International Journal of Epidemiology © International Epidemiological Association 1992
Vol. 21, No. 4 Printed in Great Britain
Body Mass Index and the Initial Manifestation of Coronary Heart Disease in Women Aged 40-59 Years C MARY BEARD. ANTHONY ORENCIA. THOMAS KOTTKE AND DAVID J BALLARD
Although several study designs have been used to address the association between obesity and coronary heart disease (CHD) in women, the strength of the association across studies has been conflicting.1"* This observation may reflect lack of uniformity across studies in the diagnostic criteria for CHD and in definition of body mass index. For example, among Framingham women without known cardiac risk factors, relative weight at initial exam has been shown to predict the subsequent development of cardiovascular disease.1 In contrast, neither current body mass index nor relative weight were significant risk factors for CHD among women less than 65 at the 16th biennial Framingham examination, after controlling for other CHD risk factors. 2 Body mass index has recently been
shown to be associated with angina, fatal and nonfatal myocardial infarction among young women.3 A study conducted in Gothenburg, Sweden, has reported an association between body mass index and myocardial infarction, but not between body mass index and angina in an age-stratified female cohort. 4 To further evaluate the association between body mass index and CHD among women aged 40-59. We also reassessed previously published studies to examine the consistency Project to evaluate the role of body mass index and CHD among women aged 40-59. We also reassessed previously published studies to examine the consistency of the magnitude of the relationship between body mass index and CHD among women. A potential problem in diagnosing CHD is the clinical impression dilemma regarding the obese.
Department of Health Sciences Research, Section of Clinical Epidemiology, Mayo Clinic and Mayo Foundation, 200 First street SW, Rochester, MN 55905, USA.
ObCSC W O m e n
n3Ve a ^ P«>pensity for being labelled by their attending physicians as having angina. Thus, a putative association of obesity and CHD may arise
656
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Beard C M (Department of Health Sciences Research, Section of Clinical Epidemiology, Mayo Clinic and Mayo Foundation, 200 First Street SW, Rochester, MN 55905, USA), Orencia A, Kottke T and Ballard D J. Body mass index and the initial manifestation of coronary heart disease in women aged 40-59 years. International Journal of Epidemiology 1992; 21: 656-664. A population-based case-control study was conducted to evaluate body mass as a coronary heart disease (CHD) risk factor among women 40-59 years of age. Cases were women aged 40-59 whose first manifestation of CHD was angina (No. = 133), nonfatal myocardial infarction (No. = 90), and sudden unexpected death (No. = 18) during 1960-1382. Two randomly selected controls were matched on age and time of the initial disease manifestation of the case. The adjusted relative risk for weight and body mass index respectively demonstrated a moderate association with all CHD as well as with angina, but no association with definite CHD (myocardial infarction or sudden unexpected death). To determine if the observed association between body mass index and angina was possibly attributable to differential misclassrfication bias (i.e. obese women were, in contrast to non-obese women, preferentially labelled as having coronary artery disease) data for angina were stratified by confirmed versus unconfirmed cardiac origin. In the unconfirmed angina analysis, the 75th percentile for weight contrasted with the 25th percentile was associated with a 50% increase in the risk of being labelled as having angina (adjusted odds ratio (OR) = 1.59, 95% confidence interval (CD : 1.11-2.28), while a similar contrast for Quetelet Index was also associated with a neariy 2-fold increase in the risk of being labelled as having angina (adjusted OR = 1.74, 95% Cl : 1.18-2.57). A 3-fold statistically significant increase in risk of being labelled as having angina was also observed for similar contrasts of weight and Quetelet Index among women with confirmed cardiac origin for their symptoms. These data from the Rochester Coronary Heart Disease project suggest that anthropometric attributes are independent risk factors for angina but not for first myocardial infarction and sudden unexpected death among women aged 40-59 years old. These findings cannot be explained by differential misclassrfication of obese women. Additional studies of the association between body mass index and CHD among women age 40-59 should precisely define CHD and obesity measures to avoid possible differential misclassification bias.
BODY MASS INDEX AND CORONARY HEART DISEASE
from the expectation of physicians. In epidemiological practice, unconfirmed CHD, specifically angina, is often used as a risk factor endpoint in assessing disease aetiology. The differential labelling hypothesis suggests that the proportion of misclassified obese women without CHD who are labelled as having disease is greater than that for non-obese women. Using the resources of the Rochester Coronary Heart Disease Project, we assessed this hypothesis by evaluating individuals with angina who had confirmed and nonconfirmed CHD.
Cases Incident cases included all female residents of Rochester, 40-59 years old, whosefirstmanifestation of CHD was sudden unexpected death (No. = 18), myocardial infarction (No. = 90), or angina (No. = 133) during 1960-1982. These diagnostic categories are mutually exclusive. Cases were identified from ongoing studies of CHD in the Rochester community.10 Details of the clinical criteria for defining cases in this project have been described elsewhere." Briefly, angina without confirmed CHD was defined by the study clinicians after careful review of the appropriate records and was based on the standard criteria for patient-reported symptoms, including substernal pressure, pain, or tightness associated with excitement or exercise, or both, and relieved by rest or nitroglycerin.
