A Flexible Method for Identifying Spatial Clusters of Breast Cancer Using Individual-Level Data

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis

Background:

• Cancer risk exhibits geographic variations, but traditional methods struggle with multiple clusters and risk factor adjustments.
• Identifying localized cancer patterns is crucial for targeted public health interventions.

Purpose of the Study:

• To develop and apply a novel statistical method for detecting multiple geographic clusters of breast cancer.
• To analyze spatial patterns of breast cancer odds in Wisconsin using a large case-control study.
• To assess the impact of geographic clustering on breast cancer risk after adjusting for known factors.

Main Methods:

• Developed a two-step penalized regression approach using the least absolute shrinkage and selection operator (Lasso) to identify spatial clusters.
• Utilized a grid-based system with overlapping circles to define potential geographic clusters.
• Incorporated cluster identification into a participant-level logistic regression model, selecting the number of clusters using BIC.
• Analyzed data from the Wisconsin Women's Health Study (16,076 cases, 16,795 controls, aged 20-79, 1988-2004).

Main Results:

• Identified 15 distinct geographic clusters, defining 23 areas with unique odds ratios for breast cancer.
• After adjusting for known risk factors, confidence intervals for odds ratios narrowed, but the ratios themselves did not change significantly.
• One additional breast cancer hotspot was identified through the refined spatial analysis.
• The method successfully discerned spatial gradients in breast cancer odds across Wisconsin.

Conclusions:

• The developed method effectively identifies multiple, overlapping spatial clusters of breast cancer.
• Spatial variations in breast cancer odds persist even after accounting for individual-level risk factors.
• This approach provides a more nuanced understanding of geographic disparities in breast cancer risk.