Blood pressure and metabolic profile after surgical menopause - Nature

Journal of Human Hypertension (2000) 14, 799–805  2000 Macmillan Publishers Ltd All rights reserved 0950-9240/00 $15.00 www.nature.com/jhh

ORIGINAL ARTICLE

Blood pressure and metabolic profile after surgical menopause: comparison with fertile and naturally-menopausal women E Casiglia1, G Ginocchio2, V Tikhonoff1, D D’Este3, A Mazza1, A Pizziol1, A Pavei1, GB Ambrosio4, A Piccoli5 and AC Pessina1 1

Department of Clinical and Experimental Medicine, University of Padova, Italy; 2Department of Cardiology, Hospital of Castelfranco Veneto, Italy; 3Department of Cardiology, Hospital of Mirano, Italy; 4 Department of Medicine, General Hospital of Venice, Italy; 5Department of Medical and Surgical Sciences, University of Padova, Italy

In 1978 a random sample (367 men and 568 women aged 18–65 years) taken from the general population of a north-eastern Italian town was screened for cardiovascular risk; 16 years later, the women were invited to a second screening. Three groups were identified at the initial screening (fertile, naturally menopausal and surgically menopausal) and four in the longitudinal study (137 remained fertile during the whole study, 205 became naturally menopausal, 56 were ovariectomised and 127 were already going through the menopause). The protocol included a questionnaire, blood pressure (BP) measurement, and blood exams. Continuous variables were adjusted for confounders. Systolic BP, prevalence of hypertension, cholesterol, glycaemia and uricaemia were similar, whereas diastolic and triglycerides (TG) were lower in surgically-menopausal than in fertile women (P ⬍ 0.001). No significant difference in 16 years’ variation from baseline was observed between

the four groups, although women who remained fertile showed the smallest increases. In particular, neither systolic or diastolic BP increases differed between the women who were oophorectimised and those who remained fertile. ‘Fertile status’ was rejected from the logistic equation of incidence of hypertension, and ‘age of menopause’ was also rejected when this analysis was repeated in ovariectomised women. New coronary artery disease (angina pectoris or myocardial infarction) was observed in one ovariectomised woman, in three naturally menopausal, and in 13 already menopausal women which seemed to reflect the age trend. No new cases were observed in women who remained fertile. In conclusion, in Italian women surgical menopause, similarly to natural menopause, is devoid of any negative prognostic effect. Journal of Human Hypertension (2000) 14, 799–805

Keywords: human; epidemiology; blood pressure; menopause; oophorectomy

Introduction About 95% of women in industrialised areas remain more than one-third of their life in menopause.1,2 If menopause were a cardiovascular risk factor, it could be defined as the most incident one. However, the question whether cardiovascular changes appearing in women in later life are attributable to their hormonal status2–12 rather than to advancing age13–16 is still debatable. Another question is whether the excess cardiovascular risk is greater after surgical than natural menopause. The general belief is that oophorectomy abolishes sexual dimorphism and causes women to develop a male pattern of blood pressure17 and risk,18,19 an effect which appears to be prevented by oestrogen-replacement therapy.20 On the other hand, Correspondence: Dr E Casiglia, Department Clinical Experimental Medicine, University of Padova Via Giustiniani, 2 35128 Padova, Italy. E-mail: casiglia얀uxl.unipd.it Received 29 March 2000; revised 19 June 2000; accepted 21 June 2000

some authors believe that surgical menopause exerts an effect only on cholesterol21 or LDL fraction,22 or no effect at all on lipids,23 and the role of sudden ovarian failure is still unclear. To address this issue, the possible association between surgical menopause, blood pressure (BP) increase, and worsening metabolic pattern was studied closely together with a comparison with the effects of natural menopause.

