WARNING: the interactive features of this website use CSS3, which your browser does not support. To use the full features of this website, please update your browser.
In this research we used data from the UK Biobank cohort to test how well 14 previously published risk models are able to identify people in the UK who go on to develop bowel cancer. We found that the ability of the risk models to identify those who developed bowel cancer in the future varied substantially. In men, four models (the QCancer10 model and models by Tao, Driver and Ma) were reasonably good at being able to identify which people were more likely to be diagnosed with bowel cancer. The QCancer10, Wells, Tao and Ma models were the best performing in women but all were slightly less good than in men. When we compared the risk calculated by the models to the actual risk among the people in the UK Biobank cohort, the risk calculated by all the models was higher than the actual risk. The risk models would therefore need to be adjusted if they were going to be used to tell people their risk of developing bowel cancer.
Development and validation of risk prediction models for cancer.
Cancer is caused by unregulated cell growth leading to forming malignant tumors. Cancers are known to cause by environmental and genetic factors. The suggested common environmental factors include diet, lack of exercise, smoking, radiation etc. For genetic factor, there are suggestions that some genes operate across various types of cancers for example DNA repair genes, genes involved in growth factor,in aging process (telomerase reverse transcriptase gene (TERT)) etc. Some cancers are preventable and early diagnosis can leads to better survival.This proposal aims to build one/more risk algorithms for cancers based on lifestyle/environmental factors and genetic. markers. This research fits UK Biobank?s stated purpose in that it is in the public interest. We will build optimised risk prediction models (include biomarkers, lifestyle/environmental factors collected within cohort). We will require data from the whole cohort including full genetic data(not samples) once available. Both cohort and nested case control methods will be used to construct a risk prediction model. We require data from the full cohort.