Who benefits most from mild therapeutic hypothermia in coronary intervention era? A retrospective and propensity-matched study

We retrospectively enrolled witnessed adult (>18 years of age) OHCA patients with cardiac causes transported to Hiroshima City Hospital, who achieved ROSC and who were comatose between September 2003 and January 2010. All OHCA patients were treated in accordance with an advanced cardiac life support protocol, and the patients who met the criteria for hypothermia treatment were treated with MTH as reported previously [10]. MTH was fundamentally induced in cardiac arrest patients with presumed cardiac origin and the following criteria [15]: age 18 to 79 years, and an estimated interval of less than 15 minutes from collapse to the first attempt at resuscitation by any person.

Before 2006, an assessment of cardiac arrest complicated by ischemic heart disease was made in patients treated with and without MTH. In patients with suspected acute coronary syndrome, emergency coronary angiography or percutaneous coronary intervention, or both, were subsequently performed. After 2006, routine emergency coronary angiography was performed in patients treated with MTH.

The present study was approved by the local ethics committee on human research and is conducted in accordance with the guidelines of the Declaration of Helsinki. All data were collected within the normal daily care routine in an anonymous fashion. The institutional review board therefore waived the need for informed patient consent.

MTH was induced in comatose survivors using a surface cooling mattress and the administration of physiological saline (4°C) as reported previously [10]. Before January 2008, the target temperature was set between 32°C and 34°C (fundamentally, 33°C) and was maintained for 48 hours followed by rewarming at 0.5°C every 12 hours. After January 2008, the core temperature was maintained for 24 hours and rewarming continued for 12 hours.

The primary endpoint was a favorable neurological outcome, which is defined as category 1 (good performance) or category 2 (moderate disability) on the five-point Pittsburgh cerebral performance scale; the other categories are 3 (severe disability), 4 (vegetative state), and 5 (death) at the time of hospital discharge [16].

Continuous variables are presented as medians (with interquartile ranges), and categorical variables are presented as numbers and percentages. Differences between groups at baseline were analyzed using the Mann-Whitney U test for continuous variables and a chi-square test or Fisher's exact test for categorical variables as appropriate. The rate of favorable neurological outcome was plotted against the CPA-ROSC every 5 minutes for patients treated with and without MTH. We used Fisher's exact test to access differences in the rate of favorable neurological outcomes every 5 minutes between the groups.

To detect favorable neurological outcome, we constructed receiver-operating characteristic curves for the CPA-ROSC.

The threshold for performing MTH was set high early in the study period. For example, we induced MTH only in comatose survivors whose initial rhythm indicated ventricular fibrillation. Because the patients were not randomly assigned to receive MTH or normothermia, potential confounding and selection biases were accounted for by developing a propensity score. The propensity for MTH or not was determined without regard to the outcome using a multivariate regression model. For this model, we chose variables we thought might increase the propensity for MTH over normothermia therapy, which included age, Pittsburgh overall performance category score before cardiac arrest, initial rhythm, the time interval from collapse to start of CPR and the CPA-ROSC. This model yielded a concordance statistic of 0.80, which indicated good discrimination. A propensity score was then calculated from the logistic equation for each patient, which indicated the probability that a patient would be treated with MTH. All study patients were pooled and sorted according to their propensity scores in the ascending order and a propensity-matched cohort was formed [17].

A logistic regression model was used to examine the association between the CPA-ROSC and favorable neurological outcome in an unadjusted model, in a model adjusted for age, gender, and initial rhythm, and in a model adjusted for age, gender, Pittsburgh overall performance category score, initial rhythm, time interval from collapse to start of CPR, hypertension, diabetes mellitus, history of heart disease, emergency coronary angiography, primary percutaneous coronary intervention, use of intra-aortic balloon pump, use of extracorporeal life support, admission after 2006 and propensity score.

The JMP statistical package (version 5.0.1 J; SAS Institute, Cary, NC, USA) and R version 2.9.2 [18] were used for statistical analyses. All tests were two-sided, and P < 0.05 was considered statistically significant.

A flow diagram of the study patients and their outcomes is depicted in Figure 1. Witnessed adult comatose survivors of OHCA (n = 400) with cardiac causes were enrolled in the study. The baseline clinical characteristics, treatment and findings for study patients are shown in Tables 1 and 2. One hundred and ten patients (28%) were treated with MTH (hypothermia group) and 290 patients were treated without MTH (normothermia group). There were many differences in baseline characteristics between these groups. The normothermia group was in a more severe condition for resuscitation.

The outcomes of the study patients are shown in Table 3. The rate of favorable neurological outcome (39% vs. 14%, P < 0.001) and 30-day survival (48% vs. 16%, P < 0.001) were higher in the hypothermia group than in the normothermia group. The CPA-ROSC in patients with a favorable neurological outcome was longer (P < 0.001) in the hypothermia group (median, 22 minutes; interquartile range, 16 to 32 minutes; mean ± standard deviation, 25 ± 13 minutes) than in the normothermia group (median, 11 minutes; interquartile range, 5 to 14 minutes; mean ± standard deviation, 12 ± 8 minutes).

