- Email: [email protected]
Lifetime estrogen exposure versus age at menopause as mortality predictor
Maturitas 43 (2002) 105 /112 www.elsevier.com/locate/maturitas
Lifetime estrogen exposure versus age at menopause as mortality predictor Sophia C. Jansen, Elisabeth H.M. Temme *, Evert G. Schouten Department of Public Health, Catholic University of Leuven, Division of Nutritional Epidemiology, Kapucijnenvoer 33 /35, B-3000 Leuven, Belgium Received 26 September 2001; received in revised form 5 June 2002; accepted 9 July 2002
Abstract Objectives: The aim of this study was to evaluate the relation between lifetime estrogen exposure and mortality and compare this with menopausal age as exposure variable. Methods: We studied a cohort of 1462 naturally postmenopausal women, aged 37 /77 at enrolment in the Belgian Interuniversity Research on Nutrition and Health study. After a follow-up time of 10 years, 181 women had died, of whom 76 of cardiovascular causes. Logistic regression analysis was used to investigate the relations between lifetime estrogen exposure (calculated as menopausal age minus menarcheal age) and death as well as the relations between age at menopause and death. Results: The risk of mortality was lower in women with a longer lifetime estrogen exposure as well as in women with higher menopausal ages. For women with a lifetime estrogen exposure of ]/40 years the odds ratio of all-cause mortality was 0.58 (95% confidence interval (CI) 0.35 /0.93) compared to women who had a lifetime estrogen exposure of 5/33 years. Women who became menopausal after the age of 53 years had a similar reduction in mortality risk (odds ratio 0.62; 95% CI 0.36 /1.03) when compared to women with menopausal ages of 5/46 years. These decreases in mortality risk were particularly due to a reduction in mortality of cardiovascular diseases. Conclusions: This study indicates that menopausal age predicts mortality. The prediction is not improved by adding the age at menarche, to obtain an estimate for lifetime estrogen exposure. # 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Lifetime estrogen exposure; Menopause; Mortality; Women
1. Introduction Menopause indicates a loss in ovarian function and subsequent deficiency of endogenous estrogens. As endogenous estrogens exert a favorable
* Corresponding author. Tel.: /32-16-336-914; fax: /32-16337-015 E-mail address: [email protected] (E.H.M. Temme).
influence on lipoproteins, haemostasis and vasomotion, menopause is associated with an unfavorable change in these factors [1]. As a result, early age at natural menopause incurs a higher risk of mortality, in particular mortality of cardiovascular diseases [2 /8]. Although the effects of early menopause on mortality risk have been well documented in many studies, there are very few studies on the effects of the duration of exposure to endogenous estrogen
0378-5122/02/$ - see front matter # 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 3 7 8 - 5 1 2 2 ( 0 2 ) 0 0 1 8 3 - 4
106
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
on mortality risk. Lifetime duration of exposure to estrogen could be a stronger predictor of mortality, because it is a more precise quantification of duration of exposure to endogenous estrogen than menopausal age. One study found that a longer exposure to endogenous estrogen in premenopausal life (lasting from menarche until menopause) is protective against coronary heart disease mortality [9]. Other studies have investigated only the relationships between lifetime exposure to endogenous estrogen and specific diseases, mainly breast cancer [10] and osteoporosis [11]. As far as we know, the relation between lifetime estrogen exposure and all-cause mortality has not been reported before. We examined the relationships between menopausal age and death (all-cause mortality, mortality of cardiovascular diseases and mortality not caused by cardiovascular diseases) as well as the relationships between lifetime estrogen exposure and death in a cohort of 1462 Belgian women who had experienced a natural menopause. The data analyzed in the present study were obtained from the Belgian Interuniversity Research on Nutrition and Health (BIRNH) study.
ceptives, number of children, menopausal status, age at menopause and type of menopause (natural, surgical or other). Dietary habits were assessed according to the 24-h food record method. A total of 1705 women reported in the questionnaires a history of natural menopause. Data for 46 women were excluded because their menopausal or menarcheal age was unknown. An additional 197 observations were excluded because of missing values for at least one of the following variables: alcohol consumption, body mass index, presence of diabetes, education level, ever smoking, marital status, number of children, occupation, previous use of oral contraceptives or systolic blood pressure. As a result, 1462 women were included in the present analysis. Ten years after enrollment of each participants vital status and cause of death was retrieved. Follow-up for cause-specific mortality was completed satisfactorily in 99%. Causes of death were ascertained from the family doctor or the doctor who completed the death certificate. Death certificate coding was based on the International Coding of Diseases, Ninth Revision (ICD-9). Any one of the codes 390/459 was used as the definition of cardiovascular mortality.
2. Subjects and methods
2.2. Statistical analysis
2.1. Study population
Lifetime estrogen exposure was calculated for each woman by subtracting the menarcheal age from the age at menopause. The lifetime estrogen exposure was categorized into 33 /34, 36/37, 38 / 39 and ]/40 years. The reference group comprised women who had a lifetime estrogen exposure of 5/ 33 years. Menopausal age was categorized into 47/48, 49 /50, 51/52 and ]/53 years, with a reference group of women who became menopausal at age 5/46 years. Differences in characteristics between groups of lifetime estrogen exposure and between groups of menopausal age were tested with a Chi-square test or a Tukey test (a /0.05). Possible confounding variables were examined by assessing their association with mortality. Relationships between lifetime estrogen exposure and mortality of all causes, mortality of cardiovascular diseases (CVD) and mortality not caused by CVD were investigated
The aims, design, methodology, and results of the BIRNH study have been described in detail elsewhere [12,13]. Briefly, a randomized sample of the population was drawn from each of the 42 counties in Belgium by using the voting lists. The general population eligible for sampling was between the ages of 25 and 74. Of the 30 964 subjects sampled, a total of 11 302 subjects (5949 men and 5353 women) agreed to participate and was screened between 1980 and 1984. Major coronary and other risk factors were measured. Questions were asked on smoking habits, socialprofessional class, educational level and marital status. Further questions concerned the use of medications and personal medical antecedents. For women, questionnaires included items on age at menarche, past and present use of oral contra-
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
using logistic regression analysis. Age-adjusted and multivariate adjusted odds ratios (OR) with 95% confidence intervals (95% CI) were calculated by lifetime estrogen exposure group. Variables used to adjust odds ratios originated from the baseline examinations, e.g. the age at filling out the questionnaire. Relationships between age at menopause and mortality were investigated in a similar way. When analyzing cause-specific mortality, seven additional observations were excluded because of missing values for ICD-9 codes. Analyses were performed using SAS software (SAS institute, Cary, NC).
