SUPPLEMENTARY TEXT

METHODS

Participants

Cross-sectional and cohort studies

The Copenhagen City Heart Study is a prospective cardiovascular study of individuals randomly selected according to the Central-Population-Register code to reflect the adult Danish population at large. Those invited were stratified into 5-year age groups ranging from 20 to 95 years; 35- to 70-year-olds were emphasized. In 1976-78, 19329 individuals were invited, of whom 74% (14223) participated. In 1981-83, the original cohort supplemented with 500 20- to 25-year-olds was invited to participate; 70% (12698) participated. Finally, in 1991-94, when DNA was collected, the cohort was further supplemented with 3000 20-49-year-olds, and of those invited 61% (10135) participated. More than 99% of participants were whites of Danish descent. Of the 10135 participants who attended the 1991-94 examination, 9259 gave blood for DNA analyses. Of these 9244 underwent genotyping for the ESR1 IVS1-397T/C polymorphism. Thus, 40% of all those ever invited to the Copenhagen City Heart Studies were included in the present study.

All-cause mortality among non-responders was approximately 70% higher than among responders (Supplementary Figure 1). Morbidity from IHD and ICVD was 20% higher among non-responders than among responders at the 1991-94 examination. Thus, as in most other studies incidence of disease endpoints is underestimated in our study; however, as this underestimate is unlikely to differ by genotype, estimated effect sizes by genotype is unlikely to be affected by differences between responders and non-responders.

Diagnoses of IHD, MI, ICVD and ischemic stroke was based on International Classification of Diseases 8th revision (ICD-8) codes 410-414, 410, 432-435 and 432-434 until the end of 1993, and thereafter based on 10th revision (ICD-10) codes I20-I25, I21-I22, I63-I64 and I63. IHD was defined as a previous MI or characteristic symptoms of stable angina pectoris. The diagnosis of MI required the presence of at least 2 of the following criteria: characteristic chest pain, elevated cardiac enzymes and electrocardiographic changes indicative of MI. Fatal events were dying within one week of a MI. ICVD was determined on the basis of sudden onset of focal neurological symptoms (ischemic stroke, transient ischemic attack or amaurosis fugax). Ischemic stroke was focal neurological symptoms lasting more than 24 hours. Hemorrhagic stroke and subarachnoidal hemorrhage diagnosed on computed tomography scan were excluded from the ICVD and ischemic stroke groups. We also gathered information on deep vein thrombosis (DVT) (ICD-8 codes 451.00, 451.08, 451.09, 451.90, 451.92, 671.01–671.09 and ICD-10 codes I80.1, I80.2, I80.3, O22.3, O87.1) and pulmonary embolism (PE) (ICD-8 codes 450.99, 673.99 and ICD-10 codes I26.0, I26.9, O88.2). The diagnostic criteria used were ultrasonography or venography in the case of DVT and ventilation–perfusion scintigraphy or pulmonary angiography in the case of PE. ICD-7 codes 170.0-170.5, 470.0-1, 970.0-1 were classified as breast cancer, codes 171.0, 171.1, 171.2, 174.0, 174.1, 971.0, 972.0, 972.1, 973.0 as uterine cancer (including both corpus and cervix uteri), codes 175.0, 175.1, 175.2, 175.3, 175.5, 176.9, 375.0, 475.0 as ovarian cancer, and codes 177.0, 477.0, 977.0 as prostate cancer. Intracapsular and intertrochanteric hip fractures were classified according to ICD-8 codes 820.00-820.02 and ICD-10 codes S72.0-S72.1.

Case-control studies

The diagnosis of IHD was obtained from experienced cardiologists based on angina pectoris plus at least one of the following criteria: stenosis/atherosclerosis on coronary angiography, a previous MI, or significant myocardial ischemia on a bicycle exercise test. Experienced neurologists and vascular surgeons diagnosed ICVD on the basis of sudden onset of focal neurological symptoms, together with carotid artery stenosis of at least 50% on the symptomatic or most stenotic side. Cerebral hemorrhage was excluded by computed tomography. The diagnosis of MI (n=1136), ischemic stroke (n=311) and breast cancer (n=1256) was established with the same criteria as in the Copenhagen City Heart Study. Cases were compared with healthy age and gender matched subjects from the 1991-94 examination of The Copenhagen City Heart Study. Matching cases to controls ranged from 1:1 to 1:4 depending on availability of controls. Case-patients who were registered as participants in the Copenhagen City Heart Study were excluded from the hospital samples.

Genotyping

In the NCBI database, SNPs in ESR1 around IVS1-397T/C (rs2234693) are as follows:

Contig position rs# Heterozygosity

56267713 rs4986936 0.041 intron A/G

56267760 rs11155815 0.498 intron C/T

56267763 rs8179176 0.495 intron C/T

56267764 rs2234693 0.497 intron C/T IVS1-397T/C

56267807 rs11155816 0.461 intron C/T

Beside the IVS1-397T/C (rs2234693) SNP, two of these other reported SNPs (rs11155815 and rs8179176) are within the PvuII recognition site as well as the TaqMan probe site (Supplementary Fig. 2). These two SNPs therefore potentially could interfere with correct genotyping of the IVS1-397T/C (rs2234693) SNP studied in the present paper. However, after sequencing 311 persons we did not detect these two SNPs in the Danish population (Supplementary Table 2).

