PREVENTIVE MEDICINE 27, 311–316 (1998)ARTICLE NO. PM970303

LEAD ARTICLE

The Correspondence between Coronary Heart Disease Mortality andRisk Factor Prevalence among States in the United States, 1991–1992

Tim Byers,*,1 Rob Anda,† David McQueen,† David Williamson,† Ali Mokdad,† Michelle Casper,†Earl Ford,† and Jim Marks†

*University of Colorado School of Medicine, Campus Box C245, 4200 East Ninth Avenue, Denver, Colorado 80262; and†National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention,

Atlanta, Georgia 30341

of increased efforts to reduce the prevalences of modi-Objective. This study aimed to examine the corre- fiable CHD risk factors in their populations to reduce

spondence between seven established risk factors for CHD mortality. q1998 American Health Foundation and Academic

coronary heart disease (CHD) and CHD mortality Press

among the states in the United States. An ecologic anal- Key Words: coronary heart disease; mortality; to-ysis relating CHD risk factor prevalences to CHD mor- bacco; obesity; physical activity.tality rates among 49 states was undertaken in 1991–1992.

Methods. Approximately 68,000 men and women ages INTRODUCTION45–74 were randomly sampled and interviewed by tele-

Coronary heart disease (CHD) is the leading causephone in surveys conducted in 49 states in 1991 andof death for both men and women in all 50 states in1992. From these interviews, we estimated state-spe-the United States [1]. Mortality rates from CHD varycific prevalences of smoking, overweight, physical in-considerably by state, however. The average CHD mor-activity, hypertension, elevated cholesterol, diabetes,

and alcohol abstinence. These seven CHD risk factors tality for the 10 states with the highest rates is aboutwere also combined to create a CHD risk index for 40% greater than for the 10 states with the lowest rateseach state. The main outcome measures were mortality [1]. Tobacco, excess body weight, physical inactivity,rates from CHD (ICD9 codes 410.0–414.9) in each of 49 hypertension, elevated cholesterol, diabetes, and alco-states in 1991–1992 for men and women ages 45–74. The hol abstinence are all modifiable factors that are estab-analysis was based on multiple linear regression and lished risk factors for CHD [2,3]. The correspondenceSpearman’s rank-order correlations between the CHD between CHD mortality and the prevalences of theserisk factor prevalences, the combined CHD risk index, risk factors among the states in the United States mightand the CHD mortality rates among the 49 states. provide an estimate of the mortality reductions that

Results. The prevalences of most of the CHD risk fac- could result from lowering the prevalences of these risktors correlated with CHD mortality rates in the expec- factors among the populations of states in the Unitedted directions, and correlations were similar for men States.and women. The CHD risk index correlated stronglywith CHD mortality for both men (r = 0.75) and women

METHODS(r = 0.80).Conclusion. About 60% of the variance in CHD mor-

For this ecologic analysis we drew on mortality datatality between the states in the United States (56% forcollected by state health agencies within the Centersmen and 64% for women) is attributable to differences

between the states in the prevalences of seven estab- for Disease Control and Prevention (CDC) standardizedlished risk factors for CHD. As state health agencies mortality reporting system [4], and data from the CDCprioritize resources for chronic disease prevention Behavioral Risk Factor Surveillance System (BRFSS)programs, they should consider the potential benefits [5]. Both the mortality data and the risk factor data

were collected by state departments of health in collabo-1 To whom reprint requests should be addressed. ration with CDC in the 2-year period 1991–1992. All

311 0091-7435/98 $25.00Copyright q 1998 by American Health Foundation and Academic Press

All rights of reproduction in any form reserved.

312 BYERS ET AL.

deaths in which the underlying cause was reported as CHD risk index computed from all risk factors. To createthe CHD risk index for each state, we summed thecoronary ischemic heart disease (ICD 9 codes 410.0–

414.9) were regarded as CHD deaths. We examined only population attributable risks (PAR) of the seven CHDrisk factors for each state8. The PAR is an expressionthe age range of 45–74 because CHD before age 45 is

due more to genetic than to behavioral factors and CHD of the proportion of a disease attributable to a particularfactor, computed as a function of the prevalence of theafter age 74 is less strongly associated with CHD risk

factors and more confounded by competing mortality factor and the relative risk independently attributableto the factor. The CHD risk index is an estimate of thethan is CHD in the 45–74 age range [2].