3. Results Results are reported for 1462 naturally postmenopausal women who were 37 /77 years of age at baseline. The mean age at natural menopause in these women was 49.4 years (range 31 /65 years), the mean age at menarche was 13.6 years (range 10 /19 years) and the mean lifetime estrogen exposure 35.8 years (range 13/51 years). The Pearson correlation coefficient of lifetime estrogen exposure with age at menopause was 0.9 and with age at menarche /0.03. After a total follow-up time of 14 000 person-years, 181 women (12.4%) had died, of whom 42% due to cardiovascular diseases, 54% due to non-cardiovascular causes and the remaining due to unknown causes. 3.1. Lifetime estrogen exposure The baseline characteristics of the population stratified by lifetime estrogen exposure group are given in Table 1. A higher lifetime estrogen exposure was statistically significantly associated with higher age, higher body mass index, higher systolic blood pressure and higher prevalence of diabetes mellitus. The menarcheal age, number of children, use of oral contraceptives and ever smoking were distributed almost similarly among groups of lifetime estrogen exposure. Table 3 shows the associations between lifetime estrogen exposure and all-cause mortality. The estimated odds ratio of mortality decreased with increasing categories of lifetime estrogen exposure.
107
After adjustment for possible confounding variables, the odds ratio of mortality for women in the highest lifetime estrogen exposure group was 0.58 (95% CI 0.35 /0.93) compared to women with the lowest lifetime estrogen exposure. A test for a linear trend across groups of lifetime estrogen exposure gave a multivariate adjusted P -value of 0.06. Further adjustments for alcohol consumption, body mass index, ever smoking, number of children, previous use of oral contraceptives, education level, marital status and occupation level did not change the odds ratio estimates substantially (data not shown). Additional analyses by cause of death are shown in Table 4. An inverse relation was observed between categories of lifetime estrogen exposure and mortality of CVD, which explained the largest part of the association between lifetime estrogen exposure and all-cause mortality. Confidence intervals however, all comprised 1.0, which could be due to the low numbers of deaths, particularly in the cause-specific analyses. 3.2. Menopausal age Table 2 gives the baseline characteristics of the population by group of menopausal age. Women with the highest age at natural menopause were significantly older and had significantly higher systolic blood pressure and higher prevalence of diabetes than women that were younger at menopause. Body mass index and number of children did also increase across categories of menopausal age, but not enough to be statistically significant. Age at menarche was slightly lower at higher ages at menopause. There was no clear trend for previous use of oral contraceptives or ever smoking. The P -value for a linear trend across groups of menopausal age was 0.05. The associations between menopausal age and all-cause mortality are shown in Table 3. Women with menopause at ages 47 /48 and 49/50 years had a slightly higher mortality risk than women in the reference group. For women with menopause above age 50, however, the mortality risk was decreased compared to the reference group. After adjustment for possible confounding variables, the odds ratio for women in the highest menopausal
108
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
Table 1 Baseline characteristics stratified by lifetime estrogen exposure (n /1462) Lifetime estrogen exposure (years) 5/33
34 /35
36 /37
38 /39
]/40
394 61 (9) 44 (4) 14 (2) 2 (2) 7 27 (4) 143 (22) 2 (4) 14 4
253 61 (8) 49 (1)1 14 (1) 2 (2) 10 27 (5) 143 (21) 2 (3) 13 2
277 61 (7) 50 (1)12 14 (1)12 2 (2) 8 27 (4) 141 (20) 2 (4) 14 3
250 61 (6) 52 (1)123 13 (1)123 2 (2) 8 28 (5) 145 (21) 2 (4) 13 4
288 63 (6)1234 54 (2)1234 13 (1)1234 3 (2) 6 28 (4)123 147 (22)3 2 (4) 13 8
Education level (%) Primary school More than primary school
65 35
63 37
63 37
58 42
62 38
Marital status (%) Married Not married Widowed Divorced
74 5 19 2
66 9 23 2
73 4 19 4
76 6 16 2
72 3 24 1
Occupation level (%) Low Intermediate High
11 60 29
15 55 30
10 64 26
17 54 28
7 63 30
9 (2)
10 (2)
10 (2)
10 (2)
10 (2)
16 8 8
15 5 10
11 4 6
10 4 5
12 6 6
n Age (years) Menopausal age (years) Menarcheal age (years) No. children Ever use of OC (%) BMI (kg/m2) SBP (mmHg) Alcohol intake (en%) Ever smoking (%) Diabetes mellitus (%)*
Follow up (years) No. deaths (/1000 py) All causes CVD Non-CVD
Values for continue variables are presented as mean (S.D.). Tukey test for difference from the first1, second2, third3 or fourth4 group (P B/0.05). OC, oral contraceptives; BMI, body mass index; SBP, systolic blood pressure; CVD, cardiovascular diseases. * P B/0.05 (Chi-square test) for differences among groups.
age group was 0.62 (95% CI 0.36 /1.03) compared to women with the lowest menopausal age. Similar patterns were seen in the associations between age at menopause and mortality of CVD, as well as mortality not caused by CVD (Table 4). When menopausal age was grouped into 5/50 and /50 years, the multivariate adjusted odds ratio for women with menopausal ages above 50 years was 0.57 (95% CI 0.40 /0.81) for all-cause mortality and 0.54 (95% CI 0.31 /0.92) for mortality of CVD, compared to women with lower menopausal ages. The odds ratio of mortality not
caused by CVD was 0.65 (95% CI 0.41 /1.002) (data not shown). These results indicate that mortality of CVD explained the major part of the associations between menopausal age and death.
4. Discussion In this prospective cohort study of 1462 naturally postmenopausal women, we found an odds ratio of 0.58 (95% CI 0.35 /0.93) for women with a
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
109
Table 2 Baseline characteristics stratified by menopausal age (n /1462) Menopausal age (years) 5/46
47 /48
49 /50
51 /52
]/53
299 61 (9) 13 (2) 30 (3) 2 (2) 6 28 (4) 144 (23) 2 (4) 14 5
217 60 (8) 14 (2)1 34 (2)1 2 (2) 8 27 (4) 140 (20) 2 (4) 14 1
382 61 (7) 14 (2) 36 (2)12 2 (2) 9 27 (4) 143 (21) 2 (3) 14 2
254 61 (6) 14 (2) 38 (2)123 3 (2) 8 28 (5) 144 (21) 2 (4) 14 6
310 63 (6)1234 14 (2) 41 (2)1234 3 (2) 7 28 (4) 146 (21)2 2 (4) 11 6
Eduction level (%) Primary school More than primary school
63 37
59 41
66 34
60 40
63 37
Marital status (%) Married Not married Widowed Divorced
74 6 18 2
72 5 21 2
69 7 20 3
76 3 19 2
73 5 21 1
Occupation level (%) Low Intermediate High
32 56 12
27 59 14
27 60 13
28 60 12
30 62 8
10 (2)
9 (2)
9 (2)
10 (1)
10 (2)
15 7 8
15 5 7
15 6 9
9 4 5
11 5 6
n Age (years) Menarcheal age (years) LEE (years) No. children Ever use of OC (%) BMI (kg/m2) SBP (mmHg) Alcohol intake (en%) Ever smoking (%) Diabetes mellitus (%)*
Follow up (years) No. deaths (/1000 py) All causes CVD Non-CVD
Values for continue variables are presented as mean (S.D.). Tukey test for difference from the first1, second2, third3 or fourth4 group (P B/0.05). LEE, lifetime estrogen exposure; OC, oral contraceptives; BMI, body mass index; SBP, systolic blood pressure; CVD, cardiovascular diseases. * P B/0.05 (Chi-square test) for differences among groups.
