177
92 (1992) 111-185 0 1992 Elsevier Scientific Publishers Ireland, Ltd. 0021-9150/92/%05.00 Printed and Published in Ireland Atherosclerosis,
ATHERO 04757
A prospective study of obesity, lipids, apolipoproteins ischaemic heart disease in women
and
M.P. Coleman’, T.J.A. Key2, D.Y. Wang3, C. Herrnon*, I.S. Fentiman3, D.S. Allen3, M. Jarvis3, M.C. Pike4 and T.A.B. Sanders’ ‘Thames Cancer Registry, Sutton (U.K.), ‘Imperial Cancer Research Fund, Cancer Epidemiology Unit, Oxford (U.K.), ‘Imperial Cancer Research Fund, London (U.K.), 4University of Southern California School of Medicine. Los Angeles (U.S. A.) and ‘King’s College, London (U.K.)
(Received 13 August, 1991) (Accepted 6 November, 1991)
Summary
The aim of the study was to examine the relationships of obesity, lipids and apolipoproteins with the risk for subsequent ischaemic heart disease in middle-aged women, using a case-control study nested within a cohort study. A total of 3634 women aged 26-88 were recruited in Guernsey between 1977 and 1985 and followed until June 1986 by abstraction of their general practitioners’ records. Fifty-one cases of incident ischaemic heart disease (11 myocardial infarction, 40 angina) were identified. For each case up to 4 controls were selected, matched for age and date at recruitment. Odds ratios for the development of ischaemic heart disease in the middle and upper thirds of the distribution for each variable in the controls, relative to the lowest third (and two-sided P-values for linear trends), were: 3.0, 2.6 (0.015) for Quetelet’s index; 3.3, 5.1 (0.003) for total cholesterol; 0.5, 0.6 (0.102) for apolipoprotein A-I; 1.8, 2.4 (0.015) for apolipoprotein B; 1.3, 2.1 (0.155) for apolipoprotein(a). The increased risks associated with increased Quetelet’s index and total cholesterol were independent of each other and these variables were more strongly related to myocardial infarction than to angina. The relationships of risk with serum cotinine, fatty acids, dehydroepiandrosterone sulphate and sex hormone binding globulin were weak and did not approach statistical significance.
Key words: Ischaemic heart disease in women; Obesity; Lipids; Apolipoproteins
Introduction
Correspondence to: T.J.A. Key, ICRF Cancer Epidemiology Unit, Gibson Building, Radcliffe Infirmary, Oxford OX2 6HE. U.K.
Recent prospective studies have clarified the effects of obesity and high serum cholesterol on the risk for ischaemic heart disease in women [1,2], but few prospective studies have yet provided in-
178 formation on the possible roles of other biochemical variables. We report here the results of a nested case-control study which investigated obesity, lipids and apolipoproteins in women who subsequently developed ischaemic heart disease (myocardial infarction or angina). Methods Subjects This study utilised a cohort of approximately 5000 women recruited in Guernsey between 1977 and 1985 for a study of the relationships of sex hormones with breast cancer [3]. During recruitment a questionnaire mostly concerned with reproductive history was completed at interview, height and weight were measured, a 60-ml blood sample was taken and serum was stored in 2-ml amounts at -20°C. Women who developed ischaemic heart disease subsequent to recruitment were identified by searching the medical records of the women in the cohort. Permission to do this had been obtained at recruitment. Trained non-medical personnel identified possible cases by looking for particular key words or phrases in the general practitioners’ records. The records of possible cases were then abstracted in detail by a physician. The general practitioners’ records for 3634 of the cohort were successfully searched. Women were classified as incident cases of ischaemic heart disease if their records showed a first diagnosis of myocardial infarction or angina after recruitment and before the close of the study (30 June 1986). Women with incident myocardial infarction during the study period but who had prevalent angina at recruitment were not classified as cases. The following criteria were accepted as diagnoses of ischaemic heart disease: coronary artery bypass graft surgery; an ECG taken during an episode of chest pain and reported either as infarction or as ischaemia; a record of myocardial infarction after entry to the study; coronary angiography identifying diseased arteries; a general practitioner’s diagnosis of angina or infarction, treated accordingly. Sixty eligible cases were identified; 12 of myocardial infarction (8 non-fatal, 4 fatal) and 48 of angina. Adequate
volumes of frozen serum were available for 51 of these cases; 11 of myocardial infarction (7 nonfatal, 4 fatal) and 40 of angina. Four controls per case were selected from among the remaining women in the cohort. Controls were chosen at random, matched by age within 2 years and by date of recruitment within 1 year of that of the case. It was possible to match 93.5% of controls within both these criteria. If suitable controls could not be found within these criteria, the matching restraints were relaxed until controls were found. For each of 3 cases, only 3 suitable controls with adequate volumes of stored serum could be found, giving a total of 201 controls. Assays Only a sample of the cohort (2 1.5%) were asked about cigarette smoking at recruitment, therefore serum cotinine concentration was measured by gas chromatography [4] and women were classified as current smokers if the serum cotinine concentration was greater than or equal to 14 rig/ml [ 51. Of the 247 women for whom serum cotinine concentration was measured, questionnaire information on smoking was available for 88. Of the 75 women classified as non-smokers by the questionnaire, 68 (91%) were classified as non-smokers by serum cotinine. Of the 13 women classified as current smokers by the questionnaire, 11 (85%) were classified as current smokers by serum cotinine. Five .of the seven women classified as current smokers by serum cotinine but not by questionnaire had a cotinine concentration above 40 @ml, a level which cannot be achieved by passive smoking and which therefore shows that their answer to the questionnaire was false. Total cholesterol was measured by enzymatic assay (c-system, Boehringer Mannheim, Lewes). Apolipoproteins A-I and B were measured using immunoturbidometric methods (Boehringer Mannheim). Apolipoprotein(a) was measured by a double-antibody liquid phase radioimmunoassay, using a kit supplied by Pharmacia Diagnostics AB, (S-75 182, Uppsala, Sweden). Plasma was trans-esterified [6] in the presence of pentadecanoic acid as an internal standard and fatty acids were analysed by capillary gas-liquid chromatography using a 25 m CpSil 88 column
179 liminary analysis, including calculation of means and Pearson correlation coefficients. Matched odds ratios (ORs) for increasing levels of each variable and tests for linear increases in ORs, were calculated by conditional logistic regression using the EGRET statistical package. Two-sided Pvalues are quoted.
