| |
Menopause Role in Heart Risk Disputed (see editorial, full text below)
|
| |
| |
Download the PDF here
Download the PDF here
MedPage Today
Published: September 06, 2011
Action Points
· Explain that heart disease mortality in women increased with age but menopause did not appear to play a role,
· Point out that in contrast, men had an accelerated risk of dying of heart disease at younger adult ages, but mortality plateaued at middle and older ages.
Heart disease mortality in women increased with age but menopause did not appear to play a role, a 50-year review of mortality trends showed, disputing the belief that premenopausal hormones protect women from heart disease, investigators said.
No spike in heart disease mortality occurred at menopausal ages in women from three large birth cohorts from the U.S., England, and Wales. Instead, mortality increased steadily across all age groups of adult women.
In contrast, men had an accelerated risk of dying of heart disease at younger adult ages, but mortality plateaued at middle and older ages, as reported online in BMJ.
"Our data show there is no big shift toward higher fatal heart attack rates after menopause," Dhananjay Vaidya, PhD, of Johns Hopkins, said in a statement.
"What we believe is going on is that the cells of the heart and arteries are aging like every other tissue in the body, and that is why we see more and more heart attacks every year as women age. Aging itself is an adequate explanation and the arrival of menopause with its altered hormonal impact does not seem to play a role."
The findings have clear implications for managing heart disease risk in women, Vaidya and coauthors believe.
"Efforts to improve cardiac health in women should focus on lifetime risk rather than risk only after menopause," they wrote in conclusion.
Scientists have long attributed the delayed onset of ischemic heart disease in women compared with its appearance in men to a protective effect of premenopausal hormones. However, that belief has little or no supporting evidence from epidemiologic or clinical studies, the authors noted in their introduction.
Mortality from heart disease does increase with age, but studies have shown no acceleration of mortality in women during or after menopause, they continued. Men's more rapid acceleration of ischemic heart disease mortality at younger ages could just as likely reflect a failure of vascular repair mechanisms in men, as opposed to hormonal protection in women.
In an effort to sort out the conflict between clinical perceptions and data, Vaidya and colleagues analyzed data on three birth cohorts, beginning in 1916 and ending in 1945. They determined the population of each birth cohort on the basis of census data from 1950 and 2000.
Overall mortality declined with each successive birth cohort at advanced ages.
The authors analyzed ischemic heart disease mortality over time as the three birth cohorts aged. None of the cohorts had a substantial upswing in the slope of heart disease log-mortality curves for women around the time of menopause. Rather, the data showed a log-linear increase in overall mortality across all ages.
Among men, the log-linear increase in mortality continued until about age 45, and then remained stable into older ages.
Consistent with knowledge that men develop heart disease at an earlier age, the ischemic heart disease mortality curve for men increased by 30.3% a year until age 45, slowing significantly to 5.2% thereafter (P=0.042). The overall change averaged 5.8%
Among women, ischemic heart disease mortality increased by 7.9% overall and did not differ significantly before, during, or after menopause (P=0.43).
The data are consistent with the authors' proposed biologic explanation that fatal ischemic heart disease events result from "loss of reparative reserve."
"Acceleration in male heart disease mortality at younger ages could explain sex differences rather than any menopausal changes in women," the authors wrote.
--------------------------------------
Editorial
Sex differences in the risk of cardiovascular disease -pdf attached
BMJ 2011 Sept 6
Nisha I Parikh assistant professor of medicine John A Burns School of Medicine, University of Hawaii and Division of Cardiology, Queen's Medical Center, Honolulu HI 96813, USA
The accelerated risk at menopause may not be as clear as previously thought
Long before the landmark Women's Health Initiative clinical trial studied the effects of hormone replacement therapy (HRT) on cardiovascular disease in postmenopausal women, several questions framed the debate about whether menopause truly represented a period of equalisation between women and men in the risk of cardiovascular disease.1 2 Was menopause a pathophysiological turning point for women, or was the acceleration in risk due to natural ageing processes, independent of the hormonal effects of menopause? Recent studies have asked whether cardiovascular disease risk factors lead to menopause rather than the other way around.3 Still, the overarching question remains: if a sex gap exists in the risk of cardiovascular disease, how can we use our understanding of what drives this difference to take better care of our patients and ultimately to close the divide?
