O R I G I N A L A R T I C L E Diabetes, Other Risk Factors, and 12-Yr Cardiovascular Mortality for Men Screened in the Multiple Risk Factor Intervention Trial JEREMIAH STAMLER, MD OLGA VACCARO, MD JAMES D. NEATON, PHD DEBORAH WENTWORTH, MPH FOR THE MULTIPLE RISK FACTOR INTERVENTION TRIAL RESEARCH GROUP OBJECTIVE — To assess predictors of CVD mortality among men with and without diabetes and to assess the independent effect of diabetes on the risk of CVD death. RESEARCH DESIGN AND METHODS— Participants in this cohort study were screened from 1973 to 1975; vital status has been ascertained over an average of 12 yr of follow-up (range 11-13 yr). Participants were 347,978 men aged 35-57 yr, screened in 20 centers for MRFIT. The outcome measure was CVD mortality. RESULTS— Among 5163 men who reported taking medication for diabetes, 1092 deaths (603 CVD deaths) occurred in an average of 12 yr of follow-up. Among 342,815 men not taking medication for diabetes, 20,867 deaths were identified, 8965 ascribed to CVD. Absolute risk of CVD death was much higher for diabetic than nondiabetic men of every age stratum, ethnic background, and risk factor level—overall three times higher, with adjust- ment for age, race, income, serum cholesterol level, sBP, and reported number of cigarettes/ day (P < 0.0001). For men both with and without diabetes, serum cholesterol level, sBP, and cigarette smoking were significant predictors of CVD mortality. For diabetic men with higher values for each risk factor and their combinations, absolute risk of CVD death increased more steeply than for nondiabetic men, so that absolute excess risk for diabetic men was progres- sively greater than for nondiabetic men with higher risk factor levels. CONCLUSIONS— These findings emphasize the importance of rigorous sustained intervention in people with diabetes to control blood pressure, lower serum cholesterol, and abolish cigarette smoking, and the importance of considering nutritional-hygienic approaches on a mass scale to prevent diabetes. FROM THE DEPARTMENT OF PREVENTIVE MEDICINE, NORTHWESTERN UNIVERSITY MEDICAL SCHOOL, CHICAGO, ILLINOIS; THE DEPARTMENT OF INTERNAL MEDICINE AND METABOLIC DISEASES, SECOND FACULTY OF MEDICINE, UNIVERSITY OF NAPLES, NAPLES, ITALY; THE DIVISION OF BIOSTATISTICS, SCHOOL OF PUBLIC HEALTH, UNIVERSITY OF MINNESOTA, MINNEAPOLIS, MINNESOTA; AND THE NATIONAL HEART, LUNG, AND BLOOD INSTITUTE, BETHESDA, MARYLAND. ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO JEREMIAH STAMLER, MD, NORTHWESTERN UNI- VERSITY MEDICAL SCHOOL, DEPARTMENT OF PREVENTIVE MEDICINE, 680 NORTH LAKE SHORE DRIVE, SUITE 1102, CHICAGO, IL 60611-4402. RECEIVED FOR PUBLICATION 11 MARCH 1992 AND ACCEPTED IN REVISED FORM 6 MARCH 1992. MRFIT, MULTIPLE RISK FACTOR INTERVENTION TRIAL; CVD, CARDIOVASCULAR DISEASE; NIDDM, NON-INSULIN-DEPENDENT DIABETES MELLITUS; IDDM, INSUUN-DEPENDENT DIABETES MELUTUS; CHD, CORONARY HEART DISEASE; BP, BLOOD PRESSURE; S B P , SYSTOLIC BLOOD PRESSURE; D B P , D1ASTOLIC BLOOD PRESSURE; ICD-9, INTERNATIONAL CLASSIFICATION OF DISEASE, NINTH REVISION; RR, RELATIVE RISK; NHANES, NATIONAL HEALTH AND NUTRITION SURVEY; CI, CONFIDENCE INTERVAL. S everal clinical and epidemiological studies have addressed the problem of mortality and survival in popu- lations consisting predominantly or ex- clusively of people with NIDDM. Be- cause of the great variation in data collection and study design and the problem of different diagnostic criteria, comparison between studies is not straightforward. However, the data con- sistently show an excess risk of mortality in diabetic individuals of both sexes in all age-groups (1-6). Generally, CVD ac- counts for the majority of these deaths, diabetes being an independent risk factor for both CVD and CHD death (7-12). Much less information is available on the influence of other established cardiovas- cular risk factors on mortality in people with diabetes (12-19). The relatively small size of previously studied co- horts—ranging from 200 to 497 individ- uals with diabetes in the cited studies— has limited analyses of this problem. Clarification of the problem, however, has relevance for both theoretical and practical reasons. Optimal approaches to reducing cardiovascular risk in people with diabetes is an issue of major impor- tance given their inordinate risk. This study takes advantage of the large cohort—361,662 men—screened for MRFIT (20) and compares the rela- tionships of sBP, serum total cholesterol, and cigarette smoking to CVD mortality in men with and without diabetes. The increased risk of CVD mortality associ- ated with diabetes is also estimated. RESEARCH DESIGN AND METHODS Cohort and baseline assessment The study cohort consisted of the 361,662 men aged 35-57 yr who were seen as potential participants at the ini- tial screening visit of MRFIT (20-22). Examinations took place between 1973 and 1975 at 20 screening centers in 18 U.S. cities. Participation in screening 434 DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993
Transcript
O R I G I N A L A R T I C L E
Diabetes, Other Risk Factors,and 12-Yr CardiovascularMortality for Men Screened inthe Multiple Risk FactorIntervention Trial
JEREMIAH STAMLER, MD
OLGA VACCARO, MD
JAMES D. NEATON, PHD
DEBORAH WENTWORTH, MPH
FOR THE MULTIPLE RISK FACTOR
INTERVENTION TRIAL RESEARCH GROUP
OBJECTIVE — To assess predictors of CVD mortality among men with and withoutdiabetes and to assess the independent effect of diabetes on the risk of CVD death.