(Chrompack, London, U.K.). Fatty acids were identified by comparison with authentic standards and quantified using an integrator. The total esteritied fatty acids were calculated by reference to the internal standard and the proportions of individual fatty acids were calculated. Serum sulphate dehydroepiandrosterone (DHEAS) concentration was measured by solidphase radioimmunoassay, using a kit supplied by Diagnostic Products Corp. (Los Angeles, CA, USA). Serum sex hormone binding globulin (SHBG) was measured by liquid-phase immunoradiometric assay [7]. Due to the small volume of serum available for some subjects, some assays were not completed on all 252 subjects. There were 5 missing values for cotinine (all controls), 11 for apolipoprotein(a) (2 cases, 9 controls) and 6 for DHEAS (all controls). These subjects were excluded from the relevant analyses.
Results
Mean values in cases and controls for questionnaire data, anthropometry and serum cotinine are given in Table 1. The mean heights of the cases and controls were similar, but the cases were on average 5% heavier than the controls and had a 5% higher Quetelet’s index (QI, kg/m2). Parity was similar in the two groups, but more of the cases than controls had had their menopause (natural or surgical) before age 50 and more cases than controls had had a hysterectomy with unknown age at cessation of ovarian function. A smaller percentage of cases than of controls had a serum cotinine concentration of 14 ng/ml or above.
Statistical methods The SPSS statistical package was used for preTABLE 1 QUESTIONNAIRE
DATA, ANTHROPQMETRY,
AND SERUM COTININE AT RECRUITMENT
QI, Quetelet’s index. Variable
Age (years) Height (m) Weight (kg)
QI (kg/m*)
Parity 0 Parity 1 Parity 2 Parity 3+ Menopausal status Natural menopause or ovariectomy 50+a Natural menopause or ovariectomy < 50 Premenopausal, aged < 50 Hysterectomy without ovariectomy, or unknown age at natural menopause Cotinine 14+ ng/ml
Cases (n = 51)
Controls (n = 201)
Mean
(SD.)
Mean
(S.D.)
59.2 1.58 67.1 26.9
(9.5) (0.06) (10.8) (3.9)
59. I 1.59 63.9 25.5
(9.2)
Number
(“h)
Number
(%I)
11 9 14 17
(21.6) (17.6) (27.5) (33.3)
40 45 63 53
(19.9) (22.4) (31.3) (26.4)
13 20 5 13
(25.5) (39.2) (9.8) (25.5)
73 70 22 36
(36.3) (34.8) (11.0) (17.9)
9
(17.6)
45b
(23.0)
aIncludes women who were aged 50+ and premenopausal at recruitment. by = 196.
(0.06) (9.7) (3.7)
180 Mean values for the other biochemical variables are given in Table 2. In comparison with controls, cases had higher mean serum concentrations of cholesterol (9% higher), apolipoprotein B (8%), apolipoprotein(a) (400/o),total esterified fatty acids (5%) and percent linoleic acid (1%). Apolipoprotein(a) had a strongly skewed distribution; geometric mean values in cases and controls were 19.3 and 12.0 mg/dl, respectively. Cases had lower serum concentrations of apolipoprotein A-I (8% lower), percent eicosapentaenoic acid (5%), percent docosahexaenoic acid (8%), DHEAS (12%) and SHBG (7%). The correlations between the biochemical variables, age and QI are given in Table 3. Table 4 gives the ORs for ischaemic heart disease for the 2 upper levels of each variable relative to the lowest level, together with the P-value for a linear trend with each variable in its original units. The three levels are approximate thirds of the distribution of each variable among controls, with the exception of cotinine for which the cut points are 14 ng/ml (distinguishing non-smokers from
TABLE 2 BIOCHEMICAL VARIABLES Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein B; Ape(a), apolipoprotein(a); DHA, docosahexaenoic acid; DHEAS, dehydroepiandrosterone sulphate; EPA, eicosapentaenoic acid; SHBG, sex hormone binding globulin. Variable
Total cholesterol (mmolil) Apo A-I (mg/dl) Apo B (mg/dl) Ape(a) (mg/dl) Total esterified fatty acids (fl) % Linoleic % EPA % DHA DHEAS (amoY1) SHBG (nmol/l) “n = 49. bn = 192. cn = 195.
Cases (n = 51)
Controls (n = 201)
Mean
(S.D.)
Mean
(SD.)
6.3
(1.2)
5.8
(1.2)
83 105 40.2a 4.6
(29) (24) (46.1) (1.5)
90 97 28.gb 4.4
(26)
23.27 0.91 0.80 2.9 62.8
(4.07) (0.31) (0.32) (2.4) (29.5)
2293 0.96 0.87 3.3c 67.5
(4.31) (0.50) (0.37) (2.4) (30.0)
(23) (41.2) (1.2)
light smokers) and 200 ng/ml (distinguishing light smokers from heavy smokers). A QI greater than 23.7 kg/m2 was associated with a highly significant 2.8-fold increase in risk. Cotinine was not significantly associated with risk, but the highest risk (OR 1.4) was in the group with the highest cotinine levels. Total cholesterol was strongly positively associated with risk. Apolipoprotein A-I was weakly and non-significantly inversely associated with risk, while apolipoprotein B was quite strongly and significantly positively associated with risk. Apolipoprotein(a) was positively associated with risk; this trend was not statistically significant, but the elevated risk in the highest third was nearly significant at the 5% level (OR 2.1, P = 0.066). The 2 upper levels of total esterified fatty acids and of linoleic acid were associated with ORs above 1 and the 2 upper levels of eicosapentaenoic and docosahexaenoic acid with ORs below 1, but none of these differences was statistically significant. Both DHEAS and SHBG were weakly inversely related to risk, but these trends did not approach statistical significance. The relationships of parity and of menopausal status with risk were also examined. Risk was not related to parity. Relative to women with a menopause at ages 50 and above (natural, ovariectomy or premenopausal at ages 50 and above), the ORs were 1.5 for women with menopause (natural or ovariectomy) before age 50, 1.1 for women who were aged less than 50 and premenopausal at recruitment and 2.0 for women who had had a hysterectomy (without ovariectomy) or who had had a natural menopause at an unknown age, but none of these differences was statistically significant. Repeating these comparisons in women aged 50 or above at recruitment, to reduce any confounding due to age, gave similar results. In univariate analysis, therefore, clear increases in risk were seen with increased QI and with increasing cholesterol and apolipoprotein B. Table 5 shows the effect on other risk factors of adjusting for QI and cholesterol. QI and cholesterol themselves appear to be independent risk factors, since adjusting one for the other leaves the ORs almost unchanged. The protective effect of high levels of apolipoprotein A-I was slightly increased and became statistically significant after adjusting
3
BETWEEN
VARIABLES
-8 -II -2 -7 -32
-16 5 13 -41
-1
coefficients,
SHBG
“Pearson
Ape(a) % Linoleic % EPA % DHA DHEAS 11
-5 12 11 -17
16 61 17
Two-sided
Cholesterol
as percentages.