lifetime estrogen exposure of 40 or more years compared to women who had a lifetime estrogen exposure of 5/33 years. The decrease in the risk of mortality of CVD was (although not statistically significant) of the same magnitude. Kleijn et al. grouped lifetime estrogen exposure into 5/25 and /35 years, and found that the latter group had a 20% reduced risk of mortality of CHD (ageadjusted hazard ratio 0.80, 95% CI 0.68 /0.96) [9]. Our findings indicate a stronger inverse relationship between lifetime estrogen exposure
and mortality of CVD. As other studies have investigated only the relationships between lifetime estrogen exposure and specific diseases, we are not able to compare our results of the association between lifetime estrogen exposure and mortality with others. Considering the associations between age at menopause and death in the same population, the odds ratio of all-cause mortality was 0.62 (95% CI 0.36 /1.03) for women who became menopausal after the age of 53 years compared to women with
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
110
Table 3 Estimated odds ratio of all-cause mortality stratified by lifetime estrogen exposure (LEE) or menopausal age (MA), with 95% confidence interval n
Age-adjusted odds ratio
Multivariate adjusted* odds ratio
LEE (years) 5/33 34 /35 36 /37 38 /39 ]/40
394 253 277 250 288
1.00 0.96 0.74 0.63 0.64
(0.60 /1.54) (0.45 /1.20) (0.37 /1.06) (0.40 /1.02)
1.00 1.00 0.75 0.62 0.58
(0.61 /1.60) (0.45 /1.23) (0.35 /1.04) (0.35 /0.93)
MA (years) 5/46 47 /48 49 /50 51 /52 ]/53
299 217 382 254 310
1.00 1.16 1.06 0.68 0.64
(0.68 /1.97) (0.67 /1.67) (0.39 /1.18) (0.38 /1.06)
1.00 1.29 1.14 0.66 0.62
(0.75 /2.21) (0.72 /1.81) (0.36 /1.16) (0.36 /1.03)
*
Ajusted for age, presence of diabetes mellitus and systolic blood pressure.
menopausal ages of 5/46 years. This reduction was particularly due to a reduction in mortality of CVD. Other studies have shown associations of similar magnitude, for example relative risk ratios 1.3 or 1.6 for women with very early natural menopause (i.e. before age 40), compared to women aged 49/51 [7] or ]/50 [6] at menopause. Most studies investigating the relationships between age at menopause and death used menopausal age intervals of 5 years, with women aged
50/54 or ]/50 as reference group. We calculated 2-year intervals to have equally distributed numbers of subjects in each menopausal age group. Van der Schouw et al. found that each year’s delay in natural menopause gives a decrease in risk of mortality of CVD of 2% (age-adjusted hazard ratio 0.982, 95% CI 0.968 /0.996) [4]. This would incur a decrease of at least 14% for women with menopausal ages ]/53 years, compared to women who had their menopause before 46 years. Our
Table 4 Estimated odds ratio of cause-specific mortality stratified by lifetime estrogen exposure (LEE) or menopausal age (MA), with 95% confidence interval n
Age-adjusted odds ratio
Multivariate adjusted* odds ratio
CVD
non-CVD
CVD
non-CVD
LEE (years) 5/33 394 34 /35 253 36 /37 277 38 /39 250 288 ]/40
1.00 0.62 0.59 0.57 0.69
(0.29 /1.26) (0.27 /1.21) (0.24 /1.21) (0.35 /1.30)
1.00 1.34 0.89 0.70 0.72
(0.74 /2.39) (0.47 /1.66) (0.33 /1.37) (0.38 /1.33)
1.00 0.65 0.59 0.54 0.57
(0.30 /1.34) (0.27 /1.22) (0.23 /1.16) (0.28 /1.10)
1.00 1.34 0.90 0.69 0.71
(0.74 /2.40) (0.47 /1.67) (0.33 /1.36) (0.37 /1.31)
MA (years) 5/46 299 47 /48 217 49 /50 382 254 51 /52 ]/53 310
1.00 0.89 0.87 0.61 0.61
(0.40 /1.91) (0.45 /1.67) (0.25 /1.34) (0.29 /1.23)
1.00 1.10 1.20 0.70 0.74
(0.54 /2.18) (0.67 /2.16) (0.33 /1.44) (0.38 /1.41)
1.00 1.09 1.00 0.54 0.56
(0.48 /2.39) (0.51 /1.95) (0.22 /1.24) (0.26 /1.15)
1.00 1.12 1.21 0.70 0.73
(0.55 /2.23) (0.68 /2.19) (0.32 /1.43) (0.38 /1.41)
*
Ajusted for age, presence of diabetes mellitus and systolic blood pressure. CVD, cardiovascular diseases.