We estimated state-specific prevalences of seven proportion of CHD mortality that is attributable to thecombined effect of all seven risk factors, computed fromCHD risk factors from interviews conducted in the

BRFSS, a standardized telephone survey that state the PARS as follows:health agencies operate in collaboration with CDC [5].In each month of each year the BRFSS survey staff

o7

i51PARi 5 Pi(RRi 2 1)/[1 1 Pi(RRi 2 1)],select by random digit dialing a sample of adults ages

18 and older in each participating state. Those agreeingto participate are asked questions about behaviors that

where i is the designator for each of seven CHD riskrelate to disease risk. In 1991–1992, 49 states (all ex-factors estimated in this study, PARi is the populationcept Wyoming) participated in the BRFSS. Approxi-attributable risk for risk factor i, RRi is the independentmately 82% of sampled adults agreed to complete therelative risk for CHD mortality for risk factor i, and Pitelephone survey. The present report is based on re-is the state-specific prevalence of risk factor i.sponses from the 28,195 male and 39,674 female re-

Three sets of assumptions about the relative risk forspondents in 1991 and 1992 who were ages 45–74.CHD mortality independently associated with each riskThe BRFSS defines the seven risk factors as follows:factor were used—both the upper and the lower bounds(1) smoker: having smoked more than 100 cigarettesof reasonable estimates for each risk factor and whatduring lifetime and currently smoking regularly, (2)we regarded to be our “best estimate” of the probableoverweight: body mass index based on self-reportedindependent relative risks for CHD for each risk factor:height and weight that exceeds the sex-specific stan-smoking, 1.75 [9]; overweight, 1.50 [10,11]; physicaldards for body mass index (BMI.27.3 kg/m2 for womeninactivity, 1.25 [12–14]; hypertension, 1.25 [15]; ele-and 27.8 kg/m2 for men) [7], (3) physically inactive:vated cholesterol, 1.10 [16]; diabetes, 1.75 [17,18]; andreporting no regular leisure-time physical activity, (4)alcohol abstinence, 1.10 [19–21]. Some of these valueshypertensive: told by a health professional on two orare lower than estimates reported in the literature be-more occasions that blood pressure was elevated or re-cause they reflect the need to use PARs that expressporting currently taking blood pressure medicine, (5)only the risk independently associated with each factor.elevated cholesterol: having ever been tested for choles-We assumed the same relative risks for men andterol and told by a health professional that the choles-women, but used sex-specific risk factor and mortalityterol level was high, (6) diabetic: having been diagnoseddata for the analyses.by a doctor with diabetes mellitus, (7) alcohol abstainer:

not having consumed any alcohol-containing beveragesRESULTSin the past month (beer, wine, or hard liquor).

To account for the complex survey sampling design,we used SUDAAN to compute prevalence estimates of Both the prevalence of CHD risk factors and the CHD

mortality rates varied considerably among the statesthe seven CHD risk factors for each of three age groups(45–54, 55–64, and 65–74) for each gender in each state (Table 1). Between-state variability was greater for

physical activity and alcohol abstinence than for the[6]. Both the risk factor prevalences and the CHD mor-tality rates were then age-adjusted by the direct method other risk factors. The rank-order correlations were in

the expected direction between the seven CHD riskto the 1990 standard U.S. population (weighting fac-tors: 45–54, 0.39; 55–64, 0.33; 65–74, 0.28). The vari- factors and the CHD mortality for most of the factors

(Table 2). Univariate correlations were the highest forability in risk factor prevalences among states was ex-pressed as the interquartile range (the range from the physical inactivity in both sexes, for smoking in men,

and for overweight in women. Among men, correlations25th to the 75th percentiles).The risk factor prevalence estimates were then re- were the lowest for those conditions that require a medi-

cal diagnosis and are amenable to medical treatment—lated to CHD mortality rates among the 49 states bythree methods—univariate rank-order correlations for hypertension, diabetes, and elevated cholesterol (for

which the correlation was actually slightly negative).each risk factor; multivariate linear regression, inwhich all seven risk factors were regressed on CHD Among women, correlations were higher for diabetes

and hypertension, but as among men, the correlationmortality; and univariate rank-order correlations for a

CORONARY HEART DISEASE MORTALITY AND RISK PREVALENCE 313

risk index based on our “best estimate” of relative risksTABLE 1were 0.75 for men and 0.80 for women. Thus, aboutCHD Mortality Rates, Risk Factor Prevalences (%), and CHD Risk60% of the variance in CHD mortality between statesIndex for Adults Ages 45–79 in 49 States in the United States,

1991–1992a in the United States (56% for men and 64% for women)is attributable to differences between the states in the

Men Womenprevalences of seven risk factors for CHD. Figure 1

Inter- Inter- depicts as scatter plots the correspondence between thisquartile quartile CHD risk index and CHD mortality for men and women

Median range Median rangein the 49 states.

