Background: Lifestyle factors such as smoking, diet and physical activity play a major role in cancer incidence. However, individual cancer risk factors are not independent of each other. Factor analysis ascertains relationships between risk factors to elucidate “latent factors” that summarise risk factor patterns. Our aim was to identify latent factors and test their association with cancer incidence, specifically testing for non-additive effects of associated latent factors.
Methods: Factor analysis was applied to up to 36 self-reported baseline variables (measuring smoking, alcohol intake, diet, physical activity, reproduction, medication) for 120,771 female and, separately, 100,891 male participants of the NSW 45 and Up Study, all with no history of cancer registration. All participants were followed up until 31 December 2013 via record linkage with the NSW Cancer Registry (carried out by the Centre for Health Record Linkage) to identify incident cancers. Associations between the latent factors and cancer incidence (in 4-8 years of follow-up) were tested using logistic regression. We adjusted for socioeconomic covariates, baseline health rating, BMI, and age at recruitment. Associations were confirmed using sensitivity analysis and competing risk regression.
Results: We identified 8 latent factors each for women (largely representing smoking, alcohol, vigorous exercise, age at childbirth, hormone replacement therapy (HRT), number of children, sitting/standing, fruit and vegetables) and men (largely representing smoking, alcohol, vigorous exercise, medications, moderate exercise, sitting/standing, fruit and vegetables, meat and BMI). After correction for multiple testing, we identified 10 significant associations between some latent factors and cancer incidence for women (e.g. breast cancer and alcohol factor, OR 1.10, 95%CI 1.06-1.18, p<0.00125). For men, we identified 9 significant associations between some latent factors and cancer incidence (e.g. bowel cancer and alcohol factor, OR 1.17, 95%CI 1.08-1.26, p<0.00125). We found suggestive evidence for four interaction effects, including between alcohol and HRT latent factors for breast cancer incidence in women (OR 1.07, 95%CI 1.01-1.14, p<0.05).
Conclusions: This study elucidates the associations between lifestyle factors and cancer incidence in a large-scale, systematic approach.