4 -1 -2 -8 -11 16 -5 14 16
Cotinine
expressed
6 -5 21 -2
correlation
-10
Cholesterol Apo A-I Apo B
Cotinine
29 10 33 16
QI
11 -23
QI
Age
significance
11
12 10 25 -4
4 -1
Apo A-I
Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein B; Ape(a), apolipoprotein(a); taenoic acid, QI, Quetelet’s index; SHBG, sex hormone binding globulin.
CORRELATIONSa
TABLE
6
2 -I 3 -II
Ape(a)
7
-19 14 11
6
39 -6
% EPA
dehydroepiandrosterone
% Linoleic
acid; DHEAS,
level: r = 16%. P = 0.01
-12
-21 4 -1 -23
8
Apo B
DHA, docosahexaenoic
5
-5
% DHA
sulphate;
-2
DHEAS
EPA, eicosapen-
TABLE 4 ODDS RATIOS FOR INCIDENT ISCHAEMIC HEART DISEASE Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein B; Ape(a), apolipoprotein(a); DHA, docosahexaenoic acid; DHEAS, dehydroepiandrosterone sulphate; EPA, eicosapentaenoic acid; QI, Quetelet’s index; SHBG, sex hormone binding globulin. Cut points”
Variable
QI
Wd
Odds ratios in levels (95% confidence limits)
23.1, 21.1 14, 200
Cotinine (rig/ml) Cholesterol (mmohl)
5.24, 6.31
Apo A-I (mg/dl)
15.3, 97.1
Apo B (mg/dl)
88.0, 105.7
Apob) (m&W
6.9, 20.7
Total esteritied fatty acids (g/l) % Linoleic
3.7, 4.8
% EPA
0.72, 0.98
% DHA
0.69, 0.97
DHEAS (pmolfl)
1.85, 3.98
SHBG (nmohl)
50.3, 79.0
20.8, 24.1
P-value for linear trendb
2
3
3.0 (1.2-7.3) 0.2 (0.1-1.1) 3.3 (1.2-8.8) 0.5 (0.2-1.1) 1.8 (0.7-4. I) 1.3 (0.5-3.0) 1.5 (0.7-3.6) 1.8 (0.9-3.9) 0.9 (0.4-I .8) 0.8 (0.4- 1.6) 0.7 (0.3-1.5) 0.8 (0.4-1.8)
2.6 (1.0-6.5)
0.015 0.296
fof6-3.7) 5.1 (1.9-14.0) 0.6 (0.3-1.3) 2.4 (1.o-5.9) 2.1 (I .o-4.7) I.5 (0.7-3.4) 1.3 (0.6-2.9) 0.9 (0.4-2.0) 0.8 (0.4-I .8) 0.6 (0.2-1.3) 0.7 (0.3-1.6)
0.003 0.102 0.015 0.155 0.264 0.604 0.555 0.199 0.226 0.309
YJsed to define approximate thirds of the distribution in controls, with the exception of cotinine (see text). bTwo-sided test for linear trend with continuous variables. TABLE 5 ODDS RATIOS FOR INCIDENT ISCHAEMIC HEART DISEASE, WITH ADJUSTMENT FOR QIJETELET’S INDEX AND FOR SERUM CHOLESTEROL CONCENTRATION Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein Variable
QI Cholesterol
B; Ape(a), apolipoprotein(a); Odds ratios in levels
Adjustment
None Cholesterol None
.
QI Cotinine Apo A-I Apo B Ape(a)
None QI, cholesterol None QI, cholesterol None QI, cholesterol None QI, cholesterol
“Two-sided test for linear trend with continuous variables.
2
3
3.0 3.0 3.3 3.1 0.2 0.2 0.5 0.5 I.8 1.1 1.3 I.1
2.6 2.6 5.1 5.5 1.4 1.4 0.6 0.5 2.4 1.1 2.1 1.9
QI, Quetelet’s index. P-value for linear trend”
0.015 0.022 0.003 0.005 0.296 0.304 0.102 0.041 0.015 0.691 0.155 0.150
183 TABLE 6 ODDS RATIOS FOR INCIDENT MYOCARDIAL INFARCTION OR ANGINA Apo A-I, apolipoprotein Ape(a), apolipoprotein(a);
A-I; Apo B, apolipoprotein QI, Quetelet’s index.
B;
Variable
Condition
Odds ratio per unit increase
P-value for linear trend”
QI
Infarction Angina Infarction Angina Infarction Angina Infarction Angina Infarction Angina Infarction Angina
1.230 1.084 0.989 1.002 2.311 1.477 0.979 0.992 1.040 I.013 0.989 1.009
0.046 0.079 0.128 0.108 0.043 0.019 0.135 0.295 0.021 0.122 0.201 0.025
(kg/m*)
Cotinine (@ml) Cholesterol (mmol/l) Apo A-I (mgdl) Apo B (mg/dl) Ape(a) (mgdl)
“Two-sided test for linear trend with continuous variables.
