In addition, because arterial blood gas measurements of PaO2 and blood pH may be affected by blood temperature (18) and cooling to 33°C would be expected to alter heart rate, urine output, and respiratory rate, without necessarily altering illness severity, ANZROD with temperature, PaO2, pH, heart rate, urine output, and respiratory rate components removed were calculated.
The principal comparison was the pre-TTM trial cohort with the post-TTM trial cohort. We chose to compare these cohorts because we reasoned that a comparison of data from 3 years before to 3 years after publication of the TTM trial would be less subject to confounding due to temporal changes than a comparison between the extended pre-TTM trial cohort and the post-TTM trial cohort. We chose a 3-year period of time because 3 years of data were available from the post-TTM trial period.
Comparisons between groups were performed using chi-square tests for proportions, Student t tests for normally distributed data and Wilcoxon rank-sum tests otherwise, with results reported as n with percentages, mean ± SDs, or median (interquartile range), respectively. Multivariable comparisons were performed using hierarchical logistic regression with patients nested within site and site treated as a random effect with results reported as odds ratio (OR), 99% CIs. To test the robustness of findings, and as a form of sensitivity analysis, three risk-adjustment models were used: 1) adjusting for ANZROD, 2) adjusting for ANZROD with the temperature component removed, and 3) adjusting for ANZROD with the temperature-affected components removed (temperature, respiratory rate, heart rate, urine output, pH, and PaO2).
Temporal trends from January 2005 until December 2016 were evaluated using segmented linear regression analysis with data aggregated at a monthly level. Because the TTM trial was published in December 2013, we defined the end of December 2013 as the break point in this analysis. Segmented regression was used to evaluate whether there was a stepwise change in the value of each outcome of interest associated with TTM publication and whether there was a difference in the slope or rate of change before versus after TTM publication. Additional details of the adjusted analyses and segmented regression analyses are described in the supplementary methods (supplementary materials, Supplemental Digital Content 1, http://links.lww.com/CCM/D847).
The associations between admission before versus after publication of the TTM trial and lowest body temperature in the first 24 hours in ICU and inhospital mortality were evaluated in three prespecified subgroup pairs. These subgroup pairs were 1) patients admitted to tertiary ICUs versus not, 2) patients admitted to New Zealand versus Australian ICUs, and 3) patients less than the 60 years versus those 60 years old or older. To determine if temperature management differed between subgroups, temperatures between periods (minimum, maximum, difference between highest and lowest temperature) were analyzed using hierarchical mixed modeling fitting main effects for period and subgroup and an interaction between the two. A similar analysis was undertaken to evaluate associated mortality changes in prespecified subgroups. All analyses were performed using SAS Version 9.4 (SAS Institute, Cary, NC). To increase the robustness of the analysis, a two sided p value of less than 0.01 was used to indicate statistical significance. No additional adjustment for multiple comparisons was performed.
Of 1,292,663 ICU admission episodes between January 2005 and the December 2016, 17,788 from 140 hospitals fulfilled the inclusion criteria. Of these, 1,536 were excluded with because they were admitted to ICU for palliative care or with treatment limitations in place (Fig. 1).
The pre-TTM trial cohort included 4,450 patients, and the post-TTM trial cohort included 5,184 patients. Time between hospital admission and intensive care admission was statistically significantly longer in pre-TTM trial patients than post-TTM trial patients. Pre-TTM trial patients were statistically significantly older, more likely to have chronic cardiac disease, and less likely to be admitted to a tertiary ICU than post-TTM trial patients (Table 1).
Illness severity based on ANZROD was significantly higher in the pre-TTM trial cohort than the post-TTM trial cohort; however, when ANZROD was recalculated by either removing the temperature component or removing all temperature-affected variables, there was no difference in illness severity between groups (Table S1, Supplemental Digital Content 1, http://links.lww.com/CCM/D847). There were statistically significant differences between the pre-TTM trial cohort and the post-TTM trial cohort in all physiologic variables which were prespecified as likely to be affected by mild induced hypothermia (i.e., temperature, respiratory rate, heart rate, PaO2, urine output, and pH) (Table S1, Supplemental Digital Content 1, http://links.lww.com/CCM/D847).