Statistical analyses

We used left truncation as a statistical term for the fact that a period of ignorance exists before an individual enters into the study. In The Copenhagen City Heart Study individuals have the genetic risk (the genotype in question) already at birth (or even at conception), but first enter the study many years later. In Stata this is implemented by setting origin=birthday and enter=date of the entry into the study. Then, the individuals´ age, instead of follow-up time, is used as time-scale to estimate risk of disease. The individuals included in the analyses must be free of the disease in question prior to the commencement of follow-up period, i.e. no failures during the truncation period. Age was used in defining the risk set because age is a strong predictor of the outcomes (cardiovascular disease, cancer and hip fracture). With age as time-scale, we cannot study the effects of age itself. We therefore tested age-interaction with genotype by using the follow-up time to define the risk set without the origin term.

An updated meta-analysis on risk of MI in men according to ESR1 IVS1-397T/C CC genotype versus TC or TT genotype combined was conducted with 8 studies (6 previously published and the present 2). Studies were identified via searches in PubMed. The inclusion criterion was data for men on ESR1 IVS1-397T/C genotype and fatal and/or non-fatal MI. There were no exclusion criteria. Fixed-effect summary measures were calculated as the inverse-variance weighted average of the log odds ratios. Subsidiary analyses included random-effects model. Stratification on study design and fatal versus non-fatal MI was decided a priori. Heterogeneity was examined using I² statistics. Publication bias was examined with Begg’s funnel plot and Begg’s statistical test.

RESULTS

Meta-analysis

We included 8 studies with a total of 4799 male cases with MI (4516 non-fatal, 283 fatal) and 12190 male controls. With fixed-effect model the overall odds ratio of non-fatal MI in CC vs. CT/TT genotype was 1.08(0.97-1.21)(Figure 4); the corresponding random-effects odds ratio was 1.20(0.98-1.46). The overall model for non-fatal MI showed evidence for study heterogeneity (p=0.009). When The Framingham Heart Study was excluded from the model, evidence for study heterogeneity disappeared (P=0.15), and the equivalent overall odds ratio of non-fatal MI was 1.04 (0.93-1.17) and 1.09 (0.93-1.27) with fixed-effects and random-effects model, respectively. Overall odds ratio for fatal MI with fixed-effects and random-effects model was 0.81(0.59-1.12) and 0.75(0.47-1.22).

When stratifying by study design, overall odds ratio for non-fatal MI was 1.42 (1.12-1.78) and 1.46 (1.00-2.14) in cohort studies using fixed-effects and random-effects model, respectively. Equivalent overall odds ratio was 1.00 (0.88-1.13; p=0.49) and 1.01 (0.87-1.19) in case-control studies using fixed-effects and random-effects model, respectively.

Odds ratios for individual studies derived in our analyses differ slightly from those reported in the previous meta-analysis by Shearman et al.. The reason is that in that meta-analysis the odds ratios were adjusted; however, we did not have this possibility and used unadjusted odds ratios. There was no evidence for publication bias in the studies.

DISCUSSION

Though Framingham, Rotterdam, and Copenhagen Studies are all prospective cohorts of Caucasians of European descent there are also differences, for example, the Rotterdam Study participants had a mean baseline age of almost 70 years and were followed for 7 years, while Framingham has a mean baseline age of around 40 years and was followed for 27 years. Though Framingham had a relatively young baseline age only subjects that survived until the examination cycle at which blood samples were taken for DNA extraction were included in the subset used for genetic association studies. In The Copenhagen City Heart Study, the mean baseline age was 45 years and follow-up time was 24 years. Only subjects surviving until the 1991-94 examination, when blood was drawn for the DNA extraction, were included in the study. In these respects, our study is similar to The Framingham Heart Study and dissimilar to The Rotterdam Study. This could potentially explain why our results differ from The Rotterdam Study, but not why they differ from The Framingham Heart Study. Furthermore, we confirmed our findings on MI from the prospective cohort study in a case-control study. The Framingham Heart Study found increased risk of MI in men with CC versus TT genotype. In our efforts to mimic analyses of The Framingham and Rotterdam Studies with respect to entry age and follow-up time (Figure 3), our data suggested increased risk of MI in younger men with CC versus TT genotype (equivalent to decreased risk in TT versus CC genotype). However, a recently published meta-analysis (including The Framingham and Rotterdam Studies) suggested increased risk of non-fatal MI for older, but not younger men with CC versus CT/TT genotype.


LEGENDS

Supplemantary Figure 1. Total mortality and incidence of ischemic heart disease and ischemic cerebrovascular disease after the 1991-94 examination in responders versus non-responders of the Copenhagen City Heart Study.

Supplementary Figure 2. Part of intron 1 of the estrogen receptor alpha gene.

Word count: 1593

8