Abstract / Summary
Abstract Purpose Kidney cancer incidence continues to rise globally. Population-wide screening is not feasible due to low disease prevalence and risk of overdiagnosis, highlighting the need for targeted, risk-stratified approaches. We aimed to develop and internally validate a lifestyle-based model to predict 10-year risk of kidney cancer. Materials and Methods Data were obtained from the Sax Institute’s 45 and Up Study, a prospective cohort of adults aged ≥ 45 years recruited between 2006 and 2009 with linkage to the NSW Cancer Registry through to December 2019. Incident kidney cancer (ICD-10 C64) was defined as the first diagnosis after study enrolment. Candidate predictors included demographic, lifestyle, and clinical variables. A multivariable logistic regression model was developed using a 70% training set and evaluated in a 30% validation set. Model performance was assessed using discrimination (C-statistic), calibration, and decision curve analysis. Results Among 265,712 participants (mean follow-up 10.9 years), 1,010 incident kidney cancer cases were identified. Increasing age, male sex, obesity, hypertension, and diabetes were associated with higher risk, while vigorous physical activity was associated with lower risk. The model demonstrated moderate discrimination (C-statistic 0.68) and good calibration. At a predicted 10-year risk threshold of ≥ 1.0%, high-risk subgroups comprised approximately 4–5% of the population, with a 2.2- to 3.8-fold higher incidence. An eight-variable model incorporating smoking exposure demonstrated the strongest decision-analytic performance. Conclusions This population-based model using routinely available variables enables risk stratification in a low-incidence setting. Following external validation, it may support targeted prevention and risk-stratified screening strategies.