Materials and methods Subjects In 1978, a random sample of all the inhabitants of Mirano (VE) aged from 18 to 60 years was identified and screened for cardiovascular risk factors; 2000 randomly-chosen subjects were taken from the general population (obtained from the Registrar’s Office); 1887 of them (952 men and 935 women) had given their informed consent prior to the study and took part in the investigation.24 Sixteen years later,

Surgical menopause in general population E Casiglia et al

800

the whole baseline population was invited to a second screening.25 In the meantime, some of the women had moved, or their data were incomplete, and very few had been placed on hormonal replacement therapy, thereby leaving 568 subjects to participate in the second branch of the study. The general characteristics of the 367 excluded and the 568 women included did not differ significantly (Table 1). Cross-sectional and a 16-year longitudinal analyses were then performed. Three groups were identified at the initial screening (fertile, naturally menopausal, and surgically menopausal) and four in the longitudinal study (the women who remained fertile during the whole study, those who became naturally menopausal, those who became surgically menopausal, and those who were already in menopause at baseline).

Table 2 Cut-off values and definitions of disease

Data collection All women were examined both at the initial screening and 16 years later according to a standardised protocol24 after overnight fasting at Mirano General Hospital, where a special outpatient clinic had been made available. The visit included a blood sample to evaluate biochemical parameters, anthropometrics, a detailed questionnaire, and three BP readings at 5-min intervals after an initial 15-min rest in the sitting position (diastolic phase 5). The mean of the last two BP readings was employed for an analysis of data. The questionnaire used for the interview26 investigated personal and familial medical history, current occupation, smoking habits, diet, consumption of beverages containing caffeine and alcohol, intake of medication including antihypertensive drugs, and the presence vs absence of angina pectoris. Menstrual status, type (surgical or natural) and length of menopause and the use of oral contraceptives were also recorded. Body mass index (BMI) was calculated as the ratio between weight (kg) and squared height (metres). Standard electrocardiogram was performed and interpreted with the Minnesota code by a cardiologist who did not know the protocol and the aims of the study.

Table 1 Baseline characteristic of women recruited to initial screening Items

Women excludeda (n = 367)

Women included (n = 568)

Height (cm) 160.4 ± 6.3 160.1 ± 5.6 Weight (kg) 62.9 ± 11.1 64.1 ± 11.6 Systolic BP (mm Hg) 134.9 ± 20.2 137.3 ± 21.6 Diastolic BP (mm Hg) 81.7 ± 12.3 83.2 ± 12.3 Total cholesterol (mg/dl) 205.2 ± 45.4 208.1 ± 42.9 Triglycerides (mg/dl) 92.3 ± 62.4 91.7 ± 5.2 Blood glucose (mg/dl) 95.6 ± 14.7 94.7 ± 15.5 Uric acid (mg/dl) 45.3 ± 17.2 44.8 ± 10.8 Blood urea (mg/dl) 34.1 ± 29.8 32.9 ± 7.7 a

P values

NS NS NS NS NS NS NS NS NS

Women who had moved in the meantime, with incomplete data, or on hormonal replacement therapy.

Journal of Human Hypertension

Arterial hypertension

Borderline: systolic BP ⭓140 and ⬍160 mm Hg, or diastolic ⭓90 and ⬍95 mm Hg, and untreated. Sustained: systolic ⭓160 or diastolic ⭓95 mm Hg or treated

Diabetes mellitus

Blood glucose ⭓126 mg/dl or antidiabetic treatment

Left ventricular hypertrophy

Minnesota code 8.1 or 8.3

Historical myocardial infarction

Minnesota code 1.1, 1.2 or 1.3 (being absent 6.4.1) or positive history in hospital files or akinesia/diskinesia at echocardiogram or angiography

Historical angina pectoris

Typical chest pain with Minnesota code 4.1, 4.2, 5.1 or 5.2 (being absent 6.4.1, 7.1.1, 7.2.1, 7.4), or a positive scintigraphy or angiography

History of stroke

Cerebrovascular episode with permanent outcomes (motor or sensor loss, aphasia, dysarthria)

