This study was aimed at ascertaining whether extracorporeal photopheresis (ECP) is an effective treatment for pediatric patients with steroid resistant graft versus host disease (GvHD). Fifteen patients with acute GvHD (aGvHD) and 10 patients with chronic GvHD (cGvHD) were enrolled in the study. At the start of the ECP protocol, aGvHD was staged as II (n=7), III (n=4), and IV (n=4). The response rate was 100% for aGvHD II, 75% for aGvHD III, and finally 0% for aGvHD IV (P=0.02). In multivariate analysis, the strongest predictor for ECP response was the aGvHD severity: aGvHD II 100%, aGvHD III-IV 30% [relative risk (RR) 5.071, confidence interval (CI) 95% 2.2-5.5, P=0.0016], this translates in a higher risk of transplant-related mortality for ECP nonresponders (RR 5.26, CI 95% 3.4-6.2, P=0.02). cGvHD was diagnosed as limited n=3, and extensive n=7; the response rate was 100% and 28% for limited or extensive cGvHD, respectively (P=0.03). For cGvHD the strongest predictor for ECP response was the absence of visceral organ involvement (RR 5.17, CI 95% 2-4.9, P=0.001), and the highest risk of transplant-related mortality was among patients not responding to ECP (RR 12.4, CI 95%, P=0.02). Our results suggest that ECP can rescue good-risk GvHD-patients, whereas for advanced, poor-risk GvHD patients, new therapies are required.