Angina was considered confirmed if one or more of the following criteria were met: a) coronary atherosclerosis demonstrated by angiogram defined as greater than 50% occlusion, b) coronary artery bypass surgery, c) angioplasty, or d) myocardial infarction subsequent to the angina dianosis. The diagnosis of myocardial infarction was based upon a clinical history indicative of acute myocardial ischaemia, the presence of serial electrocardiographic changes, and increase in the level of serum enzymes considered to be diagnostic (serum glutamic oxaloacetic transaminase, introduced in 1955; lactate hydrogenase, introduced in 1963; and creatine kinase, introduced in 1964). Sudden unexpected death diagnosis was made: a) if the patient with no prior clinical evidence of CHD died within 24 hours of onset of symptoms strongly indicating acute myocardial infarction and b) if after extensive review of all clinical and autopsy data, no other cause of death was found. In most instances, death occurred within a few minutes. Autopsies had been performed on 90% of the sudden unexpected death cases. The index (reference) date was the date of diagnosis of initial CHD manifestation. Control Subjects Two Rochester women were selected as controls for each case from lists of residents who were registered for medical care in the year the case was diagnosed. Of the control subjects, 93 160/95 mm Hg, or b) antihypertensive medications were prescribed before the date of CHD diagnosis or the corresponding index date for controls. The diagnosis of diabetes before the index date was based on criteria from a long-term study of the incidence of diabetes among Rochester residents.13 These diagnostic criteria used for diabetes mellitus have been reported in detail previously.14 This required fasting hyperglycaemia—more than 120 mg/dl (FolinWu method) for the period 1945-1958 or more than 110 mg/dl (AutoAnalyzer ferrocyanide reductase technique) for the period 1959-1969, on two whole venous blood determinations. Blood glucose values from other institutions where Rochester patients sought medical care were interpreted in light of the method being used. An oral glucose tolerance test was generally performed when an equivocal fasting or postprandial blood glucose determination was obtained. At the Mayo Clinic, the oral glucose tolerance test was performed by administering 1 g of glucose/kg body weight and determining blood glucose concentrations at 0, 1,2 and 3 hours after the loading dose. Only the 1- and 2-hour values were used for interpretation of the oral glucose tolerance test, and both values had to be increased in comparison with age-specific standards14 for diabetes mellitus to be diagnosed.15 Steroidal oestrogen use (predominantly conjugated oestrogens) was defined as any prescription for these drugs documented in the medical records. Oral contraceptives were not included. Serum cholesterol values were available for some of the cases, but few controls had such determinations in
659
BODY MASS INDEX AND CORONARY HEART DISEASE
TABLE 1 Risk factor prevalence among 241 cases and 481 matched controls with coronary heart disease among women 40-59 years of age in Rochester, Minnesota, 1960-1982
No (ft)
Cases No. - 241 Yes (ft)
Unknown (ft)
No ("ft)
Diabetes
210 (87.1)
31 (12.9)
—
468 (97.3)
Steroidal oestrogen
181 (75.1)
60 (24.9)
-
338 (70.3)
143 (29.7)
Hypertension
118 (48.9)
112 (46.5)
11 (4.6)
365 (75.9)
110 (22.9)
6 (1.2)
63 (26.1)
161 (66.8)
17 (7.1)
219 (45.5)
225 (46.8)
37 (7-7)
13
(2.7)
Unknown (ft) — -
"One control who bad agenesis of the legs was excluded from the analysis of body mass. TABLE 2 Relative risk and 95% confidence interval (CI) of coronary heart disease (CHD), definite coronary heart disease (myocardial infarction and sudden unexpected death) and angina in relationship to weight among women 40-59 years of age In Rochester, Minnesota, 1960-1982 Contrast
Total CHD" (95 ft CI)
Definite CHD (95ft CI)
Total angina" (95 ft CI)
75ft versus 25ft
1.43 (1.13- 1.81)
1.00 (0.67- 1.49)
1.79 (1.31- 2.46)
Diabetes
Yes/No
5.52 (2.24-13.61)
4.96 (1.14-21.57)
4.21 (1.24-14.22)
Smoking
Yes/No
3.16 (2.01- 4.96)
4.57 (2.10- 9.97)
2.55 (1.45- 4.49)
Hypertension
Ye»/No
3.07 (1.98- 4.75)
4.38 (2.15- 8.95)
2.26 (1.24- 4.10)
Risk factor
Weight in kgc
Cases and controls (No.) d
n, = 614
241
n 3 - 373
•Coronary heart disease included definite coronary heart disease, confirmed and unconfirmed angina. "includes angina with or without confirmed coronary artery disease. c For weight measured as a continuous variable, the odds ratio contrasts the change in weight from the 75th percentile relative to the 25th percentile of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variable mentioned in the Table. 'Vj - the available 85ft of 722 cases and controls were used in the analysis. nj - the available 75ft of 323 cases and controls were used in the analysis. n. = the available 93ft of 399 cases and controls were used in the analysis.
