We assessed the retrospective applicability and prognostic value of the National Institutes of Health (NIH) classification of chronic graft versus host disease (cGVHD) in 159 consecutive patients after allogeneic hematopoietic stem cell transplant (HSCT). Seventy-four patients (46.5%) were affected by late-acute GVHD (n = 19; 25.7%), classic cGVHD (n = 44; 59.4%) and overlap syndrome (n = 11; 14.9%). Overall, patients with NIH-defined cGVHD (i.e. classic cGVHD and overlap syndrome) had better 10-year overall survival (OS) as compared to patients without GVHD (76.9% vs. 47.4%, p = 0.0002) or with late-acute GVHD (47.4%, p = 0.001). Relapse mortality (RM) was lower in patients with NIH-defined cGVHD than in patients without GVHD (14.5% vs. 38.7%, p = 0.001), but comparable to that of late-acute type (19.4%, p = 0.31). Non-relapse mortality (NRM) was lower in patients with NIH-defined cGVHD as compared to late-acute GVHD (10.0% vs. 41.1%, p = 0.0005), as well as patients without GVHD (22.2%, p = 0.045). At multivariate analysis, NIH-defined cGVHD remained independently predictive for lower RM, but not for NRM. Thus, the new NIH classification identifies two subtypes of GVHD (late-acute and chronic) with different long-term outcomes and impact on RM and NRM.