The characteristics and outcomes of the 11,068 patients included in analyses of temporal trends from January 2005 until December 2013 are shown in Tables S2-S4 (Supplemental Digital Content 1, http://links.lww.com/CCM/D847).
Mean ± SD lowest body temperature in the first 24 hours in ICU in pre-TTM patients and post-TTM patients was 33.77 ± 1.71°C and 34.74 ± 1.39°C, respectively (absolute difference, 0.98°C [99% CI, 0.89–1.06°C]; p < 0.001) (Table 2). Segmented regression analysis demonstrated a statistically significant stepwise change in lowest body temperature in the first 24 hours in ICU associated with publication of the TTM trial (absolute change, 0.86°C [99% CI, 0.56–1.16°C]; p < 0.001) (Fig. 2) with a corresponding decrease in the proportion of patients with a lowest temperature less than 34°C (Fig. S1 and Table S5, supplementary materials, Supplemental Digital Content 1, http://links.lww.com/CCM/D847). Rate of change of lowest temperature over time was –0.08°C (99% CI, –0.11°C to –0.05°C) per year from January 2005 until December 2013 and +0.12°C (99% CI, 0.02–0.22°C) per year from January 2014 until December 2016. After accounting for autocorrelation, the change in slope between periods was +0.23°C per year (99% CI, 0.08–0.38; p < 0.001).
In 568 of 4,450 pre-TTM trial patients (12.8% [99% CI, 11.5–14.1%]) and 853 of 5,184 post-TTM trial patients (16.5% [99% CI, 15.2–17.8%]), a fever of greater than 38°C was recorded in the first 24 hours in ICU (OR post-TTM trial publication vs pre-TTM trial publication 1.35 [99% CI, 1.16–1.57]; p < 0.001) (Table 2).
Two-thousand three-hundred thirty-one (52.4%) of 4,450 pre-TTM trial patients and 2,769 (53.4%) of 5,184 post-TTM trial patients died in hospital (OR post-TTM trial vs pre-TTM trial 1.04 [99% CI, 0.94–1.16]; p = 0.31). When adjusted for ANZROD, admission in the post-TTM trial period was associated with a statistically significant increased mortality risk (adjusted OR, 1.27 [99% CI, 1.13–1.43]; p < 0.001). However, the strength of this association decreased when the temperature component of ANZROD was removed (adjusted OR, 1.06 [99% CI, 0.95–1.20]; p = 0.17) and when all temperature-affected variables of ANZROD were removed (adjusted OR, 1.06 [99% CI, 0.95–1.19]; p = 0.18).
In the segmented regression analysis, inhospital mortality decreased by 1.3 (99% CI, –1.8 to –0.9) percentage points per year from January 2005 until December 2013 and increased by 0.6 (99% CI, –1.4 to 2.6) percentage points per year from January 2014 until December 2016 (change in slope 1.9 percentage points per year [99% CI, –0.6 to 4.4]; p = 0.05) (Fig. 3). There was no statistically significant stepwise change in mortality associated with admission after December 2013 (stepwise change in mortality 2.5 percentage points [99% CI, –2.4 to 7.4 percentage points]; p = 0.20). Additional data on mortality trends are presented in Figs. S2 and S3 (Supplemental Digital Content 1, http:links.lww.com/CCM/D847). An interactive representation of study data showing detailed temperature and mortality data over time is presented online (https://doi.wellingtonicu.com/cardiac_arrest_temperatures/).
ICU and Hospital Length of Stay.