Proteinuria

Frank: ⬎200 mg/l, traces: 30–200 mg/l in morning sample

Statistics Continuous variables were averaged and expressed as mean and standard error. In the cross-sectional analysis, a between-group comparison was performed with an analysis of covariance after adjustment for age and BMI due to the differences in these two parameters in the different groups. In the longitudinal analysis, the difference between the baseline and the 16th year was further adjusted for the baseline values. Multivariate incidence of events was compared with logistic regression,27 including age, BP, total cholesterol (TC) and triglycerides (TG), BMI, serum uric acid, history of angina pectoris or myocardial infarction, history of stroke, left ventricular hypertrophy, proteinuria, diabetes and age of menopause as covariates (see Table 2 for criteria and cutoff values). As TC and TG were directly correlated (r = 0.38, P ⬍ 0.0001), the bivariate distribution of TC and TG were considered together as a bivariate gaussian random vector (vector {TC, TG}). The 95% confidence intervals (95% confidence ellipses) of the {TC, TG} vector for women who remained fertile during the follow-up and those who were ovariectomised were calculated according to formulae included in the Appendix. Statistical testing for significance between {TC, TG} vectors was performed with Hotelling T2 test.28

Results Cross-sectional analysis of the initial survey At the initial screening, 398 women were fertile and 127 in menopause (102 natural and 25 surgical). A


Table 3 General characteristics of the population at the initial survey. Only the 525 women having paired data at baseline and at the end of follow-up are included. Mean (standard errors) Fertile women (n = 398) Age (years) BMI (kg/m2) Systolic BP (mm Hg)a Diastolic BP (mm Hg)a Serum total cholesterol (mg/dl)a Serum triglycerides (mg/dl)a Blood glucose: fasting (mg/dl)a after 75-g load (mg/day)a Prevalence of diabetes mellitus (%) Serum uric acid (mg/dl)a Coffee consumption (cups/day) Alcohol consumption (ml/day) Cigarette smoking (%)

38.9 24.6 137.5 83.6 206.1 90.2

(0.4) (0.2) (0.9) (0.5) (1.9) (2.5)

95.7 (0.7) 97.7 (1.4) 1.8 4.48 (0.08) 2.5 (0.1) 27 (0.4) 18.4

Surgically menopausal women (n = 25)

Naturally menopausal women (n = 102)

49.8 (1.1)* 25.5 (0.7) 133.5 (3.7) 8.16 (1.2) 196.4 (7.7) 79.0 (10.1)*

53.1 (0.4)*† 26.6 (0.4)* 137.8 (2.1) 82.9 (2.1)* 218.2 (4.4)* 97.8 (5.7)*†

94.3 (2.9) 103.4 (6.1) 0 4.55 (0.12) 2.0 (0.1) 26 (1.2) 4

92.7 (1.7) 103.1 (3.0) 2.9† 4.28 (0.21)*† 2.0 (0.2) 26 (0.5) 12.7†

801

Variables adjusted for confounders. *P ⬍ 0.001 vs fertile, †P ⬍ 0.001 vs surgically menopausal. BMI: body mass index. Age and BMI are expressed as crude values, the other continuous variables age adjusted for age and BMI.

a

further 43 women had died during the follow-up period; as the present study is based on paired data at baseline and at the end of follow-up, these 43 women were not included in the analysis. There was not one single drop out. In Table 3, the general characteristics of the whole population according to fertility status are shown. Age was lowest in fertile and highest in naturally-menopausal women, while in the surgically menopausal it was intermediate. BMI paralleled age. The prevalence of hypertension was similar in fertile and oophorectomised women, while it was significantly higher in naturally-menopausal women than in the other two categories (Figure 1). Age- and BMI-adjusted systolic BP was similar in the three classes; age (P ⬍ 0.0001) and BMI (P ⬍ 0.0001) explained all the apparent differences seen during univariate analysis. Diastolic BP was even significantly lower in surgically-menopausal than in fertile

Figure 1 Cross-sectional analysis. Prevelance of borderline and sustained arterial hypertension among fertile (䊐, n = 441), naturally menopausal (䊏, n = 102) and surgically menopausal (a, n = 25). See text for definition of borderline and sustained hypertension. Numbers indicate P values.

