There were 490,321 UK Biobank participants without a history of MI at study baseline. After further removal of 10,524 participants whose oestradiol and testosterone reason for missingness was not recorded 479,797 participants [264,282 (55.1%) women] were included in the present study. The median age at recruitment was 57 years for women and 58 for men (Table 1). Over 12.5 years of follow-up, there were 4908 MI events in women and 10,517 in men.

Oestradiol was detectable in 21.9% of women (in 64.1% of premenopausal women and 5.3% of postmenopausal women) and in 8.2% of men (Additional file 1: Table S1). Testosterone was detectable in 79.7% of women (in 86.9% of premenopausal and 78.0% of postmenopausal women) and 94.7% of men. Women with detectable oestradiol levels were typically younger [Median (IQI)] 47 years (44, 51) than those with undetectable 60 years (55, 64) levels (Additional file 1: Table S2 and Fig. S1). Oestrogen levels for women with detectable levels were highest in the < 45, 45–50, 50–55 year age groups (560 pmol/L, 581 pmol/L, 515 pmol/L, respectively), and lower for those 55–60 years (381 pmol/L), and 60–65 years (345 pmol/L) (Fig. 1). Similar patterns were observed by menopause status (Additional file 1: Fig. S2 and S3).

Rates of myocardial infarction and levels of sex hormones, with 95% confidence intervals, by sex and age group. Blue Men, Pink/red Women

Data for both oestradiol and testosterone, and thus the O/T were available in 19.9% of women and 8.2% of men. The median (IQI) O/T ratio was 0.379 (0.237, 0.631) nmol/L in women and 0.016 (0.013, 0.021) nmol/L in men (Additional file 1: Fig. S4). There were some variations in the age-adjusted mean risk factor levels according to quarters of O/T ratio (Additional file 1: Table S3).

SHBG was detectable in 85.8% of women and 87.6% of men. Median (IQI) SHBG levels were 56.5 (40.1, 77.4) nmol/L in women and 36.9 (27.9, 48.2) nmol/L in men. Corresponding levels were 62.4 (44.4, 84.8) nmol/L in premenopausal women and 54.9 (39.3, 74.5) nmol/L in postmenopausal women. FAI was detectable in 72.1% of women and 87.1% of men, median (IQI) levels were 1.80 (1.14, 2.85) nmol/L in women, and was similar by menopause status, and 31.32 (25.21, 39.28) nmol/L in men. SHBG was higher with higher age in men, compared with FAI that was lower with older age. Both SHBG and FAI tended to be lower with higher age in women (Additional file 1: Fig. S5).

For women, in the unadjusted analyses, a unit higher in log-transformed oestradiol was associated with a lower risk MI: HR [95% confidence interval (CI)] 0.70 (0.60; 0.82) (Additional file 1: Table S4). After adjusting for age, the HR became 0.93 (0.79, 1.08), and after adjusting for traditional CVD risk factors, it was 1.02 (0.87, 1.19) (Fig. 2). Multivariable-adjusted HR (95% CI) for MI for undetectable levels of oestradiol compared with detectable levels were 0.94 (0.91, 0.98) for women and 0.96 (0.93, 0.98) for men.

Association of per unit higher of log transformed sex hormones with MI by sex. Multiple adjusted hazard ratios (HR) with 95% confidence intervals: Blue Men, Pink/red Women. Multiple adjusted models adjusted for age, smoking status, total cholesterol, BMI, systolic blood pressure, Townsend deprivation score, diabetes, any anti-hypertensive or lipid lowering medication. O/T, oestradiol/testosterone; FAI, free androgen index, SHBG sex hormone–binding globulin

For women, a unit higher in log-transformed testosterone was not associated with MI after multiple adjustment: HR (95% CI) 0.97 (0.91, 1.04). For men, in unadjusted and age adjusted analyses (per unit higher in log-transformed) testosterone was associated with a lower risk of MI: the corresponding HR’s (95% CI) were 0.72 (0.68, 0.76), 0.79 (0.75, 0.83), this no longer held after multiple adjustment 1.02 (0.946, 1.08).

