Classification and regression tree analysis of 1000 consecutive patients with unknown primary carcinoma

Area of Science:

• Oncology
• Biostatistics
• Clinical Research

Background:

• Clinical features and survival times for unknown primary carcinoma (UPC) are highly variable.
• Effective prognostic factor identification is crucial for managing UPC patients.

Purpose of the Study:

• To apply a novel analytical method, classification and regression tree (CART) analysis, to a cohort of UPC patients.
• To identify novel prognostic factors, explore clinical variable interactions, and illustrate their impact on survival.
• To segregate UPC patients into homogeneous subgroups for enhanced clinical trial stratification.

Main Methods:

• Multivariate survival analyses were performed using classification and regression tree (CART) analysis on 1000 UPC patients.
• CART analysis was utilized to identify prognostic variables and create distinct patient subgroups based on clinical features and survival.
• Data were collected from patients referred to the University of Texas M. D. Anderson Cancer Center between 1987 and 1994.

Main Results:

• The median survival for all 1000 UPC patients was 11 months.
• CART analysis identified 10 terminal subgroups with median survival ranging from 40 months to 5 months.
• Key prognostic variables included liver involvement, metastatic organ sites, histology, and age, which effectively segregated patients into groups with similar clinical features and survival.

Conclusions:

• Classification and regression tree (CART) analysis is a valuable method for dissecting complex clinical situations in unknown primary carcinoma (UPC).
• CART effectively identified previously unappreciated patient subsets and stratified patients into homogeneous groups.
• This approach facilitates the identification of suitable patient populations for future clinical trials and improves prognostic accuracy.