Background The reliability of somatosensory evoked potential (SSEP) to predict a poor outcome of cardiac arrest patients after targeted temperature management (TTM) has been questioned due to self-fulfilling prophecy. Methods This was a multicentre, prospective, registry-based study. Data were collected from the Korean Hypothermia Network (KORHN)-pro registry between November 2015 and December 2018. We excluded cases with possible bias (inappropriate SSEP recordings and patients who decided on the withdrawal of life-sustaining therapy [WLST]) and calculated the sensitivities and false positive rates (FPRs) for an absent N20 and an absent brainstem reflex. A poor outcome was defined as a cerebral performance category score of 3–5 after 6 months. Results A total of 262 patients were analysed: 83 in the good outcome group and 179 in the poor outcome group. A bilaterally absent N20 was found in 127 patients and predicted a poor outcome with a sensitivity of 71.0% (95% confidence interval [CI], 63.7–77.5) and an FPR of 0.0% (95% CI, 0.0–4.3). Among the patients with absent brainstem reflexes (n = 103), 3 had a good outcome, with an FPR of 4.3% (95% CI, 0.9–12.2). The absence of one or both N20 and brainstem reflex had a sensitivity of 84.2% (95% CI, 77.4–89.6) and an FPR of 4.3% (95% CI, 0.9–12.2). Conclusions Our results provide further evidence that SSEP exactly predicts poor neurological outcome in these patients and suggest that caution be taken when the brainstem reflex is used as a single test to make decisions regarding WLST.