A Power-Law Framework for Characterizing Global Cancer Burden: Incidence–Mortality Scaling, Severity Ranking, and Cumulative Risk Patterns Using GLOBOCAN 2020
Senyefia Bosson-Amedenu *
Department of Mathematics, Statistics and Actuarial Science, Takoradi Technical University, Sekondi–Takoradi, Ghana.
Eric Justice Eduboah
Department of Economics Education, University of Education, Winneba, Ghana.
Noureddine Ouerfelli
Institut Supérieur des Technologies Médicales de Tunis, LR13SE07, Laboratoire de Biophysique et Technologies Médicales, University of Tunis El Manar, Tunis, Tunisia.
*Author to whom correspondence should be addressed.
Abstract
Background: Global cancer burden is unevenly distributed across cancer sites, and several malignancies show mortality levels that are disproportionate to incidence. Quantifying the scaling relationship between incidence and mortality may clarify comparative severity, identify outlying cancer sites, and support population-level prioritisation.
Methods: This study used GLOBOCAN 2020 data for 36 major cancer sites worldwide. The association between incidence (Nc) and mortality (Nd) was assessed using linear, log-log power-law, and quadratic models. Model performance was evaluated using R², adjusted R², RMSE, MAE, AIC, BIC, residual diagnostics, ten-fold cross-validation, outlier exclusion, and subset restriction to the top 20 cancers. A rank-based severity index (σ) was examined against fatality ratio, age-standardised rates, and cumulative death risk.
Results: The log-log power-law model provided the strongest overall representation of the incidence-mortality relationship across all cancer sites (R² = 0.853; adjusted R² = 0.848), with substantially lower AIC (63.67) and BIC (68.33) than the competing models. The estimated scaling exponent was α = 1.031 (95% CI: 0.879-1.182), indicating near-proportional mortality scaling with incidence. Cross-validation supported predictive stability, and the Breusch-Pagan test indicated reduced heteroscedasticity after logarithmic transformation (p = 0.452). The proposed severity index showed positive associations with fatality ratio, ASMR/ASRI, ASMR, and cumulative death risk. Prediction-interval analysis identified lung, liver, stomach, oesophageal, and pancreatic cancers as having higher mortality than expected from incidence alone, whereas breast, thyroid, prostate, testicular, and melanoma cancers showed lower-than-expected mortality.
Conclusions: The findings support a power-law framework for describing global incidence-mortality scaling and suggest that rank-based severity may provide a complementary comparative indicator of cancer burden.
Keywords: GLOBOCAN 2020, cancer incidence, cancer mortality, incidence-mortality scaling, power-law model, rank-based severity, cumulative risk, age-standardised rates, fatality ratio, global cancer burden