Differentiated effect of hypothermia versus normothermia after cardiac arrest by neurological status and clinical correlates of ischemic burden

INTRODUCTION: The optimal temperature target after out-of-hospital cardiac arrest (OHCA) remains uncertain, as major randomized trials have not demonstrated a clear benefit of lower-target hypothermia compared with normothermia. Despite this, patients with differing degrees of ischemic burden or early neurological injury may hypothetically respond differently to temperature targets, raising the possibility of heterogeneous treatment effects. This study aimed to determine whether outcomes differ after OHCA with 33 °C versus controlled normothermia (36.0-37.7 °C) according to correlates of ischemic burden and neurological findings at hospital admission. METHODS: This secondary analysis of the TTM and TTM2 trials employed effect-based machine learning to probe for heterogeneous treatment effects. Causal forests and Bayesian additive regression trees (BART) estimated individualized treatment effects (ITE) based on ischemia duration, prehospital factors, and neurological examination findings at hospital admission. Treatment effect modifier variable importance parameter (TEM-VIP) analysis ranked the baseline variables most likely to modify effect. The primary outcome was six-month all-cause mortality, and the secondary outcome was poor functional outcome (modified Rankin scale score 4-6). Primary analyses evaluated potential heterogeneity within each trial, while secondary analyses pooled all participants. MEASUREMENTS AND MAIN RESULTS: All 2789 adult OHCA patients from the trials' primary analyses were included (median age 65 years (IQR 56-73), 79.9% male). Death occurred in 49.4% of patients in the 33 °C group, and in 47.9% in the controlled normothermia group. Variation in six-month mortality ITEs estimated from ischemic burden correlates and initial neurological findings was minimal, and no significant heterogeneity of treatment effect was identified across all methods and analysis populations. In the pooled analysis, estimated conditional mean mortality risk difference ranged from - 1.2% to 2.1% in causal forest (p = 0.99), and from - 0.5% to 1.5% in BART, with posterior probabilities for hypothermia benefit ranging from 42% to 53%. Absolute subgroup estimates in TEM-VIP analysis were small and non-significant. Results were consistent for the six-month functional outcome. CONCLUSIONS: Among TTM and TTM2 participants randomized to 33 °C versus controlled normothermia post-OHCA, no evidence of detectable heterogeneity of treatment effect based on available clinical correlates of ischemic burden or initial neurological findings was identified, with minimal estimated variation in treatment effect.