CHD in contrast to angina, we examined the hypothesis that our findings for angina may be due to a differential misclassificatio'n of angina status associated with obesity. For this purpose, subgroup analyses were limited to angina cases and controls with or without confirmed CHD (Tables 4 and 5). The difference in the 75th versus the 25th percentile comparison of weight (Table 4) was independently associated with a 3-fold increase in the odds of confirmed CHD (95% CI : 1.47-6.65). Similarly, the 75th versus 25th percentile comparison for Quetelet Index (Table 5) was associated with an approximate 3-fold increase in the odds of confirmed CHD (95% CI : 1.47-5.85). For the 75th versus 25th percentile comparison in weight (Table 4), the risk of unconfirmed CHD was 59% higher among the heavier women (95% CI :
1.11-2.28). Similarly, for the 75th versus 25th percentile Quetelet Index comparison (Table 5), the risk of unconfirmed CHD was approximately 2-fold greater among women at the higher Quetelet Index level (95% CI : 1.18-2.57). DISCUSSION There was no association detected between definite CHD (myocardial infarction and sudden unexpected death) and body mass index in the present study. When total CHD and total angina endpoints respectively, were evaluated, a significant association with body mass index was observed. After angina cases were divided into subcategories representing confirmed versus unconfirmed CHD, independent associations with weight and Quetelet Index were noted for both. Thus, these results do not support the hypothesis that
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Smoking
Controls No. = 481* Yes (ft)
660
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 3 Relative risk and 95% confidence interval (Cl) of coronary heart disease (CHD), definite coronary heart disease (myocardial infarction and sudden unexpected death) and angina in relationship to body mass index among women 40-59 years of age in Rochester, Minnesota, 1960-1982 Risk factor
Totil CHD* (95% CO
Definite CHD (95% Q )
Total angina'' (95% CI)
75% versus 25%
1.49 (1.17- 1.91)
0.97 (0.64- 1.47)
1.95 (1.40- 2.71)
Diabetes
Yes/No
5.84 (2.28-14.95)
6.21 (1.17-32.95)
3.97 (1.16-13.53)
Smoking
Yes/No
3.13 (1.99- 4.93)
4.39 (1.99- 9.69)
2.69 (1.51- 4.78)
Hypertension
Yes/No
3.15 (2.02- 4.90)
4.66 (2.23- 9.76)
2.22 (1.21- 4.07)
n, = 608
nj = 235
n 3 = 373
Quetelet indexc
Cases and controls (No.) d
"Coronary heart disease included definite coronary heart disease, confirmed and unconfirmed angina. includes angina with or without confirmed coronary artery disease. c For Quetdet Index measured as a continuous variable, the odds ratio contrasts the change from the 75th percentQe relative to the 25th percentfle of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variables mentioned in the Table. 'Vj - the available 84% of 722 cases and controls were used in the analysis. n 2 = the available 73% of 323 cases and controls were used in the analysis. n 3 - the available 93% of 399 cases and controls were used in the analysis.
differential misclassification was responsible for the observed association between body mass index and angina found in this study. If definite CHD was the primary outcome investigated, ignoring angina cases would have led to the conclusion that CHD was not associated with body mass. The Rochester Coronary Heart Disease Project, however, provided the opportunity to evaluate confirmed versus unconfirmed angina. This analysis revealed an even stronger association between body mass and confirmed angina than between body mass and unconfirmed angina. The prevalence of diabetes in the control group was similar to that expected based on prevalence for the community for 1 January 1970, and 1 January 1980, in the age groups studied.14 Similarly, data from other Rochester studies suggest that the Rochester Coronary Heart Project estimates of the prevalence of hypertension, smoking, and steToidal oestrogen exposure in the control group were crudely representative of the prevalence of exposure in the community.18"21 The Rochester Coronary Heart Disease Project has adopted an alternative approach compared to others in ascertaining CHD among women: outcomes were ascertained at first or initial manifestation of disease. Angina, myocardial infarction, sudden unexpected death were recorded at its first clinically ascertained manifestation. In the present study, CHD outcomes were stratified for comparative purposes to evaluate the association between body mass index and CHD using the following subgroups: a) definite or confirmed angina, b) unconfirmed angina, c) definite CHD, and d) total CHD.