women (P ⬍ 0.0001), despite the confounding effect of age (P ⬍ 0.0001) and BMI (P ⬍ 0.0001); the difference was biologically irrelevant, however (2%). The same was true for serum TG, which were 12% lower in ovariectomised than in fertile women (effect of age: P ⬍ 0.005; effect of BMI: P ⬍ 0.0001). Serum TC, blood glucose and serum uric acid were similar in both oophorectomised and fertile females. Longitudinal study Information on menopausal status: During the 16 years of follow-up, 137 women remained fertile, 56 were ovariectomised (mean length of menopause 8.2 years), 205 went through natural menopause (mean length 19.7 years, P ⬍ 0.05 vs ovariectomised) and 127 were already in menopause at the initial survey (mean length 20 years, P ⬍ 0.05 vs ovariectomised). As mentioned above, a further 43 women had died and consequently were not included in the longitudinal study; none of them was in the group of the oophorectimised. Trend of BP and other items during the followup: No significant difference in 16 years BP variation from baseline was observed among the four groups, although women who were still fertile showed the lowest increments (Figure 2). No BP difference was found between baseline and the end of follow-up in the 56 women who were oophorectimised. In the 56 women who underwent oophorectomy, TC and TG were higher at the 16th year than at the initial survey (Table 4). This could be due both to the effect of menopause or, more simply, to the effect of advancing age. As paired data do not allow age-adjustment, to clarify this point two separate analyses were performed: (1) a multivariate regression adjusted for age between the 16-year increment of lipids and the length of menopause, and (2) a vectorial analysis comparing the displacement of the confidence ellipses of lipid pattern {TC, TG} built separately for the women remaining fertile Journal of Human Hypertension


802

Figure 2 Longitudinal analysis. Variation (% from initial survey) of systolic (SBP) and diastolic blood pressure (DBP), total cholesterol (TC), blood glucose (BG) and serum uric acid in women who remained fertile (n = 137), in those who went through oophorectomy (n = 56), in those who went through natural menopause (n = 205) and in those who were already menopausal at baseline (n = 127).

Table 4 Blood pressure and lipid values at baseline and at the end of follow-up in the 56 women who were oophorectimised Items Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Total cholesterol (mg/dl) Triglycerides (mg/dl)

Baseline

Follow-up

136.1 ± 20.1

141.1 ± 16.2

85.1 ± 11.2

86.8 ± 7.7

197.8 ± 37.9 89.7 ± 37.4

239.2 ± 39.7* 111.6 ± 51.8*

*P ⬍ 0.001 vs baseline.

and for those who were oophorectomised (Figure 3). The regression showed no correlation (TC: r = 0.14, NS; TG: r = −0.04, NS), provisionally indicating that menopause was not a determinant of lipid increase. On the contrary, the vectorial analysis showed a significantly greater displacement of the confidence ellipse towards higher values of both TC and TG in oophorectomised women than in those remaining fertile; this could indicate that a significant effect of menopause on lipids was added to that of age (see discussion). Analysis of new cases of disease: At univariate analysis, the incidence of new cases of arterial hypertension was apparently higher in ovariectomised (40%) than in fertile women (26.8%). However, this simply reflected the confounding effect of age. In fact, ‘fertile status’ as an independent variable (fertility, oophorectomy, natural menopause) was rejected from the logistic equation of incidence of hypertension, whereas age (␤ = 0.39, P ⬍ 0.001) and BMI (␤ = 0.89, P ⬍ 0.05) were the only multivariate predictors. ‘Age of menopause’ as a continuous variable was also rejected when this analysis was repeated in oophorectomised women. New cases of coronary artery disease (angina pectoris or myocardial infarction) were observed in one woman who underwent oophorectomy, in three who Journal of Human Hypertension

Figure 3 Vector analysis of the ‘total cholesterol (TC), tryglicerides (TG)’ vector. The 95% ellipses of confidence of the oophorecomised women (clear) and of those who remained fertile (dashed) are not overlapping at baseline (B) and at the end of follow-up (F), indicating a univariate effect of surgical menopause directed towards a mild increase of both TC and TG.

became naturally menopausal and in 13 who were already menopausal at the initial screening (10.2%). Once again, this simply reflected the trend of age. No new cases were observed in women who remained fertile.

