Smoking prevalence as a predictor of preventable cancer burden: A mediation analysis with health systems performance in 167 nations (2008–2023).

Abstract

e22678 Background: Smoking prevalence and cancer mortality relationships vary significantly across healthcare systems. Healthcare access may mediate smoking's effects on malignancy mortality, but this pathway remains incompletely characterized. We examine smoking prevalence as a predictor of age-standardized cancer mortality while quantifying direct effects and indirect pathways through health system capacity. Methods: A 16-year prospective study of 167 nations (2008–2023) integrated: adult smoking prevalence (%) from WHO, age-standardized cancer mortality (per 100,000) from Global Burden of Disease 2023, health system efficiency index (HSEI, 0–100), and healthcare expenditure (% GDP). Mixed-effects regression with regional random intercepts modeled cancer mortality as a function of lagged smoking prevalence (t–2 years), controlling for HSEI, expenditure, and income. Mediation analysis decomposed effects into direct (smoking to mortality) and indirect (smoking to health systems to mortality) pathways using natural effects estimators. Two-way prevalence*HSEI interactions were tested with variance estimation and country-level clustering. Results: Among 2,672 country-year observations, baseline smoking prevalence (mean 24.3%, SD 11.8%) predicted cancer mortality (mean 96.4/100,000, SD 54.2). A 10-point prevalence increase independently predicted 8.7 additional cancer deaths per 100,000 (95% CI: 6.2–11.3; p < 0.001). Direct effect: 64% (5.6/100,000; 95% CI: 3.8–7.4); indirect effect: 36% (3.1/100,000; 95% CI: 1.9–4.3). HSEI significantly modified associations: systems with HSEI ≥75 (n = 48) showed 2.9/100,000 effect (95% CI: 0.4–5.4), while HSEI < 50 (n = 67) showed 11.2/100,000 (95% CI: 8.1–14.3)—a 3.9-fold difference. Regional intercepts ranged –12.4 (Asia-Pacific) to +18.6 (sub-Saharan Africa). Seventy-three percent of nations showed significant prevalence-mortality associations (p < 0.10). Model R² was 0.58 (within-country) and 0.84 (between-country). Conclusions: Smoking prevalence exerts substantial independent effects on cancer burden; approximately one-third operates through health system deficiencies. Strengthening oncology and primary prevention infrastructure may reduce smoking-attributable cancer mortality by 30–45% independent of prevalence reductions, particularly in resource-limited settings. Health system efficiency improvements offer potential for 3.8 per 100,000 annual mortality reduction despite unchanged smoking rates.