Coding of covariates
Socio-economic status in THIN is coded by the patient postcode-based Townsend Deprivation Index which is constructed from 4 census variables: households without a car, overcrowded households, households not owner-occupied and persons unemployed. It is scaled from 1 to 5, where the first quintile represents the least deprived and the fifth represents the most deprived group.31 In the final analysis, patients within quintiles 1 and 2 were re-coded as low, 3 as medium, 4 and 5 as high level of deprivation. BMI was categorized as healthy weight and overweight compared to obese. To classify hypertension, measurements of systolic and diastolic blood pressure (SBP≥140 mmHg and DBP≥90 mmHg) were used in conjunction with Read codes, as previous research showed that using only Read codes to select hypertensive patients in THIN underestimates the actual prevalence of hypertension in the UK.32 Depending on the use of anti-hypertensive drugs at baseline, hypertensive patients were categorized as treated or untreated. Uterine and ovarian status was grouped as: intact (no history of removal of uterus and ovaries), hysterectomy with oophorectomy (hysterectomy and at least one ovary removed), and oophorectomy only (one or both ovaries removed). A very small group of women with hysterectomy without oophorectomy was not included in the model. Birth year was grouped into four decade-long cohorts.