New Model for Predicting Cancer Outcomes

Predicting the long-term outcomes and risk of progression in patients with cancer is an important component of care that can drive treatment strategies. Unfortunately, current risk prediction models are lacking and focus primarily on pretreatment factors such as tumor phenotype and patient characteristics, which are not always indicative of long-term outcome. The need for improved prognostic tools is particularly high in tumors like diffuse large B-cell lymphoma (DLBCL), where the majority of patients can be cured by systemic therapy, but a subset requires more intensified treatment regimens for best outcomes. These patients are difficult to identify, and current risk stratification strategies fall short.

In a study published in Cell, a novel risk assessment tool was examined as a potential solution for current risk assessment challenges. The Continuous Individualized Risk Index (CIRI) relies on serial tumor sampling to develop a dynamic risk model that estimates the risk for an individual patient over the course of their disease. This risk assessment tool was inspired by win-probability assessments used to predict the outcome of soccer matches. The study was divided into two parts. In the first part, a naive Bayesian framework was used to retrospectively estimate the probability of a certain clinical outcome in patients with DLBCL included in 11 previously published studies. In the second part of the study, a personalized probability of survival over time model based on Cox proportional hazard modeling was used to predict the outcomes for patients in the DLBCL population as well as patients enrolled in 3 phase III clinical trials for chronic lymphocytic leukemia (CLL) and a study of neoadjuvant chemotherapy for breast cancer.

Using a combination of 6 clinical and molecular risk factors commonly used to assess patients with DLBCL, the CIRI model was able to more accurately predict event-free survival (EFS) at 24 months than any single factor alone (P < .05). In the CLL cohort, CIRI was able to accurately predict progression-free survival (PFS) more accurately than the current standards, the International Prognostic Index for Chronic Lymphocytic Leukemia (CLL-IPI) and minimal residual disease (MRD) status, at 12, 24,36, and 48 months (P < .05). Similarly, among patients with breast cancer, the CIRI model was more accurate than other risk assessment models at predicting relapse-free survival (RFS) at 12, 24, 36, 48, and 60 months (P < .05).

The investigators concluded that these results support the CIRI tool as an accurate method to predict outcomes for patients with cancer, highlighting the ability to update predictions over time as a particular strength of this approach. The tool is now available online for use, though the authors cautioned that it should only be used in coordination with other risk assessment tools as these results have not been validated by an independent study.

Read more about this on Medscape Oncology

Cell. 2019 July 4. [Epub ahead of print.]

Clinical Opinion Poll

When would you recommend dabrafenib + trametinib combo for a patient with advanced, symptomatic, PD-L1–low, lung adenocarcinoma and BRAF V600 mutation?