The disease-specific survival (DSS) of 151 patients with chronic graft-versus-host disease (cGVHD) was studied in an attempt to stratify patients into risk groups and to form a basis for a new grading of cGVHD. The data included the outcome and 23 variables at the diagnosis of cGVHD and at the primary treatment failure (PTF). Eighty-nine patients (58%) failed primary therapy for cGVHD. Nonrelapse mortality was 44% after a median follow-up of 7.8 years. The probability of DSS at 10 years after diagnosis of cGVHD (DSS1) and after PTF (DSS2) was 51% (95% confidence interval [CI] = 39%, 60%) and 38% (95% CI = 28%, 49%), respectively. According to multivariate analysis, extensive skin involvement (ESI) more than 50% of body surface area; hazard ratio (HR) of 7.0 (95% CI = 3.6-13.4), thrombocytopenia (TP) (< 100 000/μL; HR, 3.6; 95% CI = 1.9-6.8), and progressive-type onset (PTO) (HR, 1.7; 95% CI = 0.9-3.0) significantly influenced DSS1. These 3 factors and Karnofsky Performance Score of less than 50% at PTF were significant predictors for DSS2. The DSS1 at 10 years for patients with prognostic factor score (PFS) at diagnosis of 0 (none), less than 2 (ESI only or TP and/or PTO), 2 to 3.5 (ESI plus either TP or PTO), and more than 3.5 (all 3 factors) was 82%, 68%, 34%, and 3% (P = .05, < .001, < .001), respectively. The DSS2 at 5 years for patients with PFS at PTF of 0, 2 or less, 2 to 3.5, and more than 3.5 were 91%, 71%, 22%, and 4% (P = .2, .005, and < .001), respectively. It was concluded that these prognostic models might be useful in grouping the patients with similar outcome.