A Prediction Model For The Absolute Risk Of Death From Cancer And Other Causes

David MULLER, Imperial College London, United Kingdom
MURPHY N. 1 , LASSALE C. 1 , JOHANSSON M. 2 , GUNTER M. 3,1 , RIBOLI E. 1 , BRENNAN P. 2 , TZOULAKI I. 1

1 School of Public Health, Imperial College London, United Kingdom
2 Genetic Epidemiology Group, International Agency for Research on Cancer, Lyon, France
3 Nutritional Epidemiology Group, International Agency for Research on Cancer, Lyon, France

Background: There are numerous modifiable risk factors for premature death, with tobacco smoking, overweight and obesity, high alcohol consumption, poor diet, and physical inactivity being among the most important. Whilst the importance of these factors is widely known in the general public, the joint magnitude of their influence on the absolute risk of death is less well appreciated. We sought to build a model to calculate the absolute risk of death as a function of these modifiable factors.

Methods: We analysed data from the European Prospective Investigation into Cancer and Nutrition study, which is a prospective cohort of over 500,000 European adults. We restricted our analysis to the approximately 330,000 participants with complete covariate data follow-up information. Absolute risk of death was estimated using flexible parametric survival models including sex, smoking status, BMI, a composite indicator of quality of diet, alcohol intake, and physical activity.

Results: Among the available cohort of 330,000 participants, there were approximately 25,000 deaths that occurred during follow-up. A model including all covariates could discriminate well between those individuals at high and low risk of death within 5 years (c-statistic 0.769; 95% confidence interval [0.763, 0.775]). This represented a modest improvement over a model including only age and sex (c-statistic 0.747; 95% confidence interval [0.741, 0.7553]). We will present risk profiles according to various patterns of risk factors, and show how these factors influence the absolute risk of death from cancer and other causes for individuals of various ages.

Summary: Whilst age and sex alone can discriminate well between those at high and low risk of death, the absolute risk of death varies strongly with several modifiable risk factors, most notably tobacco smoking. This model could be used as an aid to public heath communication, or to motivate the maintenance of healthy habits.