Discussion The question whether menopause is per se a cardiovascular risk factor is still under debate. The majority of studies have tackled this topic according to a cross-sectional analysis. Some authors have found higher levels of BP,18,29 serum lipids,18,21


circulating adrenaline18 and plasma fibrinogen,30 as well as endothelial dysfunction,31 in naturallymenopausal in comparison to fertile women. In other surveys, on the contrary, no unfavourable effect on blood lipids25,32 or BP21,32,33 have been observed. Being as, at a population level, the naturally-menopausal women are generally older than the fertile ones, there is a possibility that the presumptive unfavourable effects attributed to menopause are simply due to the more advanced age of menopausal women compared with fertile women; this is certainly the case for Italian women.25,33 Based on the results of the present study (in which the women taking replacement therapy were purposely excluded to avoid any exogenous intereference), we can add the information that this is also true for surgical menopause. In fact, the prevalence of hypertension showed no difference in oophorectomised or fertile women, and correction for age and BMI eliminated any apparent worsening effect of oophorectomy on BP, atherogenic lipids, and blood glucose. The few longitudinal studies on natural menopause that are available were unable demonstrate any direct effect of the decreased ovarian function on cardiovascular mortality23,25 or morbidity,25 nor on the development of arterial hypertension25,32 or hypercholesterolemia.25,32 Two epidemiological population-based experiences have been conducted in Italy on natural menopause: one published a few years ago by our group25 and one by Portaluppi et al.33 In the former, the univariate risk excess observed in naturally-menopausal women when compared to fertile women was simply due to the higher age of the former, and disappeared after age-correction or age-matching.25 In the latter, the unfavourable risk profile of naturallymenopausal women was related to age (not to the fertility status or length of menopause), and was no longer detectable after age-correction.33 The experience of Matthews et al34 is in keeping with this evidence which showed that after statistical adjustment for age, fertile and menopausal women no longer differed in relation to TC level. The data outlined herein demonstrate that not only naturally-occurring, but also surgicallyinduced menopause have no influence whatsoever on cardiovascular risk. In fact, no significant difference in BP was detectable during the 16 years of follow-up between fertile and oophorectomised women, and the difference in lipids (similar to that described in Japanese women)21 was probably only apparent, as it did not correlate with length of menopause. At this point of the analysis, age increase was the best candidate for explaining the worsening of the lipid pattern. In fact, a comparable increase in plasma lipids was observed in women who remained fertile during the same follow-up period (Figure 2), demonstrating that age increase – rather then menopause – was responsible for it. Only the vectorial analysis (Figure 3) seemed to demonstrate an additional effect of menopause on lipids, in addition to that of age. In fact, when serum TC and TG were considered together, the displacement toward higher levels of the {TC, TG} confidence

ellipses was more pronounced in the oophorecomised women (clear ellipses) than in those who remained fertile during the follow-up (dashed ellipses); this was entirely attributable to TC (+25 mg/dl), whereas for TG no further increment due to oophorectomy in comparison to age increase was found (Figure 3). Unfortunately, the vectorial analysis does not allow any covariance, eg, adjustment for age. Stevenson et al35 recently found that the lipid increase associated to menopause was independent of age. However, their work was not population-based, as only women who had volunteered to act as controls for studies of oral contraceptives were studied, so that any comparison with our data is impossible. In any case, even if the worsening of lipid pattern was due in part to surgical menopause, it is unlikely that it was clinically relevant. We were unable to analyse mortality, simply because none of the ovariectomised women had died during the follow-up, but the analysis of morbidity showed a trend reflecting that of age, with a low incidence of new cases of coronary artery disease in the women who were oophorectomised during the follow up. This is in agreement with the belief that the passage through surgical climaterium is devoid of any effect on prognosis. Accordingly, Cooper and Sandler23 found no age-related increase in mortality among the women of the NHANES who had bilateral oophorectomy. Only the Scandinavian researchers operating in Malmo36,37 found an increase in the frequency of coronary heart disease in women after surgical menopause. In conclusion, surgical menopause in Italian women, similarly to natural menopause, is probably devoid of any direct negative prognostic effect. It does not worsen BP or glycaemic control, while its effects on lipid pattern are uncertain, perhaps due to age advancement, and far from being confirmed.