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
results suggest that the decrease in the risk of mortality of CVD, caused by a delay in natural menopause, might be even stronger. The risk estimates obtained in our study for the lifetime estrogen exposure groups were not substantially different from those of the menopausal age groups. This could be due to the low variability in age at menarche (95% between 11 and 16 years), compared to the variability in age at menopause (95% between 42 and 55 years), which makes the age at menopause the most powerful determinant of the lifetime estrogen exposure. The low correlation between lifetime estrogen exposure and age at menarche (Pearson correlation coefficient of /0.03) also points at a weak influence of menarcheal age. Our hypothesis was that lifetime estrogen exposure could be a stronger predictor of mortality than menopausal age, because it is a more precise quantification of duration of exposure to endogenous estrogen. However, to make an accurate assessment of the total duration of estrogen exposure, one should probably also take into account duration of breast feeding, number of abortions and miscarriages, length and regularity of menstrual cycles, and use of hormonal replacement therapy [14]. We had no information on these issues, and used only the age at menarche and menopause to calculate the lifetime estrogen exposure. We had information on the number of children and the use of oral contraceptives, but they were similar between the groups of lifetime estrogen exposure. Therefore, any misclassification seems to be non-differential. One could debate that the term lifetime estrogen exposure is inappropriate in the context of our study, because we had no adequate information to make a really accurate assessment of the cumulative exposure to endogenous estrogen. However, we followed the terminology commonly used in the literature. Most studies investigating the relationship between lifetime estrogen exposure and specific diseases used the formula age at menopause minus age at menarche [11], or the time from menarche to menopause minus the duration of pregnancies and breastfeeding [10]. Our study indicates that, when assessing mortality risk, these
111
measures are not superior to menopausal age as such. Biases could have led to incorrect effect estimates. Our study sample could have been healthier than the general Belgian population of postmenopausal women, because women had to be healthy enough to attend baseline examinations. However, this would probably have led to an underestimation of a possible inverse relationship between menopausal age or lifetime estrogen exposure and mortality. It should be noted that the participation rate in the BIRNH study was only 37%. Therefore, an additional random sample of 10% of the nonrespondents was screened. Reproductive history was not checked, but in dietary habits no significant differences were found between respondents and non-respondents [15]. Furthermore, the BIRNH study used a questionnaire to assess reproductive history, which could have led to biased reporting of age at menopause and menarche. But, as reproducibility and validity studies have indicated, most women are able to report the age at menopause and even the age at menarche with a high degree of accuracy [4,16,17]. Possible confounding variables were evaluated in our analysis by assessing their association with mortality. Since age is a major risk factor for mortality, all analyses were adjusted for age. Although systolic blood pressure could be involved in the biological mechanism of the effect of estrogens, the higher systolic blood pressure in women with a higher menopausal age or longer lifetime estrogen exposure can not be explained by this mechanism. When there would be an effect of estrogen, women with a higher menopausal age or longer lifetime estrogen exposure would rather have a lower systolic blood pressure [18]. The higher systolic blood pressure in these women might therefore be due to their higher age. Smoking reduces the age at natural menopause [19]. However, in our study, the percentage of women who had ever smoked did not significantly differ among groups of menopausal age and among groups of lifetime estrogen exposure. We could not evaluate the possible confounding effect of physical activity, stress and family history of CVD or cancer, because information on these variables was not available. Adjustments for alcohol con-
112
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
sumption, body mass index, ever smoking, menarcheal age, number of children, previous use of oral contraceptives, education level, marital status and occupation level did only marginally influence the odds ratios estimates. Although we had information on lipid concentrations, it would be inappropriate to adjust for them, because they could be involved in the biological mechanism of the effect of estrogens. In conclusion, this cohort study of 1462 naturally postmenopausal women indicates that menopausal age predicts mortality, especially mortality of CVD. The prediction is not improved by adding the age at menarche, to obtain an estimate for lifetime estrogen exposure.
Acknowledgements This study was supported by a grant from the Unilever Chair in Nutritional Epidemiology.
References [1] Kuller LH, Meilahn EN, Cauley JA, Gutai JP, Matthews KA. Epidemiologic studies of menopause: changes in risk factors and disease. Exp Gerontol 1994;29:495 /509. [2] Lapidus L, Bengtsson C, Lindquist O. Menopausal age and risk of cardiovascular disease and death. Acta Obstet Gynecol Scand Suppl 1985;130:37 /41. [3] Snowdon DA, Kane RL, Beeson WL, et al. Is early natural menopause a marker of health and aging? Am J Public Health 1989;79:709 /14. [4] Schouw van der YT, Graaf van der Y, Steyerberg EW, Eijkemans MJC, Banga JD. Age at menopause as a risk factor for cardiovascular mortality. Lancet 1996;347:714 / 8. [5] Jacobsen BK, Nilssen S, Heuch I, Kvale G. Does age at natural menopause affect mortality from ischemic heart disease? J Clin Epidemiol 1997;50:475 /9. [6] Cooper GS, Sandler DP. Age at natural menopause and mortality. Ann Epidemiol 1998;8:229 /35.
[7] Jacobsen BK, Knutsen SF, Fraser GE. Age at natural menopause and total mortality and mortality from ischemic heart disease: the Adventist Health Study. J Clin Epidemiol 1999;52:303 /7. [8] Cooper GS, Baird DD, Weinberg CR, Ephross SA, Sandler DP. Age at menopause and childbearing patterns in relation to mortality. Am J Epidemiol 2000;151:620 /3. [9] Kleijn de MJ, Schouw van der YT, Peeters PH, Banga JD, Verbeek AL, Graaf van der Y. Number of menstrual cycles is associated with coronary heart disease mortality in postmenopausal women (abstract). Circulation 1999;99:1111. [10] Adami HO, Bergstrom R, Lund E, Meirik O. Absence of association between reproductive variables and the risk of breast cancer in yong women in Sweden and Norway. Br J Cancer 1990;62:122 /6. [11] Nguyen TV, Jones G, Sambrook PN, White CP, Kelly PJ, Eisman JA. Effects of estrogen exposure and reproductive factors on bone mineral density and osteoporotic fractures. J Clin Endocrinol Metab 1995;80:2709 /14. [12] Kesteloot H. Nutrition and health: the conclusions of the BIRNH study. For the BIRNH study group. Acta Cardiol 1989;44:183 /94. [13] De Backer G. Regional differences in dietary habits, coronary risk factors and mortality rates in Belgium. I. Design and methodology. Acta Cardiol 1984;39:285 /92. [14] Kleijn de MJJ, Schouw van der YT, Graaf van der Y. Reproductive history and cardiovascular disease risk in postmenopausal women. A review of the literature. Maturitas 1999;33:7 /36. [15] Kornitzer M, Dramaix M. The Belgian Interuniversity Research on Nutrition and Health (BIRNH): general introduction for the BIRNH study group. Acta Cardiol 1989;44:89 /99. [16] Livson N, McNeil D. The accuracy of recalled age at menarche. Hum Biol 1962;34:218 /21. [17] Bean JA, Leeper JD, R.B. W, Sherman BM, Jagger H. Variations in the reporting of menstrual histories. Am J Epidemiol 1979;109:181 /185. [18] Staessen JA, Ginocchio G, Thijs L, Fagard R. Conventional and ambulatory blood pressure and menopause in a prospective population study. J Hum Hypertens 1997;11:507 /14. [19] Willet W, Stampfer MJ, Bain C, Lipnick R, Speizer FE, Rosner B. Cigarette smoking, relative weight, and menopause. Am J Epidemiol 1983;117:651 /8.