The linked study by Vaidya and colleagues(below) substantially extends our understanding of the epidemiology of sex differences in death from ischaemic heart disease.4 The authors used longitudinal data in men and women from the United States and from England and Wales to quantitatively test whether ischaemic heart disease mortality significantly changes (either accelerates or decelerates) around the estimated time of menopause (45-54 years of age). They found that, unlike deaths from breast cancer, rates of death from ischaemic heart disease did not rise around the estimated time of menopause in women. Interestingly, a deceleration in the rate of death from ischaemic heart disease was seen in men at age 45, which the authors argue might account for the perceived equalisation of risk in women at the time of menopause. The use of longitudinal data, two large geographically separate cohorts, and statistical methods to test for the ischaemic heart disease mortality rate changes strengthen the authors' observations.
While Vaiyda and colleagues' results seem to exonerate menopause as the sole driver of sex differences in death from ischaemic heart disease, the story may not be that simple. Current data from clinical trials do not support a protective effect of maintaining a pre-menopausal hormonal milieu; in fact, HRT in postmenopausal women was associated with modest increases in cardiovascular disease compared with placebo.5 However, debate continues about whether HRT is efficacious in younger, recently postmenopausal women compared with older, remotely menopausal women.6 With the so called "timing hypothesis" as yet unresolved, evidence based practice guidelines appropriately advise that HRT should not be used primarily for the prevention of cardiovascular disease in postmenopausal women.
Vaidya and colleagues' study confirms previous reports of a deceleration in the rate of ischaemic heart disease among men at age 45 years. The concept of a male climacteric or "andropause" has been popularised in the lay press7 and has been referred to in the scientific literature as late onset hypogonadism or male testosterone deficiency. Testosterone levels fall steadily in men between their 30s and 90s.8 However, testosterone deficiency has not been convincingly linked to incident cardiovascular disease in men and experts have called for more scientifically rigorous clinical trials of androgen replacement.9 Biological processes aside from those related to sex hormones might account for the increased ischaemic heart disease risk among younger men, and this would be an interesting area for further study.
Vaidya and colleagues show that ischaemic heart disease in women is a life course disease that steadily marches forward, showing no midlife acceleration. However, this does not mean that cardiovascular disease risk factors in women are entirely "sex neutral." Increasing evidence shows that factors related to pregnancy-such as a history of pre-eclampsia,10 gestational diabetes,11 preterm delivery,10 and having babies with low birth weight10-increase the risks of long term cardiovascular disease in women. Indeed, for the first time, guidelines on preventing cardiovascular disease in women from the American Heart Association advocate that doctors take a reproductive history when stratifying the risk of cardiovascular disease.12 Knowledge of pregnancy related risk factors may help physicians identify women who are at risk earlier than midlife. Whether these factors confer an independent risk of cardiovascular disease above and beyond classic risk factors in women is still uncertain, but is an important area for further study.
Although use of epidemiological data to answer clinical questions is not without its faults, a major strength is that it generally offers a bird's eye view rather than a keyhole vantage point. To answer the challenging question of what drives differences in cardiovascular disease in men and women over their lifespan, a combination of science from both far and near perspectives will be needed.
---------------------------------------------
Ageing, menopause, and ischaemic heart disease mortality in England, Wales, and the United States: modelling study of national mortality data
BMJ 2011 September 2011
Abstract
Objectives To use changes in heart disease mortality rates with age to investigate the plausibility of attributing women's lower heart disease mortality than men to the protective effects of premenopausal sex hormones.
Design Modelling study of longitudinal mortality data with models assuming (i) a linear association between mortality rates and age (absolute mortality) or (ii) a logarithmic association (proportional mortality). We fitted models to age and sex specific mortality rates in the census years 1950 to 2000 for three birth cohorts (1916-25, 1926-35, and 1936-45).
Data sources UK Office for National Statistics and the US National Center for Health Statistics.
Main outcome measure(s) Fit of models to data for England and Wales and for the US.
