Interactive Use of Cox Model-Predicted Survival Curves: An Application Using ACS10 Score and Age to Personalize Treatment of Pediatric AML.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
PURPOSE: Analyses in oncology frequently include Cox proportional hazards models for survival outcomes, but reported results typically only include hazard ratios and respective inference. Cox model predictions of survival functions may help to provide a greater clinical meaning from fitted Cox models for precision oncology research. PATIENTS AND METHODS: We created a publicly available software library, shinyCox, that can generate user-friendly interactive applications to visualize survival outcome predictions of fitted Cox models. To illustrate the benefits of this software, we analyzed data from AML02 and AML08, randomized clinical trials for pediatric patients with AML. Building on a recent article, we assessed how baseline factors and a pharmacogenomics score (ACS10) can affect the predicted overall survival (OS) and event-free survival (EFS) between patients assigned to induction regimens of clofarabine plus cytarabine or daunorubicin and etoposide combined with low-dose cytarabine or high-dose cytarabine. RESULTS: Our model outcome prediction visualization application highlights previously reported associations of ACS10 with EFS and OS, while also providing a better understanding of how other prognostic factors amplify or mitigate prognostic implications of ACS10. It is informative to better understand the practical clinical outcomes in terms of predicted survival probabilities to complement the insights gained from hazard ratio tables. CONCLUSION: Using shinyCox, we generated visualization applications that let us identify complex relationships between the ACS10 score, age, and the predicted OS and EFS probabilities after AML therapy. This article shows how shinyCox will facilitate model interpretation and accelerate the development of personalized therapies in oncology.