LAUSR

selected Danish patients as modeled by the CPH model and estimated using the Kaplan‐Meier method. The spread of the CPH curves within each of the IPI groups and overlap between IPI groups illustrates the limitation of the group‐based prognostic models for individual outcome assessment. Conclusions: Prognostic assessment of newly diagnosed DLBCL can be substantially improved by predicting survival probabilities from a model, based on continuous variables instead of allocating patients to predefined groups. This study also shows that, when appropriately used, simple clinical information will (likely in combination with molecular genetics) be an important resource for better individual outcome prediction in lymphoma. Figure 1A. The time‐varying AUC corresponding with the risk groups of the IPI‐like indices and a Cox proportional hazards model with the IPI variables in continuous form. Figure 1B. Survival estimates of 250 randomly selected patients in the Danish cohort predicted from the Cox proportional hazards model and estimated using the Kaplan‐Meier method for each IPI class. There is a notable spread of the modeled‐based survival estimates within each risk group and a certain overlap of estimated survival curves between risk groups.