Ovarian cancer is the 6th most common cause of death from cancer in UK women. Most ovarian cancers are not picked up until the disease is advanced when it is harder to treat. Earlier diagnosis may improve outcomes including survival but this remains challenging. The majority of women with ovarian cancer are diagnosed after visiting their general practitioner (GP) with possible symptoms of the disease. However, these symptoms are often non-specific and it can be difficult to determine which women have ovarian cancer and should be referred urgently and which women can be reassured. A blood test - Cancer Antigen 125 (CA125) - is currently used by GPs to investigate women with symptoms of possible ovarian cancer. We have developed models (Ovatools) which predict the probability of ovarian cancer and all cancers based on a woman’s CA125 level and age, in those being tested in General Practice. These could be used to help make individual informed decisions about the need for further investigation and could help inform guidelines on cancer detection. Before using these models in clinical practice, it is important to ensure their reliability. In this study, we will use anonymised primary care (CPRD) and linked cancer registry (NCRAS) hospital (HES), imaging (DID) deprivation (Townsend) and mortality (ONS) data from women who have had a CA125 test in English general practice to evaluate the Ovatools models and determine how well they perform. We will also consider the health economic implications of implementing these models within the diagnostic pathway.
The majority of women with ovarian cancer are diagnosed at an advanced stage which contributes to poor outcomes. Earlier diagnosis may improve outcomes including survival, but this remains challenging. Most women with the ovarian cancer are diagnosed after visiting their GP with possible symptoms of the disease. However, symptoms are generally non-specific, so it is difficult to determine which patients should be referred urgently and which can be reassured. Cancer-Antigen-125 is recommended by the National Institute for Health and Care Excellence (NICE) as the first line test for ovarian cancer in primary care, with a threshold of ≥35U/ml considered abnormal. However, the risk of ovarian cancer varies markedly based on a woman’s CA125 level and age. We recently developed models (Ovatools) to estimate the probability of ovarian cancer and any cancer diagnosis within 12-months of a primary care CA125 test based on these variables. These models could be used to inform individual decisions on the need for further investigation and to select high-risk women for expedited referral.
In this study, we will use anonymised primary care (AURUM), cancer registry (NCRAS), hospital (HES), imaging (DID), deprivation (Townsend) and mortality (ONS) data for women who have had a CA125 test in English GP between 2011-2017 to externally validate the Ovatools models. We will calculate measures of model calibration (slope and intercept) and discrimination (area under the curve). Accuracy at a range of model cut-offs, including one equating to the NICE advocated 3% risk-threshold for urgent cancer investigation, will be determined. We will develop and perform model-based health economic analyses to assess the value (relative additional cost per quality-adjusted life-year gained) of different Ovatools ‘action thresholds’ in addition to the recommended NICE CA125 cut-off. This work will help determine how the Ovatools models might be used to best effect within the diagnostic pathway.
The primary clinical outcome is a diagnosis of ovarian cancer, as recorded in NCRAS data, within the 12 months following a primary care CA125 test. A sub-analysis will use diagnosis of an invasive ovarian cancer as its outcome. The secondary clinical outcomes is diagnosis of any cancer (excluding non-melanoma skin cancer) within 12 months as recorded in NCRAS data.
Further outcomes of interest to inform the health economic analysis will include cancer stage at diagnosis (for both ovarian and other cancers); benign mass; and diagnostic tests (including CT) and treatments for these; cancer diagnosis at any time, cancer recurrences and mortality.
Fiona Walter - Chief Investigator - Queen Mary University of London
Garth Funston - Corresponding Applicant - University of Cambridge
Borislava Mihaylova - Collaborator - Barts and the London Queen Mary's School of Medicine and Dentistry
Emma Atakpa - Collaborator - Queen Mary University of London
Joseph Hart - Collaborator - Queen Mary University of London
Kirsten Arendse - Collaborator - Queen Mary University of London
Runguo Wu - Collaborator - Barts and the London Queen Mary's School of Medicine and Dentistry
Stephen Duffy - Collaborator - Queen Mary University of London
Tyler Saunders - Collaborator - Queen Mary University of London
HES Admitted Patient Care;HES Diagnostic Imaging Dataset;NCRAS Cancer Registration Data;No additional NCRAS data required;ONS Death Registration Data;Patient Level Townsend Index