Statistical Methods in Cancer Epidemiology Research in Low- and Middle-Income Countries: A Systematic Review

Statistical methods in cancer epidemiology research

Abstract

Cancer epidemiology research in LMICs has grown with the increasing global cancer burden, but the statistical methods used remain diverse and insufficiently evaluated. This review assessed methodological trends, dominant analytical approaches, and the use of advanced statistical methods. A PRISMA 2020–guided systematic review identified 46 eligible studies on cancer epidemiology in LMICs. Data on study characteristics, cancer types, data sources, and statistical methods were extracted and synthesized descriptively. The included studies covered sub-Saharan Africa, Asia, and Latin America, with cervical and breast cancers being the most frequently investigated malignancies. Most included studies employed observational designs, particularly cross‑sectional and cohort approaches, relying mainly on hospital-based datasets, cancer registries, and population surveys. Conventional statistical methods predominated, especially descriptive statistics and regression-based analyses, with logistic regression being the most applied inferential method. Survival analysis techniques, including Kaplan–Meier estimation and Cox proportional hazards models, were increasingly used, whereas multilevel modeling, simulation approaches, and machine learning (ML) methods remained comparatively limited. Across studies, consistent patterns included late-stage diagnosis, low screening uptake, poor survival outcomes, and marked socioeconomic and health-system inequalities influencing access to care and treatment outcomes. Cancer epidemiology research in LMICs still relies mainly on traditional methods, with limited use of advanced analytical approaches. Strengthening data systems, research infrastructure, and analytical capacity is crucial for improving evidence generation and cancer control strategies.