803

References 1 Zachariasen RD. Oral manifestations of menopause. Compend Contin Educ Dent 1995; 12: 1584 –1591. 2 Sitruk-Ware R, Ibarra de Palacios P. Estrogen replacement therapy and cardiovascular disease in postmenopausal women: a review. Maturitas 1989; 11: 259–274. 3 Ayalon D, Pines A. Cardiovascular disease and hormone replacement therapy: a review. Isr J Med Sci 1993; 29: 660–663. 4 Ettinger B. Hormone replacement therapy and coronary heart disease. Obstet Gynecol Clin North Am 1990; 17: 741–757. 5 Belchetz PE. Hormonal treatment of postmenopausal women. N Engl J Med 1994; 330: 1062–1071. 6 Ravn SH, Rosenberg J, Bostofte E. Postmenopausal hormone replacement therapy – clinical implications. Eur J Obstet Gynecol Reprod Biol 1994; 53: 81–93. 7 Knopp RH. Cardiovascular effects of endogenous and exogenous sex hormones over a women’s lifetime. Am J Obstet Gynecol 1988; 158: 1630–1643. 8 Radwanska E. The role of reproductive hormones in vascular disease and hypertension. Steroids 1993; 58: 605–610. 9 Lindquist O. Influence of the menopause on ischaemic heart disease and its risk factors and on bone mineral Journal of Human Hypertension


804

10

11

12

13 14

15

16 17 18

19 20 21 22 23 24

25

26 27 28 29

30

content. Acta Obstet Gynecol Scand 1982; 110 (Suppl): 7–32. Lindquist O, Bengtsson C. Serum lipids, arterial BP and body weight in relation to the menopause: results from a population study of women in Go¨teborg, Sweden. Scan J Clin Lab Invest 1980; 40: 629–636. Chabanel A, Chien S. Blood viscosity as a factor in human hypertension. In: Laragh JH, Brenner BM (eds). Hypertension: Pathophysiology, Diagnosis, and Management. Vol. 1. Raven Press: New York, 1990, pp 329–337. Yarnell JWG et al. Fibrinogen, viscosity, and white blood cell count are major risk factors for ischemic heart disease. The Caerphilly and Speedwell Collaborative Heart Disease Studies. Circulation 1991; 83: 836–844. Hjortland MC, Mc Namara PM, Kannel WB. Some atherogenic concomitants of menopause: the Framingham study. Am J Epidemiol 1976; 103: 304 –311. Lindquist O. Intraindividual changes of BP, serum lipids, and body weight in relation to menstrual status: results from a prospective population study of women in Go¨teborg, Sweden. Prev Med 1982; 11: 162–172. Van Beresteyn ECH, Van’t Hof MA, De Waard H. Contributions of ovarian failure and aging to BP in normotensive perimenopausal women: a mixed longitudinal study. Am J Epidemiol 1989; 129: 947–955. Van Beresteijn ECH et al. Habitual dietary calcium intake and BP change around the menopause: a longitudinal study. Int J Epidemiol 1992; 21: 683–689. Calhoun DA, Oparil S. High blood pressure in women. Int J Fertil Womens Med 1997; 42: 198–205. Stoney CM, Owens JF, Guzick DS, Matthews KA. A natural experiment on the effects of ovarian hormones on cardiovascular risk factors and stress reactivity: bilateral salpingo oophorectomy versus hysterectomy only. Health Psychol 1997; 16: 349–358. Witteman JC et al. Increased risk of atherosclerosis in women after the menopause. BMJ 1989; 298: 642–644. Colditz GA et al. Menopause and the risk of coronary heart disease in women. N Engl J Med 1987; 316: 1105–1110. Akahoshi M et al. Effects of menopause on trends of serum cholesterol, blood pressure, and body mass index. Circulation 1996; 94: 61–66. Farish E, Fletcher CD, Hart DM, Smith ML. Effects of bilateral oophorectomy on lipoprotein metabolism. Br J Obstet Gynaecol 1990; 97: 78–82. Cooper GS, Sandler DP. Age at natural menopause and mortality. Ann Epidemiol 1998; 8: 229–235. Gruppo di Ricerca ATS-RF2. I fattori di rischio dell’arteriosclerosi in Italia. La Fase A del Progetto Finalizzato del C.N.R. Medicina preventiva – Aterosclerosi – RF2. G It Cardiol 1980; 10: 1–184. Casiglia E et al. Lack of influence of menopause on blood pressure and cardiovascular risk profile: a 16year longitudinal study concerning a cohort of 568 women. J Hypertens 1996; 14: 729–736. Rose GA, Blackburn H. Cardiovascular Survey Methods. WHO, Geneva, 1968. Cox DR. Regression models and life tables. J Roy Statist Soc 1972; 34 (Series B): 187–220. Morrison DF. Multivariate Statistical Methods. McGraw-Hill: New York, 1967 (page 120). 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: 507–514. Punnonen R et al. Impaired ovarian function and risk factors for atherosclerosis in premenopausal women. Maturitas 1997; 27: 231–238.