Two important cohort studies have opted to utilize a different strategy in ascertaining CHD; hierarchial classification was adopted. Angina cases were reclassified as myocardial infarction or sudden unexpected death if those events occurred within the time frame of the Nurses Health Study.3 Patients with angina who had subsequent myocardial infarction were classified to the later event in the Framingham Study, while patients with angina who died of CHD were retained as angina.22 Because of the population-based nature of the Rochester Study, where complete incidence cases could be documented in this dynamic cohort, the study should be relatively free of selection bias. Given that a true association exists, underascertaining cases may have biased the association downward if the odds of being selected as a control exceeds that of the case group. This problem seems not to be so in the present study. To diminish effects of selective survival, only incidence cases were included. Ideal controls were obtained because they are representative of the community where cases emanated, thus reducing the selection forces of hospitalization. Although we were able to evaluate some CHD risk factors in this study, it was not possible to study the confounding effects of cholesterol. Because cholesterol values were not generally obtained for 40-59 year old women in Rochester during 1960-1982, the effects of this confounding variable could not be examined. The presence of such a value in a medical record may indicate a suspicion of disease, while lack of a measurement predicates lack of suspicion. Furthermore, we
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Contrast
661
BODY MASS INDEX AND CORONARY HEART DISEASE
TABLE 4 Relative risk and 95 % confidence interval (CI) of angina (confirmed coronary heart disease) and unconfirmed angina in relationship to weight among women 40-59 years of age in Rochester, Minnesota, 1960-1982 Risk factor
Contrast
Confirmed angina (93% CO
Unconfirmed angina (95% CI)
75% versus 25%
3.13 (1.47- 6.65)
1.59 (1.11- 2.28)
Diabetes
Yes/No
6.06 (0.82-45.02)
3.50 (0.64-19.09)
Smoking
Yes/No
11.61 (2.38-56.62)
1.73 (0.91- 3.26)
Hypertension
Yes/No
4.72 (1.50-14.80)
1.77 (0.80- 3.91)
Weight in kg*
143
= 230
*For weight as a continuous variable, the contrast is the difference between the 75th peTcentfle and the 25th percentile of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variables mentioned in the Table. %! = the available 93% of 153 cases and controls were used in the analysis, n^ = the available 93% of 246 cases and controls were used in the analysis. TABLE 5 Relative risk and 95% confidence interval (CI) of angina (confirmed coronary heart disease) and unconfirmed coronary heart disease in relationship to body mass index among women 40-59 years of age in Rochester, Minnesota, 1960-1982 Risk factor
Quetelet Index*
Contrast
Confirmed angina (95% CI)
Unconfirmed angina (95% CI)
75% versus 25%
2.87 (1.47- 5.85)
1.74 (1.18- 2.57)
Diabetes
Yes/No
6.00 (0.77-16.96)
3.34 (0.62-18.09)
Smoking
Yes/No
10.14 (2.16-47.48)
1.83 (0.96- 3.49)
Hypertension
Yes/No
4.80 (1.51-15.25)
1.70 (0.77- 3.79)
Cases and controls (No.) b
143
230
*For Quetdet index measured as a continuous variable, the contrast was the difference between the 75th percentile and the 25th percentile of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variables mentioned in the Table. b n, = the available 93% of 153 cases and controls were used in the analysis, n^ - the available 93% of 246 cases and controls were used in the analysis.
were not able to assess the impact of socioeconomic status on the association between body mass index and CHD. Comparison with the Published Literature Using arbitrarily defined categories of Quetelet Index, the Nurses Health Study reported an association for angina with increasing Quetelet Index.3 However, when adjusted for smoking, age, menopausal status, use of postmenopausal oestrogens, hypertension, serum cholesterol, and diabetes, the relative risks were significantly elevated only for those in the highest Quetelet Index category. Table 6a and 6b portrays selected results from cohort and case-control studies assessing the weight-CHD, or body mass index-CHD hypotheses. Results across several studies have ex-
amined different age groups. Definitions for weight including the cutoff points examined to classify obesity are not standardized across those studies. Evidence of an association between obesity and CHD among women, including this study, show inconsistent results with different magnitudes of effect that may be due, at least in part, to different definitions of obesity and of CHD, and to selection of different age groups for study. CONCLUSION AND RECOMMENDATIONS The results of the Rochester Coronary Heart Project suggest that anthropometric attributes are independent risk factors for angina but not for definite CHD among Rochester women aged 40-59 years old. We were unable to attribute these divergent findings to
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Cases and controls (No.) b
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
TABLE 6a
Weight and body mass index (Quetelet Index) as related to coronary heart disease (selected published cohort studies)'
Principal author, year of publication Manson 19903
Sample
Population source
n = 115 886 age 30-55 in 1976
• nurses from 11 states with greatest population density. • free of CHD, stroke and cancer in 1976.
