Prognostic impact of sarcopenia in patients with diffuse large B‐cell lymphoma treated with rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone

Abstract Background Sarcopenia is known to be related to an increased risk of chemotherapy toxicity and to a poor prognosis in patients with malignancy. We assessed the prognostic role of sarcopenia in patients with diffuse large B‐cell lymphoma (DLBCL). Methods In total, 187 consecutive patients with DLBCL treated with induction rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R‐CHOP) immunochemotherapy were reviewed. Sarcopenia was defined as the lowest sex‐specific quartile of the skeletal muscle index, calculated by dividing the pectoralis muscle area by the height. Clinical outcomes were compared between the sarcopenic and non‐sarcopenic groups. A nomogram was constructed from the Cox regression model for overall survival (OS). Results Treatment‐related mortality (21.7 vs. 5.0%, P = 0.002) and early discontinuation of treatment (32.6 vs. 14.9%, P = 0.008) were more common in the sarcopenic group than in the non‐sarcopenic group. The 5 year progression‐free survival (PFS) rates were 35.3% in the sarcopenic group and 65.8% in the non‐sarcopenic group ( P < 0.001). The 5 year OS rates were 37.3% in the sarcopenic group and 68.1% in the non‐sarcopenic group ( P < 0.001). Sarcopenia and the five variables of the International Prognostic Index (IPI) were independent prognostic factors in a multivariate analysis for PFS and OS and were used to construct the nomogram. The calibration plot showed good agreement between the nomogram predictions and actual observations. The c index of the nomogram (0.80) was higher than those of other prognostic indices (IPI, 0.77, P = 0.009; revised‐IPI, 0.74, P < 0.001; National Comprehensive Cancer Network‐IPI, 0.77, P = 0.062). Conclusions Sarcopenia is associated with intolerance to standard R‐CHOP chemotherapy as well as a poor prognosis. Moreover, sarcopenia itself can be included in prognostic models in DLBCL.