RESEARCH DESIGN AND METHODS— Participants in this cohort study werescreened from 1973 to 1975; vital status has been ascertained over an average of 12 yr offollow-up (range 11-13 yr). Participants were 347,978 men aged 3 5 - 5 7 yr, screened in20 centers for MRFIT. The outcome measure was CVD mortality.RESULTS— Among 5163 men who reported taking medication for diabetes, 1092 deaths(603 CVD deaths) occurred in an average of 12 yr of follow-up. Among 342,815 men nottaking medication for diabetes, 20,867 deaths were identified, 8965 ascribed to CVD.Absolute risk of CVD death was much higher for diabetic than nondiabetic men of every agestratum, ethnic background, and risk factor level—overall three times higher, with adjust-ment for age, race, income, serum cholesterol level, sBP, and reported number of cigarettes/day (P < 0.0001). For men both with and without diabetes, serum cholesterol level, sBP, andcigarette smoking were significant predictors of CVD mortality. For diabetic men with highervalues for each risk factor and their combinations, absolute risk of CVD death increased moresteeply than for nondiabetic men, so that absolute excess risk for diabetic men was progres-sively greater than for nondiabetic men with higher risk factor levels.CONCLUSIONS— These findings emphasize the importance of rigorous sustainedintervention in people with diabetes to control blood pressure, lower serum cholesterol,and abolish cigarette smoking, and the importance of considering nutritional-hygienicapproaches on a mass scale to prevent diabetes.
FROM THE DEPARTMENT OF PREVENTIVE MEDICINE, NORTHWESTERN UNIVERSITY MEDICAL SCHOOL,
CHICAGO, ILLINOIS; THE DEPARTMENT OF INTERNAL MEDICINE AND METABOLIC DISEASES, SECOND
FACULTY OF MEDICINE, UNIVERSITY OF NAPLES, NAPLES, ITALY; THE DIVISION OF BIOSTATISTICS, SCHOOL
OF PUBLIC HEALTH, UNIVERSITY OF MINNESOTA, MINNEAPOLIS, MINNESOTA; AND THE NATIONAL HEART,
LUNG, AND BLOOD INSTITUTE, BETHESDA, MARYLAND.
ADDRESS CORRESPONDENCE AND REPRINT REQUESTS TO JEREMIAH STAMLER, MD, NORTHWESTERN UNI-
VERSITY MEDICAL SCHOOL, DEPARTMENT OF PREVENTIVE MEDICINE, 680 NORTH LAKE SHORE DRIVE,
SUITE 1102, CHICAGO, IL 60611-4402.
RECEIVED FOR PUBLICATION 11 MARCH 1992 AND ACCEPTED IN REVISED FORM 6 MARCH 1992.
NON-INSULIN-DEPENDENT DIABETES MELLITUS; I D D M , INSUUN-DEPENDENT DIABETES MELUTUS; C H D ,
CORONARY HEART DISEASE; B P , BLOOD PRESSURE; SBP, SYSTOLIC BLOOD PRESSURE; DBP, D1ASTOLIC BLOOD
PRESSURE; ICD-9, INTERNATIONAL CLASSIFICATION OF DISEASE, NINTH REVISION; RR, RELATIVE RISK;
NHANES, NATIONAL HEALTH AND NUTRITION SURVEY; CI, CONFIDENCE INTERVAL.
Several clinical and epidemiologicalstudies have addressed the problemof mortality and survival in popu-
lations consisting predominantly or ex-clusively of people with NIDDM. Be-cause of the great variation in datacollection and study design and theproblem of different diagnostic criteria,comparison between studies is notstraightforward. However, the data con-sistently show an excess risk of mortalityin diabetic individuals of both sexes in allage-groups (1-6). Generally, CVD ac-counts for the majority of these deaths,diabetes being an independent risk factorfor both CVD and CHD death (7-12).Much less information is available on theinfluence of other established cardiovas-cular risk factors on mortality in peoplewith diabetes (12-19). The relativelysmall size of previously studied co-horts—ranging from 200 to 497 individ-uals with diabetes in the cited studies—has limited analyses of this problem.Clarification of the problem, however,has relevance for both theoretical andpractical reasons. Optimal approaches toreducing cardiovascular risk in peoplewith diabetes is an issue of major impor-tance given their inordinate risk.
This study takes advantage of thelarge cohort—361,662 men—screenedfor MRFIT (20) and compares the rela-tionships of sBP, serum total cholesterol,and cigarette smoking to CVD mortalityin men with and without diabetes. Theincreased risk of CVD mortality associ-ated with diabetes is also estimated.
RESEARCH DESIGN ANDMETHODS
Cohort and baseline assessmentThe study cohort consisted of the361,662 men aged 35-57 yr who wereseen as potential participants at the ini-tial screening visit of MRFIT (20-22).