Lifetime estrogen exposure versus age at menopause as mortality predictor Sophia C. Jansen, Elisabeth H.M. Temme *, Evert G. Schouten Department of Public Health, Catholic University of Leuven, Division of Nutritional Epidemiology, Kapucijnenvoer 33 /35, B-3000 Leuven, Belgium Received 26 September 2001; received in revised form 5 June 2002; accepted 9 July 2002
Abstract Objectives: The aim of this study was to evaluate the relation between lifetime estrogen exposure and mortality and compare this with menopausal age as exposure variable. Methods: We studied a cohort of 1462 naturally postmenopausal women, aged 37 /77 at enrolment in the Belgian Interuniversity Research on Nutrition and Health study. After a follow-up time of 10 years, 181 women had died, of whom 76 of cardiovascular causes. Logistic regression analysis was used to investigate the relations between lifetime estrogen exposure (calculated as menopausal age minus menarcheal age) and death as well as the relations between age at menopause and death. Results: The risk of mortality was lower in women with a longer lifetime estrogen exposure as well as in women with higher menopausal ages. For women with a lifetime estrogen exposure of ]/40 years the odds ratio of all-cause mortality was 0.58 (95% confidence interval (CI) 0.35 /0.93) compared to women who had a lifetime estrogen exposure of 5/33 years. Women who became menopausal after the age of 53 years had a similar reduction in mortality risk (odds ratio 0.62; 95% CI 0.36 /1.03) when compared to women with menopausal ages of 5/46 years. These decreases in mortality risk were particularly due to a reduction in mortality of cardiovascular diseases. Conclusions: This study indicates that menopausal age predicts mortality. The prediction is not improved by adding the age at menarche, to obtain an estimate for lifetime estrogen exposure. # 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Lifetime estrogen exposure; Menopause; Mortality; Women
1. Introduction Menopause indicates a loss in ovarian function and subsequent deficiency of endogenous estrogens. As endogenous estrogens exert a favorable
* Corresponding author. Tel.: /32-16-336-914; fax: /32-16337-015 E-mail address: [email protected] (E.H.M. Temme).
influence on lipoproteins, haemostasis and vasomotion, menopause is associated with an unfavorable change in these factors [1]. As a result, early age at natural menopause incurs a higher risk of mortality, in particular mortality of cardiovascular diseases [2 /8]. Although the effects of early menopause on mortality risk have been well documented in many studies, there are very few studies on the effects of the duration of exposure to endogenous estrogen
0378-5122/02/$ - see front matter # 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 3 7 8 - 5 1 2 2 ( 0 2 ) 0 0 1 8 3 - 4
106
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
on mortality risk. Lifetime duration of exposure to estrogen could be a stronger predictor of mortality, because it is a more precise quantification of duration of exposure to endogenous estrogen than menopausal age. One study found that a longer exposure to endogenous estrogen in premenopausal life (lasting from menarche until menopause) is protective against coronary heart disease mortality [9]. Other studies have investigated only the relationships between lifetime exposure to endogenous estrogen and specific diseases, mainly breast cancer [10] and osteoporosis [11]. As far as we know, the relation between lifetime estrogen exposure and all-cause mortality has not been reported before. We examined the relationships between menopausal age and death (all-cause mortality, mortality of cardiovascular diseases and mortality not caused by cardiovascular diseases) as well as the relationships between lifetime estrogen exposure and death in a cohort of 1462 Belgian women who had experienced a natural menopause. The data analyzed in the present study were obtained from the Belgian Interuniversity Research on Nutrition and Health (BIRNH) study.
ceptives, number of children, menopausal status, age at menopause and type of menopause (natural, surgical or other). Dietary habits were assessed according to the 24-h food record method. A total of 1705 women reported in the questionnaires a history of natural menopause. Data for 46 women were excluded because their menopausal or menarcheal age was unknown. An additional 197 observations were excluded because of missing values for at least one of the following variables: alcohol consumption, body mass index, presence of diabetes, education level, ever smoking, marital status, number of children, occupation, previous use of oral contraceptives or systolic blood pressure. As a result, 1462 women were included in the present analysis. Ten years after enrollment of each participants vital status and cause of death was retrieved. Follow-up for cause-specific mortality was completed satisfactorily in 99%. Causes of death were ascertained from the family doctor or the doctor who completed the death certificate. Death certificate coding was based on the International Coding of Diseases, Ninth Revision (ICD-9). Any one of the codes 390/459 was used as the definition of cardiovascular mortality.
2. Subjects and methods
2.2. Statistical analysis
2.1. Study population
Lifetime estrogen exposure was calculated for each woman by subtracting the menarcheal age from the age at menopause. The lifetime estrogen exposure was categorized into 33 /34, 36/37, 38 / 39 and ]/40 years. The reference group comprised women who had a lifetime estrogen exposure of 5/ 33 years. Menopausal age was categorized into 47/48, 49 /50, 51/52 and ]/53 years, with a reference group of women who became menopausal at age 5/46 years. Differences in characteristics between groups of lifetime estrogen exposure and between groups of menopausal age were tested with a Chi-square test or a Tukey test (a /0.05). Possible confounding variables were examined by assessing their association with mortality. Relationships between lifetime estrogen exposure and mortality of all causes, mortality of cardiovascular diseases (CVD) and mortality not caused by CVD were investigated
The aims, design, methodology, and results of the BIRNH study have been described in detail elsewhere [12,13]. Briefly, a randomized sample of the population was drawn from each of the 42 counties in Belgium by using the voting lists. The general population eligible for sampling was between the ages of 25 and 74. Of the 30 964 subjects sampled, a total of 11 302 subjects (5949 men and 5353 women) agreed to participate and was screened between 1980 and 1984. Major coronary and other risk factors were measured. Questions were asked on smoking habits, socialprofessional class, educational level and marital status. Further questions concerned the use of medications and personal medical antecedents. For women, questionnaires included items on age at menarche, past and present use of oral contra-
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
using logistic regression analysis. Age-adjusted and multivariate adjusted odds ratios (OR) with 95% confidence intervals (95% CI) were calculated by lifetime estrogen exposure group. Variables used to adjust odds ratios originated from the baseline examinations, e.g. the age at filling out the questionnaire. Relationships between age at menopause and mortality were investigated in a similar way. When analyzing cause-specific mortality, seven additional observations were excluded because of missing values for ICD-9 codes. Analyses were performed using SAS software (SAS institute, Cary, NC).
