Multivariate analysis of prognostic factors in stage IV follicular low-grade lymphoma: a risk model.

Abstract

We analyzed the records of 96 previously untreated patients with stage IV follicular low-grade lymphoma (FLGL) uniformly treated with cyclophosphamide, doxorubicin, vincristine, prednisone, and bleomycin (CHOP-Bleo) chemotherapy from 1972 to 1982. The overall complete remission (CR) rate was 77%. At a median follow-up of 138 months, the 10-year cause-specific survival rate was 42% with a median survival of 100 months. Failure-free survival (FFS) was 15% at 10 years with a median FFS of 30 months. Multivariate analysis showed peripheral lymph node size (LN), degree of marrow involvement, and sex, in that order, to be important for FFS, while the number of extranodal sites (#ENS), LN, sex, and degree of marrow involvement were important for cause-specific survival. We devised a tumor burden (TB) model, incorporating #ENS, LN, and degree of marrow involvement. Three groups were identified with statistically significant differences in cause-specific survival and FFS. Those with low TB (one ENS exclusive of extensive marrow and nodal disease less than 5 cm) had a 10-year cause-specific survival of 73% compared with 24% for patients with high TB (greater than or equal to two ENS and nodal disease greater than or equal to 5 cm) (P less than .001) and 40% for those with intermediate TB (either greater than or equal to 2 ENS, or extensive marrow only, or nodal disease greater than 5 cm) (P = .050). Patients with low TB had a 10-year FFS rate of 32%, while the intermediate and high TB groups had 10% and 9% FFS, respectively (P = .003). Because sex was a very strong prognostic variable, we created a risk model for survival and FFS based on TB and sex. Females with low TB had the best prognosis (92% survival and 50% FFS at 10 years) and males with high TB had the worst outlook (median survival and FFS, 43 and 12 months, respectively). Other TB-sex combinations defined two groups with statistically significant differences in survival but comparable FFS. This model should aid in the design and analysis of future trials.