Examinations took place between1973 and 1975 at 20 screening centers in18 U.S. cities. Participation in screening
434 DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993
Stamler and Associates
for MRFIT was essentially on a voluntarybasis; the most common recruitmentprocedure was to offer screening to em-ployee groups or communities. Detailson recruitment procedures have been de-scribed previously (22).
Eligibility for the trial was as-sessed on the basis of the individual'scardiovascular risk factor profile. There-fore, the first screening visit includedmeasurement of BP, serum cholesterolconcentration, smoking habits, and as-sessment of conditions that met exclu-sion criteria: that is, age, reported drugtreatment for diabetes mellitus, and pre-vious hospitalization of >2 wk for myo-cardial infarction (21). The need to sur-vey hundreds of thousands of men, torecruit at least 12,000 that met MRFITeligibility criteria, produced major logis-tical and cost considerations. Therefore,data collection at first screening was lim-ited to the foregoing recruitment-relatedmeasurements. Sera were not analyzedfor glucose or lipid fractions. No anthro-pometric measurements were made; nourinalyses were done; and no ECGs wererecorded. For men reporting previousdrug treatment for diabetes, no historywas taken as to diet treatment, type ofdrug treatment, or duration of diabetes.Thus, in this study, men who indicatedthey were not taking medication for dia-betes are identified as men without dia-betes.
BP was measured according to astandardized protocol with the partici-pant seated (23). A standard mercurysphygmomanometer was used. dBP wasrecorded as fifth Korotkoff sound. Threereadings were taken, and the average ofthe second and third readings was usedin the analysis. Blood was drawn for se-rum cholesterol determination, whichwas performed in 1 of 14 laboratoriesunder the supervision of the MRFIT cen-tral laboratory in San Francisco and theLipid Standardization Laboratory of theCenters for Disease Control in Atlanta(24).
Death ascertainmentThe vital status of participants screenedfor MRFIT is being ascertained on an acontinuing basis through the NationalDeath Index. Prior to 1979, a data file ofknown deaths provided by the U.S. So-cial Security Administration was used.Details on the death ascertainment pro-cedure and its validation have been pub-lished previously (25). Death certificateswere obtained for 94% of identified de-cedents. Causes of death were coded by atrained nosologist according to ICD-9(26). Coronary deaths were defined asICD codes 410-414 plus 429.9, strokeas codes 430-438, and total CVD ascodes 390-459. Data given are an aver-age of 12 yr follow-up.
Data analysisMortality data are given as age-adjustedrates per 10,000 person-yr. Direct stan-dardization was used to adjust for differ-ences in age distribution between menwith and without diabetes (27); the totalgroup of men screened was used as thestandard population.
The effect of cardiovascular riskfactors on mortality was analyzed formen with and without diabetes acrosslevels of serum cholesterol, sBP, and re-ported number of cigarettes smoked/day.
To evaluate the combined effectsof the three risk factors on mortality inmen with and without diabetes, and theindependent influence of diabetes onrisk of CVD death, participants weregrouped according to serum cholesterol(<200 and >200 mg/dl), sBP (<120and >120 mmHg, also <140 and >140mmHg), and cigarette use (no or yes),yielding eight strata for comparison. Inaddition, age-adjusted CVD death ratesfor men with and without diabetes werecompared according to the presence ofone, two, or three risk factors.
Proportional hazards regressionanalyses on the relationship of diabetesand other risk factors to CVD mortalityalso were done, stratified by clinical cen-ter, to estimate the independent effect ofdiabetes on risk of CVD death and to
compare the association of sBP, serumcholesterol level, and cigarettes/day withrisk of CVD death for men with andwithout diabetes (28). To estimate RR,multivariate proportional hazards regres-sion coefficients were exponentiated. Forexample, for sBP higher by 20 mmHg,for nondiabetic men, RR - e
a a n 4 x P - ° ,where e is the base of natural logarithms(= 2.7182. . . ) and 0.0234 is the multi-variate regression coefficient for the rela-tionship of sBP and time to CVD death;RR = 1.60. RR values for lower com-pared with higher levels of a risk factorcan be regarded as estimates of relativecapacity for prevention of death—e.g.,
death rate estimated to be lower by 37%.
RESULTS
Baseline descriptive statisticsAltogether, 361,662 men aged 35-57 yrwere screened in 1973-1975 for eligibil-ity for the MRFIT. The cohort consistedof 5625 men who reported being treatedat that time for diabetes mellitus; the356,037 men free of this condition madeup the cohort without diabetes; 380 men(7.2%) of the former group and 5060(1.4%) of the latter reported a previoushospitalization for myocardial infarctionand were excluded from analysis. A fur-ther 82 men with diabetes (1.6%) and8162 men without diabetes (2.3%) wereexcluded because of missing sBP. Of theremaining 347,978 men serving as thebasis for this study, numbers screened bythe 20 MRFIT centers in 18 cities acrossthe U.S. ranged from 11,241 (Miami) to29,518 (Minneapolis). These large num-bers and their identification across thecountry, along with the population-oriented approaches to recruiting themfor screening, yielded a cohort withmean levels and distributions for the ma-jor risk factors similar to those recordedfor U.S. middle-aged male respondentsto the random sampling efforts of theU.S. National Center for Health Statistics,including NHANES I and NHANL\S II inthe early and late 1970s (29-33).
DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993 435
Diabetes and cardiovascular mortality in MRFIT
2.72
35-39 40-44 45-49
Age (years)
50-54 55-59
Figure 1—Percent (by age-group) of men screened for MRFIT taking medication for diabetes.
For the defined cohort of347,978 men, the percentage who re-ported then-current treatment for diabe-tes was progressively higher with age(Fig. 1). Table 1 lists the characteristicsfor men with and without diabetes at theinitial screening visit. Those with diabe-tes were on average 3 yr older and hadhigher sBP (5.8 mmHg greater on aver-age); a smaller difference was observedfor dBP (1.9 mmHg). The higher meanBPs for diabetic compared with nondia-betic men remained with adjustment forage. Serum cholesterol level and numberof cigarettes/day were similar in the twogroups of men; however, a slightlyhigher percentage of men with diabetescompared with those without diabetesreported smoking of cigarettes. Amongmen without diabetes, 10% were non-white; among those with diabetes, 18%were nonwhite. This difference in pro-portion of nonwhite participants was pri-marily the result of an excess of blacks inthe diabetes group.
Overall mortality dataDeaths in the average 12-yr follow-upperiod numbered 1092 (160.1/10,000person-yr) among men with diabetes and
20,867 (53.2/10,000 person-yr) amongmen without diabetes. Distribution ofCVD causes of death is given in Table 2.Mortality from CVDs, particularly coro-nary disease, largely accounted for theexcess mortality observed in men withdiabetes. Cardiovascular deaths repre-sented 55% of total deaths in men with
diabetes and 43% in men without diabe-tes.
Crude CHD and CVD death rateswere approximately five times higher inmen with diabetes compared with menwithout diabetes. After adjustment forage, race, sBP, serum cholesterol level,and cigarettes/day, RR estimates for CHDand CVD were 3.2 (P < 0.0001) and 3.0(P < 0.0001), respectively. A significantindependent association of diabetes withCVD mortality (P < 0.0001) over andabove the effects of the other risk factorswas present for each of five age-groups(35-39, 40-44, 45-49, 50-54, and55-57 yr) (Table 3). Multivariate-adjusted RR estimates for diabetic com-pared with nondiabetic men ranged from2.4 (age 45-49 yr) to 3.3 (age 50-54yr); these RR were not significantly dif-ferent from each other (P = 0.84). Ab-solute excess risk of CVD death was pro-gressively higher with age for diabeticcompared with nondiabetic men—26.2(age 35-39 yr), 39.8 (age 40-44 yr),51.5 (age 45-49 yr), 97.7 (age 50-54yr), and 124.4 (age 55-57 yr) per10,000 person-yr.
With stratification by ethnicity,risk of CVD death again was consistently
Table 1—Age, ethnicity, and risk factor levels for men with and without diabetes atinitial screening for the MRFIT
AGE (YR)*
ETHNICITY* (%)
WHITE
BLACK
HISPANIC
ASIAN
AMERICAN INDIAN
OTHER
SERUM CHOLESTEROL (MG/DL)
sBP (MMHG)*
DBP (MMHG)*
CIGARETTE SMOKERS (%)*
CIGARETTES/DAY
MEN WITH DIABETES
(N = 5163)
49.1 ± 5.7
82.013.62.51.40.20.4
213.2 ± 39.5135.8 ± 19.285.8 ± 11.7
39.025.7 ± 14.7
MEN WITHOUT DIABETES
(N = 342,815)
45.8 ± 6.4
90.16.41.91.20.10.4
214.5 ± 49.7130.0 ± 15.883.9 ± 10.7
36.725.8 ± 13.2
Data are means ± SD or %.*P < .0001 for differences between men with and without diabetes.
436 DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993
Stamler and Associates
Table 2—Number of deaths by cause and age-adjusted death rate for men with (n = 5163) and without (n = 342,815) diabetes atinitial screening for the MRFIT
* Adjusted for age, race, income, serum cholesterol level, sBP, and number of cigarettes/day.
higher for diabetic compared with non-diabetic men, with RR ranging from 1.8(Hispanic) to 4.1 (other); these riskswere significantly different from eachother (P < 0.001) (Table 3). For three ofthe four ethnic groups (whites, blacks,other) the higher multivariate-adjustedRR for diabetic compared with nondia-betic men was statistically significant.(The other strata include American In-dian, 6%; Asian, 71%; and other, 23%;deaths among these groups were too fewfor separate analyses.) Absolute excessrisks for diabetic compared with nondi-
abetic men were 45.8 (Hispanic), 52.9(black), 62.5 (white), and 95.7 (other)per 10,000 person-yr.
Baseline risk factors andmortality—univariate analysesSerum cholesterol. A significant posi-tive relationship of serum cholesterol toCVD mortality was observed for bothdiabetic and nondiabetic men (Table 4).CVD death rates increased markedlyfrom lowest (<180 mg/dl) to highest(>280 mg/dl) serum cholesterol levels
for both cohorts. At every level of serumcholesterol, CVD death rate was severaltimes higher for diabetic than nondia-betic men, and the increase in CVD mor-tality rate with higher serum cholesterollevel tended to be disproportionatelygreater—i.e. steeper—for diabetic thannondiabetic men. Therefore, higher se-rum cholesterol level was associated withgreater absolute excess risk of CVD deathfor diabetic compared with nondiabeticmen; absolute excess risk for diabeticmen ranged from 47.9/10,000 person-yrwith serum cholesterol <180 mg/dl to
Table 3—Age-specific CVD death rates and age-adjusted CVD death rates by ethnicity for men with and without diabetes at initialscreening for MRFIT
*From proportional hazards regression model stratified by clinical center and with covariates corresponding to age, race, income, serum cholesterol level, sBP,and cigarettes/day.