ICU length of stay was significantly longer in pre-TTM trial patients than post-TTM trial patients (2.94 d [99% CI, 2.82–3.07 d] pre-TTM trial vs 2.75 d [99% CI, 2.64–2.85 d] post-TTM trial; ratio of geometric means post-TTM trial to pre-TTM trial 0.93 [99% CI, 0.88–0.99]; p = 0.002). This effect was primarily driven by a difference in ICU length of stay among survivors (Table 2). There were no significant differences in hospital length of stay between the pre-TTM trial and post-TTM trial periods (Table 2). The associations between admission in the pre-TTM trial and post-TTM trial periods and ICU and hospital length of stay were similar in adjusted analyses (Table S7, Supplemental Digital Content 1, http://links.lww.com/CCM/D847).
There was no significant heterogeneity between subgroup pairs based on country of admission, admission to a tertiary versus nontertiary ICU, or patient age with respect to mortality risk associated with admission in the pre-TTM trial period versus the post-TTM trial period (Table S6, Supplemental Digital Content 1, http://links.lww.com/CCM/D847).
Similar proportions of pre-TTM trial and post-TTM trial patients were discharged home (Table S8 and Fig. S4, Supplemental Digital Content 1, http://links.lww.com/CCM/D847). Of surviving patients, 1,375 (64.9%) of 2,119 pre-TTM trial patients and 1,651 (68.4%) of 2,415 post-TTM trial patients were discharged home (OR post-TTM trial vs pre-TTM trial 1.17 [99% CI, 0.99–1.38]; p = 0.01). Of surviving patients, 307 (14.5%) of 2,119 pre-TTM trial patients and 300 (12.4%) of 2,415 post-TTM trial patients were discharged to rehabilitation (OR post-TTM trial vs pre-TTM trial 0.84 [99% CI, 0.67–1.05]; p = 0.04) (Table S8 and Fig. S5, Supplemental Digital Content 1, http://links.lww.com/CCM/D847).
We conducted a retrospective cohort study to evaluate translation of knowledge into practice in relation to temperature management in post-OHCA patients admitted to ANZ ICUs. We demonstrated a highly statistically significant change in temperature management following the publication of the TTM trial with an increase in lowest body temperature in the first 24 hours in ICU and a corresponding fall in the proportion of patients with a lowest temperature less than 34°C. Changes in temperature management were evident both in Australia and New Zealand; in tertiary, metropolitan, and rural hospitals; and occurred irrespective of patient age. Contrary to our a priori hypothesis, admission after publication of the TTM trial was not associated with a decrease in mortality over time. We observed a temporal decline in mortality in the years prior to publication of the TTM trial but not after its publication.
Admission after publication of the TTM trial was associated with higher risk of fever. Fever is associated with mortality risk in OCHA patients (5). Although this association does not necessarily indicate that fever causes mortality, this finding is of potential concern. An increased frequency of fever was not observed in patients allocated to the 36°C group in the TTM trial (6). Thus, the increased frequency of fever observed in our study may be a consequence of targeting 36°C outside the context of a clinical trial or may reflect less stringent adherence to any form of temperature control after publication of the TTM trial.
Our study has a number of limitations. Because our database only included the highest and lowest temperature in the first 24 hours in ICU, we are unable to report important confounding variables in relation to temperature management. Such variables include whether TTM was initiated prior to ICU admission, how hypothermia was induced and maintained, whether target temperatures were achieved, how rewarming was performed, whether shivering occurred, and how sedation and paralysis were used. Additionally, we are not able to report the frequency of fever occurring beyond 24 hours, and this might vary according to the temperature management strategy used (13). We conducted our study in accordance with a prespecified analysis plan in order to avoid the risk of analysis bias but acknowledge that other methods to evaluate temporal trends are possible (19). We observed statistically significantly higher mortality associated with admission after publication of the TTM trial in the illness-severity adjusted analysis adjusting for ANZROD. However, conventional ICU illness severity scoring systems only offer moderately accurate prediction of mortality risk in the OHCA setting (20). Furthermore, because this analysis was confounded by the influence of mild induced hypothermia on markers of illnesses severity such as body temperature, heart rate, pH, and PaO2 (18), its significance is uncertain. Our database did not capture predictors of outcome in OHCA patients including whether the patients received bystander cardiopulmonary resuscitation (CPR), their initial rhythm, or time to return of spontaneous circulation. In many parts of the world (21), including Australia (22), the proportion of patients receiving bystander CPR is increasing with time. This and other improvements in prehospital care, such as increased public access to defibrillators (23), would be expected to result in mortality declining with time as has been demonstrated in a number of previous OHCA studies (9–12 , 24 , 25). These background temporal trends toward reducing mortality would be expected to bias against a finding of increased in mortality risk associated with ICU admission at a later time point. Thus, given the mortality trends seen in our study, the possibility of increased mortality risk associated with implementation of less aggressive temperature management cannot be excluded. Nevertheless, the observation that a greater proportion of OHCA patients are being discharged home over time and that this trend continued unabated after publication of the TTM trial, provides some reassurance that the proportion of patients surviving with a good neurologic outcome has not decreased despite the increased occurrence of fever observed in the post-TTM period in our study.