CHD mortality 385.2 359.6–452.2 146.9 125.8–186.2(per 100,000) DISCUSSION

CHD risk factorsSmoking 23.0 20.9–25.7 20.5 19.0–22.4

This analysis was not conducted to confirm the etio-Overweight 30.7 28.0–32.9 31.3 28.8–34.0logic importance of these already well-established CHDPhysical inactivity 33.1 27.2-41.5 33.1 28.2-39.2

Alcohol abstinence 47.3 41.1-57.0 65.9 56.6-77.3 risk factors; rather, it was conducted to assess the de-Hypertension 32.0 29.7-35.2 33.9 32.0-36.4 gree of correspondence between the prevalence of theseElevated cholesterol 32.8 29.4-35.1 37.0 35.2-38.0 risk factors and the CHD mortality rates among theDiabetes 8.2 7.4-9.2 8.5 7.4-9.7

states in the United States. We found that CHD mortal-CHD risk indexb 0.58 0.52-0.61 0.58 0.53-0.62ity rates among states in the United States correspond

a All values in the table are age-adjusted by the direct method, surprisingly strongly to modifiable behaviors and condi-using the 1990 U.S. standard population, ages 45–74. tions that are known risk factors for CHD (r = 0.75 forb See text for details. This is the “best estimate” risk index.

men, r = 0.80 for women).In the Seven Countries Study, the variation in CHD

mortality between countries corresponded to differ-with CHD mortality was also lowest for elevated choles-terol. Multivariate regression also demonstrated the ences between countries in the prevalences of CHD risk

factors [22]. More recently, however, in the ecologicgreater strength of the behaviorally related risk factors(smoking, physical activity, overweight, and alcohol ab- analysis of the baseline data from the MONICA study,

smoking, blood pressure, and cholesterol were lessstinence) than for the risk factors based on medicaldiagnoses (elevated cholesterol, hypertension, and dia- strongly associated with CHD [23]. Examination of the

correspondence between CHD mortality and risk fac-betes). Smoking was less associated with mortalityamong women than among men. Age-specific analyses tors on a regional level in Britain [24] showed correla-

tions similar to those seen between the U.S. states infor three age groups (45–54, 55–64, and 65–74) showedcorrelations that were similar to the age-adjusted corre- this study. Interestingly, just as we observed, correla-

tions were also the lowest for cholesterol in Britain. Alations (data not shown).The correlations between CHD mortality rates and study of four communities in Sweden also showed a

correspondence between both behavioral and physicalCHD risk indices were similar for three indices basedon different assumptions about the sizes of the indepen- risk factors and CHD prevalence [25]. In addition to

cross-sectional studies, examinations of temporaldent relative risks for the seven risk factors (Table 3).The correlations between CHD mortality and the CHD trends of CHD mortality have been found to correspond

TABLE 2

Univariate Spearman Correlations and Multiple Linear Regression Coefficients Relating Age-Adjusted CHD Risk Factor Prevalences(%) and CHD Mortality Rates (per 100,000) for Men and Women Ages 45–74 in 49 States in the United States, 1991–1992

Men Women

Univariate Spearman Multivariate Univariate Spearman Multivariatecorrelation* regression coefficient correlation* regression coefficient

Smoking 0.64* 6.58* 0.34* 1.57Overweight 0.49* 5.06* 0.65* 3.33*Physical inactivity 0.73* 3.28* 0.72* 1.92*Alcohol abstinence 0.55* 0.59 0.52* 0.13Hypertension 0.18 22.23 0.56* 1.36Elevated cholesterol 20.13 1.35 0.18 20.48Diabetes 0.28 1.73 0.40* 0.98[R2 for full model] [0.66] [0.70]

* P value ,0.05.

