Abstract 505P
Background
Metastatic colorectal cancer still is a lethal disease. Survival, however, is increasing due to rapidly growing treatment options, including systemic and surgical treatment. Due to that and the magnitude of prognostic factors and their unclear interactions, prediction of mortality is difficult but essential for clinical practice inside and outside clinical trials. The aim of the study is to provide a clinical model supporting prognostication at 24 and 36 months for all clinical treatment scenarios (BSC, multimodal treatment, systemic treatment). By that the majority of patients may be covered as median survival is around 30 months in the literature.
Methods
2915 Patients were treated at three different cancer centers from 2006 to 2019 and documented in monitored cancer databases. Prognostic factors were identified by a stepwise backward method and a nomogram was constructed. Performance of the model was evaluated by C-Index and cumulative dynamic time dependent AUC. Calibration of the nomogram was performed by bootstrap strategy for different risk groups.
Results
1104 patients with metastatic adenocarcinoma met inclusion criteria. Age, primary location, number of organs with metastases, lung as only site of metastases, BRAF mutation status and treatment modality were prognostic variables. A nomogram allows the prediction of survival at 24 and 36 months. Treatment modality showed to have the most prominent influence on survival, followed by BRAF mutation. Validation showed high discrimination with a cumulative dynamic time dependent AUC of 78% (C-Index 72%). Calibration by bootstrapping (k=2000) of 5 groups with differing survival probability showed a reliable accordance between predicted and observed survival at given time points.
Conclusions
We developed a validated nomogram based on a real-world population including patients with all conceivable treatment scenarios, ranging from BSC to multimodal treatment. This broad spectrum of patients and the clinical focus with a earily available and plausible prognostic set of factors may support clinical prognostication for treatment decisions and communication in daily practice.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
Holger Rumpold.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.