103.8 for diabetic men in the 260-279mg/dl range.
RR of CVD mortality for diabeticcompared with nondiabetic men rangedfrom 2.83 to 4.46 at varying levels ofserum cholesterol (Table 4). Unlike ab-solute excess risk, RR was lower at higherserum cholesterol levels.
For nondiabetic men, absoluteexcess risk of CVD death with the high-est compared with the lowest serum cho-lesterol level was 32.28/10,000 per-son-yr (46.12 - 13.84); for diabeticmen, it was 68.71 (130.43 - 61.72) ormore than double (Table 4). RR of CVDdeath for men in the highest comparedwith men in the lowest serum cholesterollevel was 3.33 for nondiabetic (46.12/13.84) and 2.11 for diabetic (130.43/61.72) participants.
The greater RR for nondiabeticthan diabetic men with higher serumcholesterol is also evident from the pro-portional hazards regression coefficientscited at the bottom of Table 4; the coef-ficient is approximately twice as large formen without diabetes compared withmen with diabetes (0.0061 vs. 0.0030).Based on these coefficients, with a serumcholesterol higher by 40 mg/dl (e.g. 250compared with 210 mg/dl), RR of CVD
death was 1.28 for nondiabetic and 1.13for diabetic men. Again, because abso-lute CVD death rates were much higherfor diabetic than nondiabetic men at ev-ery level of baseline serum cholesterol,the lower RR for diabetic compared withnondiabetic men translated into a muchhigher (approximately twofold) absoluteexcess risk.
Alternatively, as an estimate ofthe potential for CVD prevention, RRwith a 40 mg/dl lower serum total cho-lesterol can readily be estimated: 0.78 fornondiabetic and 0.89 for diabetic men.Again, given the much higher CVD deathrates for diabetic than nondiabetic men,these RRs translate into a greater esti-mated potential for preventing deathsover a specified time for diabetic thannondiabetic men.sBP. sBP was positively related to risk ofCVD death with a significant trend inboth cohorts (P < 0.001) (Table 5).Thus, for men with diabetes, CVD deathrates increased from 53.6 to 242.6deaths/10,000 person-yr, and for menwithout diabetes, from 12.2 to 128.7. Aswith serum cholesterol, at every level ofsBP, CVD death rate was much greaterfor diabetic than nondiabetic men. Withhigher sBP, CVD mortality rate increased
more steeply for diabetic than nondia-betic men, hence absolute excess risktended to be greater for diabetic men thehigher their sBP (113.96/10,000 per-son-yr with sBP >200 mmHg comparedwith 41.42 with sBP <120 mmHg).
RR for diabetic compared withnondiabetic men at varying sBP levelsranged from 1.89 to 4.40, with lower RRat higher sBP levels (Table 5).
Absolute excess risk of CVDdeath for men in the highest comparedwith those in the lowest sBP stratum was116.46/10,000 person-yr for nondiabeticand 189.00/10,000 person-yr for dia-betic participants, i.e., 1.62 times greaterfor the latter (Table 5). RR for men in thehighest compared with those in the low-est sBP stratum was 4.52 for diabetic and10.55 for nondiabetic participants. Cor-respondingly, based on the multivariateregression coefficients, a 20 mmHghigher sBP was associated with an RR of1.60 for nondiabetic men and 1.41 fordiabetic men; a 20 mmHg lower sBP,with an RR of 0.63 and 0.71, respec-tively. Given the high absolute risks andabsolute excess risks for diabetic menwith sBP levels above optimal, the lowerRR with lower sBP translates into a largeestimated absolute potential for preven-
438 DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993
Stamler and Associates
Table 5—Age-adjusted CVD death rates by sBP level for men with and without diabetes at initial screening for MRFIT
SBP LEVEL (MMHG)
<120120-139140-159160-179180-1992*200COEFFICIENT*
MEN (N)
7572316142149413144
MEN WITH
CVDDEATHS (
522032061022713
0.0172 ±
DIABETES
RATE (PER
10,000[N) PERSON-YR)
53.6165.47
108.15158.71155.65242.61
0.0019
MEN WITHOUT DIABETES
MEN (N)
86,702175,82664,44412,827
2356660
CVDDEATHS ( N )
111237452794
952253109
RATE (PER
10,000PERSON-YR)
12.1919.0734.1856.4779.27
128.650.0234 ± 0.0005
R R F O R
DIABETIC/NONDIABETIC
4.403.433.162.811.961.89
ABSOLUTE EXCESS RISK
FOR DIABETIC MINUS
NONDIABET1C (PER
10,000 PERSON-YR)
41.42
46.4073.97
102.2476.38
113.96
* From porportional hazards regression model stratified by clinical center and with covariates corresponding to age, race, income, serum cholesterol level, sBP,and cigarettes/day.
tion of CVD death by control of elevatedsBP, greater for diabetic than nondiabeticmen.Cigarette smoking. A significant,graded increase in CVD mortality alsowas observed in men across increasinglevels of cigarettes smoked/day for bothgroups (Table 6). As with serum choles-terol and sBP, at every level of cigaretteuse CVD death rate was several timeshigher for diabetic than nondiabeticmen. With cigarette smoking, the CVDmortality rate increased more steeply fordiabetic than nondiabetic men, henceabsolute excess risk was greater for dia-
betic heavy smokers than diabetic non-smokers (89.64/ vs. 56.28/10,000 per-son-yr).
