Purpose To obtain initial data on the effect of different levels of targeted temperature management (TTM) in out-of-hospital cardiac arrest (OHCA). Methods We designed a multicentre pilot trial with 1:1:1 randomization to either 32 °C ( n  = 52), 33 °C ( n  = 49) or 34 °C ( n  = 49), via endovascular cooling devices during a 24-h period in comatose survivors of witnessed OHCA and initial shockable rhythm. The primary endpoint was the percentage of subjects surviving with good neurologic outcome defined by a modified Rankin Scale (mRS) score of ≤ 3, blindly assessed at 90 days. Results At baseline, different proportions of patients who had received defibrillation administered by a bystander were assigned to groups of 32 °C (13.5%), 33 °C (34.7%) and 34 °C (28.6%; p  = 0.03). The percentage of patients with an mRS ≤ 3 at 90 days (primary endpoint) was 65.3, 65.9 and 65.9% in patients assigned to 32, 33 and 34 °C, respectively, non-significant (NS). The multivariate Cox proportional hazards model identified two variables significantly related to the primary outcome: male gender and defibrillation by a bystander. Among the 43 patients who died before 90 days, 28 died following withdrawal of life-sustaining therapy, as follows: 7/16 (43.8%), 10/13 (76.9%) and 11/14 (78.6%) of patients assigned to 32, 33 and 34 °C, respectively (trend test p  = 0.04). All levels of cooling were well tolerated. Conclusions There were no statistically significant differences in neurological outcomes among the different levels of TTM. However, future research should explore the efficacy of TTM at 32 °C. Clinical trial registration ClinicalTrials.gov unique identifier: NCT02035839 ( http://clinicaltrials.gov ).