University of Manchester

Academic Institute

Sarah Kitson - Chief Investigator - University of ManchesterSarah Kitson - Corresponding Applicant - University of ManchesterArtitaya Lophatananon - Collaborator - University of ManchesterDarren Ashcroft - Collaborator - University of ManchesterEmma Crosbie - Collaborator - University of ManchesterEvangelos Kontopantelis - Collaborator - University of ManchesterGlen Martin - Collaborator - University of ManchesterHelena O'Flynn - Collaborator - University of Manchester

CPRD880

Womb (endometrial) cancer is the fourth most common female cancer. The number of cases is rising quickly, probably because of increasing levels of obesity, with which it is strongly associated. Being diagnosed with womb cancer is not only upsetting, but treatment can be unpleasant and even dangerous for some. When found early, womb cancer can be cured with a hysterectomy (surgical removal of the womb) but when it has spread, the outlook is very poor.

This project aims to develop an accurate endometrial cancer risk prediction model to be used in primary care and to identify potential prevention strategies for investigation in future clinical trials. We aim to use the CPRD to externally validate our flexible parametric survival model that we will develop separately, which utilises age, body mass, reproductive and family history and measures of insulin resistance to quantify an individual woman’s endometrial cancer risk. The CPRD has been specifically chosen for model validation as it contains data on a patient cohort representative of the general population. The dataset will be restricted to female patients aged 45-60 years registered with either a CPRD Gold or Aurum practice between 1/1/2000 and 31/12/2016 and eligible for linkage to HES APC, NCRAS and ONS mortality records. HES data will be used to determine diagnoses and operations and combined with cancer registry data to ascertain endometrial cancer cases. Validation of the model will be performed by applying the prediction model (exactly as derived) to women with an intact uterus, with calibration quantified through flexible calibration plots (plotting the observed number of events against predicted risks for groups across the risk prediction spectrum) and estimation of calibration intercept and slope. The ability of the model to discriminate at baseline between individuals who do and do not go on to develop endometrial cancer will be assessed using the concordance (c) statistic. Should the model perform poorly at external validation, the coefficients will be updated using data from new individuals within the CPRD.

15/01/2021

Release