Results For England-Wales data, proportional increases in ischaemic heart disease mortality fitted the data better than absolute increases (improvement in deviance statistics: women, 58 logarithmic units; men, 37). We identified a deceleration in male mortality after age 45 years (decreasing from 30.3% to 5.2% per age-year, P=0.042), although the corresponding difference in women was non-significant (P=0.43, overall trend 7.9% per age-year, P<0.001). By contrast, female breast cancer mortality decelerated significantly after age 45 years (decreasing from 19.3% to 2.6% per age-year, P<0.001). We found similar results in US data.
Conclusions Proportional age related changes in ischaemic heart disease mortality, suggesting a loss of reparative reserve, fit longitudinal mortality data from England, Wales, and the United States better than absolute age related changes in mortality. Acceleration in male heart disease mortality at younger ages could explain sex differences rather than any menopausal changes in women.
Introduction
The relative delay in the onset of ischaemic heart disease in women compared with men, on a population level, has been attributed largely to the putative protective effects of the premenopausal hormonal milieu, but little epidemiological or clinical evidence supports this.1 Circulating lipids have been shown to change levels during the menopausal transition,2 but interpretation of these changes is made complex by discrepant changes in the levels of cholesterol and apolipoprotein.3 Ischaemic heart disease mortality increases with age, but cross sectional analyses of age specific mortality have not shown any sudden proportional acceleration at the age of menopause in women, either in data from England and Wales,4 the United States,5 or Japan.6 By contrast, the proportional increase in breast cancer mortality with age abruptly decelerates at menopausal age among women.1 7
Two issues need resolution with respect to previous cross-sectional analyses. Firstly, it is not known why age related mortality should be examined on the proportional (or logarithmic) scale for ischaemic heart disease mortality, rather than an absolute age related increase in mortality, which accelerates throughout adult life in both men and women. By contrast, the accumulation of multiple injuries has been suggested as a mechanistic explanation for a proportional increase in cancer related mortality with age.7 Secondly, ischaemic heart disease has a long latency period before death, and in view of changing lifestyles and treatment environments over a person's lifetime, it is difficult to interpret cross sectional associations between age and mortality.
The dual goal of this analysis was to challenge existing paradigms regarding causality attribution to menopause as it relates to the risk of ischaemic heart disease in women, by providing a mechanistic hypothesis for proportional increases in mortality, and applying it to longitudinal vital statistics data. To do this, we aimed to show that a proportional change in mortality is biologically and numerically meaningful, possibly representing a constant probability of the failure of tissue reparative cells across a lifetime.8 9 Indeed, the length of telomeres in circulating leucocytes, which can be taken as a proxy for other bone marrow cells that replenish the circulatory system, reduces linearly during adult life.10 Because telomere length shortens linearly with every cell division, the linear loss of telomere length over adult life10 indicates that, on average, cells divide at a constant rate over time. Loss of telomere length limits a cell's capacity to replicate.11 Therefore, regeneration of bone marrow cells is limited in a linear fashion with increasing age. This explanation is a generalisation of the cancer specific process proposed by Pike and colleagues7 to all senescent chronic disease, and is based on age related heart disease mortality among birth cohorts (born 1916-45) in England and Wales12 and the United States.13
Methods
Mortality models as a function of age
Model 1 assumes a linear association between mortality rates and age (absolute mortality). Indeed, this model is implicit in the statement of the current hypothesis that the increase in absolute heart disease mortality accelerates at menopause. Such a model would suggest that the upward slope of absolute mortality versus age is roughly constant before menopause, and would change to a different slope around the time of menopause in women.
Mortality rate=ß0+ß1xage in years+ß2x(time since menopause in years)
(where ß0=intercept, ß1=linear slope with respect to age, and ß2=change in slope after menopause; if there is no change in slope at menopause, ß2 estimate=0)
Model 2 assumes a proportional (or logarithmic) association between mortality and age. Survival of an organism results from an ongoing repair of tissue injury that is dependent on a pool of repair mechanisms (for example, regenerative repair cells). Ageing is thought to be a result of loss of this pool. If the probability of removal of these reparative cells due to cell division is constant over time, the reparative reserve is lost at a constant proportion with age. Using the derivation in web appendix 1, we obtained the following model:
log(mortality rate)=ß0+ß1xage in years+ß2x(time since menopause in years)
(where ß0=intercept, ß1=linear slope with respect to age, and ß2=change in slope after menopause; if there is no change in slope at menopause, ß2 estimate=0)
Therefore, we can interpret modelling changes in proportional mortality in model 2, rather than absolute mortality as in model 1, as changes in reparative demands on a decreasing reserve of repair mechanisms (for example, stem cells).