314 BYERS ET AL.

TABLE 3

Spearman Correlations between CHD Mortality and Three CHDRisk Indices Based on Different Assumptions about the Relative Risksfor Seven CHD Risk Factors for Men and Women Ages 45–74 in 49States in the United States, 1991–1992a

CHD Risk Index computed byassuming

“Upper “Lowerbounds” bounds” “Best

estimates estimates estimates”for relative for relative for relative

FIG. 1. The relationship between CHD mortality (per 100,000)risks risks risksand the CHD risk index among 49 states in the United States for

Relative risk assumptions men and women ages 45–74, in 1991–1992. The CHD risk index isSmoking 2.50 1.50 1.75 based on the “best estimate” for the independent relative risk forOverweight 2.00 1.25 1.50 seven CHD risk factors (see text for details). Both the mortality dataPhysical inactivity 1.50 1.10 1.25 and the CHD risk factor prevalences were age-adjusted to the 1990Alcohol abstinence 1.25 1.00 1.10 U.S. standard population, ages 45–74. The linear regression formulasHypertension 2.00 1.00 1.25 are as follows:Elevated cholesterol 1.50 1.00 1.10

CHD (men) = 2155 + 955 (CHD index),Diabetes 2.50 1.50 1.75CHD (women) = -153 + 528 (CHD index).Spearman correlationsb

Men 0.72 0.78 0.75Women 0.78 0.78 0.80

a See text for details of the computation of the CHD risk index. other used prospective methods [32]. In neither analy-The correlations were computed based on risk factor prevalences and sis did CHD risk factors seem to account for the gener-mortality rates that were age-adjusted by the direct method to the ally lower CHD mortality rates of the group of western1990 U.S. standard population of adults ages 45–79.

states, but the national sampling frame of NHANES 1b P value ,0.0001 for all correlations.precludes any state-specific analyses of the type wereport here.

Drawing causal inference from an ecologic study re-to temporal trends in CHD risk factor prevalences. In quires caution, particularly when the analysis is cross-Finland, declining CHD mortality rates have correlated sectional. Analyses using a latency design, in whichwith improvements in CHD risk factors over a 20-year risk factor prevalences from several years earlier areperiod [26]. In the Minneapolis area, the decline in both related to later mortality rates, will be useful in futureCHD mortality and incidence has been found to corre- years, but the BRFSS system has included nearly alllate with improving trends in CHD risk factors over states only in recent years, so those analyses will nottime [27,28]. be possible for all states for several more years. Some

We are unaware of any other studies that have re- factors, however, such as smoking may be just as rele-lated the prevalences of multiple CHD risk factors to vant to CHD risk cross-sectionally as prospectively.CHD mortality among states in the United States, but Many possible confounding factors could account forthere have been ecologic studies of single risk factors. correlations between suspected etiologic factors andCigarette sales in 1960 from 44 states were correlated disease rates [33]. However, all of the factors consideredwith CHD mortality (r = 0.55) [29]. That correlation, in this analysis are established risk factors for CHD,likely due mostly to smoking by men in that era, was and we think it is unlikely that other risk factors forsimilar to the correlation we observed for smoking as CHD would confound this association. Although thereported by men in the current study (r = 0.64). Physical prevalences of most self-reported behavioral risk fac-inactivity as assessed by the BRFSS in 37 states in tors for diseases are underestimated by telephone sur-1988 was shown to correlate to CHD mortality (r = 0.58) veys [34], we think it is unlikely that self-reports would[30]. The correlation between physical inactivity and be appreciably biased by the CHD mortality rates ofCHD we observed in this study (r = 0.73) was higher, one’s state of residence. The correlations between mor-probably because of the wider age range used in the tality and the prevalences of hypertension, diabetes,earlier study (ages 20–74). Two analyses of data from and elevated cholesterol in men were small and notthe first National Health and Nutrition Examination statistically different from 0 (P . 0.05), as was theSurvey (NHANES 1) attempted to relate the preva- correlation between elevated cholesterol and mortalitylences of CHD risk factors to CHD mortality rates in for women (P . 0.05). Self-reporting may be an inade-two large areas (the western states vs all other states). quate way to assess the true impact of medically man-

ageable CHD risk factors such as elevated cholesterol,One analysis used cross-sectional methods [31], and the

CORONARY HEART DISEASE MORTALITY AND RISK PREVALENCE 315

2. Kannel W, Doyle J, Ostfeld A, et al. Optimal resources for primaryhypertension, and diabetes, however, as the respon-prevention of atherosclerotic diseases. Circulation 1984;70:155a–dent’s awareness of the presence of the condition likely205a.also corresponds to measures for successful medical