RR for diabetic compared withnondiabetic men ranged from 2.38(smokers of 16-25 cigarettes/day) to4.56 (nonsmokers) (Table 6).
The absolute excess risk of CVDdeath for men with the greatest dailycigarette use compared with nonsmokerswas 26.16/10,000 person-yr for nondia-betic and 59.52 for diabetic participants,i.e., more than double for diabetic men(Table 6). RR was 2.65 for nondiabeticheavy smokers and 1.83 for diabetic
heavy smokers. Correspondingly, the re-gression coefficient for cigarettes/day forthose without diabetes was approxi-mately twice the size of that estimated formen with diabetes (0.0230 vs. 0.0127).Based on these coefficients, for smokersof 20 cigarettes/day compared with non-smokers, risk of CVD death was 1.58times higher for nondiabetic men and1.29 times higher for diabetic men; fornonsmokers compared with smokers ofone pack/day, RR was 0.63 for nondia-betic men and 0.78 for diabetic men.Because CVD mortality rates and abso-lute excess risks were high for men with
Table 6—Age-adjusted CVD death rates by reported number of cigarettes/day for men with and without diabetes at initial screeningfor MRFIT
CIGARETTES PER DAY
01-1516-25>26COEFFICIENT*
MEN (N)
3284472587820
MEN WITH
CVD
DEATHS ( N )
3475967
1300.0127 ±
DIABETES
RATE (PER 10,000
PERSON-YR)
72.11100.0085.40
131.630.0024
MEN (N)
218,06826,23742,29456,216
MEN WITHOUT DIABETES
CVDDEATHS ( N )
4112733
16392481
0.0230 ±
RATE (PER 10,000
PERSON-YR)
15.8325.8735.8141.99
: 0.0006
R R F O R
DIABETIC/NONDIABETIC
4.563.872.383.13
ABSOLUTE EXCESS
RISK FOR DIABETIC
MINUS NONDIABETIC
(PER 10,000PERSON-YR)
56.2874.1349.5989.64
*From proportional hazards regression model stratified by clinical center and with covariates corresponding to age, race, income, serum cholesterol level, sBP,and cigarettes/day.
DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993 439
Diabetes and cardiovascular mortality in MRFIT
Table 7—Baseline major risk factors and age-adjusted CVD mortality in men and without diabetes at initial screening for MRFIT
MAJOR RISK FACTORS
SERUM
CHOLESTEROL CIGARETTES/
(MG/DL) DAY
<200
<200
<200
200+
ANY
<200
200+
200+
ANY
200+
No
No
YES
No
ONE ONLY
YES
No
YES
TWO ONLY
YES
sBP(MMHG)
<120120+<120<120
120+120+<120
120+
MEN
MEN
WITH
DIABETES
2161093
137240
1470
6311735
164
2530
947
(N)
MEN
WITHOUT
DIABETES
25,51756,39513,26230,189
99,846
31,637105,967
17,734
155,338
62,114
CVD DEATHS (N)
MEN
WITH
DIABETES
9941017
121
84227
16
327
146
MEN
WITHOUT
DIABETES
144827178343
1348
9332798
447
4178
3295
AGE-ADJUSTED
DEATH RATE (PER
10,000 PERSON-YR)
MEN
WITH
DIABETES
30.6860.3357.1252.17
58.82
102.7187.0386.01
90.87
125.23
MEN
WITHOUT
DIABETES
6.0212.9614.339.99
12.21
28.5020.5923.48
22.41
47.38
R R F O R
DIABETIC/
NONDIABETIC
5.104.663.995.22
4.82
3.604.233.66
4.05
2.64
ABSOLUTE EXCESS RISK
FOR DIABETIC MINUS
NONDIABETIC (PER
10,000 PERSON-YR)
24.6647.3742.7942.18
46.61
74.2166.4462.53
68.46
77.85
diabetes for each stratum of smokers,this latter RR translates into a large po-tential for prevention of CVD deaththrough cessation of smoking.
Baseline risk factors andmortality—multivariate analysesAs described in METHODS, participants
with and without diabetes were stratifiedaccording to baseline serum cholesterol,sBP, cigarette use, and diabetes status.The cutoff point of 120 mmHg was usedfor sBP in these analyses because risk ofCVD death was greater for men at everyhigher level (Table 5). Analyses with sBPdichotomized at 140 mmHg or with dBPdichotomized at 80 or 90 mmHg yieldedqualitatively similar results (data notshown). The nonsmokers with sBP < 120mmHg and serum cholesterol < 2 0 0mg/dl made up a small proportion ofboth cohorts—4.2 and 7.4%, respec-tively (Table 7). Most men screened hadone or more risk factors. For men bothwith and without diabetes, only one ofthe three risk factors was present in 28and 29%, respectively; 49 and 45% had
a combination of any two of these traits,and 18% in both groups had all three.
