Binary genetic algorithm for optimal joinpoint detection: Application to cancer trend analysis

Area of Science:

• Biostatistics
• Computational Statistics
• Epidemiology

Background:

• Joinpoint regression models (JRM) identify trend changes using joinpoints (changepoints).
• Existing grid search (GS)-based methods for joinpoint detection are computationally intensive, limiting the number of detectable joinpoints.
• Accurate detection of multiple joinpoints is crucial for analyzing complex trends.

Purpose of the Study:

• To develop a computationally efficient genetic algorithm-based joinpoint (GAJP) model for optimal joinpoint detection.
• To overcome the limitations of existing grid search methods in terms of computational demand and the number of detectable joinpoints.
• To evaluate the accuracy and performance of the GAJP model compared to GS-based methods.

Main Methods:

• Developed a GAJP model embedding a binary genetic algorithm into the JRM.
• Represented joinpoint combinations as binary chromosomes and used genetic operations to minimize fitness functions (Bayesian information criterion - BIC, BIC3).
• Evaluated model performance through intensive simulations and comparison with GS-based methods using BIC, BIC3, and permutation tests.

Main Results:

• The GAJP model demonstrated outstanding computational efficiency in detecting multiple joinpoints.
• The proposed method is practically feasible for detecting multiple joinpoints without preassigning their number.
• The GAJP model showed suitability for analyzing large cancer incidence datasets, as demonstrated with colorectal cancer data.

Conclusions:

• The GAJP model offers a computationally efficient and feasible approach for detecting multiple joinpoints.
• This method is easily extendable to analyze trends in large datasets, such as cancer incidence data.
• The GAJP model provides a valuable tool for epidemiological trend analysis and public health research.