We used data from the Risk of Arterial Thrombosis in Relation to Oral Contraceptives (RATIO) follow-up study, based on subjects who participated in the RATIO case-control study [12]. For the present study two groups of patients, subjects who survived either a myocardial infarction or an ischaemic stroke, were included. Patient selection within the RATIO case-control study has been described in detail previously [13‐15]. In summary, the RATIO study included young women (between 18 and 49 years old) with either a first myocardial infarction or ischaemic stroke (index events) who consecutively presented to one of the 16 participating hospitals in the Netherlands between 1990 and 1995. Exclusion criteria were TIA (defined as an event lasting < 24 h), haemorrhagic stroke, venous sinus thrombosis, carotid artery dissection, history of cardiovascular or cerebrovascular diseases, severe illness, aphasia, or cognitive impairment interfering with the questionnaire and not speaking Dutch. Myocardial infarction was diagnosed by the presence of clinical symptoms, elevated cardiac enzyme levels, and corresponding electrocardiographic changes. Clinical symptoms of ischaemic stroke were confirmed by either computed tomography or magnetic resonance imaging. Ischaemic stroke of cardioembolic origin was excluded by the presence of atrial fibrillation or suggestive cardiac ultrasound imaging. All cases were included from 1995 to 1998 [16]. The population-based group of control women was identified by random digit dialing and matched according to age, area of residence, and year of event. Eligible as controls were women without prior myocardial infarction and ischaemic stroke, who were aged 18 to 49 years and who did not meet the exclusion criteria that were used for selecting patients. All participants were asked to fill in the same structured questionnaire comprising questions on medical history of cardiovascular risk factors. Smoking was defined as having regularly smoked in the year before the index date. Alcohol consumption was defined as drinking at least one unit of alcohol per week. Body mass index (BMI) was calculated as body weight (kg) divided by height squared (m²). Women were classified as hypertensive, diabetic or hypercholesterolaemic when they reported a physician’s diagnosis or were taking medication for these conditions before the index date. Family history of cardiovascular disease was defined as the presence of myocardial infarction, stroke or peripheral artery disease below 60 years of age in first degree relatives.

Blood samples and buccal swabs were collected for blood measurement of clotting factors and DNA analyses in a subset of women who initially participated in the first phase of the RATIO study (205 participants with myocardial infarction, 125 with ischaemic stroke and 638 controls). To compensate for the loss of statistical power in the ischaemic stroke group, we additionally recruited 50 women between 1996 and 2001, leading to a small difference in median time between event and blood sampling between the two case groups (69 months, range 38 to 117 months, for myocardial infarction cases and 95 months, range 23 to 146 months, for ischaemic stroke cases). The interval between the initial event and blood sampling minimises the risk of reverse causation as well as the mixing of short and long-term effects of hypercoagulability on cardiovascular risk. Blood draw procedures and laboratory procedures are described in detail in the previous publications [13‐15]. All participants gave informed consent and the study was approved by the ethics committees of the participating hospitals [13].

In 2013, data on participants of the RATIO case-control study were linked to the Dutch Register of death certificates and to the Dutch Hospital Data register by the Central Bureau of Statistics of the Netherlands. The first provides both primary and secondary causes of deaths coded according to the International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10) classification. The second provides electronic coverage of data on all hospital admissions in the Netherlands since 1995. The data collected concern the date of admission and discharge, diagnoses, and surgical procedures from virtually all university and general hospitals and most specialized clinics. Diagnosis are encoded according to ICD 9th Revision (ICD-9). Almost all personal (age, sex, date of birth and postal code) and administrative data (date of admission, discharge and death), and 84% of principal diagnoses were found to be correctly encoded by a previous comparison of a random sample of hospital admissions in the Dutch Hospital Data register with information from hospital records [17]. For myocardial infarction, the percentage of correctly encoded diagnosis has since been found to be almost 100% [18].

Data from the RATIO case-control study were linked to these registries through date of birth, sex, and postal code of each RATIO participant. Individuals with information leading to more than one person (e.g., twins or individuals with the same date of birth in the same postal area) or to nobody at all, were excluded from the analyses. The original RATIO case control study included 1376 women (248 myocardial infarction cases, 203 ischaemic stroke cases and 925 controls). Of these women, 1168 (87%) were successfully linked to the registries.

For this study prespecified outcomes were the first occurrence of an acute cardiovascular event, either myocardial infarction or ischaemic stroke, whichever occurred first.

The outcome myocardial infarction includes the diagnoses defined by the ICD-9-CM codes 4100 to 4109 and the cause of death defined by the ICD-10 code I21. The outcome ischaemic stroke includes the diagnosis defined by the ICD-9-CM codes 433, 4330 to 4333, 4338, 4339, 434, 4340, 4341, and 4349 and codes 4350- to 4359 (transient ischaemic stroke), and the cause of death defined by the ICD-10 codes I63 and I64. We chose to include these two acute cardiovascular diseases, as these will most likely lead to hospitalization and therefore captured by the Dutch Hospital Data register.

Follow-up ended on the date of a first incident cardiovascular event of interest, the date of death, or 31st December 2012, whichever came first. Median follow-up was calculated using the estimates of the censoring distribution for overall mortality.

To explore the influence of an increased coagulation propensity on the risk of a recurrent cardiovascular event we compiled an individual prothrombotic score (coagulation score). The score was constructed with coagulation markers measured in both patient groups and in control subjects, being: 1) tissue factor/tissue plasminogen activator induced clot-lysis time (CLT) as a measure of fibrinolytic potential [19]; 2) antigen levels of coagulation factors of the intrinsic coagulation system (factor XII, FXII, and FXI and prekallikrein) [20]; 3) inhibitor complexes of the serine proteases of the intrinsic coagulation system (C1 esterase inhibitor for FXIIa, FXIa, Kallikrein, and antitrypsin inhibitor) as measures of the activation of intrinsic coagulation factors [21]; 4) a disintegrin-like and metalloprotease with thrombospondin type 1 motif, member 13 (ADAMTS13) antigen levels [22]; 5) von Willebrand factor (VWF) antigen levels [22]; 6) antiphospholipid antibody tests (presence of lupus anticoagulant detected with dilute Russell’s viper venom time (dRVVT) reagents, IgG anticardiolipin antibody concentrations, IgG anti-β2-glycoprotein I concentrations and anti-prothrombin antibodies concentrations) [9]; 7) presence of factor V Leiden (either heterozygous or homozygous) [23, 24]; 8) presence of G20210A mutation in the prothrombin gene (either heterozygous or homozygous) [23, 24]; 9) presence of the homozygous form of the methylenetetrahydrofolate reductase (MTHFR) C677T [23, 24]. The score was based on the beta coefficients, adjusted for matching variables, obtained from logistic regression models for the association between each coagulation marker and index ischaemic stroke. These analyses were based on dichotomous exposures, with the 90th percentile of the control group distribution as cut off value, or on the presence of the genetic variant, or on the test positivity, whichever was the most appropriate. These beta coefficients were summed so that the compiled score represents the coagulation weighted prothrombotic potential for each patient, with higher values of the score corresponding to a higher levels of thrombotic propensity.

To investigate whether high values of the coagulation score increased the risk of a recurrent event, we applied a Cox proportional hazard model based on the quartiles of the prothrombotic score distribution as exposure categories, with the lowest quartile as reference. The model was adjusted for age, sex, body mass index, alcohol consumption, smoking history, diabetes mellitus, hypertension, hyperlipidaemia and family history of a cardiovascular event.