Cardiovascular health in the menopause transition: a longitudinal study of up to 3892 women with up to four repeated measures of risk factors – BMC…

Main findings

Our results suggest that reproductive age (reflecting the menopausal transition) does not independently influence change in sub-clinical atherosclerosis (CIMT) or risk factors (e.g. SBP, non-HDL-cholesterol and triglycerides) strongly associated with atherosclerosis, as shown in randomised trials and/or Mendelian randomisation studies to causally influence coronary heart disease [17, 27, 28]. By contrast reproductive age may increase adiposity and risk of diabetes, albeit modestly, as suggested by stronger positive linear associations with reproductive age than chronological age for BMI, fat mass, and fasting glucose. HRT may not identically reflect endogenous hormonal and other changes associated with a natural menopause. However, it is notable that our findings have some consistency with randomised controlled trials of HRT, which have shown no protection, or a possible increased risk for coronary heart disease and reduced risk for type 2 diabetes [29].

To our knowledge, this is the largest prospective study to date with two repeat CIMT measures and up to four repeated cardiovascular risk factor measures that spans the late reproductive period, from menopausal transition into post menopause. The average 5-year follow-up period with up to four repeat measures in women of different baseline ages allowed the description of associations from 4years before to 16years after the menopause, a longer postmenopausal period than described in previous studies.

We used multilevel models, which allow all women with at least one measurement occasion to be included in the analysis under the MAR assumption, i.e. missingness depends on observed data and therefore associations do not differ in women (with the same characteristics) who have fewer repeat measures. Furthermore, sensitivity analysis restricted to women who had three or four repeat measures showed similar results to those with at least one repeat measure. We had to restrict our main analyses to women in whom we could calculate their FMP, meaning only those who has at least 12months since their last period could be included. This could introduce selection bias. However, consistency of our main analysis findings with those of the associations of change in outcome with chronological age by strata of menopausal status suggests our findings are not substantially biased by selection. We do however note that a womans menopausal stage will reflect her chronological age (i.e. at the time of baseline assessment, women who are pre-menopausal will be on average younger than those who are postmenopausal). We therefore need to be cautious in our interpretation and in the magnitude of associations which are likely to be driven by differences in the age distribution across the groups.

We fit models which included time since FMP and chronological age to separate the influence of both chronological and reproductive age. However, given that chronological age is the sum of age at menopause and time since FMP, we could have instead analysed time since FMP and age at menopause only, a reparameterisation of time since FMP and chronological age. As such, in mutually adjusted models, the coefficients of chronological age are equivalent to that of age at menopause, whilst time since FMP in the model including age at menopause is the sum of time since FMP in the model including chronological age and additionally the coefficient of chronological age.

As reproductive age is a self-reported measure and measured with more error than chronological age, it may be that this causes some bias towards the null for reproductive age, and correspondingly away from the null for chronological age.

Distributions of outcomes and confounders were similar between women included and excluded from the main analysis (Additional file 3: Tables S9S10).

Our study is predominantly of White European origin women, and previous studies have shown ethnic differences in cardiovascular risk factors [30], so our findings might not be generalisable to women of other race/ethnic groups. As our study recruited women during an index pregnancy and only followed those with a live birth from that pregnancy, all participants had at least one live birth and we cannot assume that our findings would generalise to women with no previous pregnancies or live births. As we know the risk of cardiovascular disease increases with an increase in live births [31], the association between reproductive age and cardiovascular health may differ in studies that also include nulliparous women. Vasomotor symptom severity and duration arealso known to associate with HRT use (the most effective treatment for these symptoms) and CVD risk. Censoring those who use exogenous hormones because we could not determine age at a natural menopause could induce some collider bias [32] if there is residual confounding between HRT and CVD. However, given the key confounders of HRT-CVD effects are the same as those for time to FMP and CVD (e.g. age, BMI, education) which we already adjust for, we anticipate that any bias would be small.

When restricting our sample to women with a time since FMP greater than 0, 71% of the sample, the median time (IQR) since FMP was 5.7years (4.28.8). We believe this time is long enough to observe any differences in CVD risks possibly related to the menopause. However, it may be possible that the longer women are followed up after menopause, evidence of associations become apparent, or the observed associations become larger in magnitude. Furthermore, given only 12% (203/1702) of the sample experienced early menopause, it is possible that women at the very low end of the age at menopause distribution are indeed at increased risk and we were not able to pick this up. These analyses are in unselected women in mid-life and only 20 (1.2%) had evidence of plaque or atherosclerosis, highlighting the need for further follow-up into older ages.

We were able to identify ten papers published up to December 2021 that either explored change in cardiovascular outcomes by reproductive age [8, 12, 15, 16, 33,34,35] or change with chronological age within strata of menopausal status [18, 36, 37]. We have summarised these in Additional file 4: Table S11 [8, 12, 15, 16, 18, 33,34,35,36,37] including number of women, number of repeated measures, sample characteristics and key results. With one exception, these included fewer than 500 women [18, 36, 37]. The one exception was the SWAN which included between 249 to 2659 women in different publications [8, 12, 15, 16, 18, 33, 35].