We used the deviance metric (loge(likelihood2reference/likelihood2model)), with the naive model as the reference, to determine which model made the better fit to the data. For likelihood calculations to be comparable, we also specified the residuals from the models in the same manner-that is, on the absolute mortality rate scale.
Modelling of data from England, Wales, and the United States
We analysed data abstracted from the United Kingdom Office for National Statistics 20th century mortality data for England and Wales,12 US census and vital statistics databases, and the US National Center for Health Statistics.13 We used the age and sex specific mortality rates for the actual census years 1950 to 2000. We used the methods proposed by Frost14 to follow the mortality of three decade-long birth cohorts (1916-25, 1926-35, and 1936-45). We chose these birth cohorts because they included women who reached postmenopausal ages within the 20th century, and these cohorts were relatively unaffected by large scale immigration waves into the United States. We chose the same birth cohorts for the England-Wales data for comparability.
We followed US cohorts in time thus: the census counts and mortality rates for the 1936-45 cohort corresponded with the 5-14 year age group (cohort mid-age rounded to 10 years) in the 1950 census and vital statistics reports, and the 15-24 year age group (cohort mid-age 20 years) in the 1960 reports. We continued this method up to the 75-84 year age group (cohort mid-age 80 years) in the 2000 reports. Similarly, we traced the mortality rates for all three cohorts. We analysed all cause mortality and disease specific mortality rates for heart disease (in men and women) and breast cancer (only in women).
We abstracted mortality data for ischaemic heart disease and breast cancer for the England-Wales cohorts, using International Classification of Diseases codes (table 1↓). For the US data source,13 longitudinal mortality rates over the full analysis period were available for total heart disease rather than for ischaemic heart disease. The National Center for Health Statistics collected breast cancer mortality data only for women older than 25 years.13 We excluded death rates for individuals older than 85 years because the National Center for Health Statistics placed these deaths into a single group in these datasets. Therefore, we could not assign these individuals to any particular decadal birth cohort. Web appendix 2 shows the full abstracted dataset from the vital statistics sources used to graph and model the data.
We analysed data separately for men and women. We used the cohort mid-age as the age variable for plotting and regression analysis. We regarded rates for women in cohorts when they aged to 45-54 years (cohort mid-age 50 years) and older to be postmenopausal. The vast majority of menopause occurs during the 45-54 year age period. We also compared the different age-mortality slopes before and after 45 years of age in models 1 and 2, to determine whether the age-mortality trend would change at menopause in women, and at a similar age in men.
We calculated the longitudinal death rates in the cohorts using generalised mixed models. For comparability of model fit between the linear age dependence and the log-linear age dependence, we calculated the explained variance in the mortality rate data on the original scale (Gaussian residual errors). Detailed model equations are presented in web appendix 1. If we saw a secular trend by calendar year, we would model such a trend as secondary analysis.
To compare the models, we tabulated deviance statistics and the difference in the Akaike information criterion15 with the naive linear model as the reference. We considered a reduction of at least four logarithmic units in the Akaike information criterion to be a significant improvement, and a reduction of at least 10 logarithmic units to be a highly significant improvement in model fit (simplified from Burnham and Anderson's multiple level criteria16).
Model fitting was undertaken using the generalised linear latent and mixed models software implemented in Stata (version 10, StataCorp, College Station, TX, USA). Web appendix 1 also describes the sensitivity analysis.
Results
Table 2↓ shows the census counts of men and women of the three birth cohorts (1916-25, 1926-35, and 1936-45) for 1950 and 2000, by country. Figure 1↓ shows the ischaemic heart disease mortality of the three birth cohorts in England and Wales as they aged in the 20th century, either on the absolute scale or the logarithmic scale.
