A perspective on prognostic models in chronic lymphocytic leukemia in the era of targeted agents

Abstract

Despite the increase in the number of prognostic models currently available for evaluating patients with chronic lymphocytic leukemia (CLL), their current application and utilization in clinical practice in the era of targeted agents is unclear. A critical reappraisal of recently developed prognostic models is presented in this review. The underlying CLL's genetic instability and changes in the host's health and comorbidities can all contribute to the acquisition of additional risk factors for adverse outcomes during the course of the disease. Therefore, available risk models solely based on pretreatment variables only partially predict patients' clinical outcome. A dynamic prognostic model that takes into account changes in the risk profile over time could indeed be useful in routine clinical practice. The next generation of risk assessment models should incorporate post-treatment and response biomarkers such as minimal residual disease. Finally, recent advances in the field of machine learning present novel opportunities to generate models capable of providing an individualized estimation of clinical outcomes in CLL. However, in the era of improved prognostic models, it is important to remember that these indices should supplement but not replace clinical expertise and medical decision-making.

Keywords: CLL; dynamic prognostic models; machine learning; post-treatment biomarkers; prognostic models; systematic review.