3. Results Results are reported for 1462 naturally postmenopausal women who were 37 /77 years of age at baseline. The mean age at natural menopause in these women was 49.4 years (range 31 /65 years), the mean age at menarche was 13.6 years (range 10 /19 years) and the mean lifetime estrogen exposure 35.8 years (range 13/51 years). The Pearson correlation coefficient of lifetime estrogen exposure with age at menopause was 0.9 and with age at menarche /0.03. After a total follow-up time of 14 000 person-years, 181 women (12.4%) had died, of whom 42% due to cardiovascular diseases, 54% due to non-cardiovascular causes and the remaining due to unknown causes. 3.1. Lifetime estrogen exposure The baseline characteristics of the population stratified by lifetime estrogen exposure group are given in Table 1. A higher lifetime estrogen exposure was statistically significantly associated with higher age, higher body mass index, higher systolic blood pressure and higher prevalence of diabetes mellitus. The menarcheal age, number of children, use of oral contraceptives and ever smoking were distributed almost similarly among groups of lifetime estrogen exposure. Table 3 shows the associations between lifetime estrogen exposure and all-cause mortality. The estimated odds ratio of mortality decreased with increasing categories of lifetime estrogen exposure.
107
After adjustment for possible confounding variables, the odds ratio of mortality for women in the highest lifetime estrogen exposure group was 0.58 (95% CI 0.35 /0.93) compared to women with the lowest lifetime estrogen exposure. A test for a linear trend across groups of lifetime estrogen exposure gave a multivariate adjusted P -value of 0.06. Further adjustments for alcohol consumption, body mass index, ever smoking, number of children, previous use of oral contraceptives, education level, marital status and occupation level did not change the odds ratio estimates substantially (data not shown). Additional analyses by cause of death are shown in Table 4. An inverse relation was observed between categories of lifetime estrogen exposure and mortality of CVD, which explained the largest part of the association between lifetime estrogen exposure and all-cause mortality. Confidence intervals however, all comprised 1.0, which could be due to the low numbers of deaths, particularly in the cause-specific analyses. 3.2. Menopausal age Table 2 gives the baseline characteristics of the population by group of menopausal age. Women with the highest age at natural menopause were significantly older and had significantly higher systolic blood pressure and higher prevalence of diabetes than women that were younger at menopause. Body mass index and number of children did also increase across categories of menopausal age, but not enough to be statistically significant. Age at menarche was slightly lower at higher ages at menopause. There was no clear trend for previous use of oral contraceptives or ever smoking. The P -value for a linear trend across groups of menopausal age was 0.05. The associations between menopausal age and all-cause mortality are shown in Table 3. Women with menopause at ages 47 /48 and 49/50 years had a slightly higher mortality risk than women in the reference group. For women with menopause above age 50, however, the mortality risk was decreased compared to the reference group. After adjustment for possible confounding variables, the odds ratio for women in the highest menopausal
108
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
Table 1 Baseline characteristics stratified by lifetime estrogen exposure (n /1462) Lifetime estrogen exposure (years) 5/33
34 /35
36 /37
38 /39
]/40
394 61 (9) 44 (4) 14 (2) 2 (2) 7 27 (4) 143 (22) 2 (4) 14 4
253 61 (8) 49 (1)1 14 (1) 2 (2) 10 27 (5) 143 (21) 2 (3) 13 2
277 61 (7) 50 (1)12 14 (1)12 2 (2) 8 27 (4) 141 (20) 2 (4) 14 3
250 61 (6) 52 (1)123 13 (1)123 2 (2) 8 28 (5) 145 (21) 2 (4) 13 4
288 63 (6)1234 54 (2)1234 13 (1)1234 3 (2) 6 28 (4)123 147 (22)3 2 (4) 13 8
Education level (%) Primary school More than primary school
65 35
63 37
63 37
58 42
62 38
Marital status (%) Married Not married Widowed Divorced
74 5 19 2
66 9 23 2
73 4 19 4
76 6 16 2
72 3 24 1
Occupation level (%) Low Intermediate High
11 60 29
15 55 30
10 64 26
17 54 28
7 63 30
9 (2)
10 (2)
10 (2)
10 (2)
10 (2)
16 8 8
15 5 10
11 4 6
10 4 5
12 6 6
n Age (years) Menopausal age (years) Menarcheal age (years) No. children Ever use of OC (%) BMI (kg/m2) SBP (mmHg) Alcohol intake (en%) Ever smoking (%) Diabetes mellitus (%)*
Follow up (years) No. deaths (/1000 py) All causes CVD Non-CVD
Values for continue variables are presented as mean (S.D.). Tukey test for difference from the first1, second2, third3 or fourth4 group (P B/0.05). OC, oral contraceptives; BMI, body mass index; SBP, systolic blood pressure; CVD, cardiovascular diseases. * P B/0.05 (Chi-square test) for differences among groups.
age group was 0.62 (95% CI 0.36 /1.03) compared to women with the lowest menopausal age. Similar patterns were seen in the associations between age at menopause and mortality of CVD, as well as mortality not caused by CVD (Table 4). When menopausal age was grouped into 5/50 and /50 years, the multivariate adjusted odds ratio for women with menopausal ages above 50 years was 0.57 (95% CI 0.40 /0.81) for all-cause mortality and 0.54 (95% CI 0.31 /0.92) for mortality of CVD, compared to women with lower menopausal ages. The odds ratio of mortality not
caused by CVD was 0.65 (95% CI 0.41 /1.002) (data not shown). These results indicate that mortality of CVD explained the major part of the associations between menopausal age and death.
4. Discussion In this prospective cohort study of 1462 naturally postmenopausal women, we found an odds ratio of 0.58 (95% CI 0.35 /0.93) for women with a
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
109
Table 2 Baseline characteristics stratified by menopausal age (n /1462) Menopausal age (years) 5/46
47 /48
49 /50
51 /52
]/53
299 61 (9) 13 (2) 30 (3) 2 (2) 6 28 (4) 144 (23) 2 (4) 14 5
217 60 (8) 14 (2)1 34 (2)1 2 (2) 8 27 (4) 140 (20) 2 (4) 14 1
382 61 (7) 14 (2) 36 (2)12 2 (2) 9 27 (4) 143 (21) 2 (3) 14 2
254 61 (6) 14 (2) 38 (2)123 3 (2) 8 28 (5) 144 (21) 2 (4) 14 6
310 63 (6)1234 14 (2) 41 (2)1234 3 (2) 7 28 (4) 146 (21)2 2 (4) 11 6
Eduction level (%) Primary school More than primary school
63 37
59 41
66 34
60 40
63 37
Marital status (%) Married Not married Widowed Divorced
74 6 18 2
72 5 21 2
69 7 20 3
76 3 19 2
73 5 21 1
Occupation level (%) Low Intermediate High
32 56 12
27 59 14
27 60 13
28 60 12
30 62 8
10 (2)
9 (2)
9 (2)
10 (1)
10 (2)
15 7 8
15 5 7
15 6 9
9 4 5
11 5 6
n Age (years) Menarcheal age (years) LEE (years) No. children Ever use of OC (%) BMI (kg/m2) SBP (mmHg) Alcohol intake (en%) Ever smoking (%) Diabetes mellitus (%)*
Follow up (years) No. deaths (/1000 py) All causes CVD Non-CVD
Values for continue variables are presented as mean (S.D.). Tukey test for difference from the first1, second2, third3 or fourth4 group (P B/0.05). LEE, lifetime estrogen exposure; OC, oral contraceptives; BMI, body mass index; SBP, systolic blood pressure; CVD, cardiovascular diseases. * P B/0.05 (Chi-square test) for differences among groups.
