Widlansky ME, Sesso HD, Rexrode KM, Manson JE, Gaziano JM. Body Mass Index and Total and Cardiovascular Mortality in Men With a History of Cardiovascular Disease. Arch Intern Med. 2004;164(21):2326-2332. doi:10.1001/archinte.164.21.2326
Copyright 2004 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2004
Previous studies designed to identify an association between body mass index (BMI) (calculated as weight in kilograms divided by the square of height in meters) and cardiovascular or total mortality in populations with known atherosclerotic disease have shown conflicting results. In this study, we used the Physicians’ Health Study enrollment cohort to examine the risk of total and cardiovascular mortality among men reporting a history of myocardial infarction or stroke, excluding those who reported a history of cancer.
Cause-specific death was ascertained for 5010 men during a mean follow-up of 5.0 years. End points were classified as total deaths and deaths due to cardiovascular causes. Four BMI categories (<22.0, 22.0-24.9 [referent], 25.0-27.9, and ≥28.0) were created a priori. We used proportional hazards models to calculate age and multivariate-adjusted relative risks (RRs) for each BMI category for each end point.
Compared with men with a BMI of 22.0 to 24.9, men with a BMI of 28.0 or greater had an age-adjusted RR of 1.11 (95% confidence interval [CI], 0.91-1.36), a multivariate RR of 1.04 (95% CI, 0.84-1.28) in a model that did not include biological mediators of obesity, and a multivariate RR of 1.06 (95% CI, 0.78-1.44) in a model that included these mediators. The RRs for cardiovascular mortality were similar, at 1.07 (95% CI, 0.85-1.35), 1.01 (95% CI, 0.79-1.29), and 1.01 (95% CI, 0.71-1.43), respectively. A BMI of less than 22.0 was associated with a small increased risk of total mortality and cardiovascular mortality.
These findings indicate that elevated BMI may not be strongly associated with total or cardiovascular mortality among men with previously manifested coronary artery disease.
The relationship between body mass index (BMI) (calculated as weight in kilograms divided by the square of height in meters) and total and cardiovascular mortality has been extensively studied. Elevated BMI is associated with an increased risk of death in several studies.1- 7 Cardiovascular morbidity and mortality have alsobeen associated with elevated BMI in several studies in populations without cardiovascular disease.8- 12 However, few studies have specifically examined this relationship in populations with preexisting cardiovascular disease. Those studies that included subjects with previously diagnosed cardiovascular events have not separated this population when calculating their relative risks (RRs) for subsequent fatal events.6,9,10
Studies have specifically examined BMI and the rate of repeat myocardial infarction (MI) with conflicting results.13- 16 Ness et al13 examined 2033 men in the Diet and Reinfarction Trial and found no relationship between elevated BMI and recurrent MI in multivariate models, including among nonsmokers. On the contrary, Rea et al14 found that those with a BMI of greater than 30 had an elevated RR for recurrent MI, independent of diabetes, hypertension, and dyslipidemia. Kaplan et al15 found a modest increased RR for recurrent cardiovascular events in patients with a BMI of greater than 32 compared with a reference group composed of mildly overweight patients. Using the Physicians’ Health Study enrollment cohort, we examined the association of BMI with total and cardiovascular mortality among men reporting a history of MI or stroke.
The Physicians’ Health Study was a randomized, double-masked, placebo-controlled, primary prevention trial testing whether 325 mg of aspirin taken on alternate days decreases mortality due to cardiovascular disease, and whether 50 mg of beta carotene taken on alternate days decreases the incidence of cancer.17 In 1982 and 1983, we sent letters of invitation, informed consent forms, and baseline questionnaires to 261 248 men listed on an American Medical Association mailing tape. By December 31, 1983, 104 353 physicians had answered the enrollment questionnaire. Our analyses focus on the 5010 men excluded from the trial because of a history of MI and/or stroke, but with no baseline history of cancer and with a BMI between 15 and 45.
On the baseline questionnaire, respondents provided self-reported age, height, weight, and disease history, including history of MI and stroke. Body mass index was calculated from self-reported weight and height. Data on coronary risk factors included smoking status (never, past, or current, and the number of cigarettes smoked daily), use of antihypertensive medication (never, past, or current), total cholesterol concentration, history of diabetes mellitus and angina pectoris, and frequency of vigorous exercise (2-6 times/wk, 1 time/wk, 1 time/mo, or rarely/never). Current hypertension was defined as the use of antihypertensive medications as indicated by the questionnaire, systolic blood pressure of 140 mm Hg or greater, and/or diastolic blood pressure of 90 mm Hg or greater. Elevated cholesterol level was defined as a baseline reported total cholesterol level of 240 mg/dL or greater (≥6.21 mmol/L). The men also gave information about alcohol intake (rarely/never, 1-3 times/mo, 1-6 times/wk, or ≥1 time/d) and use of multivitamins (never, past, or current), ascorbic acid (vitamin C) (never, past, or current), and aspirin (yes/no).