Within each stratum homoge-neous for cardiovascular risk, CVD mor-
tality was considerably higher for thosewith diabetes, including for the stratumwith optimal profile of the three CVDrisk factors (Fig. 2 and Table 7). Age-
140-1
120-
100-
80-
60-
40-
=5- 2 0 -
one omy two omy
Figure 2—Age-adjusted CVD death rates by presence of number of risk factors for men screened
for MRFIT, with and without diabetes at baseline.
440 DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993
Stamler and Associates
adjusted CVD death rates progressivelyincreased with increasing number of thethree major risk factors, in men withdiabetes from 30.7 to 125.2/10,000 per-son-yr, and in men without diabetesfrom 6.0 to 47.4. Presence of risk factorssingly or in combination (any two or allthree) was associated with steeper in-crease in CVD mortality for diabetic thannondiabetic participants; hence, absoluteexcess risk for diabetic compared withnondiabetic men was progressivelygreater with more adverse risk factor sta-tus—24.7/10,000 person-yr for menwith none of the three risk factors, 46.6for those with any one only, 68.5 forthose with any two only, and 77.8 forthose with all three. On average, mortal-ity rates increased in men with diabetesmore than expected based on an additiveeffect of diabetes and the other risk fac-tors.
Age-adjusted RR for men with di-abetes compared with those without di-abetes ranged from 2.6 to 5.2 (Table 7).RR estimates associated with increasingnumber of risk factors were less for menwith diabetes than men without diabetes,but these translated into much greaterabsolute excess risks given the muchhigher CVD mortality rates for diabeticcompared with nondiabetic participants.
Based on the data in Table 7, es-timates can be made of the proportion ofCVD deaths that are excess deaths attrib-utable to diabetes, elevated serum cho-lesterol and sBP, and cigarette use. If the5163 men with diabetes had a CVDdeath rate of 6.02/10,000 person-yr (therate for nonsmoking men without diabe-tes and without elevated serum choles-terol and sBP), one would have observed—37 CVD deaths in 12 yr, instead of603, or 94% fewer. If all those with di-abetes had a CVD death rate of 30.68/10,000 person-yr (the rate for nonsmok-ing men with diabetes and withoutelevated serum cholesterol and sBP), onewould have observed 190 CVD deathsinstead of 603, or 68% fewer.
Similar calculations for the342,815 men without diabetes in the
MRFIT cohort yield the estimate that—70% of all CVD deaths among themwere excess deaths attributable to ele-vated serum cholesterol, sBP, and ciga-rette use recorded for most of them atbaseline.
CONCLUSIONS— The findings ofMRFIT confirm that diabetes is a strong,independent risk factor for CVD mortal-ity over and above the effect of serumcholesterol, BP, and cigarette use. In ad-dition, the MRFIT results clearly indicatethat serum cholesterol, sBP, and cigarettesmoking are significant, strong, indepen-dent predictors of mortality in men withand without diabetes. Previous studieson this issue have yielded inconsistentresults for people with diabetes (12-19),probably because of small sample sizeand thus low statistical power to detectthese relationships: in contrast to the5163 men with diabetes in the MRFITcohort, other investigations have in-volved 200-497 diabetic individuals.Although the MRFIT findings, based onmen 35—57 yr of age with a history ofdrug treatment for diabetes, cannot beextrapolated directly in a quantitativeway to other groups (e.g., persons withpresumptively milder diabetes, women,older persons with diabetes), it is a rea-sonable inference that the pattern of re-lationships reported here is generalizablequalitatively. The MRFIT data collectiondid not include information that wouldpermit classification into IDDM andNIDDM. However, it has been estimatedfrom surveys of representative U.S. pop-ulation samples that >90% of diabeticindividuals >35 yr of age are likely to benon-insulin- dependent (3 4).
For diabetic men in the MRFITcohort, compared with nondiabetic men,12-yr CVD mortality rates were muchhigher at every level of the three majorrisk factors (serum cholesterol, sBP, cig-arette smoking) considered singly and incombination. This was the finding alsofor the small stratum with optimal levelsof all three risk factors. With progres-
sively more unfavorable risk factor sta-tus, the CVD mortality rate rose muchmore steeply for diabetic men than fornondiabetic men, so that absolute excessrisk of CVD death became progressivelygreater for diabetic than nondiabetic menthe worse the baseline risk factors. Thisresult, underscoring the importance andpotential for prevention of prematureCVD death in diabetic individualsthrough sustained control of the majorCVD risk factors, is unequivocal in the12-yr MRFIT data reported here. Thisfact merits emphasis because the MRFIT6-yr follow-up data, previously re-ported, were equivocal in this regard(12,35). That is, although they clearlyshowed higher coronary and CVD deathrates for diabetic than nondiabetic menat each level of the three risk factorsconsidered together, and significant pro-gressive increases in CVD death rates forboth diabetic and nondiabetic men withhigher risk factor levels at baseline, the6-yr data were equivocal as to whetherCVD excess risk was greater for diabeticthan nondiabetic men with higher riskfactor levels. With the 12-yr follow-updata—and the evidence of many moredeaths—the answer on this matter isclearly affirmative.