lifetime estrogen exposure of 40 or more years compared to women who had a lifetime estrogen exposure of 5/33 years. The decrease in the risk of mortality of CVD was (although not statistically significant) of the same magnitude. Kleijn et al. grouped lifetime estrogen exposure into 5/25 and /35 years, and found that the latter group had a 20% reduced risk of mortality of CHD (ageadjusted hazard ratio 0.80, 95% CI 0.68 /0.96) [9]. Our findings indicate a stronger inverse relationship between lifetime estrogen exposure
and mortality of CVD. As other studies have investigated only the relationships between lifetime estrogen exposure and specific diseases, we are not able to compare our results of the association between lifetime estrogen exposure and mortality with others. Considering the associations between age at menopause and death in the same population, the odds ratio of all-cause mortality was 0.62 (95% CI 0.36 /1.03) for women who became menopausal after the age of 53 years compared to women with
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
110
Table 3 Estimated odds ratio of all-cause mortality stratified by lifetime estrogen exposure (LEE) or menopausal age (MA), with 95% confidence interval n
Age-adjusted odds ratio
Multivariate adjusted* odds ratio
LEE (years) 5/33 34 /35 36 /37 38 /39 ]/40
394 253 277 250 288
1.00 0.96 0.74 0.63 0.64
(0.60 /1.54) (0.45 /1.20) (0.37 /1.06) (0.40 /1.02)
1.00 1.00 0.75 0.62 0.58
(0.61 /1.60) (0.45 /1.23) (0.35 /1.04) (0.35 /0.93)
MA (years) 5/46 47 /48 49 /50 51 /52 ]/53
299 217 382 254 310
1.00 1.16 1.06 0.68 0.64
(0.68 /1.97) (0.67 /1.67) (0.39 /1.18) (0.38 /1.06)
1.00 1.29 1.14 0.66 0.62
(0.75 /2.21) (0.72 /1.81) (0.36 /1.16) (0.36 /1.03)
*
Ajusted for age, presence of diabetes mellitus and systolic blood pressure.
menopausal ages of 5/46 years. This reduction was particularly due to a reduction in mortality of CVD. Other studies have shown associations of similar magnitude, for example relative risk ratios 1.3 or 1.6 for women with very early natural menopause (i.e. before age 40), compared to women aged 49/51 [7] or ]/50 [6] at menopause. Most studies investigating the relationships between age at menopause and death used menopausal age intervals of 5 years, with women aged
50/54 or ]/50 as reference group. We calculated 2-year intervals to have equally distributed numbers of subjects in each menopausal age group. Van der Schouw et al. found that each year’s delay in natural menopause gives a decrease in risk of mortality of CVD of 2% (age-adjusted hazard ratio 0.982, 95% CI 0.968 /0.996) [4]. This would incur a decrease of at least 14% for women with menopausal ages ]/53 years, compared to women who had their menopause before 46 years. Our
Table 4 Estimated odds ratio of cause-specific mortality stratified by lifetime estrogen exposure (LEE) or menopausal age (MA), with 95% confidence interval n
Age-adjusted odds ratio
Multivariate adjusted* odds ratio
CVD
non-CVD
CVD
non-CVD
LEE (years) 5/33 394 34 /35 253 36 /37 277 38 /39 250 288 ]/40
1.00 0.62 0.59 0.57 0.69
(0.29 /1.26) (0.27 /1.21) (0.24 /1.21) (0.35 /1.30)
1.00 1.34 0.89 0.70 0.72
(0.74 /2.39) (0.47 /1.66) (0.33 /1.37) (0.38 /1.33)
1.00 0.65 0.59 0.54 0.57
(0.30 /1.34) (0.27 /1.22) (0.23 /1.16) (0.28 /1.10)
1.00 1.34 0.90 0.69 0.71
(0.74 /2.40) (0.47 /1.67) (0.33 /1.36) (0.37 /1.31)
MA (years) 5/46 299 47 /48 217 49 /50 382 254 51 /52 ]/53 310
1.00 0.89 0.87 0.61 0.61
(0.40 /1.91) (0.45 /1.67) (0.25 /1.34) (0.29 /1.23)
1.00 1.10 1.20 0.70 0.74
(0.54 /2.18) (0.67 /2.16) (0.33 /1.44) (0.38 /1.41)
1.00 1.09 1.00 0.54 0.56
(0.48 /2.39) (0.51 /1.95) (0.22 /1.24) (0.26 /1.15)
1.00 1.12 1.21 0.70 0.73
(0.55 /2.23) (0.68 /2.19) (0.32 /1.43) (0.38 /1.41)
*
Ajusted for age, presence of diabetes mellitus and systolic blood pressure. CVD, cardiovascular diseases.