3. McGinnis J, Foege W. Actual causes of death in the United States.control. This probably explains the weaker correlations JAMA 1993;270:2207–12.we observed between CHD mortality and these biologi- 4. National Center for Health Statistics. Mortality data tapes, 1991cal conditions that are amenable to clinical control fol- and 1992. Hyattsville (MD): Centers for Disease Control and

Prevention, National Center for Health Statistics, 1993.lowing detection than for the strictly behaviorally re-5. Remington P, Smith M, Williamson D, Anda R, Gentry E, Hogelinlated risk factors, such as physical inactivity, smoking,

G. Design, characteristics, and usefulness of state-based behav-and overweight.ioral risk factor surveillance: 1981–1987. Public Health RepOther studies have shown that CHD risk factors are 1988;103:366–75.

not independent. Those who smoke or are physically 6. Najjar M, Rowland M. Anthropometric reference data and preva-sedentary, for example, are more likely to have addi- lence of overweight, United States, 1976–80. Rockville (MD).tional CHD risk factors, including diets high in fat and U.S. Department of Health and Human Services, Public Health

Service, National Center for Health Statistics. 1987. [Vitallow in fruits and vegetables [35,36]. However, we usedHealth Stat, Series 11, No. 238]conservative estimates of the relative risks in this anal-

7. Shah B, Barnwell B, Hunt P, LaVange C. SUDAAN user’s man-ysis to reflect the independent effects of the variousual, release 5.50. Research Triangle Park (NC): 1991.factors, and we found that the correlations between the

8. Last J, editor. A dictionary of epidemiology. New York: OxfordCHD risk index and CHD mortality did not change Univ. Press, 1983.much across a wide range of alternative relative risk as- 9. U.S. Department of Health and Human Services. Reducing thesumptions. health consequences of smoking: 25 years of progress. A report

of the Surgeon General. Atlanta: U.S. Department Health andIn the United States, about 60% of the variance inHuman Services, Centers for Disease Control and Prevention,CHD mortality between states (56% for men and 64%1989. [Publication No.(CDC) 89-8411]for women) is attributable to the prevalences of seven

10. Willett W, Manson J, Stampfer M, et al. Weight, weight change,established risk factors for CHD. Therefore, we con-and coronary heart disease in women. JAMA 1995;273:461–5.

clude that an important potential exists to reduce CHD11. Rimm E, Stampfer M, Giovannucci E, et al. Body size and fat

mortality further in the United States by concerted distribution as predictors of coronary artery disease among mid-actions to reduce the risk factors for CHD in the popula- dle-aged and older US men. Am J Epidemiol 1995;141:117–27.tions of the states. Each state public health agency has 12. Powell K, Thompson P, Caspersen C, Kendrick J. Physical activity

and the incidence of coronary heart disease. Annu Rev Publicthe primary responsibility for the public health of itsHealth 1987;8:253–87.citizens. Managed care organizations are also begin-

13. Paffenbarger R, Hyde R, Wing A, Lee I, Jung D, Kampert J. Thening to assume responsibility for both disease treat-association of changes in physical-activity level and other lifestylement and prevention for increasingly large proportions characteristics with mortality among men. N Engl J Med

of the populations in states. In setting priorities for 1993;328:538–45.disease prevention programs, both public health agen- 14. Blair S, Kohl H, Barlow C, Paffenbarger R, Gibbons L, Maceracies and managed care organizations need to carefully C. Changes in physical fitness and all-cause mortality—a pro-

spective study of healthy and unhealthy men. JAMA 1995;consider approaches that reduce the prevalences of the273:1093–8.underlying behavioral causes of chronic diseases within

15. Stamler J, Stamler R, Neaton J. Blood pressure, systolic andtheir populations [3]. Of particular note is our observa-diastolic and cardiovascular risks. Arch Intern Med 1993;tion that physical inactivity, overweight, and tobacco 153:598–615.

use are the three strongest correlates of CHD mortality 16. Anderson K, Castelli W, Levy D. Cholesterol and mortality: 30among the states. Although clinical preventive services years of follow-up from the Framingham Study. JAMA

1989;257:2176–80.to identify and control hypertension, elevated choles-17. Kleinman J, Donahue R, Harris M, Finucane F, Madans J, Brockterol, and diabetes are important medical priorities for