Adjusted relative risks (RR) and 95% confidence limits (CI)
Outcome: nonfatal MI and fatal CHD (combined) Model Follow-up 1984 BMI 21 21-
Vol. 21, No. 4 Printed in Great Britain
Body Mass Index and the Initial Manifestation of Coronary Heart Disease in Women Aged 40-59 Years C MARY BEARD. ANTHONY ORENCIA. THOMAS KOTTKE AND DAVID J BALLARD
Although several study designs have been used to address the association between obesity and coronary heart disease (CHD) in women, the strength of the association across studies has been conflicting.1"* This observation may reflect lack of uniformity across studies in the diagnostic criteria for CHD and in definition of body mass index. For example, among Framingham women without known cardiac risk factors, relative weight at initial exam has been shown to predict the subsequent development of cardiovascular disease.1 In contrast, neither current body mass index nor relative weight were significant risk factors for CHD among women less than 65 at the 16th biennial Framingham examination, after controlling for other CHD risk factors. 2 Body mass index has recently been
shown to be associated with angina, fatal and nonfatal myocardial infarction among young women.3 A study conducted in Gothenburg, Sweden, has reported an association between body mass index and myocardial infarction, but not between body mass index and angina in an age-stratified female cohort. 4 To further evaluate the association between body mass index and CHD among women aged 40-59. We also reassessed previously published studies to examine the consistency Project to evaluate the role of body mass index and CHD among women aged 40-59. We also reassessed previously published studies to examine the consistency of the magnitude of the relationship between body mass index and CHD among women. A potential problem in diagnosing CHD is the clinical impression dilemma regarding the obese.
Department of Health Sciences Research, Section of Clinical Epidemiology, Mayo Clinic and Mayo Foundation, 200 First street SW, Rochester, MN 55905, USA.
ObCSC W O m e n
n3Ve a ^ P«>pensity for being labelled by their attending physicians as having angina. Thus, a putative association of obesity and CHD may arise
656
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Beard C M (Department of Health Sciences Research, Section of Clinical Epidemiology, Mayo Clinic and Mayo Foundation, 200 First Street SW, Rochester, MN 55905, USA), Orencia A, Kottke T and Ballard D J. Body mass index and the initial manifestation of coronary heart disease in women aged 40-59 years. International Journal of Epidemiology 1992; 21: 656-664. A population-based case-control study was conducted to evaluate body mass as a coronary heart disease (CHD) risk factor among women 40-59 years of age. Cases were women aged 40-59 whose first manifestation of CHD was angina (No. = 133), nonfatal myocardial infarction (No. = 90), and sudden unexpected death (No. = 18) during 1960-1382. Two randomly selected controls were matched on age and time of the initial disease manifestation of the case. The adjusted relative risk for weight and body mass index respectively demonstrated a moderate association with all CHD as well as with angina, but no association with definite CHD (myocardial infarction or sudden unexpected death). To determine if the observed association between body mass index and angina was possibly attributable to differential misclassrfication bias (i.e. obese women were, in contrast to non-obese women, preferentially labelled as having coronary artery disease) data for angina were stratified by confirmed versus unconfirmed cardiac origin. In the unconfirmed angina analysis, the 75th percentile for weight contrasted with the 25th percentile was associated with a 50% increase in the risk of being labelled as having angina (adjusted odds ratio (OR) = 1.59, 95% confidence interval (CD : 1.11-2.28), while a similar contrast for Quetelet Index was also associated with a neariy 2-fold increase in the risk of being labelled as having angina (adjusted OR = 1.74, 95% Cl : 1.18-2.57). A 3-fold statistically significant increase in risk of being labelled as having angina was also observed for similar contrasts of weight and Quetelet Index among women with confirmed cardiac origin for their symptoms. These data from the Rochester Coronary Heart Disease project suggest that anthropometric attributes are independent risk factors for angina but not for first myocardial infarction and sudden unexpected death among women aged 40-59 years old. These findings cannot be explained by differential misclassrfication of obese women. Additional studies of the association between body mass index and CHD among women age 40-59 should precisely define CHD and obesity measures to avoid possible differential misclassification bias.
BODY MASS INDEX AND CORONARY HEART DISEASE
from the expectation of physicians. In epidemiological practice, unconfirmed CHD, specifically angina, is often used as a risk factor endpoint in assessing disease aetiology. The differential labelling hypothesis suggests that the proportion of misclassified obese women without CHD who are labelled as having disease is greater than that for non-obese women. Using the resources of the Rochester Coronary Heart Disease Project, we assessed this hypothesis by evaluating individuals with angina who had confirmed and nonconfirmed CHD.
Cases Incident cases included all female residents of Rochester, 40-59 years old, whosefirstmanifestation of CHD was sudden unexpected death (No. = 18), myocardial infarction (No. = 90), or angina (No. = 133) during 1960-1982. These diagnostic categories are mutually exclusive. Cases were identified from ongoing studies of CHD in the Rochester community.10 Details of the clinical criteria for defining cases in this project have been described elsewhere." Briefly, angina without confirmed CHD was defined by the study clinicians after careful review of the appropriate records and was based on the standard criteria for patient-reported symptoms, including substernal pressure, pain, or tightness associated with excitement or exercise, or both, and relieved by rest or nitroglycerin.