Our secondary finding that post-TTM trial patients, who had higher body temperatures, had a shorter ICU length of stay than pre-TTM trial patients is consistent with previous literature. This literature indicates that clearance of sedatives and neuromuscular blocking agents is delayed in patients receiving mild induced hypothermia (26 , 27). This finding might also be attributable to the greater time it takes to rewarm patients to normothermia from 33°C compared with 36°C.
In conclusion, our findings strongly suggest widespread practice change in ANZ ICUs following publication of the TTM trial with clinicians adopting higher temperature targets. The change in temperature management was associated with an increased frequency of fever not seen in the TTM trial. Given the uncertainty about whether the increased risk of fever observed in our study affected mortality rates, further research is required to establish the optimal approach to temperature management in adults who are comatose after OHCA.
1. Hypothermia after Cardiac Arrest
Study Group: Mild therapeutic hypothermia
to improve the neurologic outcome after cardiac arrest
. N Engl J Med 2002; 346:549–556
2. Bernard SA, Gray TW, Buist MD, et alTreatment of comatose survivors of out-of-hospital cardiac arrest
with induced hypothermia. N Engl J Med 2002; 346:557–563
3. Peberdy MA, Callaway CW, Neumar RW, et alAmerican Heart Association: Part 9: Post-cardiac arrest
care: 2010 American Heart Association Guidelines for cardiopulmonary resuscitation and emergency cardiovascular care. Circulation 2010; 122:S768–S786
4. Nielsen N, Friberg H, Gluud C, et alHypothermia after cardiac arrest
should be further evaluated–a systematic review of randomised trials with meta-analysis and trial sequential analysis. Int J Cardiol 2011; 151:333–341
5. Zeiner A, Holzer M, Sterz F, et alHyperthermia after cardiac arrest
is associated with an unfavorable neurologic outcome. Arch Intern Med 2001; 161:2007–2012
6. Nielsen N, Wetterslev J, Cronberg T, et alTTM Trial Investigators: Targeted temperature management at 33°C versus 36°C after cardiac arrest
. N Engl J Med 2013; 369:2197–2206
7. Callaway CW, Donnino MW, Fink EL, et alPart 8: Post-cardiac arrest
care: 2015 American Heart Association guidelines update for cardiopulmonary resuscitation and emergency cardiovascular care. Circulation 2015; 132:S465–S482
8. Stow PJ, Hart GK, Higlett T, et alANZICS Database Management Committee: Development and implementation of a high-quality clinical database: The Australian and New Zealand Intensive Care Society adult patient database. J Crit Care 2006; 21:133–141
9. Nolan JP, Ferrando P, Soar J, et alIncreasing survival after admission to UK critical care units following cardiopulmonary resuscitation. Crit Care 2016; 20:219
10. Fugate JE, Brinjikji W, Mandrekar JN, et alPost-cardiac arrest
mortality is declining: A study of the US National Inpatient Sample 2001 to 2009. Circulation 2012; 126:546–550
11. Reinikainen M, Oksanen T, Leppänen P, et alFinnish Intensive Care Consortium: Mortality in out-of-hospital cardiac arrest
patients has decreased in the era of therapeutic hypothermia
. Acta Anaesthesiol Scand 2012; 56:110–115
12. van der Wal G, Brinkman S, Bisschops LL, et alInfluence of mild therapeutic hypothermia
after cardiac arrest
