The National Institutes of Health (NIH) classification of graft-versus-host disease (GVHD) is a significant improvement over prior classifications, and has prognostic implications. We hypothesized that the NIH classification of GVHD would predict the survival of patients with GVHD treated with extracorporeal photopheresis (ECP). Sixty-four patients with steroid refractory/dependent GVHD treated with ECP were studied. The 3-year overall survival (OS) was 36% (95% confidence interval [CI] 13-59). Progressive GVHD was seen in 39% of patients with any acute GVHD (aGVHD) (classic acute, recurrent acute, overlap) compared to 3% of patients with classic chronic GVHD (cGVHD) (P=.002). OS was superior for patients with classic cGVHD (median survival, not reached) compared to overlap GVHD (median survival, 395 days, 95% CI 101 to not reached) and aGVHD (delayed, recurrent or persistent) (median survival, 72 days, 95% CI 39-152). In univariate analyses, significant predictors of survival after ECP included GVHD subtype, bilirubin, platelet count, and steroid dose. In multivariate analyses overlap plus classic cGVHD was an independent prognostic feature predictive of superior survival (hazard ratio [HR] 0.34, 95% CI 0.14-0.8, p=.014). This study suggests that NIH classification can predict outcome after ECP for steroid refractory/dependent GVHD.