for QI and cholesterol, whereas adjusting for these variables eliminated the elevated risk associated with increasing levels of apolipoprotein B. Adjusting for QI and cholesterol had little effect on the ORs for cotinine or for apolipoprotein(a). The ORs for the main variables of interest were recalculated with the cases subdivided into myocardial infarction and angina (Table 6). For QI, cholesterol, apolipoprotein A-I and apolipoprotein B the trends in ORs were clearly stronger in relation to infarction than angina. Discussion
Follow-up by examination of the general practitioners’ notes was achieved for only 3634 (73%) of the 5000 women in the cohort, so that the number of cases identified was small and the estimates of risks are imprecise. We also had no information on blood pressure, an important risk factor for ischaemic heart disease [8]. However, this is still one of the few prospective studies of ischaemic heart disease in women and provides, we believe,
the first prospective data on apolipoprotein(a) in women. We found that increased QI was strongly associated with an increase in the risk for ischaemic heart disease. These results are in agreement with those from the US Nurses’ Health Study [l] and as in that study we found that the risks for both myocardial infarction and angina were increased and that a significantly increased risk was found even for mildly to moderately overweight women. The risk associated with increased QI was not confounded by cholesterol or by other risk factors measured, but it is likely that part of the increase in risk was due to an increase in blood pressure. Despite the established increase in risk for ischaemic heart disease caused by cigarette smoking in women [9], we failed to find a significant increase in risk. We assesse,d smoking habit by serum cotinine; this is certainly a better measure of cigarette smoking during the previous 24 h than is a questionnaire [5], but may be a poorer measure of long-term smoking habits. A relationship between smoking and ischaemic heart disease may therefore have been reduced by exposure misclassification. Furthermore, the relationship of cigarette smoking with angina in women has been observed to be considerably weaker than its relationship with myocardial infarction [9]. Most of our cases were of angina, so that the weak association between serum cotinine and risk is consistent with other studies on angina and the absence of an association with infarction may be due to chance. The low prevalence of smoking in our population also reduces the power of the study to detect relationships with smoking. Finally, smoking is associated with a number of other factors such as alcohol intake and social class, which could confound the relationship of smoking with ischaemic heart disease. There was a strong positive relationship between cholesterol and risk for ischaemic heart disease, in agreement with other studies [2]. The risk associated with cholesterol was stronger than that associated with apolipoprotein B, the major apolipoprotein of low-density lipoprotein (LDL) cholesterol. Apolipoprotein A-I, the major apolipoprotein of high-density lipoprotein (HDL) cholesterol, was inversely related to risk and this
184 relationship became stronger after adjusting for QI and cholesterol. Apolipoprotein(a), the apolipoprotein of lipoprotein(a) [lo], was positively associated with risk for ischaemic heart disease; this association was not statistically significant despite the large difference in means between cases and controls, probably because of the strongly skewed distribution of values. Wood et al. [I l] suggested that low levels of linoleic acid, at least in adipose tissue, may be a risk factor for coronary heart disease. We measured the fatty acid composition of serum total esterified fatty acids rather than of adipose tissue, but this measure has been shown to correlate well with dietary intake of linoleic acid [12]. Serum percentage linoleic acid was a little higher in cases than in controls, providing no support for the hypothesis, but there was a weak negative correlation between percentage linoleic acid and apolipoprotein B, itself a risk factor for ischaemic heart disease. Serum percentages of eicosapentaenoic and docosahexaenoic acids were lower in cases than in controls. These differences were small and were not statistically significant, but are consistent with the hypothesized protective effect of omega 3 fatty acids [13]. We measured DHEAS because Barrett-Connor et al. [ 141reported an inverse relationship between DHEAS and death from ischaemic heart disease in men, although they failed to find any evidence for this in elderly women [1.5]. Our finding of a weak, between non-significant inverse relationship DHEAS and risk suggests that this hormone may deserve further study. Lapidus et al. [ 161suggested that SHBG may be a predictor of ischaemic heart disease in women, but we found only very weak evidence for such a relationship. In conclusion, the strongest risk factors for ischaemic heart disease in this study were obesity and high serum cholesterol. Both these risk factors can be changed by comprehensive lifestyle changes 1171.
Foulds, Claudine Paluch, Dr. Julia Peet, Su Williams-Yeagers and Hayley Willson for help with data collection on Guernsey; Professor D.A. Wood for help in the assessment of case records and ECG traces and Helen Powell for typing the manuscript. References 1
2
3
4
(1990) 882. Bush, T.L., Fried, L.P. and Barrett-Connor, E., Cholesterol, lipoproteins and coronary heart disease in women, Clin. Chem., 34 (1988) B60. Moore, J.W., Clark, G.M.G., Hoare, S.A., Millis, R.R., Hayward, J.L., Quinlan, M.K., Wang, D.Y. and Bulbrook, R.D., Binding of oestradiol to blood proteins and aetiology of breast cancer, Jnt. J. Cancer, 38 (1986) 625. Feyerabend, C. and Russell, M.A.H., A rapid gas-liquid chromatographic determination of cotinine in biological fluids, Analyst, 105 (1980) 993.
5
Jarvis, M.J., Tunstall-Pedoe, H., Feyerabend, C., Vesey, C. and Saloojee, Y., Comparison of tests used to distinguish smokers from nonsmokers, Am. J. Public Health, 77 (1987) 1435.
6
1
Lepage, G. and Roy, C.C., Direct transesteritication of all classes of lipids in a one-step reaction, J. Lipid Res., 27 (1986) 114. Hammond, G.L., Langley, M.S. and Robinson, P.A., A
8
liquid-phase immunoradiometric assay (IRMA) for human sex hormone binding globulin (SHBG), J. Steroid B&hem., 23 (1985) 451. MacMahon, S., Pete, R., Cutler, J., Collins, R., Sorlie, P., Neaton, J., Abbott, R., Godwin, J., Dyer, A. and Stamler, J., Blood pressure, stroke and coronary heart disease. Part 1, prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution
9
10
bias, Lancet, 335 (1990) 765. Willett, W.C., Green, A., Stampfer, M.J., Speizer, F.E., Colditz, GA., Rosner, B., Monson, R.R., Stason, W. and Hennekens, C.H., Relative and absolute excess risks of coronary heart disease among women who smoke cigarettes, N. Engl. J. Med., 317 (1987) 1303. Utermann, G., The mysteries of lipoprotein(a), Science, 246 (I 989) 904.
II
Wood, D.A., Butler, S., Riemersma, R.A., Thomson, M. and Oliver, M.F., Adipose tissue and platelet fatty acids and coronary heart disease in Scottish men, Lancet, ii (1984) 117.
12
Houwelingen, A.C.V., Kester, A.D.M., Kromhout, D. and Hornstra, G., Comparison between habitual intake of polyunsaturated fatty acids and their concentrations in serum lipid fractions, Eur. J. Clin. Nutr., 43 (1989) I I.
Acknowledgements
We are most grateful to: the women volunteers on Guernsey who made the study possible; the GPs and staff at the general practices on the island for allowing us access to their practices; Sue
Manson, J.E., Colditz, GA., Stampfer, M.J., Willett, W.C., Rosner, B., Monson, R.R., Speizer, F.E. and Hennekens, C.H., A prospective study of obesity and risk of coronary heart disease in women, N. Engl. J. Med., 322
185 I3 14
I5
I6
C. and Gredmark,
Leaf, A. and Weber, P.C., Cardiovascular effects of n-3 fatty acids, N. Engl. J. Med., 318 (1988) 549. Barrett-Connor, E., Khaw, K-T. and Yen, S.S.C., A prospective study of dehydroepiandrosterone sulfate, mortality and cardiovascular disease, N. Engl. J. Med., 315 (1986) 1519. Barrett-Connor, E. and Khaw, K-T., Absence of an inverse relation of dehydroepiandrosterone sulfate with cardiovascular mortality in postmenopausal women, N. Engl. J. Med. 317 (1987) 711. Lapidus, L., Lindstedt, G., Lundberg, P-A., Bengtsson,
T., Concentrations
of sex-hormone
binding globulin and corticosteroid binding globulin in serum in relation to cardiovascular risk factors and to 12-year incidence of cardiovascular disease and overall mortality in postmenopausal women, Clin. Chem., 32 17
(1986) 146. Ornish, D., Brown, SE., Scherwitz, L.W., Billings, J.H.. Armstrong, W.T., Ports, T.A., McLanahan, S.M.. Kirkeeide, R.L., Brand, R.J. and Gould, K.L., Can lifestyle changes reverse coronary heart disease? Lancet. 336 (1990) 129.