Deaths through January 31, 1989, were identified using the National Death Index, after which we obtained death certificates from state agencies. Deaths were classified by trained nosologists according to the first revision of the International Classification of Diseases, Ninth Revision, in conjunction with the Automated Classification of Medical Entities Decision Tables to manually select underlying cause of deaths. Primary end points for the present study were total, cardiovascular, and noncardiovascular mortality. The mean follow-up for the men was 5.0 years.
We created 4 categories of BMI (<22.0, 22.0-24.9, 25.0-27.9, and ≥28.0) a priori to ensure a balanced distribution of subjects with a sufficient number of events in the leanest and heaviest categories. We calculated the means or proportions of the baseline characteristics and risk factors for each category of BMI and compared them with the reference group (BMI, 22.0-24.9). Age- and multivariate-adjusted RRs for total and cause-specific mortality were calculated for each BMI category using Cox proportional hazards. The first multivariate analysis adjusted for age; use of aspirin, alcohol, multivitamins, and ascorbic acid; and exercise level. The second multivariate model added possible biological mediators or the effects of obesity (diabetes, current hypertension, and elevated cholesterol level) to the first multivariate model to quantify to what extent these sequelae of obesity mediated any association. We further stratified subjects by smoking status (never, past, and current; also considered as never and past/current). We also examined men who reported only previous MI (without history of stroke) to evaluate whether a history of MI alone had a stronger relationship of elevated BMI with mortality than the combination of previous MI and/or stroke. Too few fatal events occurred among participants who reported only previous stroke (not MI) in the leanest and heaviest BMI categories to evaluate an association in the stroke-only population appropriately. All RRs are presented with 95% confidence intervals (CIs).
Seven thousand two hundred twenty-five men reported a history of stroke or MI, of whom 2211 were excluded because of reported history of cancer, and another 4 because of extreme BMI values. The remaining 5010 men served as the population for analysis. Overall total and cardiovascular mortality rates (deaths/1000 person-years) were 36.2 and 27.9, respectively. The mean BMI for the cohort was 25.1 (SD, 3.2; median, 24.8) with an average follow-up time of 5.0 years. Their clinical characteristics are reported in Table 1. Those in the highest and lowest BMI categories exercised significantly less than those with BMIs of 22.0 to 24.9 (reference category). Those with the smallest BMI also had a significantly higher prevalence of current smoking (at the expense of the number of past smokers) than the reference group. Those with a BMI of greater than 28.0 had a significantly higher prevalence of hypertension and diabetes mellitus. We calculated the crude incidence rates of total, cardiovascular, and stroke mortality for the 5010 men according to BMI category. Men with a BMI of less than 22.0 had incidence rates of total and cardiovascular mortality of 57.7 and 43.1, respectively, which were higher than those in the reference category (36.2 and 28.0, respectively). Crude total mortality rate for those with a BMI of greater than 28.0 was 34.4, with a cardiovascular mortality rate of 25.6.
Table 2 displays the RR for mortality due to any cause in this population, based on age- and multivariate-adjusted models. The multivariable models differ in only the exclusion (model 1) or inclusion (model 2) of possible biological mediators of the effects of obesity (diabetes, hypertension, and elevated cholesterol level). A BMI of 28.0 or greater did not place the subjects at any greater risk of death than the index population, regardless of smoking status (never, past, or current). Age- and multivariate-adjusted RRs (models 1 and 2) for the entire population with a BMI of 28.0 or greater were 1.11 (95% CI, 0.91-1.36), 1.04 (95% CI, 0.84-1.28), and 1.06 (95% CI, 0.78-1.44), respectively. A BMI of less than 22.0 was associated with increased risk of death due to any cause in the total and nonsmoker populations. Although men in the group with a BMI of 25.0 to 27.9 tended toward protection from total mortality across never smokers, past smokers, and past and current smokers combined, these results were not statistically significant.
Risk factors mediated by obesity, including diabetes (RR, 2.02 [95% CI, 1.59-2.55]) and hypertension (RR, 1.36 [95% CI, 1.11-1.68]), imparted increased risk of total mortality in the entire population in the second multivariate model. Vigorous exercise was also associated with significant reductions in total mortality. An elevated total cholesterol level of 240 mg/dL or greater (≥6.21 mmol/L) did not impart any increased RR of total mortality in the overall population or by smoking status.