For the England-Wales birth cohorts, all absolute mortality curves showed an upward-bent profile, whereas the log-mortality curves did not show a prominent upward bend. Among women, none of the cohorts showed a substantial upward shift in slope in the log-mortality curves due to ischaemic heart disease around the time of menopause (45-55 years), and there was a log-linear increase in mortality throughout all ages. Among men, the log-linear increase in mortality was strongly blunted after age 45 years in all birth cohorts. For both men and women, we saw a reduction in the overall mortality of successive birth cohorts at advanced ages. US data for total heart disease mortality also showed the same pattern (fig 2↓).
Figure 3↓shows female breast cancer mortality on the absolute and logarithmic scales in the England-Wales and the US birth cohorts. As with previous cross sectional analyses,1 we observed a distinct difference in slope gradients before and after age 45 years in the log-mortality rate plots. We did not see any cohort-wide reductions in breast cancer mortality for later cohorts, especially at young ages. However, there is a suggestion that the log-mortality curves could have flattened in the 1990s for all cohorts (that is, at different ages for each of the cohorts). This finding could be a secular phenomenon related to the adoption of breast cancer screening programmes in the early 1990s.17 18 19
Table 3 shows the association of mortality with age by longitudinal data analysis for the various models and their goodness of fit. For all cause mortality and heart disease mortality, the log-mortality models fit the data much better than the absolute mortality models. Interestingly, women showed a non-significant deceleration in the log-mortality curves after age 45 years, whereas men showed a significant deceleration after age 45 years. For breast cancer mortality in women, the log-mortality model showed the best fit, with a significant deceleration of age related increase in mortality after age 45 years. We also identified a strong secular trend for reduced breast cancer mortality after 1990 (fig 3). After adjusting for birth cohort and age, breast cancer mortality was 25% lower after 1990. Web appendix 3 shows similar results of the analyses of total heart disease and breast cancer mortality in US data.
Discussion
We have shown that analysis using log-linear mortality rates is consistent with the plausible biological mechanism of reparative reserve. By using the most comprehensive and complete longitudinal mortality dataset available for England and Wales as well as for the United States, in the 20th century, we have shown that log-linear models fit the mortality data much better than a simple linear model.
Comparison with previous findings from cross sectional analyses
Our longitudinal cohort analysis confirms previously reported cross sectional age-mortality associations.4 5 However, our findings are more convincing than those of previous analyses, because age trends are best studied by following cohorts of individuals over their lifetime rather than comparing different individuals of the same and different ages in a cross section of the population.
Issues of unaccounted secular trends and cohort effects have clouded the interpretation of cross sectional mortality analyses, and are resolved by our longitudinal analysis. Cross sectional analyses would not have detected our finding that cohorts born later in the 20th century have lower cardiovascular mortality at higher ages than cohorts born earlier in the 20th century. In addition, our analysis showed a secular change in breast cancer mortality after the 1990s, which could not have been detected with cross sectional analysis.
We showed that there was no sharp increase in slope or step increase in ischaemic heart disease mortality among women at menopausal ages. The mortality steadily increased across all ages in adult women, whereas the corresponding increase was blunted at older ages in men. These observations also applied to three birth cohorts in different countries, presumably with varying secular life experiences in terms of lifestyle, knowledge of health risks (such as smoking), and disease and risk factor treatments. These data strengthen the conclusions of previous cross sectional analyses.5 In our longitudinal birth cohort analyses, the age effect is not confounded by the decreasing overall heart disease risk of successive birth cohorts. Therefore, the flattening of the curves at higher ages in men should be justifiably considered a biological phenomenon rather than an artefact of cross sectional analysis.
Potential explanations for age-mortality trends
The steady log-linear increase in age related heart disease mortality in men and women suggests cumulative effects of lifelong vascular injury and failure of repair mechanisms. However, this interpretation is based on mathematical derivation and published biological studies in small samples, and not measurements in these national cohorts.