D. Mortality among diabetics in a national sample. Am J Epide-health care providers, population-level interventions tomiol 1988;128:389–402.increase physical activity, control body weight, and

18. Manson J, Colditz G, Stampfer M, et al. A prospective studyeliminate tobacco might well have much larger effectsof maturity-onset diabetes mellitus and risk of coronary heart

on reducing future CHD mortality. Further reductions disease and stroke in women. Arch Intern Med 1991;151:1141–7.in the rates of mortality from CHD in the United States 19. Serdula M, Koong S, Williamson D, et al. Alcohol intake andwill depend in large part on our ability to create and subsequent mortality: findings from the NHANES I follow-up

study. J Stud Alcohol 1995;56:233–9.sustain improvements within populations of the states20. Stampfer M, Colditz G, Willett W, Speizer F, Hennekens C. Ain the modifiable behavioral risk factors for CHD.

prospective study of moderate alcohol consumption and the riskof coronary disease and stroke in women. N Engl J MedREFERENCES1988;319:267–73.

21. Rimm E, Giovannucci E, Willett W, et al. Prospective study of1. U.S. Bureau of the Census. Statistical abstracts of the UnitedStates: 1994. 114th ed. Washington: U.S. Bureau of the Cen- alcohol consumption and risk of coronary disease in men. Lan-

cet 1991;338:464–68.sus, 1994.

316 BYERS ET AL.

22. Keys A, Aravanis C, Blackburn H, et al. Seven countries: a multi- 29. Friedman G, Hyg S. Cigarette smoking and geographic variationin coronary heart disease mortality in the United States. J Chronvariate analysis of death and coronary heart disease. Boston:

Harvard Univ. Press, 1980. Dis 1967;20:769–79.23. World Health Organization MONICA Project. Ecological analysis 30. Yeager K, Anda R, Macera C, Donehoo R, Eaker E. Sedentary

of the association between mortality and major risk factors of lifestyle and state variation in coronary heart disease mortality.cardiovascular disease. Int J Epidemiol 1994;23:505–16. Public Health Rep 1995;110:100–2.

24. Shaper A, Elford J. Regional variations in coronary heart disease 31. Kleinman J, DeGruttola V, Cohen B, Madans J. Regional andin Great Britain: risk factors and changes in environment. In urban–suburban differentials in coronary heart disease mortalityMarmot M, Elliott P, editors, Coronary heart disease epidemiol- and risk factor prevalence. J Chron Dis 1981;34:11–9.ogy. New York: Oxford Press, 1992; Chap. 9 32. Garg R, Madans J. Kleinman J. Regional variation in ischemic

25. Nerbrand C, Olsson L, Svardsudd K, Kullman S, Tibblin G. Are heart disease incidence. J Clin Epidemiol 1992;45:149–56.regional variations in ischaemic heart disease related to differ-

33. Morgenstern H. Uses of ecologic analysis in epidemiologic re-ences in coronary risk factors? Eur Heart J 1991;12:309–14.search. Am J Public Health 1982;72:1336–44.

26. Vartiainen E, Puska P, Pekkanen J, Tuomilehto J, Jousilahti P.34. Stein A, Lederman R, Shea S. The Behavioral Risk Factor Sur-Changes in risk factors explain changes in mortality from isch-

veillance System questionnaire: its reliability in a statewide sam-emic heart disease in Finland. Br Med J 1994;309:23–7.ple. Am J Public Health 1993;83:1768–72.27. Sprafka J, Burke G, Folsom A, Luepker R, Blackburn H. Contin-

35. Simoes E, Byers T, Coates R, Serdula M, Mokdad R, Heath G. Theued decline in cardiovascular disease risk factors: results of theassociation between leisure-time physical activity and dietary fatMinnesota Heart Survey. Am J Epidemiol 1990;132:489–500.in American adults. Am J Public Health 1995;85:240–4.28. McGovern P, Pankow J, Shahar E, Doliszny K, Folson A, Black-

burn H, et al. Recent trends in acute coronary heart disease: 36. Serdula M, Byers T, Mokdad A, Simoes E, Mendlein J, CoatesR. The association between fruit and vegetable intake and chronicmortality, morbidity, medical care, and risk factors. N Engl J

Med 1996;334:884–90. disease risk factors. Epidemiology 1996;7:161–5.