on hospital mortality. Crit Care Med 2011; 39:84–88
13. Casamento A, Minson A, Radford S, et alA comparison of therapeutic hypothermia
and strict therapeutic normothermia after cardiac arrest
. Resuscitation 2016; 106:83–88
14. Bray JE, Stub D, Bloom JE, et alChanging target temperature from 33°C to 36°C in the ICU management of out-of-hospital cardiac arrest
: A before and after study. Resuscitation 2017; 113:39–43
15. Salter R, Bailey M, Bellomo R, et alTemperature Management in Ventilated Adults Admitted to Australian and New Zealand ICUs Following Out of Hospital Cardiac Arrest
: Statistical Analysis Plan. 2017. Available at: http://wellingtonicu.com/Data/Trials/OHCATempSAP.pdf
. Accessed December 10, 2017
16. Pilcher D, Paul E, Bailey M, et alThe Australian and New Zealand Risk of Death (ANZROD) model: Getting mortality prediction right for intensive care units. Crit Care Resusc 2014; 16:3–4
17. Paul E, Bailey M, Pilcher DRisk prediction of hospital mortality for adult patients admitted to Australian and New Zealand intensive care units: Development and validation of the Australian and New Zealand Risk of Death model. J Crit Care 2013; 28:935–941
18. Burnett RW, Noonan DCCalculations and correction factors used in determination of blood pH and blood gases. Clin Chem 1974; 20:1499–1506
19. Juurlink DN, Mamdani MM, Lee DS, et alRates of hyperkalemia after publication of the Randomized Aldactone Evaluation Study. N Engl J Med 2004; 351:543–551
20. Skrifvars MB, Varghese B, Parr MJSurvival and outcome prediction using the Apache III and the out-of-hospital cardiac arrest
(OHCA) score in patients treated in the intensive care unit (ICU) following out-of-hospital, in-hospital or ICU cardiac arrest
. Resuscitation 2012; 83:728–733
21. Daya MR, Schmicker RH, Zive DM, et alResuscitation Outcomes Consortium Investigators: Out-of-hospital cardiac arrest
survival improving over time: Results from the Resuscitation Outcomes Consortium (ROC). Resuscitation 2015; 91:108–115
22. Ambulance Victoria: Victorian Ambulance Cardiac Arrest
Registry 2015–2016 Annual Report. 2017, pp Doncaster, VIC, Australia, Ambulance Victoria, 30
23. Caffrey SL, Willoughby PJ, Pepe PE, et alPublic use of automated external defibrillators. N Engl J Med 2002; 347:1242–1247
24. Wong MK, Morrison LJ, Qiu F, et alTrends in short- and long-term survival among out-of-hospital cardiac arrest
patients alive at hospital arrival. Circulation 2014; 130:1883–1890
25. Okubo M, Kiyohara K, Iwami T, et alNationwide and regional trends in survival from out-of-hospital cardiac arrest
in Japan: A 10-year cohort study from 2005 to 2014. Resuscitation 2017; 115:120–128
26. Polderman KHApplication of therapeutic hypothermia
in the intensive care unit. Opportunities and pitfalls of a promising treatment modality–Part 2: Practical aspects and side effects. Intensive Care Med 2004; 30:757–769
27. Tortorici MA, Kochanek PM, Poloyac SMEffects of hypothermia on drug disposition, metabolism, and response: A focus of hypothermia-mediated alterations on the cytochrome P450 enzyme system. Crit Care Med 2007; 35:2196–2204
cardiac arrest; implementation science; intensive care medicine; knowledge translation; therapeutic hypothermia
Supplemental Digital Content
Copyright © by 2018 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.