Angina was considered confirmed if one or more of the following criteria were met: a) coronary atherosclerosis demonstrated by angiogram defined as greater than 50% occlusion, b) coronary artery bypass surgery, c) angioplasty, or d) myocardial infarction subsequent to the angina dianosis. The diagnosis of myocardial infarction was based upon a clinical history indicative of acute myocardial ischaemia, the presence of serial electrocardiographic changes, and increase in the level of serum enzymes considered to be diagnostic (serum glutamic oxaloacetic transaminase, introduced in 1955; lactate hydrogenase, introduced in 1963; and creatine kinase, introduced in 1964). Sudden unexpected death diagnosis was made: a) if the patient with no prior clinical evidence of CHD died within 24 hours of onset of symptoms strongly indicating acute myocardial infarction and b) if after extensive review of all clinical and autopsy data, no other cause of death was found. In most instances, death occurred within a few minutes. Autopsies had been performed on 90% of the sudden unexpected death cases. The index (reference) date was the date of diagnosis of initial CHD manifestation. Control Subjects Two Rochester women were selected as controls for each case from lists of residents who were registered for medical care in the year the case was diagnosed. Of the control subjects, 93 160/95 mm Hg, or b) antihypertensive medications were prescribed before the date of CHD diagnosis or the corresponding index date for controls. The diagnosis of diabetes before the index date was based on criteria from a long-term study of the incidence of diabetes among Rochester residents.13 These diagnostic criteria used for diabetes mellitus have been reported in detail previously.14 This required fasting hyperglycaemia—more than 120 mg/dl (FolinWu method) for the period 1945-1958 or more than 110 mg/dl (AutoAnalyzer ferrocyanide reductase technique) for the period 1959-1969, on two whole venous blood determinations. Blood glucose values from other institutions where Rochester patients sought medical care were interpreted in light of the method being used. An oral glucose tolerance test was generally performed when an equivocal fasting or postprandial blood glucose determination was obtained. At the Mayo Clinic, the oral glucose tolerance test was performed by administering 1 g of glucose/kg body weight and determining blood glucose concentrations at 0, 1,2 and 3 hours after the loading dose. Only the 1- and 2-hour values were used for interpretation of the oral glucose tolerance test, and both values had to be increased in comparison with age-specific standards14 for diabetes mellitus to be diagnosed.15 Steroidal oestrogen use (predominantly conjugated oestrogens) was defined as any prescription for these drugs documented in the medical records. Oral contraceptives were not included. Serum cholesterol values were available for some of the cases, but few controls had such determinations in
659
BODY MASS INDEX AND CORONARY HEART DISEASE
TABLE 1 Risk factor prevalence among 241 cases and 481 matched controls with coronary heart disease among women 40-59 years of age in Rochester, Minnesota, 1960-1982
No (ft)
Cases No. - 241 Yes (ft)
Unknown (ft)
No ("ft)
Diabetes
210 (87.1)
31 (12.9)
—
468 (97.3)
Steroidal oestrogen
181 (75.1)
60 (24.9)
-
338 (70.3)
143 (29.7)
Hypertension
118 (48.9)
112 (46.5)
11 (4.6)
365 (75.9)
110 (22.9)
6 (1.2)
63 (26.1)
161 (66.8)
17 (7.1)
219 (45.5)
225 (46.8)
37 (7-7)
13
(2.7)
Unknown (ft) — -
"One control who bad agenesis of the legs was excluded from the analysis of body mass. TABLE 2 Relative risk and 95% confidence interval (CI) of coronary heart disease (CHD), definite coronary heart disease (myocardial infarction and sudden unexpected death) and angina in relationship to weight among women 40-59 years of age In Rochester, Minnesota, 1960-1982 Contrast
Total CHD" (95 ft CI)
Definite CHD (95ft CI)
Total angina" (95 ft CI)
75ft versus 25ft
1.43 (1.13- 1.81)
1.00 (0.67- 1.49)
1.79 (1.31- 2.46)
Diabetes
Yes/No
5.52 (2.24-13.61)
4.96 (1.14-21.57)
4.21 (1.24-14.22)
Smoking
Yes/No
3.16 (2.01- 4.96)
4.57 (2.10- 9.97)
2.55 (1.45- 4.49)
Hypertension
Ye»/No
3.07 (1.98- 4.75)
4.38 (2.15- 8.95)
2.26 (1.24- 4.10)
Risk factor
Weight in kgc
Cases and controls (No.) d
n, = 614
241
n 3 - 373
•Coronary heart disease included definite coronary heart disease, confirmed and unconfirmed angina. "includes angina with or without confirmed coronary artery disease. c For weight measured as a continuous variable, the odds ratio contrasts the change in weight from the 75th percentile relative to the 25th percentile of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variable mentioned in the Table. 'Vj - the available 85ft of 722 cases and controls were used in the analysis. nj - the available 75ft of 323 cases and controls were used in the analysis. n. = the available 93ft of 399 cases and controls were used in the analysis.
