The impact of health behaviors on incident and recurrent cancers: a population based analysis

This thesis investigated the role of health behaviours in both incident cancer diagnosis and cancer survivors. First, we evaluated the validity of self-reported cancers in the Lifelines cohort; we found that self-reported cancer in the Lifelines cohort have a moderate to good agreement with pathologically reported cancer diagnoses. We further explored the associations of health behaviours to cancer outcomes using traditional statistical approaches and machine learning algorithms. We found that machine-learning algorithms did not outperform traditional statistical approaches for predicting incident cancer cases, nor for classifying cancer survivors by using health behaviours. It was observed that lifestyle behaviour in cancer survivors and the whole cohort needs much improvement. Moreover, interventions to improve diet quality could be based on the American Cancer Society score, as this score was the one showing better diet benefits for cancer survivors and the general population in the Lifelines cohort.