A systematic selection method for the development of cancer staging systems.

Area of Science:

• Oncology
• Biostatistics
• Clinical Informatics

Background:

• The tumor-node-metastasis (TNM) staging system is crucial for cancer management.
• Meaningful clinical use requires condensing TNM categories based on time-to-event outcomes.
• Selecting optimal TNM groupings is a complex statistical challenge.

Purpose of the Study:

• To develop a novel method for selecting the best groupings of TNM categories.
• To enhance the prognostic accuracy of cancer staging systems.
• To address the cutpoint selection problem for censored data with ordered covariates.

Main Methods:

• Proposed a novel bootstrap cutpoint/model selection method.
• Maximized bootstrap estimates of statistical criteria for optimal grouping.
• Utilized prognostic ability measures: explained variation, ROC curve area, and generalized Harrell's c-index.

Main Results:

• The proposed bootstrap method effectively identifies optimal TNM category groupings.
• Demonstrated improved prognostic ability compared to standard methods.
• Successfully applied the method to colorectal cancer staging.

Conclusions:

• The novel bootstrap cutpoint selection method enhances cancer staging.
• This approach provides a statistically rigorous way to condense TNM categories.
• The method has practical utility in improving cancer prognosis and treatment strategies.