Background: Several studies have compared cooling methods in therapeutic hypothermia (TH) after return of spontaneous circulation (ROSC). However, it is limited because of small and uncontrolled data, and could not demonstrate a difference between surface cooling and endovascular cooling. Methods: We performed a multicenter retrospective registry-based study of adult out of cardiac arrest (OHCA) patients treated with TH presenting to 24 hospitals across South Korea during the period from January 2007 to December 2012. Patients who received TH only by surface cooling methods such as hydrogel pads, body wraps, and other mattress were assigned to surface cooling group, and those who received TH by endovascular cooling catheter were assigned to another. We compared neurological outcomes (CPC at discharge), efficacy and adverse event of each method in the overall cohort and in separate subgroups according to cooling methods. To adjust for differences in baseline characteristics of each group, we had performed one-to-one matching by the propensity. Finally, in the propensity score matched cohort, neurological outcomes, efficacy and adverse event were evaluated. Results: Of a total of 930 OHCA patients entered in the registry, 803 were included in the analysis, of which 559 underwent only surface cooling and remaining 224 underwent endovascular cooling. The baseline characteristics of the study patients according to the cooling method are shown in Table 1. Patients undergoing endovascular cooling were significantly more witnessed and had a higher bystander CPR rate, and shockable rhythm. More patients in the endovascular cooling group than in the surface cooling group had spontaneous breathing and motor response immediately after ROSC. Endovascular cooling were mainly undergone in high TH volume hospital. More adverse events such as overcooling or rebound hypothermia occurred in the surface cooling group, which was significantly associated with poor neurological outcome (cerebral performance category 3-5) at hospital discharge. After propensity-score matching was performed for the entire population, there were 180 matched pairs of patients (Table 2). In the matched cohorts, there was no longer any significant difference between each group for any covariate, according to the use of statistical methods appropriate for matched data. In outcome analysis, we found no significant difference in rates of good neurological outcome between the surface cooling and endovascular cooling groups. The odd ratios of surface cooling for overcooling in cooling phase, hyperthermia, hypoglycemia, and hypotension in rewarming phase were 2.564 (95% CI=1.317 to 4.990, p=0.006), 2.486 (95% CI=1.145 to 5.399, p=0.021), 2.023 (95% CI=1.038 to 3.942), and 1.677 (95% CI=1.036 to 2.714, p=0.035), respectively. Conclusion: In the overall matched cohort, we found no significant difference in rates of good neurological outcome between the surface cooling and endovascular cooling groups. However, in the adverse event of cooling methods, the rates of the overcooling in cooling phase and hyperthermia, hypoglycemia and hypotension in rewarming period were significantly higher in the group that was treated with surface cooling than in the group that was treated with endovascular cooling.