92 (1992) 111-185 0 1992 Elsevier Scientific Publishers Ireland, Ltd. 0021-9150/92/%05.00 Printed and Published in Ireland Atherosclerosis,
ATHERO 04757
A prospective study of obesity, lipids, apolipoproteins ischaemic heart disease in women
and
M.P. Coleman’, T.J.A. Key2, D.Y. Wang3, C. Herrnon*, I.S. Fentiman3, D.S. Allen3, M. Jarvis3, M.C. Pike4 and T.A.B. Sanders’ ‘Thames Cancer Registry, Sutton (U.K.), ‘Imperial Cancer Research Fund, Cancer Epidemiology Unit, Oxford (U.K.), ‘Imperial Cancer Research Fund, London (U.K.), 4University of Southern California School of Medicine. Los Angeles (U.S. A.) and ‘King’s College, London (U.K.)
(Received 13 August, 1991) (Accepted 6 November, 1991)
Summary
The aim of the study was to examine the relationships of obesity, lipids and apolipoproteins with the risk for subsequent ischaemic heart disease in middle-aged women, using a case-control study nested within a cohort study. A total of 3634 women aged 26-88 were recruited in Guernsey between 1977 and 1985 and followed until June 1986 by abstraction of their general practitioners’ records. Fifty-one cases of incident ischaemic heart disease (11 myocardial infarction, 40 angina) were identified. For each case up to 4 controls were selected, matched for age and date at recruitment. Odds ratios for the development of ischaemic heart disease in the middle and upper thirds of the distribution for each variable in the controls, relative to the lowest third (and two-sided P-values for linear trends), were: 3.0, 2.6 (0.015) for Quetelet’s index; 3.3, 5.1 (0.003) for total cholesterol; 0.5, 0.6 (0.102) for apolipoprotein A-I; 1.8, 2.4 (0.015) for apolipoprotein B; 1.3, 2.1 (0.155) for apolipoprotein(a). The increased risks associated with increased Quetelet’s index and total cholesterol were independent of each other and these variables were more strongly related to myocardial infarction than to angina. The relationships of risk with serum cotinine, fatty acids, dehydroepiandrosterone sulphate and sex hormone binding globulin were weak and did not approach statistical significance.
Key words: Ischaemic heart disease in women; Obesity; Lipids; Apolipoproteins
Introduction
Correspondence to: T.J.A. Key, ICRF Cancer Epidemiology Unit, Gibson Building, Radcliffe Infirmary, Oxford OX2 6HE. U.K.
Recent prospective studies have clarified the effects of obesity and high serum cholesterol on the risk for ischaemic heart disease in women [1,2], but few prospective studies have yet provided in-
178 formation on the possible roles of other biochemical variables. We report here the results of a nested case-control study which investigated obesity, lipids and apolipoproteins in women who subsequently developed ischaemic heart disease (myocardial infarction or angina). Methods Subjects This study utilised a cohort of approximately 5000 women recruited in Guernsey between 1977 and 1985 for a study of the relationships of sex hormones with breast cancer [3]. During recruitment a questionnaire mostly concerned with reproductive history was completed at interview, height and weight were measured, a 60-ml blood sample was taken and serum was stored in 2-ml amounts at -20°C. Women who developed ischaemic heart disease subsequent to recruitment were identified by searching the medical records of the women in the cohort. Permission to do this had been obtained at recruitment. Trained non-medical personnel identified possible cases by looking for particular key words or phrases in the general practitioners’ records. The records of possible cases were then abstracted in detail by a physician. The general practitioners’ records for 3634 of the cohort were successfully searched. Women were classified as incident cases of ischaemic heart disease if their records showed a first diagnosis of myocardial infarction or angina after recruitment and before the close of the study (30 June 1986). Women with incident myocardial infarction during the study period but who had prevalent angina at recruitment were not classified as cases. The following criteria were accepted as diagnoses of ischaemic heart disease: coronary artery bypass graft surgery; an ECG taken during an episode of chest pain and reported either as infarction or as ischaemia; a record of myocardial infarction after entry to the study; coronary angiography identifying diseased arteries; a general practitioner’s diagnosis of angina or infarction, treated accordingly. Sixty eligible cases were identified; 12 of myocardial infarction (8 non-fatal, 4 fatal) and 48 of angina. Adequate
volumes of frozen serum were available for 51 of these cases; 11 of myocardial infarction (7 nonfatal, 4 fatal) and 40 of angina. Four controls per case were selected from among the remaining women in the cohort. Controls were chosen at random, matched by age within 2 years and by date of recruitment within 1 year of that of the case. It was possible to match 93.5% of controls within both these criteria. If suitable controls could not be found within these criteria, the matching restraints were relaxed until controls were found. For each of 3 cases, only 3 suitable controls with adequate volumes of stored serum could be found, giving a total of 201 controls. Assays Only a sample of the cohort (2 1.5%) were asked about cigarette smoking at recruitment, therefore serum cotinine concentration was measured by gas chromatography [4] and women were classified as current smokers if the serum cotinine concentration was greater than or equal to 14 rig/ml [ 51. Of the 247 women for whom serum cotinine concentration was measured, questionnaire information on smoking was available for 88. Of the 75 women classified as non-smokers by the questionnaire, 68 (91%) were classified as non-smokers by serum cotinine. Of the 13 women classified as current smokers by the questionnaire, 11 (85%) were classified as current smokers by serum cotinine. Five .of the seven women classified as current smokers by serum cotinine but not by questionnaire had a cotinine concentration above 40 @ml, a level which cannot be achieved by passive smoking and which therefore shows that their answer to the questionnaire was false. Total cholesterol was measured by enzymatic assay (c-system, Boehringer Mannheim, Lewes). Apolipoproteins A-I and B were measured using immunoturbidometric methods (Boehringer Mannheim). Apolipoprotein(a) was measured by a double-antibody liquid phase radioimmunoassay, using a kit supplied by Pharmacia Diagnostics AB, (S-75 182, Uppsala, Sweden). Plasma was trans-esterified [6] in the presence of pentadecanoic acid as an internal standard and fatty acids were analysed by capillary gas-liquid chromatography using a 25 m CpSil 88 column
179 liminary analysis, including calculation of means and Pearson correlation coefficients. Matched odds ratios (ORs) for increasing levels of each variable and tests for linear increases in ORs, were calculated by conditional logistic regression using the EGRET statistical package. Two-sided Pvalues are quoted.