Only two of these explored associations with CIMT [16, 18]. El Khoudary et al. [18] included 249 participants, (122 premenopausal, 115 early peri-menopausal, 4 late peri-menopausal and 8 postmenopausal at baseline) and in line with our results found that CIMT increased in post-menopause (0.024mm/year, p-value 0.03) compared to pre-menopause, adjusting for age at baseline and ethnicity. Similarly, the recent SWAN paper [16] included 890 women with CIMT measures and suggested that older age at menopause was associated with an increase in CIMT.

Consistent with our results, Greendale et al., in a sub study of SWAN with N=1246 [15], found an independent association between reproductive ageing and gain in fat mass and loss of lean mass until 2years after the FMP in women who had an average age at FMP of 52years. Our findings, with larger numbers, add to this evidence in suggesting that reproductive age, independent of chronological age, increases body fat.

Unlike our findings, Derby et al. [8] found increases in triglycerides with reproductive age, having adjusted for chronological age; however, this change was small. As in our study, Matthews et al. [12] found increases in triglycerides in midlife were small and largely related to chronological age rather than reproductive age or menopausal status. A weak positive linear change in non-HDL-c with reproductive age, consistent with our results, was also shown in that study. In a previous analysis of the same cohort (ALSPAC) using a metabolomic and largely lipids platform, Wang et al. found important changes in many lipids across the menopausal transition, taking into account chronological age [11], however, data were available for only two time points.

Reproductive and chronological age were weakly positively associated with fasting glucose in our study whilst the SWAN studies found neither or a negative association [12, 33, 37]. However, our study was considerably larger than the others. Furthermore, the decrease with reproductive or chronological age would be surprising given in general populations diabetes increases with age [14].

Some studies have looked specifically at the association of early or premature menopause as a risk factor for CVD [38,39,40]. Daan et al. compared 83 women previously diagnosed with POI (i.e. loss of ovarian function before 40years of age) to 266 premenopausal women, all aged >45years, and found an association of POI with higher adiposity and higher CRP levels [40]. Similarly, Honigberg et al. in a study with 144260 postmenopausal women (natural or surgical menopause) found that premature menopause was associated with a small but increased risk for a composite of different CVD [38]. Our study, whilst analysing different parameters, has some consistency with those findings in suggesting that reproductive age associates with intermediate risk factors of CVD, such as adiposity and higher CRP and glucose levels, which could be relevant for later CVD.

Our findings are broadly in line with the narrative review behind the recently published American Heart Association (AHA) statement on menopausal transition and CVD [41]. In that review consistent with our findings, they do not find strong evidence that menopausal transition influences blood pressure or CIMT beyond chronological age and that there is evidence of an increase in fat mass through the menopausal transition, independent of chronological age, as well as fasting glucose, as we also find. They note that non-HDL-c increases across the menopause transition, which we also observed. Notably, they do not discuss in detail magnitudes of change and our review of key papers for this study suggest that these are modest (as in our study). They conclude that guidelines for CVD prevention should have specific reference to the menopause. They highlight the importance of early age at menopause as a risk factor for CVD and that those with surgical menopause, early menopause, and vasomotor symptoms should be considered for exogenous hormone replacement therapy. Previous cohort studies show that premature menopause is associated with CVD after adjustment for age and other CVD risk factors such as high blood pressure [38, 39]. The main aim of our paper adds to this work by using detailed repeat measures of established risk traits to show how these vary in relation to chronological and reproductive age. Whilst we show that chronological age seems to be more important for some risk factors, it is possible that the impact of reproductive age is influenced by those with premature menopause or early menopause. The previous studies were very large (N=144,000 and 301,000) to have power to compare risk of different cardiovascular diseases between premature menopause and menopause aged 5051 [39] or postmenopausal women without premature menopause [38]. Though the cited studies have much bigger sample sizes, we have repeat data and are able to separate the influence of both chronological and reproductive age. Furthermore, we did not find any evidence of non-linearity between reproductive or chronological age and many outcomes, suggesting that those with an earlier menopause did not appear to over influence our results. We do however note that it may not have been possible to pick this up in our sample. Furthermore, previous studies have found that changes in CVD risk factors over time were similar in women with natural and surgical menopause [34, 35], which supports our findings that chronological age might influence CVD risk more than reproductive age. In relation to the menopausal transition, they note that firm conclusions are difficult to make on the basis of current evidence but suggest supporting women to make behavioural changes (e.g. diet and physical activity) to maintain a healthy weight across mid-life would be potentially beneficial.

Here is the original post:
Cardiovascular health in the menopause transition: a longitudinal study of up to 3892 women with up to four repeated measures of risk factors - BMC...