Table 3 displays the RRs of cardiovascular mortality in this population. There was no increased risk of cardiovascular events in the group with a BMI of 28.0 or greater compared with the reference group in any smoking subcategory and in the population as a whole. Age- and multivariate-adjusted RRs (models 1 and 2) of cardiovascular death for the entire population were 1.07 (95% CI, 0.85-1.35), 1.01 (95% CI, 0.79-1.28), and 1.01 (95% CI, 0.71-1.43), respectively. An association was found between a BMI of less than 22.0 and increased mortality due to cardiovascular events in the multivariate models for all men as well as those who never smoked. Again, as for total mortality, increased risk for cardiovascular mortality was imparted on subjects with diabetes mellitus and hypertension. A lower RR of cardiovascular mortality was associated with vigorous exercise greater than 1 time per week, consistent with previous studies as stated earlier. No significantly increased risk was associated with total cholesterol levels of 240 mg/dL or greater (≥6.21 mmol/L) in the entire cohort, regardless of smoking status (data for these non-BMI risk factors are not shown).
We evaluated whether BMI might have a different association among those who previously had only MI, but not strokes, in Table 4. A higher BMI was not associated with mortality, regardless of smoking status. We also recategorized the men with BMI of 25.0 or greater according to the recommendations of the World Health Organization18 to examine whether this categorization would reveal any association between obesity and total or cardiovascular mortality in Table 5. A BMI of 30.0 or greater was associated with multivariate-adjusted RRs of 1.01 (95% CI, 0.75-1.37) and 1.11 (95% CI, 0.72-1.70) for total mortality and 0.94 (95% CI, 0.66-1.34) and 1.02 (95% CI, 0.62-1.68) for cardiovascular mortality.
We reanalyzed the population excluding 318 men with less than 2 years of follow-up to limit the effects of prevalent severe illness on the association between the leanest and heaviest BMI categories and mortality, leaving 4692 men for analysis. Although this eliminated the association between a BMI of less than 22.0 and increased risk of total and cardiovascular mortality, no association was found between a BMI of 28.0 or greater and total or cardiovascular mortality. Out of concern for missing data in the total cholesterol level variable, we repeated the second multivariate model and omitted elevated total cholesterol level as a variable, but found results similar to those of the original analysis.
In our cohort of middle-aged and older male physicians with a history of cardiovascular disease, elevated BMI was not strongly associated with the risk of total or cardiovascular mortality. This contrasts with primary prevention studies that implicate BMI as a strong risk factor for total and cardiovascular mortality. Furthermore, a BMI of less than 22.0 appeared to portend a poor outcome in these men compared with the reference group. These data raise questions about the impact of BMI on cardiovascular events in those who have had an MI or a stroke.
The associations between increased total and cardiovascular mortality with a BMI of less than 22.0, which are eliminated by excluding those with less than 2 years of follow-up, provide evidence of the associations among comorbid illness, severity of illness, low BMI, and mortality. These associations have been established in primary prevention populations6,11,19 and in congestive heart failure.20 Smoking has also been shown to be associated with an attenuation of the increased risk of cardiovascular disease in those with an elevated BMI in a healthy population.21 Our data are consistent with findings in other populations that prevalent illness confounds the association between low BMI and mortality, precluding a conclusion from this study that an independent association exists between low BMI and mortality in this population.
The question of the level at which an elevated BMI becomes dangerous remains disputed. The latest recommendations from the World Health Organization18 and the National Institutes of Health22 classify a BMI of 25.0 or greater as overweight, 30.0 or greater as class I obesity, and 35.0 or greater as class II obesity. When we used these classifications, we still found no association between BMI and mortality. The present study, however, was limited by the relatively small number of men with a BMI of 35.0 or greater, diminishing the ability to obtain significant findings relevant to these BMI categories.
Many of the early studies of risk factors for subsequent cardiovascular events in patients with previous MI did not focus on BMI as a risk factor.23- 26 Schlant et al26 considered body weight, which was not significantly related. Wong et al27 followed up 697 men and women in the Framingham study after first MI for an average of 9.7 years. Obesity treated as a continuous variable and as measured by the Metropolitan Relative Weight Scale was inversely related to mortality, likely because those who are underweight may be sicker.
Rea et al14 followed up 2341 men and women after discharge from the hospital after their first MI, looking for risk factors for recurrent MI. A significantly elevated risk of reinfarction was found in those patients with BMIs ranging from 30.0 to 34.9 and of 35.0 or greater compared with a BMI of 24.9 or less. That study, however, was designed to look at the risk of reinfarction, not total and cardiovascular mortality. Ness et al13 followed up 2033 men in the DART trial and found no relationship between elevated BMI and mortality in multivariate analyses. However, their reference group, with a BMI ranging from 16 to 24, likely contained patients with medically debilitating disease leading to low BMI, biasing the RRs. A recent study by Kaplan et al15 followed up 2677 men and women for an average of 3.4 years after a 1.5-year run-in period to limit confounding due to prevalent illness. Using a BMI quintile of 26.2 to 28.6 as the reference BMI, Kaplan et al15 found elevated risk of a combined cardiovascular outcome for those with BMI of greater than 32.