(Chrompack, London, U.K.). Fatty acids were identified by comparison with authentic standards and quantified using an integrator. The total esteritied fatty acids were calculated by reference to the internal standard and the proportions of individual fatty acids were calculated. Serum sulphate dehydroepiandrosterone (DHEAS) concentration was measured by solidphase radioimmunoassay, using a kit supplied by Diagnostic Products Corp. (Los Angeles, CA, USA). Serum sex hormone binding globulin (SHBG) was measured by liquid-phase immunoradiometric assay [7]. Due to the small volume of serum available for some subjects, some assays were not completed on all 252 subjects. There were 5 missing values for cotinine (all controls), 11 for apolipoprotein(a) (2 cases, 9 controls) and 6 for DHEAS (all controls). These subjects were excluded from the relevant analyses.
Results
Mean values in cases and controls for questionnaire data, anthropometry and serum cotinine are given in Table 1. The mean heights of the cases and controls were similar, but the cases were on average 5% heavier than the controls and had a 5% higher Quetelet’s index (QI, kg/m2). Parity was similar in the two groups, but more of the cases than controls had had their menopause (natural or surgical) before age 50 and more cases than controls had had a hysterectomy with unknown age at cessation of ovarian function. A smaller percentage of cases than of controls had a serum cotinine concentration of 14 ng/ml or above.
Statistical methods The SPSS statistical package was used for preTABLE 1 QUESTIONNAIRE
DATA, ANTHROPQMETRY,
AND SERUM COTININE AT RECRUITMENT
QI, Quetelet’s index. Variable
Age (years) Height (m) Weight (kg)
QI (kg/m*)
Parity 0 Parity 1 Parity 2 Parity 3+ Menopausal status Natural menopause or ovariectomy 50+a Natural menopause or ovariectomy < 50 Premenopausal, aged < 50 Hysterectomy without ovariectomy, or unknown age at natural menopause Cotinine 14+ ng/ml
Cases (n = 51)
Controls (n = 201)
Mean
(SD.)
Mean
(S.D.)
59.2 1.58 67.1 26.9
(9.5) (0.06) (10.8) (3.9)
59. I 1.59 63.9 25.5
(9.2)
Number
(“h)
Number
(%I)
11 9 14 17
(21.6) (17.6) (27.5) (33.3)
40 45 63 53
(19.9) (22.4) (31.3) (26.4)
13 20 5 13
(25.5) (39.2) (9.8) (25.5)
73 70 22 36
(36.3) (34.8) (11.0) (17.9)
9
(17.6)
45b
(23.0)
aIncludes women who were aged 50+ and premenopausal at recruitment. by = 196.
(0.06) (9.7) (3.7)
180 Mean values for the other biochemical variables are given in Table 2. In comparison with controls, cases had higher mean serum concentrations of cholesterol (9% higher), apolipoprotein B (8%), apolipoprotein(a) (400/o),total esterified fatty acids (5%) and percent linoleic acid (1%). Apolipoprotein(a) had a strongly skewed distribution; geometric mean values in cases and controls were 19.3 and 12.0 mg/dl, respectively. Cases had lower serum concentrations of apolipoprotein A-I (8% lower), percent eicosapentaenoic acid (5%), percent docosahexaenoic acid (8%), DHEAS (12%) and SHBG (7%). The correlations between the biochemical variables, age and QI are given in Table 3. Table 4 gives the ORs for ischaemic heart disease for the 2 upper levels of each variable relative to the lowest level, together with the P-value for a linear trend with each variable in its original units. The three levels are approximate thirds of the distribution of each variable among controls, with the exception of cotinine for which the cut points are 14 ng/ml (distinguishing non-smokers from
TABLE 2 BIOCHEMICAL VARIABLES Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein B; Ape(a), apolipoprotein(a); DHA, docosahexaenoic acid; DHEAS, dehydroepiandrosterone sulphate; EPA, eicosapentaenoic acid; SHBG, sex hormone binding globulin. Variable
Total cholesterol (mmolil) Apo A-I (mg/dl) Apo B (mg/dl) Ape(a) (mg/dl) Total esterified fatty acids (fl) % Linoleic % EPA % DHA DHEAS (amoY1) SHBG (nmol/l) “n = 49. bn = 192. cn = 195.
Cases (n = 51)
Controls (n = 201)
Mean
(S.D.)
Mean
(SD.)
6.3
(1.2)
5.8
(1.2)
83 105 40.2a 4.6
(29) (24) (46.1) (1.5)
90 97 28.gb 4.4
(26)
23.27 0.91 0.80 2.9 62.8
(4.07) (0.31) (0.32) (2.4) (29.5)
2293 0.96 0.87 3.3c 67.5
(4.31) (0.50) (0.37) (2.4) (30.0)
(23) (41.2) (1.2)
light smokers) and 200 ng/ml (distinguishing light smokers from heavy smokers). A QI greater than 23.7 kg/m2 was associated with a highly significant 2.8-fold increase in risk. Cotinine was not significantly associated with risk, but the highest risk (OR 1.4) was in the group with the highest cotinine levels. Total cholesterol was strongly positively associated with risk. Apolipoprotein A-I was weakly and non-significantly inversely associated with risk, while apolipoprotein B was quite strongly and significantly positively associated with risk. Apolipoprotein(a) was positively associated with risk; this trend was not statistically significant, but the elevated risk in the highest third was nearly significant at the 5% level (OR 2.1, P = 0.066). The 2 upper levels of total esterified fatty acids and of linoleic acid were associated with ORs above 1 and the 2 upper levels of eicosapentaenoic and docosahexaenoic acid with ORs below 1, but none of these differences was statistically significant. Both DHEAS and SHBG were weakly inversely related to risk, but these trends did not approach statistical significance. The relationships of parity and of menopausal status with risk were also examined. Risk was not related to parity. Relative to women with a menopause at ages 50 and above (natural, ovariectomy or premenopausal at ages 50 and above), the ORs were 1.5 for women with menopause (natural or ovariectomy) before age 50, 1.1 for women who were aged less than 50 and premenopausal at recruitment and 2.0 for women who had had a hysterectomy (without ovariectomy) or who had had a natural menopause at an unknown age, but none of these differences was statistically significant. Repeating these comparisons in women aged 50 or above at recruitment, to reduce any confounding due to age, gave similar results. In univariate analysis, therefore, clear increases in risk were seen with increased QI and with increasing cholesterol and apolipoprotein B. Table 5 shows the effect on other risk factors of adjusting for QI and cholesterol. QI and cholesterol themselves appear to be independent risk factors, since adjusting one for the other leaves the ORs almost unchanged. The protective effect of high levels of apolipoprotein A-I was slightly increased and became statistically significant after adjusting
3
BETWEEN
VARIABLES
-8 -II -2 -7 -32
-16 5 13 -41
-1
coefficients,
SHBG
“Pearson
Ape(a) % Linoleic % EPA % DHA DHEAS 11
-5 12 11 -17
16 61 17
Two-sided
Cholesterol
as percentages.