CHD in contrast to angina, we examined the hypothesis that our findings for angina may be due to a differential misclassificatio'n of angina status associated with obesity. For this purpose, subgroup analyses were limited to angina cases and controls with or without confirmed CHD (Tables 4 and 5). The difference in the 75th versus the 25th percentile comparison of weight (Table 4) was independently associated with a 3-fold increase in the odds of confirmed CHD (95% CI : 1.47-6.65). Similarly, the 75th versus 25th percentile comparison for Quetelet Index (Table 5) was associated with an approximate 3-fold increase in the odds of confirmed CHD (95% CI : 1.47-5.85). For the 75th versus 25th percentile comparison in weight (Table 4), the risk of unconfirmed CHD was 59% higher among the heavier women (95% CI :
1.11-2.28). Similarly, for the 75th versus 25th percentile Quetelet Index comparison (Table 5), the risk of unconfirmed CHD was approximately 2-fold greater among women at the higher Quetelet Index level (95% CI : 1.18-2.57). DISCUSSION There was no association detected between definite CHD (myocardial infarction and sudden unexpected death) and body mass index in the present study. When total CHD and total angina endpoints respectively, were evaluated, a significant association with body mass index was observed. After angina cases were divided into subcategories representing confirmed versus unconfirmed CHD, independent associations with weight and Quetelet Index were noted for both. Thus, these results do not support the hypothesis that
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Smoking
Controls No. = 481* Yes (ft)
660
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 3 Relative risk and 95% confidence interval (Cl) of coronary heart disease (CHD), definite coronary heart disease (myocardial infarction and sudden unexpected death) and angina in relationship to body mass index among women 40-59 years of age in Rochester, Minnesota, 1960-1982 Risk factor
Totil CHD* (95% CO
Definite CHD (95% Q )
Total angina'' (95% CI)
75% versus 25%
1.49 (1.17- 1.91)
0.97 (0.64- 1.47)
1.95 (1.40- 2.71)
Diabetes
Yes/No
5.84 (2.28-14.95)
6.21 (1.17-32.95)
3.97 (1.16-13.53)
Smoking
Yes/No
3.13 (1.99- 4.93)
4.39 (1.99- 9.69)
2.69 (1.51- 4.78)
Hypertension
Yes/No
3.15 (2.02- 4.90)
4.66 (2.23- 9.76)
2.22 (1.21- 4.07)
n, = 608
nj = 235
n 3 = 373
Quetelet indexc
Cases and controls (No.) d
"Coronary heart disease included definite coronary heart disease, confirmed and unconfirmed angina. includes angina with or without confirmed coronary artery disease. c For Quetdet Index measured as a continuous variable, the odds ratio contrasts the change from the 75th percentQe relative to the 25th percentfle of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variables mentioned in the Table. 'Vj - the available 84% of 722 cases and controls were used in the analysis. n 2 = the available 73% of 323 cases and controls were used in the analysis. n 3 - the available 93% of 399 cases and controls were used in the analysis.
differential misclassification was responsible for the observed association between body mass index and angina found in this study. If definite CHD was the primary outcome investigated, ignoring angina cases would have led to the conclusion that CHD was not associated with body mass. The Rochester Coronary Heart Disease Project, however, provided the opportunity to evaluate confirmed versus unconfirmed angina. This analysis revealed an even stronger association between body mass and confirmed angina than between body mass and unconfirmed angina. The prevalence of diabetes in the control group was similar to that expected based on prevalence for the community for 1 January 1970, and 1 January 1980, in the age groups studied.14 Similarly, data from other Rochester studies suggest that the Rochester Coronary Heart Project estimates of the prevalence of hypertension, smoking, and steToidal oestrogen exposure in the control group were crudely representative of the prevalence of exposure in the community.18"21 The Rochester Coronary Heart Disease Project has adopted an alternative approach compared to others in ascertaining CHD among women: outcomes were ascertained at first or initial manifestation of disease. Angina, myocardial infarction, sudden unexpected death were recorded at its first clinically ascertained manifestation. In the present study, CHD outcomes were stratified for comparative purposes to evaluate the association between body mass index and CHD using the following subgroups: a) definite or confirmed angina, b) unconfirmed angina, c) definite CHD, and d) total CHD.