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
results suggest that the decrease in the risk of mortality of CVD, caused by a delay in natural menopause, might be even stronger. The risk estimates obtained in our study for the lifetime estrogen exposure groups were not substantially different from those of the menopausal age groups. This could be due to the low variability in age at menarche (95% between 11 and 16 years), compared to the variability in age at menopause (95% between 42 and 55 years), which makes the age at menopause the most powerful determinant of the lifetime estrogen exposure. The low correlation between lifetime estrogen exposure and age at menarche (Pearson correlation coefficient of /0.03) also points at a weak influence of menarcheal age. Our hypothesis was that lifetime estrogen exposure could be a stronger predictor of mortality than menopausal age, because it is a more precise quantification of duration of exposure to endogenous estrogen. However, to make an accurate assessment of the total duration of estrogen exposure, one should probably also take into account duration of breast feeding, number of abortions and miscarriages, length and regularity of menstrual cycles, and use of hormonal replacement therapy [14]. We had no information on these issues, and used only the age at menarche and menopause to calculate the lifetime estrogen exposure. We had information on the number of children and the use of oral contraceptives, but they were similar between the groups of lifetime estrogen exposure. Therefore, any misclassification seems to be non-differential. One could debate that the term lifetime estrogen exposure is inappropriate in the context of our study, because we had no adequate information to make a really accurate assessment of the cumulative exposure to endogenous estrogen. However, we followed the terminology commonly used in the literature. Most studies investigating the relationship between lifetime estrogen exposure and specific diseases used the formula age at menopause minus age at menarche [11], or the time from menarche to menopause minus the duration of pregnancies and breastfeeding [10]. Our study indicates that, when assessing mortality risk, these
111
measures are not superior to menopausal age as such. Biases could have led to incorrect effect estimates. Our study sample could have been healthier than the general Belgian population of postmenopausal women, because women had to be healthy enough to attend baseline examinations. However, this would probably have led to an underestimation of a possible inverse relationship between menopausal age or lifetime estrogen exposure and mortality. It should be noted that the participation rate in the BIRNH study was only 37%. Therefore, an additional random sample of 10% of the nonrespondents was screened. Reproductive history was not checked, but in dietary habits no significant differences were found between respondents and non-respondents [15]. Furthermore, the BIRNH study used a questionnaire to assess reproductive history, which could have led to biased reporting of age at menopause and menarche. But, as reproducibility and validity studies have indicated, most women are able to report the age at menopause and even the age at menarche with a high degree of accuracy [4,16,17]. Possible confounding variables were evaluated in our analysis by assessing their association with mortality. Since age is a major risk factor for mortality, all analyses were adjusted for age. Although systolic blood pressure could be involved in the biological mechanism of the effect of estrogens, the higher systolic blood pressure in women with a higher menopausal age or longer lifetime estrogen exposure can not be explained by this mechanism. When there would be an effect of estrogen, women with a higher menopausal age or longer lifetime estrogen exposure would rather have a lower systolic blood pressure [18]. The higher systolic blood pressure in these women might therefore be due to their higher age. Smoking reduces the age at natural menopause [19]. However, in our study, the percentage of women who had ever smoked did not significantly differ among groups of menopausal age and among groups of lifetime estrogen exposure. We could not evaluate the possible confounding effect of physical activity, stress and family history of CVD or cancer, because information on these variables was not available. Adjustments for alcohol con-
112
S.C. Jansen et al. / Maturitas 43 (2002) 105 /112
sumption, body mass index, ever smoking, menarcheal age, number of children, previous use of oral contraceptives, education level, marital status and occupation level did only marginally influence the odds ratios estimates. Although we had information on lipid concentrations, it would be inappropriate to adjust for them, because they could be involved in the biological mechanism of the effect of estrogens. In conclusion, this cohort study of 1462 naturally postmenopausal women indicates that menopausal age predicts mortality, especially mortality of CVD. The prediction is not improved by adding the age at menarche, to obtain an estimate for lifetime estrogen exposure.
Acknowledgements This study was supported by a grant from the Unilever Chair in Nutritional Epidemiology.
References [1] Kuller LH, Meilahn EN, Cauley JA, Gutai JP, Matthews KA. Epidemiologic studies of menopause: changes in risk factors and disease. Exp Gerontol 1994;29:495 /509. [2] Lapidus L, Bengtsson C, Lindquist O. Menopausal age and risk of cardiovascular disease and death. Acta Obstet Gynecol Scand Suppl 1985;130:37 /41. [3] Snowdon DA, Kane RL, Beeson WL, et al. Is early natural menopause a marker of health and aging? Am J Public Health 1989;79:709 /14. [4] Schouw van der YT, Graaf van der Y, Steyerberg EW, Eijkemans MJC, Banga JD. Age at menopause as a risk factor for cardiovascular mortality. Lancet 1996;347:714 / 8. [5] Jacobsen BK, Nilssen S, Heuch I, Kvale G. Does age at natural menopause affect mortality from ischemic heart disease? J Clin Epidemiol 1997;50:475 /9. [6] Cooper GS, Sandler DP. Age at natural menopause and mortality. Ann Epidemiol 1998;8:229 /35.
[7] Jacobsen BK, Knutsen SF, Fraser GE. Age at natural menopause and total mortality and mortality from ischemic heart disease: the Adventist Health Study. J Clin Epidemiol 1999;52:303 /7. [8] Cooper GS, Baird DD, Weinberg CR, Ephross SA, Sandler DP. Age at menopause and childbearing patterns in relation to mortality. Am J Epidemiol 2000;151:620 /3. [9] Kleijn de MJ, Schouw van der YT, Peeters PH, Banga JD, Verbeek AL, Graaf van der Y. Number of menstrual cycles is associated with coronary heart disease mortality in postmenopausal women (abstract). Circulation 1999;99:1111. [10] Adami HO, Bergstrom R, Lund E, Meirik O. Absence of association between reproductive variables and the risk of breast cancer in yong women in Sweden and Norway. Br J Cancer 1990;62:122 /6. [11] Nguyen TV, Jones G, Sambrook PN, White CP, Kelly PJ, Eisman JA. Effects of estrogen exposure and reproductive factors on bone mineral density and osteoporotic fractures. J Clin Endocrinol Metab 1995;80:2709 /14. [12] Kesteloot H. Nutrition and health: the conclusions of the BIRNH study. For the BIRNH study group. Acta Cardiol 1989;44:183 /94. [13] De Backer G. Regional differences in dietary habits, coronary risk factors and mortality rates in Belgium. I. Design and methodology. Acta Cardiol 1984;39:285 /92. [14] Kleijn de MJJ, Schouw van der YT, Graaf van der Y. Reproductive history and cardiovascular disease risk in postmenopausal women. A review of the literature. Maturitas 1999;33:7 /36. [15] Kornitzer M, Dramaix M. The Belgian Interuniversity Research on Nutrition and Health (BIRNH): general introduction for the BIRNH study group. Acta Cardiol 1989;44:89 /99. [16] Livson N, McNeil D. The accuracy of recalled age at menarche. Hum Biol 1962;34:218 /21. [17] Bean JA, Leeper JD, R.B. W, Sherman BM, Jagger H. Variations in the reporting of menstrual histories. Am J Epidemiol 1979;109:181 /185. [18] Staessen JA, Ginocchio G, Thijs L, Fagard R. Conventional and ambulatory blood pressure and menopause in a prospective population study. J Hum Hypertens 1997;11:507 /14. [19] Willet W, Stampfer MJ, Bain C, Lipnick R, Speizer FE, Rosner B. Cigarette smoking, relative weight, and menopause. Am J Epidemiol 1983;117:651 /8.