Sex differences in risk factors for myocardial infarction: cohort study of UK Biobank participants

Abstract Objectives To investigate sex differences in risk factors for incident myocardial infarction (MI) and whether they vary with age. Design Prospective population based study. Setting UK Biobank. Participants 471 998 participants (56% women; mean age 56.2) with no history of cardiovascular disease. Main outcome measure Incident (fatal and non-fatal) MI. Results 5081 participants (1463 (28.8%) of whom were women) had MI over seven years’ mean follow-up, resulting in an incidence per 10 000 person years of 7.76 (95% confidence interval 7.37 to 8.16) for women and 24.35 (23.57 to 25.16) for men. Higher blood pressure indices, smoking intensity, body mass index, and the presence of diabetes were associated with an increased risk of MI in men and women, but associations were attenuated with age. In women, systolic blood pressure and hypertension, smoking status and intensity, and diabetes were associated with higher hazard ratios for MI compared with men: ratio of hazard ratios 1.09 (95% confidence interval 1.02 to 1.16) for systolic blood pressure, 1.55 (1.32 to 1.83) for current smoking, 2.91 (1.56 to 5.45) for type 1 diabetes, and 1.47 (1.16 to 1.87) for type 2 diabetes. There was no evidence that any of these ratios of hazard ratios decreased with age (P>0.2). With the exception of type 1 diabetes, the incidence of MI was higher in men than in women for all risk factors. Conclusions Although the incidence of MI was higher in men than in women, several risk factors were more strongly associated with MI in women compared with men. Sex specific associations between risk factors and MI declined with age, but, where it occurred, the higher relative risk in women remained. As the population ages and the prevalence of lifestyle associated risk factors increase, the incidence of MI in women will likely become more similar to that in men.

Supplementary table 3 Multiple adjusted hazard ratios and women-to-men ratios of hazard ratios for the association between diabetes type II further classified according to self-reported treatment type, and incident myocardial infarction Systolic blood pressure (SBP) per 20 mmMg, diastolic blood pressure (DBP) per 10 mmMg, current smokers were compared with never smokers, participants with diabetes were compared with those without diabetes, obese and overweight participants were compared with those with BMI<25kg/m 2 , middle and lowest thirds of socioeconomic status (SES) were compared with highest third.
Supplementary Systolic blood pressure (SBP) per 20 mmMg, diastolic blood pressure (DBP) per 10 mmMg. Hypertension stages were compared to participants with normal blood pressure, smoking groups were compared with never smokers, participants with diabetes were compared with those without diabetes, overweight and obese were compared with BMI <25kg/m 2 , patients with Atrial Fibrillation (AF) were compared to those without AF.
All models were adjusted for age. Additionally, SBP, diabetes and SES were adjusted for each other as well as smoking status, BMI, use of lipid-lowering medication and anti-hypertensive medication. AF was similarly adjusted for these eight variables. DBP and AHA hypertension stages were adjusted for the same variables as SBP. The models for smoking variables included SES, and the models for BMI contained smoking status and SES.
Supplementary Hypertension stages were compared to participants with normal blood pressure, smoking groups were compared with never smokers, participants with diabetes were compared with those without diabetes, overweight and obese were compared with BMI <25kg/m 2 , patients with Atrial Fibrillation (AF) were compared to those without a AF.
All models were adjusted for age. Additionally, SBP and diabetes were adjusted for each other as well as smoking status, BMI, use of lipid-lowering medication and anti-hypertensive medication. AF was similarly adjusted for these seven variables. DBP and AHA hypertension stages were adjusted for the same variables as SBP. The models for smoking variables were only adjusted for age, and the models for BMI also contained smoking status.
Supplementary Missing data includes non-response and responses "don't know" and "prefer not to answer"

Supplementary text
Let P be the prevalence of a cluster of adverse risk factors, and assume this represents the full effect of risk factors. Assume P is the same in women and men. By definition, 0 ≤ P ≤ 1.
Let w be the risk of MI in unexposed (to the cluster) women.
Let s be the male:female risk ratio for MI in unexposed women.
Let r be the exposed:unexposed relative risk for MI in women.
Let t be the male:female ratio of relative risks for MI.

Then the risk of MI in women overall is
Prw + (1 -P)w and the risk in men is Ptrsw + (1 -P)sw.
If we consider all else to be fixed, this equates to = − 1 + − − 1 = .
For example, if r = 3 (women with the risk cluster are 3 times as likely to get MI as are those without), s =2 (men without the risk cluster are twice as likely to get MI as are their female peers) and t = 1/3 (the relative risk for women is three times that of men), women and men will have equal rates of MI if P = 0.5 (that is, if 50% of people have all the risk factors in the cluster).
If E > 0 then the overall female rate will exceed the male rate if P > E, if E < 0 then the male rate will always exceed the female rate.
So, in the example, women will have higher rates of MI if the (percentage) prevalence of the risk cluster exceeds 50%. In reality, this is extremely unlikely.