Two important cohort studies have opted to utilize a different strategy in ascertaining CHD; hierarchial classification was adopted. Angina cases were reclassified as myocardial infarction or sudden unexpected death if those events occurred within the time frame of the Nurses Health Study.3 Patients with angina who had subsequent myocardial infarction were classified to the later event in the Framingham Study, while patients with angina who died of CHD were retained as angina.22 Because of the population-based nature of the Rochester Study, where complete incidence cases could be documented in this dynamic cohort, the study should be relatively free of selection bias. Given that a true association exists, underascertaining cases may have biased the association downward if the odds of being selected as a control exceeds that of the case group. This problem seems not to be so in the present study. To diminish effects of selective survival, only incidence cases were included. Ideal controls were obtained because they are representative of the community where cases emanated, thus reducing the selection forces of hospitalization. Although we were able to evaluate some CHD risk factors in this study, it was not possible to study the confounding effects of cholesterol. Because cholesterol values were not generally obtained for 40-59 year old women in Rochester during 1960-1982, the effects of this confounding variable could not be examined. The presence of such a value in a medical record may indicate a suspicion of disease, while lack of a measurement predicates lack of suspicion. Furthermore, we
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Contrast
661
BODY MASS INDEX AND CORONARY HEART DISEASE
TABLE 4 Relative risk and 95 % confidence interval (CI) of angina (confirmed coronary heart disease) and unconfirmed angina in relationship to weight among women 40-59 years of age in Rochester, Minnesota, 1960-1982 Risk factor
Contrast
Confirmed angina (93% CO
Unconfirmed angina (95% CI)
75% versus 25%
3.13 (1.47- 6.65)
1.59 (1.11- 2.28)
Diabetes
Yes/No
6.06 (0.82-45.02)
3.50 (0.64-19.09)
Smoking
Yes/No
11.61 (2.38-56.62)
1.73 (0.91- 3.26)
Hypertension
Yes/No
4.72 (1.50-14.80)
1.77 (0.80- 3.91)
Weight in kg*
143
= 230
*For weight as a continuous variable, the contrast is the difference between the 75th peTcentfle and the 25th percentile of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variables mentioned in the Table. %! = the available 93% of 153 cases and controls were used in the analysis, n^ = the available 93% of 246 cases and controls were used in the analysis. TABLE 5 Relative risk and 95% confidence interval (CI) of angina (confirmed coronary heart disease) and unconfirmed coronary heart disease in relationship to body mass index among women 40-59 years of age in Rochester, Minnesota, 1960-1982 Risk factor
Quetelet Index*
Contrast
Confirmed angina (95% CI)
Unconfirmed angina (95% CI)
75% versus 25%
2.87 (1.47- 5.85)
1.74 (1.18- 2.57)
Diabetes
Yes/No
6.00 (0.77-16.96)
3.34 (0.62-18.09)
Smoking
Yes/No
10.14 (2.16-47.48)
1.83 (0.96- 3.49)
Hypertension
Yes/No
4.80 (1.51-15.25)
1.70 (0.77- 3.79)
Cases and controls (No.) b
143
230
*For Quetdet index measured as a continuous variable, the contrast was the difference between the 75th percentile and the 25th percentile of the entire case and control distribution adjusted for age, steroidal oestrogen use, menopause status, and other variables mentioned in the Table. b n, = the available 93% of 153 cases and controls were used in the analysis, n^ - the available 93% of 246 cases and controls were used in the analysis.
were not able to assess the impact of socioeconomic status on the association between body mass index and CHD. Comparison with the Published Literature Using arbitrarily defined categories of Quetelet Index, the Nurses Health Study reported an association for angina with increasing Quetelet Index.3 However, when adjusted for smoking, age, menopausal status, use of postmenopausal oestrogens, hypertension, serum cholesterol, and diabetes, the relative risks were significantly elevated only for those in the highest Quetelet Index category. Table 6a and 6b portrays selected results from cohort and case-control studies assessing the weight-CHD, or body mass index-CHD hypotheses. Results across several studies have ex-
amined different age groups. Definitions for weight including the cutoff points examined to classify obesity are not standardized across those studies. Evidence of an association between obesity and CHD among women, including this study, show inconsistent results with different magnitudes of effect that may be due, at least in part, to different definitions of obesity and of CHD, and to selection of different age groups for study. CONCLUSION AND RECOMMENDATIONS The results of the Rochester Coronary Heart Project suggest that anthropometric attributes are independent risk factors for angina but not for definite CHD among Rochester women aged 40-59 years old. We were unable to attribute these divergent findings to
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
Cases and controls (No.) b
Downloaded from http://ije.oxfordjournals.org/ at University of Bath Library & Learning Centre on June 12, 2015
TABLE 6a
Weight and body mass index (Quetelet Index) as related to coronary heart disease (selected published cohort studies)'
Principal author, year of publication Manson 19903
Sample
Population source
n = 115 886 age 30-55 in 1976
• nurses from 11 states with greatest population density. • free of CHD, stroke and cancer in 1976.
Adjusted relative risks (RR) and 95% confidence limits (CI)
Outcome: nonfatal MI and fatal CHD (combined) Model Follow-up 1984 BMI 21 21-