The rate of favorable neurological outcome was plotted against the CPA-ROSC every 5 minutes (Figure 2). In the hypothermia group, the rate of favorable neurological outcome decreased in a stepwise fashion when the CPA-ROSC was longer than 25 minutes. In the normothermia group, the rate of favorable neurological outcome decreased remarkably when the CPA-ROSC was longer than 15 minutes, and none of the patients whose CPA-ROSC was longer than 45 minutes had a favorable neurological outcome. The rate of favorable neurological outcome was significantly higher in the hypothermia group than in the normothermia group in patients with a CPA-ROSC of 15 to 20 minutes (64% vs. 17%, P < 0.01), 20 to 25 minutes (70% vs. 8%, P < 0.01), 25 to 30 minutes (50% vs. 7%, P = 0.02), 35 to 40 minutes (27% vs. 0%, P = 0.03) and 40 to 45 minutes (29% vs. 2%, P = 0.049). There was a trend toward a higher rate of favorable neurological outcome in the hypothermia group than in the normothermia group in patients in whom the CPA-ROSC was 30 to 35 minutes (27% vs. 0%, P = 0.08). The rate of favorable neurological outcome was similar in patients with a CPA-ROSC less than 15 minutes.

Receiver-operating characteristic curves of the CPA-ROSC to detect favorable neurological outcome are shown in Figure 3. The areas under the receiver-operating characteristic curve were 0.79 (95% confidence interval (CI) = 0.68 to 0.86) and 0.95 (95% CI = 0.92 to 0.98) in the hypothermia and normothermia groups, respectively. The cut-off values for the CPA-ROSC of 29 minutes in hypothermia group and 15 minutes in the normothermia group had the highest combined sensitivity and specificity, with accuracy of 75.4% and 92.8% for identifying favorable neurological outcomes, respectively (Table 4). A long CPA-ROSC was associated with more accurate negative predictive values.

The propensity score-matching process selected 79 patients from the hypothermia group (hypothermia-M group) and 79 patients from the normothermia group (normothermia-M group). The mean ± standard deviation propensity score was 0.41 ± 0.22 in the hypothermia-M group and was 0.41 ± 0.22 in the normothermia-M group (P = 0.94; Figure 4). Propensity-matched patient characteristics, treatments and outcomes are presented in Table 5. Baseline characteristics were similar in the two groups except for emergency coronary angiography, primary percutaneous coronary intervention, use of intra-aortic balloon pump and use of extracorporeal circulation. In patients with a favorable neurological outcome, the CPA-ROSC was significantly longer (P < 0.001) in patients in the hypothermia-M group (median, 22 minutes; interquartile range, 16 to 31 minutes; mean ± standard deviation, 25 ± 12 minutes) than those in the normothermia-M group (median, 10 minutes; interquartile range, 5 to 13 minutes; mean ± standard deviation, 11 ± 9 minutes).

The rate of favorable neurological outcome was plotted against the CPA-ROSC every 5 minutes in the propensity-matched cohort (Figure 5). An association similar to the full cohort (Figure 2) was observed in the propensity-matched cohort, but the difference was not statistically significant and might be caused from the small sample size of the study. There was no significant difference in the rate of favorable neurological outcome between the hypothermia-M group and the normothermia-M group in the entire propensity-matched cohort (30% vs. 29%, P = 0.86). In patients whose CPA-ROSC was more than 15 minutes, however, the rate of favorable neurological outcome was higher in the hypothermia-M group than in the normothermia-M group (27% vs. 4%, P < 0.001).

In the full cohort, we tested the association between the CPA-ROSC and favorable neurological outcome in multiple models. In an unadjusted model, there was a significant association between the CPA-ROSC and favorable neurological outcome (every 1 minute: odds ratio = 0.90, 95% CI = 0.88 to 0.92, P < 0.001). After adjustment for age, gender, and initial rhythm, this association was still significant (odds ratio = 0.89, 95% CI = 0.86 to 0.92, P < 0.001). Finally, after adjustment for all of the above-mentioned variables, the association between the CPA-ROSC and favorable neurological outcome remained significant (odds ratio = 0.89, 95% CI = 0.85 to 0.92, P < 0.001).

Our study was not double-blind or randomized and had the inherent limitations of any single-centre retrospective investigation. The present study was prone to biases related to unmeasured factors. We did, however, use multivariate and propensity analyses to carefully match patients in an effort to eliminate bias. It was difficult for us to estimate the time of collapse of cardiac arrest patients correctly, so we enrolled only the witnessed cardiac arrest patients in this study. CPR guidelines were changed during the study period, which might affect the outcomes. We attempted to adjust for this effect with the additional covariate of admission after 2006 [12]. Although we found statistically significant differences between the groups in our main cohort, we found no statistically significant differences between groups in our propensity-matched cohort. This difference might be caused by a sample that was small and may have been subject to type II error, and by good results for patients whose CPA-ROSC was less than 5 minutes in the normothermia-M group with no one whose CPA-ROSC was less than 5 minutes treated with hypothermia. We thought that clinical signs, such as neurological findings, gasping and spontaneous respiration, might be associated with the propensity score [22]. Although these clinical signs should have been recorded at fixed times after ROSC for proper analysis, we were unable to retrieve such data from the patients' medical records.