Assessing the Cox model assumption as a statistical tool for classifying lymphomas.

ABSTRACT

INTRODUCTION: The Cox model is widely used in medical research for comparing survival. Lymphomas might exhibit important differences in long-term cure rate despite a similar survival. METHODS: Using log-rank test, we compared event-free survival (EFS), and the survival of 64 patients with mantle cell lymphoma (MCL), 525 patients with follicular lymphoma, and 1136 patients with diffuse centroblastic lymphoma (CB). RESULTS: Although EFS and survival of MCL were significantly shorter than those of follicular lymphoma, checking the validity of the proportional hazards assumption shows that the distribution of rates of events and deaths over time did not differ in MCL and follicular lymphoma. In contrast, the ratios of hazards (events and deaths rates) did not remain constant over time in MCL and CB, because of a decrease in late events and deaths rates in the latter histological type. CONCLUSION: Checking the validity of the Cox model hypothesis might be a useful tool for assessing long-term cure rate in seldom lymphoma subtypes. Despite a short overall survival, MCL should not be considered to be an aggressive lymphoma, in which available chemotherapy may cure a subset of patients.