In the context of these previous studies, the present study has several strengths. First, it is the largest study of its kind to date, with nearly twice the number of participants than any previous study. Second, we excluded men with prevalent cancer and stratified results by smoking status to minimize confounding due to prevalent illness and smoking status. This helped to clarify the associations between an elevated BMI and mortality in those without prior cardiovascular disease who would have otherwise gone unnoticed.4,7,28 Further, analyses excluding the first 2 years of follow-up indicated that confounding minimally influenced our results for those with BMI of 28.0 or greater.
The lack of an association between elevated BMI and total and cardiovascular mortality does not eliminate the possibility of a small increase in risk. Although it is a matter of debate as to why BMI may be important to primary prevention of cardiovascular disease but less important in secondary prevention, the presence of atherosclerosis and the severity of cardiovascular disease itself may attenuate the association of elevated BMI with cardiovascular mortality. Low ejection fraction, ventricular arrhythmias, and persistent ischemia have all been identified as harbingers of increased mortality after MI.29- 31 Furthermore, several recent studies have noted that the presence of endothelial dysfunction in the context of coronary artery disease imparts increased risk of adverse cardiovascular events beyond traditional risk factors.32- 35 Men with more extensive infarctions are also more likely to develop cardiac cachexia or may be more motivated to lose weight, thus further biasing the RR of mortality in men with BMIs of 28.0 or greater and 30.0 or greater toward the null. Depression after MI may also influence mortality, along with confounding mortality’s association with BMI in this population.36 The weak RRs seen between BMI and mortality in this study support the notion that risk factors unique to the post-MI population play an important role in modulating risk in this population.
This study has several limitations. First, additional end points would have improved the power to detect a more modest association for higher levels of BMI. Insufficient end points prevented us from considering a BMI category as low as less than 20.0; however, we still observed elevated RRs among men with a BMI of less than 22.0, consistent with previous studies.6,11,19 Our study was powered to detect an unadjusted RR of total mortality of 1.27 for a BMI of 28.0 or greater relative to our reference group. However, more modest increases in the RRs cannot be reliably excluded. Second, we did not update changes in BMI during follow-up. The measurement of BMI at regular intervals during follow-up may have better reflected changes in BMI after a cardiovascular event. Furthermore, we do not have data as to the timing of self-reported mortal events in relation to the follow-up period, which may or may not affect the BMI noted at the beginning of the follow-up period. We were also unable to thoroughly evaluate confounding by diet and medical treatment. However, with physicians as subjects, medical care would be expected to be more homogeneous for their conditions. With no morbidity follow-up, no comment can be made on BMI and the incidence of subsequent nonfatal cardiovascular events, which may exist given the findings of Rea et al.14 Most strokes in the Physicians’ Health Study were ischemic; thus atherosclerotic processes paralleling that of MI may be the most likely cause. This study also relied on self-reported data. However, self-reported height and weight are thought to be reliable,37,38 and physicians more accurately self-report medical data than the general population.39 Thus, it is unlikely that the use of self-reported measures strongly biased our results. Finally, only men were included in this study, warranting a parallel study in women.
Body mass index was not strongly related to increases in the risk of total and cardiovascular mortality in this population of men with previous stroke or MI. This contrasts with studies examining BMI in populations without prior coronary artery disease. The presence of atherosclerosis and severity of underlying cardiovascular disease may attenuate any association between BMI and mortality. Further studies to examine the association of BMI in a secondary prevention setting are needed.
Correspondence: Howard D. Sesso, ScD, MPH, Division of Preventive Medicine, Brigham and Women’s Hospital, 900 Commonwealth Ave E, Boston, MA 02215-1204 (firstname.lastname@example.org).
Accepted for Publication: December 16, 2003.
Financial Disclosure: None.
Funding/Support: This study was supported by research grants CA-34944, CA-47988, CA-40368, HL-26490, HL-34595, and HL-43851 from the National Institutes of Health, Bethesda, Md.
Acknowledgment: We thank Natalya Gomelskaya and Vadim Bubes, PhD, for their assistance with the statistical analyses and the entire staff of the Physicians’ Health Study, under the leadership of Charlene Belanger, MA, as well as Mary Breen, Jean MacFadyen, Geneva McNair, David Potter, Leslie Power, Harriet Samuelson, Miriam Schvartz, Mickie Sheehey, Joanne Smith, and Phyllis Johnson Wojciechowski, for their crucial contributions.