For men, (per unit higher in log-transformed) O/T ratio was associated with a lower risk of MI after adjusting for classical CVD risk factors HR (95% CI) 0.79 (0.65, 0.95), though not for women 0.98 (0.86, 1.12). The women to men RHR (95% CI) was [1.24 (0.99, 1.56)]. There was no significant linear trend in the HRs by quarters of O/T ratio for women p = 0.952, though there was for men (p = 0.050), The multiple adjusted HR (95% CI) for Q4 vs Q1 for women was 1.02 (0.83, 1.25), and for men 0.81 (0.70, 0.93) (Additional file 1: Fig. S6).

For men, in the multivariable adjusted analyses, a unit higher in log-transformed SHBG was associated with a lower risk of MI HR (95% CI) 0.94 (0.89, 0.99), though not in women 1.02 (0.95; 1.09).

For men, in the multivariable adjusted analyses, a unit higher in log-transformed FAI was associated with a higher risk of MI: HR (95% CI) 1.09 (1.02, 1.15), though not in women 0.97 (0.92, 1.02), with a corresponding women to men RHR (95% CI) 0.89 (0.82, 0.97).

Analyses by quarters (Q4 vs Q1) of FAI yielded a HR (95% CI) of 1.13 (1.08, 1.19) for men and 0.92 (0.86, 0.99) for women, which yielded a women to men RHR (95% CI) 0.82 (0.75, 0.0.89).

In models that adjusted for both sex hormones simultaneously, there remained no association between either oestrogen and testosterone with MI in women. In men, testosterone was associated with a higher risk of MI after adjustment for oestrogen and traditional CVD risk factors: HR (95% CI) 1.39 (1.12, 1.73) (Additional file 1: Table S5).

The association of SHBG with lower risk of MI in men remained after adjustment for testosterone 0.91 (0.86, 0.97), although not after adjustment for oestradiol 0.96 (0.80, 1.14), this, however, is likely as a consequence of the model being fitted using data of those with both detectable SHBG and oestradiol levels.

There were no significant associations with MI per unit higher of log transformed oestradiol, testosterone and O/T ratio by menopause status, after adjusting for classical CVD risk factors. (Additional file 1: Table S6 and Fig. 3).

Association of per unit higher of log transformed sex hormones with MI by menopause status (for women). Multiple adjusted hazard ratios (HR) with 95% confidence intervals: dark green pre-menopause, purple post-menopause. Multiple adjusted hazard ratios (HR) with 95% confidence intervals: dark green pre-menopause, purple post-menopause. Multiple adjusted models adjusted for age, smoking status, total cholesterol, BMI, systolic blood pressure, Townsend deprivation score, diabetes, any anti-hypertensive or lipid lowering medication. O/T, oestradiol/testosterone; FAI, free androgen index; SHBG, sex hormone–binding globulin

Higher FAI (Q4 vs Q1) was associated with a higher risk of MI in pre-menopausal women after multiple adjustment HR (95% CI) 1.37 (1.11, 1.68), but a lower risk in post-menopausal women 0.88 (0.81, 0.96), with a post- vs pre-menopausal women RHR (95% CI) of 0.64 (0.52, 0.80).

Higher SHBG (Q4 vs Q1) was associated with a lower risk of MI in pre-menopausal women 0.69 (0.53, 0.89), though not in post-menopausal women 0.99 (0.91, 1.08). This result yielded a post- vs pre-menopausal women RHR (95% CI) of 1.44 (1.10, 1.89) (Additional file 1: Fig. S7).

When conducted by HRT/not results were broadly similar to pre- and post-menopausal women overall (Additional file 1: Table S7). These results should be interpreted with caution due to the small number of events and detectable hormone levels among the groups.