31 Staessen JA, Celis H, Fagard R. The epidemiology of the association between hypertension and menopause. J Hum Hypertens 1998; 12: 587–592. 32 Shelley JM et al. Relationship of endogenous sex hormones to lipids and blood pressure in mid-aged women. Ann Epidemiol 1998; 8: 39– 45. 33 Portaluppi F, Pansini F, Manfredini R, Mollica G. Relative influence of menopausal status, age, and body mass index on blood pressure. Hypertension 1997; 29: 976–979. 34 Matthwes KA et al. Menopause and risk factors for coronary heart disease. N Engl J Med 1989; 321: 641–646. 35 Stevenson JC, Crook D, Godsland IF. Influence of age and menopause on serum lipids and lkipoproteins in heathly women. Atherosclerosis 1993; 98: 83–90. 36 Svanberg L. Effects of estrogen deficiency in women ovariectomised when young. Acta Obstet Gynecol Scand Suppl 1981; 106: 11–15. 37 Johansson BW et al. On some late effects of bilateral oophorectomy in the age range 15–30 years. Acta Obstet Gynecol Scand 1975; 54: 449– 461.

Appendix For those readers who are less familiar with multivariate analysis, we report the basic definitions of multivariate confidence intervals as well as formulae for their calculations in the particular case of the assumption of the bivariate normal distribution for blood lipids vector (TG on x axis, TC on y axis). The confidence interval is the inferential statistical interval for a given parameter (such as a mean value) investigated. It is the region in the parameter space to which the probability 100 is assigned (1␣)% (␣ is some fixed probability, typically 0.05) in which the parameter vector lies within. The confidence interval of the mean of the univariate normal distribution is formed by two values (limits), while the interval of the mean vector of the multinormal distribution is an ellipsoid centred at the mean vector, which reduces to a hypersphere when the correlation coefficients between pairs of variables are zero.28 When the confidence ellipsoids of two mean vectors overlap, the null hypothesis of equality of the two mean vectors cannot be rejected with the significance level ␣ (ie, non-significant Hotelling’s T2 test). The confidence interval becomes smaller with increasing sample size, and in a very large population the confidence interval converges to the parameter vector (ie, the mean vector point). The 95% confidence ellipses of several BP mean (bivariate) vectors are depicted in Figure 3. Both approximate and exact methods are available for calculations of confidence ellipses of a bivariate normal distribution.36 Our modified version of the exact methods utilises common statistics of the simple linear correlation analysis. Given n pairs of observations x and y, with standard deviations sx and sy, and correlation coefficient r, and for a given probability level ␣ (for example, ␣ = 0.05), take the Snedecor’s F␣ value with 2 and n-2 degrees of freedom. The semi-axes L1 and L2, and the slopes b1 and b2 = −1/b1 of the axes of the 100(1−␣)% confidence


ellipses can be calculated using equations (1) and (2), respectively: L1,L2 = 冑K ·

冑(n−1) (s

2 x

(1)

+ s ) ± 冑[(n−1) (s + s )] − 4(n−1) (1−r ) s2xs2y 2 y

2 x

2 y

2

2

2

where, K = F/n·(n − 2) for confidence ellipses

805

b, −1/b = (s2y − sx2)/2r sxsy ± 冑1 + [(s2y − s2x)/2 r sxsy]2 (2) Free software for bivariate vector analysis is available by e-mail: apiccoli얀uxl.unipd.it
