How Does Cardiovascular Disease First Present in Women and Men?

Supplemental Digital Content is available in the text.


Content
All risk factors were based on data recorded in CPRD during primary care consultations in the year prior to the index date, unless otherwise specified. Where multiple blood test results were recorded, the values from the test closest in time to the index date was used.
Medications were deemed to be regular medications if patients had at least two repeat prescriptions, covering a two-month supply, in the year prior to the index date.
Patients' age was measured in years as at the index date. Socioeconomic status was based on an area measure of deprivation, the Index of Multiple Deprivation, 1 linked to their postcode (which was removed prior to receipt of the data). The most recent smoking record prior to the index date was used to classify patients as never, ex-or current smokers.
Patients identified as current smokers with no smoking record within the three years before study entry were reclassified as having missing smoking data. Never smokers who had a previous record of smoking were reclassified as ex-smokers. Body mass index (kg/m2) was calculated using weight measurement closest in time to the index date. Patients were defined as diabetic if they ever had a diagnosis of diabetes or were receiving regular prescriptions for either insulin or metformin. Baseline SBP was based on readings taken during routine primary care consultations; where more than one measurement was taken on the same day, the average measurement was used. Both total cholesterol and high density lipoprotein were defined as the level in routine blood tests, in mmol/L. Blood-pressure-lowering medications included in our definition are thiazide diuretics, betablockers, angiotensin converting enzyme-inhibitors, angiotensin receptor blockers, or calcium-channel blockers. Hormonal therapy (in women only) included combined oral contraceptives, progestogen-only oral contraceptives or hormone replacement therapy.
Additionally the following co-variates were used for the multiple imputation:

eText 3 Multiple imputation
Multiple imputation 2 was implemented using the mice algorithm in the statistical package R.
Imputation models were estimated separately for men and women and included: a) all the baseline covariates used in the main analysis (age, quadratic age, index of multiple deprivation, smoking, body mass index, diabetes, systolic blood pressure, total cholesterol, and HDL cholesterol); b) prior (between 1 and 4 years before study entry) and post (between 0 and 1 year after study entry) averages of continuous main analysis covariates and other measurements not in the main analysis (white cell count, haemoglobin, creatinine, alanine transferase); c) baseline medications (statins, blood pressure medications, aspirin, and oral contraceptives and hormone replacement therapy (in women only)); d) coexisting medical conditions (history of depression, cancer, renal disease, liver disease and chronic obstructive pulmonary disease); e) the Nelson-Aalen hazard and the event status for each endpoint analysed in the data 3 .
Non-normally distributed variables were log-transformed for imputation and exponentiated back to their original scale for analysis. Five multiply imputed datasets were generated, and Cox models fitted to each dataset. Coefficients were combined using Rubin's rules.
We checked whether the imputations were plausible by comparing plots of the distribution of observed and imputed values of all variables. HRs comparing men to women by source of endpoint data, adjusted for age and stratified by primary care practice. CHD indicates coronary heart disease; NOS, not otherwise specified.