The Million Women Study is a large UK cohort study of women’s health. We have detailed information on disease risk factors (eg smoking, weight, physical activity) from postal questionnaires and on deaths, cancers and hospital admissions through linkage to NHS databases. However we have limited information on the use of prescribed medication and on diseases managed wholly or largely in outpatient and primary care settings. General practice records are the most comprehensive source of such information.
To assess the reliability of the information that we already hold in the Million Women Study, we wish to compare our information with that held by CPRD on diagnoses of conditions and drug prescribing. Since our original application in 2012 we have conducted this comparison work using a linked dataset and have compared information on prescribing of various medications and on diagnoses of various chronic conditions.
We are currently focusing our work on risk factors for cancer and the role that chronic diseases, and their associated treatments, play in the development of cancer. We are therefore applying to CPRD to continue the comparison work on the same linked dataset with the addition of further comparisons of information on diagnoses and treatments for common chronic conditions. The information gained will enhance our current investigations into the role of chronic disease in the development of cancer.
The Million Women Study (MWS) is a prospective cohort study of women aged 50-64 years at the time of recruitment in 1996-2001. 1.2 million participants were recruited through the NHS breast screening programme in England and have since been followed through electronic linkage to NHS central records. Every 3-5 years, participants have also been sent postal questionnaires.
The aim of this study is to assess the reliability of the information that is held in the MWS from postal questionnaires and NHS hospital admission records (Hospital Episode Statistics) on diagnoses and treatments by comparing with information on diagnoses and treatments in CPRD on the same women. Since our original application in 2012, we have conducted this comparison work using a linked dataset (12-070R3).
We will continue this work using the same linked dataset with a focus on three domains: cardiovascular disease and associated risk factors, neurological and mental health conditions, and musculoskeletal conditions. We will also compare the full prescription history, laboratory requests and results, and treatments for each patient with one or more of the conditions of interest with their MWS record.
The study population of interest is MWS participants with a record of drug prescription or diagnosis in either the MWS or CPRD and matched controls with no such record in either the MWS or CPRD.
We will conduct comparisons using descriptive statistics and, where appropriate, statistical measures of agreement (e.g. positive predictive values (PPV), negative predictive values (NPV), PPV to NPV rate ratios, kappa coefficients, spearman correlations) with measures of uncertainty (e.g. confidence intervals).
These comparisons will validate and enrich the information held in the MWS dataset and reduce ascertainment bias, and are therefore extremely important for our current investigations into the role of chronic disease in the development of cancer.
We will make comparisons between CPRD records and Million Women Study records in three domains, concentrating on non-cancer outcomes: (i) cardiovascular diseases including interim conditions and risk factors, (ii) neurological and mental health conditions, and (iii) musculoskeletal conditions. To understand the possible effect of multi-morbidity and polypharmacy, we will also compare the full prescription history, laboratory requests and results, and treatments for each patient with one or more of the conditions of interest with the Million Women Study record.
Record comparisons will be made for the following conditions (record comparisons for the starred conditions have already been approved):
(i) Cardiovascular diseases including interim conditions and risk factors
hypertension, stroke*, conditions affecting antiplatelet use (stomach or duodenal ulcer), smoking-related conditions (chronic obstructive pulmonary disease and asthma), aortic stenosis*, atrial fibrillation*, diabetes with or without glaucoma and other complications*, thyroid disorders, high blood cholesterol, peripheral vascular disease, heart failure, chronic kidney disease, gallstones/ gall bladder problems, diverticular disease of intestine, chronic liver disease, systemic and organ-specific auto-immune diseases (including rheumatoid arthritis, type 1 diabetes, psoriasis/psoriatic arthritis, multiple sclerosis, systemic lupus erythematosus, inflammatory bowel disease, Addison’s disease, Graves’ disease, Sjögren’s syndrome, Hashimoto’s thyroiditis, myasthenia gravis, autoimmune vasculitis, pernicious anaemia, coeliac disease)
(ii) Neurological and mental health conditions
dementia*, Parkinson’s disease*, depression, anxiety, epilepsy, schizophrenia, anorexia/bulimia
(iii) Musculoskeletal conditions
carpal tunnel syndrome*, bone fractures*, osteoarthritis, osteoporosis
Gillian Reeves - Chief Investigator - University of Oxford
Sarah Floud - Corresponding Applicant - University of Oxford
Anna Brown - Collaborator - University of Oxford
Carol Hermon - Collaborator - University of Oxford
Keith Shaw - Collaborator - University of Oxford
TienYu Owen Yang - Collaborator - University of Oxford