4 -1 -2 -8 -11 16 -5 14 16
Cotinine
expressed
6 -5 21 -2
correlation
-10
Cholesterol Apo A-I Apo B
Cotinine
29 10 33 16
QI
11 -23
QI
Age
significance
11
12 10 25 -4
4 -1
Apo A-I
Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein B; Ape(a), apolipoprotein(a); taenoic acid, QI, Quetelet’s index; SHBG, sex hormone binding globulin.
CORRELATIONSa
TABLE
6
2 -I 3 -II
Ape(a)
7
-19 14 11
6
39 -6
% EPA
dehydroepiandrosterone
% Linoleic
acid; DHEAS,
level: r = 16%. P = 0.01
-12
-21 4 -1 -23
8
Apo B
DHA, docosahexaenoic
5
-5
% DHA
sulphate;
-2
DHEAS
EPA, eicosapen-
TABLE 4 ODDS RATIOS FOR INCIDENT ISCHAEMIC HEART DISEASE Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein B; Ape(a), apolipoprotein(a); DHA, docosahexaenoic acid; DHEAS, dehydroepiandrosterone sulphate; EPA, eicosapentaenoic acid; QI, Quetelet’s index; SHBG, sex hormone binding globulin. Cut points”
Variable
QI
Wd
Odds ratios in levels (95% confidence limits)
23.1, 21.1 14, 200
Cotinine (rig/ml) Cholesterol (mmohl)
5.24, 6.31
Apo A-I (mg/dl)
15.3, 97.1
Apo B (mg/dl)
88.0, 105.7
Apob) (m&W
6.9, 20.7
Total esteritied fatty acids (g/l) % Linoleic
3.7, 4.8
% EPA
0.72, 0.98
% DHA
0.69, 0.97
DHEAS (pmolfl)
1.85, 3.98
SHBG (nmohl)
50.3, 79.0
20.8, 24.1
P-value for linear trendb
2
3
3.0 (1.2-7.3) 0.2 (0.1-1.1) 3.3 (1.2-8.8) 0.5 (0.2-1.1) 1.8 (0.7-4. I) 1.3 (0.5-3.0) 1.5 (0.7-3.6) 1.8 (0.9-3.9) 0.9 (0.4-I .8) 0.8 (0.4- 1.6) 0.7 (0.3-1.5) 0.8 (0.4-1.8)
2.6 (1.0-6.5)
0.015 0.296
fof6-3.7) 5.1 (1.9-14.0) 0.6 (0.3-1.3) 2.4 (1.o-5.9) 2.1 (I .o-4.7) I.5 (0.7-3.4) 1.3 (0.6-2.9) 0.9 (0.4-2.0) 0.8 (0.4-I .8) 0.6 (0.2-1.3) 0.7 (0.3-1.6)
0.003 0.102 0.015 0.155 0.264 0.604 0.555 0.199 0.226 0.309
YJsed to define approximate thirds of the distribution in controls, with the exception of cotinine (see text). bTwo-sided test for linear trend with continuous variables. TABLE 5 ODDS RATIOS FOR INCIDENT ISCHAEMIC HEART DISEASE, WITH ADJUSTMENT FOR QIJETELET’S INDEX AND FOR SERUM CHOLESTEROL CONCENTRATION Apo A-I, apolipoprotein A-I; Apo B, apolipoprotein Variable
QI Cholesterol
B; Ape(a), apolipoprotein(a); Odds ratios in levels
Adjustment
None Cholesterol None
.
QI Cotinine Apo A-I Apo B Ape(a)
None QI, cholesterol None QI, cholesterol None QI, cholesterol None QI, cholesterol
“Two-sided test for linear trend with continuous variables.
2
3
3.0 3.0 3.3 3.1 0.2 0.2 0.5 0.5 I.8 1.1 1.3 I.1
2.6 2.6 5.1 5.5 1.4 1.4 0.6 0.5 2.4 1.1 2.1 1.9
QI, Quetelet’s index. P-value for linear trend”
0.015 0.022 0.003 0.005 0.296 0.304 0.102 0.041 0.015 0.691 0.155 0.150
183 TABLE 6 ODDS RATIOS FOR INCIDENT MYOCARDIAL INFARCTION OR ANGINA Apo A-I, apolipoprotein Ape(a), apolipoprotein(a);
A-I; Apo B, apolipoprotein QI, Quetelet’s index.
B;
Variable
Condition
Odds ratio per unit increase
P-value for linear trend”
QI
Infarction Angina Infarction Angina Infarction Angina Infarction Angina Infarction Angina Infarction Angina
1.230 1.084 0.989 1.002 2.311 1.477 0.979 0.992 1.040 I.013 0.989 1.009
0.046 0.079 0.128 0.108 0.043 0.019 0.135 0.295 0.021 0.122 0.201 0.025
(kg/m*)
Cotinine (@ml) Cholesterol (mmol/l) Apo A-I (mgdl) Apo B (mg/dl) Ape(a) (mgdl)
“Two-sided test for linear trend with continuous variables.