Because both men and women in our study showed upward bent curves on the absolute mortality scale, the upward bend cannot be interpreted as being attributable to menopause. In fact, men showed a steeper age-mortality slope than women as young adults, which flattened into middle age. As previously suggested,1 the accelerated risk in men at younger ages compared with women at the same ages could indicate an accelerated failure of the reparative reserve in young men, rather than the relative protection of premenopause hormonal effects in young women. Because we analysed only vital statistics data (and therefore could not analyse individual cardiovascular risk factors or the individual timings of menopause in women), this biological explanation is speculative. However, our proposed mechanism is consistent with a study reporting that the rate of telomere loss per age-year was equal in adult men (0.038 kb/age-year) and women (0.036 kb/age-year) at all ages between 25 and 75 years.20 But in that study,20 age adjusted telomere lengths were substantially shorter, by 0.28 kb, in men than women. This difference is equivalent to the telomere shortening process being shifted 7.6 age-years in men, compared with women.
Although we did not see any proportional acceleration of heart disease mortality at menopause, the absolute mortality still increases at all ages in women, including the postmenopausal period. Thus the current concern about cardiovascular disease among women in the ageing US population is fully justified, but with the focus on lifetime risk rather than primarily at menopause.
Pike and colleagues discussed that carcinogenesis, being a multistage process, would be expected to have a log-linear age-mortality association, and they showed such an association for colorectal cancer mortality.7 From new cell biological insights,8 9 10 we generalised this paradigm to all chronic senescent processes, including the multistage accumulation of carcinogenic trauma, as well as vascular damage under the general concept of the failure of reparative reserve. Our models confirm the deceleration of breast cancer mortality around the age of menopause as previously shown.5 7 This finding is consistent with our biological model that non-communicable diseases could represent a failure of the reparative reserve. The cyclical hyperplasia and hypoplasia of breast tissue (evidence reviewed by Pike et al7) before menopause is expected to strain the reparative reserve for this organ more than after menopause. The demands on the reparative reserve could vary with age for different organ systems, and for the two sexes, modulating the rate of reduction in reparative reserve over age. Similarly, known cardiovascular risk factors such as smoking, hypercholesterolaemia, and hypertension might damage vascular tissue and therefore increase the demand for repair. This would consequently alter the rate of loss of reparative reserve. We recognise that a full explanation of the age association of cause specific mortality due to chronic senescent conditions would require modelling these factors. However, the log-linear mortality-age model had a much better fit than the simple linear model.
Census and mortality data
A major strength of this analysis is the comprehensive nature of the British and US census and mortality data. However, there are specific caveats. Although we examined ischaemic heart disease mortality in the British data, cause specific mortality data were not separately available for the disease in the US data source for periods before 1980.13 However, the findings were substantially similar in the two analyses. Ischaemic heart disease constituted most of the age adjusted total heart disease mortality in the United States for years when the data were separately available-for example, of total heart mortality in 2000, 75.4% in men and 69.5% in women were due to ischaemic heart disease. Thus the US and British data were similarly interpretable. Total heart disease mortality also includes deaths due to heart failure, which is a chronic senescent condition associated with the loss of progenitor cells,21 22 and our biological discussion applies to heart failure as well.
Another consideration is that the birth cohorts were open cohorts due to immigration and emigration. However, individuals who migrated to the United States as adults in the 1900-20 wave of immigration would not greatly affect the earliest birth cohort (1916-25) of this analysis. The recent major immigration wave of 1980-2000 included mostly young adults and added little to the population aged 45 years and older,23 and should not affect our latest birth cohort. In addition, the US and Britain have different migration experiences, yet we found similar results between the two datasets, suggesting that immigration patterns are unlikely to affect the conclusions and interpretation of our analysis.
Conclusions
We have shown that each successive birth cohort in England, Wales, and the United States in 1916-45 had lower total and heart disease mortality over their lifetimes. Heart disease mortality in women increased exponentially throughout all ages, with no special step increase at menopausal ages. This finding sharply contrasts with breast cancer mortality in women, which decelerated drastically at menopause. The increase in heart disease mortality due to age in men was greater during early adulthood than that in women, and the age related increase was blunted at advanced ages.
Efforts to improve cardiac health in women should focus on lifetime risk rather than risk only after menopause. Our proposed biological model showing that fatal events are a result of a loss of reparative reserve fitted the mortality data far better than a model of linear age related increases in absolute mortality rates. This finding should serve as a point of departure for more refined models of age related mortality.
|
|
| |
| |
|
|
|