for QI and cholesterol, whereas adjusting for these variables eliminated the elevated risk associated with increasing levels of apolipoprotein B. Adjusting for QI and cholesterol had little effect on the ORs for cotinine or for apolipoprotein(a). The ORs for the main variables of interest were recalculated with the cases subdivided into myocardial infarction and angina (Table 6). For QI, cholesterol, apolipoprotein A-I and apolipoprotein B the trends in ORs were clearly stronger in relation to infarction than angina. Discussion
Follow-up by examination of the general practitioners’ notes was achieved for only 3634 (73%) of the 5000 women in the cohort, so that the number of cases identified was small and the estimates of risks are imprecise. We also had no information on blood pressure, an important risk factor for ischaemic heart disease [8]. However, this is still one of the few prospective studies of ischaemic heart disease in women and provides, we believe,
the first prospective data on apolipoprotein(a) in women. We found that increased QI was strongly associated with an increase in the risk for ischaemic heart disease. These results are in agreement with those from the US Nurses’ Health Study [l] and as in that study we found that the risks for both myocardial infarction and angina were increased and that a significantly increased risk was found even for mildly to moderately overweight women. The risk associated with increased QI was not confounded by cholesterol or by other risk factors measured, but it is likely that part of the increase in risk was due to an increase in blood pressure. Despite the established increase in risk for ischaemic heart disease caused by cigarette smoking in women [9], we failed to find a significant increase in risk. We assesse,d smoking habit by serum cotinine; this is certainly a better measure of cigarette smoking during the previous 24 h than is a questionnaire [5], but may be a poorer measure of long-term smoking habits. A relationship between smoking and ischaemic heart disease may therefore have been reduced by exposure misclassification. Furthermore, the relationship of cigarette smoking with angina in women has been observed to be considerably weaker than its relationship with myocardial infarction [9]. Most of our cases were of angina, so that the weak association between serum cotinine and risk is consistent with other studies on angina and the absence of an association with infarction may be due to chance. The low prevalence of smoking in our population also reduces the power of the study to detect relationships with smoking. Finally, smoking is associated with a number of other factors such as alcohol intake and social class, which could confound the relationship of smoking with ischaemic heart disease. There was a strong positive relationship between cholesterol and risk for ischaemic heart disease, in agreement with other studies [2]. The risk associated with cholesterol was stronger than that associated with apolipoprotein B, the major apolipoprotein of low-density lipoprotein (LDL) cholesterol. Apolipoprotein A-I, the major apolipoprotein of high-density lipoprotein (HDL) cholesterol, was inversely related to risk and this
184 relationship became stronger after adjusting for QI and cholesterol. Apolipoprotein(a), the apolipoprotein of lipoprotein(a) [lo], was positively associated with risk for ischaemic heart disease; this association was not statistically significant despite the large difference in means between cases and controls, probably because of the strongly skewed distribution of values. Wood et al. [I l] suggested that low levels of linoleic acid, at least in adipose tissue, may be a risk factor for coronary heart disease. We measured the fatty acid composition of serum total esterified fatty acids rather than of adipose tissue, but this measure has been shown to correlate well with dietary intake of linoleic acid [12]. Serum percentage linoleic acid was a little higher in cases than in controls, providing no support for the hypothesis, but there was a weak negative correlation between percentage linoleic acid and apolipoprotein B, itself a risk factor for ischaemic heart disease. Serum percentages of eicosapentaenoic and docosahexaenoic acids were lower in cases than in controls. These differences were small and were not statistically significant, but are consistent with the hypothesized protective effect of omega 3 fatty acids [13]. We measured DHEAS because Barrett-Connor et al. [ 141reported an inverse relationship between DHEAS and death from ischaemic heart disease in men, although they failed to find any evidence for this in elderly women [1.5]. Our finding of a weak, between non-significant inverse relationship DHEAS and risk suggests that this hormone may deserve further study. Lapidus et al. [ 161suggested that SHBG may be a predictor of ischaemic heart disease in women, but we found only very weak evidence for such a relationship. In conclusion, the strongest risk factors for ischaemic heart disease in this study were obesity and high serum cholesterol. Both these risk factors can be changed by comprehensive lifestyle changes 1171.
Foulds, Claudine Paluch, Dr. Julia Peet, Su Williams-Yeagers and Hayley Willson for help with data collection on Guernsey; Professor D.A. Wood for help in the assessment of case records and ECG traces and Helen Powell for typing the manuscript. References 1
2
3
4
(1990) 882. Bush, T.L., Fried, L.P. and Barrett-Connor, E., Cholesterol, lipoproteins and coronary heart disease in women, Clin. Chem., 34 (1988) B60. Moore, J.W., Clark, G.M.G., Hoare, S.A., Millis, R.R., Hayward, J.L., Quinlan, M.K., Wang, D.Y. and Bulbrook, R.D., Binding of oestradiol to blood proteins and aetiology of breast cancer, Jnt. J. Cancer, 38 (1986) 625. Feyerabend, C. and Russell, M.A.H., A rapid gas-liquid chromatographic determination of cotinine in biological fluids, Analyst, 105 (1980) 993.
5
Jarvis, M.J., Tunstall-Pedoe, H., Feyerabend, C., Vesey, C. and Saloojee, Y., Comparison of tests used to distinguish smokers from nonsmokers, Am. J. Public Health, 77 (1987) 1435.
6
1
Lepage, G. and Roy, C.C., Direct transesteritication of all classes of lipids in a one-step reaction, J. Lipid Res., 27 (1986) 114. Hammond, G.L., Langley, M.S. and Robinson, P.A., A
8
liquid-phase immunoradiometric assay (IRMA) for human sex hormone binding globulin (SHBG), J. Steroid B&hem., 23 (1985) 451. MacMahon, S., Pete, R., Cutler, J., Collins, R., Sorlie, P., Neaton, J., Abbott, R., Godwin, J., Dyer, A. and Stamler, J., Blood pressure, stroke and coronary heart disease. Part 1, prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution
9
10
bias, Lancet, 335 (1990) 765. Willett, W.C., Green, A., Stampfer, M.J., Speizer, F.E., Colditz, GA., Rosner, B., Monson, R.R., Stason, W. and Hennekens, C.H., Relative and absolute excess risks of coronary heart disease among women who smoke cigarettes, N. Engl. J. Med., 317 (1987) 1303. Utermann, G., The mysteries of lipoprotein(a), Science, 246 (I 989) 904.
II
Wood, D.A., Butler, S., Riemersma, R.A., Thomson, M. and Oliver, M.F., Adipose tissue and platelet fatty acids and coronary heart disease in Scottish men, Lancet, ii (1984) 117.
12
Houwelingen, A.C.V., Kester, A.D.M., Kromhout, D. and Hornstra, G., Comparison between habitual intake of polyunsaturated fatty acids and their concentrations in serum lipid fractions, Eur. J. Clin. Nutr., 43 (1989) I I.
Acknowledgements
We are most grateful to: the women volunteers on Guernsey who made the study possible; the GPs and staff at the general practices on the island for allowing us access to their practices; Sue
Manson, J.E., Colditz, GA., Stampfer, M.J., Willett, W.C., Rosner, B., Monson, R.R., Speizer, F.E. and Hennekens, C.H., A prospective study of obesity and risk of coronary heart disease in women, N. Engl. J. Med., 322
185 I3 14
I5
I6
C. and Gredmark,
Leaf, A. and Weber, P.C., Cardiovascular effects of n-3 fatty acids, N. Engl. J. Med., 318 (1988) 549. Barrett-Connor, E., Khaw, K-T. and Yen, S.S.C., A prospective study of dehydroepiandrosterone sulfate, mortality and cardiovascular disease, N. Engl. J. Med., 315 (1986) 1519. Barrett-Connor, E. and Khaw, K-T., Absence of an inverse relation of dehydroepiandrosterone sulfate with cardiovascular mortality in postmenopausal women, N. Engl. J. Med. 317 (1987) 711. Lapidus, L., Lindstedt, G., Lundberg, P-A., Bengtsson,
T., Concentrations
of sex-hormone
binding globulin and corticosteroid binding globulin in serum in relation to cardiovascular risk factors and to 12-year incidence of cardiovascular disease and overall mortality in postmenopausal women, Clin. Chem., 32 17
(1986) 146. Ornish, D., Brown, SE., Scherwitz, L.W., Billings, J.H.. Armstrong, W.T., Ports, T.A., McLanahan, S.M.. Kirkeeide, R.L., Brand, R.J. and Gould, K.L., Can lifestyle changes reverse coronary heart disease? Lancet. 336 (1990) 129.