The overall unfavorable cardio-vascular risk factor profile for men withand without diabetes therefore requiresparticular attention and concern (35-37). Despite their current treatment fordiabetes, only a small percentage (4%) ofmen with diabetes did not smoke andhad optimal levels of serum cholesterol(<200 mg/dl) and sBP (<120 mmHg).Although it might be argued that theunfavorable serum cholesterol and BPdistributions of those with diabetes mayat least in part reflect metabolic problemsrelated to the diabetic state, this notion isnot relevant with regard to the 39%prevalence of then-current cigarette usein the MRFIT men when screened in1973-1975. These and other data (4)indicate that at that time at least in theU.S., medical care for people with diabe-tes did not include vigorous sustained
DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993 441
Diabetes and cardiovascular mortality in MRFIT
attention to correction of unhealthy life-style choices (smoking, rich diet, physicalinactivity) related to cardiovascular risk.
Although little information existson the effects of intervention measures inpeople with diabetes, and judgments vary(38), the MRFIT results strongly supportthe concept that benefit can be expectedfrom changing risk factors in people withdiabetes, based on early detection and ef-fective correction at least to values recom-mended for the general population(33,35,39-44). This concept also is sup-ported by data from developing countriesand from Japan that have shown low ratesof atherosclerotic disease in people withdiabetes from populations whose dietshave not habitually contained high levelsof cholesterol and saturated fats and whodo not have high serum cholesterol levels(1,45-47). These cross-population find-ings are concordant with the judgmentthat a rich diet—through its unfavorableinfluences on body weight, BP, glycemia,serum lipids, and thrombogenesis(1,11,12,45-52)—needs to be viewed as apivotal trait that adds to risk for peoplewith diabetes, as has been emphasized inpolicy statements of the American DiabetesAssociation (35,53,54).
From the MRFIT data, it is alsoevident that diabetes is a strong, indepen-dent risk factor for CVD mortality over andabove the effect of BP, serum cholesterol,and cigarette smoking. The RR conferredby diabetes was high among men with thebest apparent baseline cardiovascular riskfactor status, i.e., nonsmokers with sBP< 120 mmHg and serum cholesterol < 200mg/dl, as well as among men with lessfavorable status. Reasons remain to be elu-cidated for this marked excess in CVD riskattributable to diabetes per se. Recent re-search has focused on the role of suchwide-ranging endogenous phenomena asglycosylation of arterial wall ground sub-stance, microalbuminuria, hypertriglyceri-demia, and low levels of plasma high-density lipoprotein, the latter twoattributable apparently, or at least in part,to insulin resistance and hyperinsulinemia(37,55-62). Data also are available to in-
dicate that hyperinsulinemia with insulinresistance, common in NIDDM, may be anindependent CVD risk factor (37,56,62).In efforts to resolve this important prob-lem, it seems relevant to consider the in-terplay between such endogenous meta-bolic abnormalities and habitual life-stylechoices, especially those concerning nutri-tion.
Whatever the etiopathogeneticpathways for the markedly increased CVDrisk attributable to diabetes per se, thefinding underscores the importance of en-hanced efforts to prevent this disease. Al-though much remains to be learned forthis purpose, extensive data are availablefrom population studies showing that anindividual's being overweight, particularlymarkedly overweight, is associated withhigher likelihood of having NIDDM, thecommon form of diabetes among middle-aged and older people (11,12,51,63,64).Data are also available on ability to favor-ably influence glucose intolerance by nu-tritional-hygienic means, including in-creased leisure-time isotonic exercise andfat-modified diets that are lower in calo-ries, saturated fat, and cholesterol, andhigher in fiber (64-67). Such findingspoint to the possibility of the primary pre-vention as well as the control of NIDDMon a mass scale by effective nutritional-hygienic programs that prevent and con-trol overweight, as well as high serum cho-lesterol and high BP.
Acknowledgments—This study was sup-ported by National Heart, Lung, and BloodInstitute Grant 1-RO1-HL-28715.
The principal investigators and senior staffof the MRFIT clinical, coordinating, and sup-port centers and the National Heart, Lung,and Blood Institute project office are listed inJAMA 248:1465-77, 1982.
References
1. West KM: Epidemiology of Diabetes and ItsVascular Lesions. New York, Elsevier,1978
2. Krolowski AS, Warram JH, ChristiliebAR: Onset, course, complications and
prognosis of diabetes mellitus. In JoslinClinic Diabetes Mellitus. 12th ed. MarbleA, Krall LP, Bradley RF, Christilieb AR,Soldner J, Eds. Philadelphia, PA, Lea &Febiger, 1985, p. 251-75
4. Kleinman JC, Donahue PR, Harris MI,Finucane FF, Madans JH, Dwight BB:Mortality among diabetics in a nationalsample. Am J Epidemiol 128:389-401,1988
5. Laakso M, Rannemaa T, Pyorala K, KallioV, Puukka P, Penttila I: Atheroscleroticvascular disease and its risk factors innon-insulin-dependent and non-dia-betic subjects in Finland. Diabetes Care11:449-63, 1988
6. Waugh NR, Dallas JH, Jung RT, NewtonRW: Mortality in a cohort of diabeticpatients: causes and relative risks. Diabe-tologia 32:103-104, 1989
7. Reunanen A: Mortality in type 2 diabetes.Ann Clin Res 15 (Suppl.) 37:26-28,1983
8. Barrett-Connor E, Wingard DL: Sex dif-ferential in ischemic heart disease mor-tality in diabetics: a prospective popula-tion based study. Am J Epidemiol 118:489-96, 1983
9. Butler WJ, Ostrander LD Jr, Carman WJ,Lamphiear DE: Mortality from coronaryheart disease in the Tecumseh study:Long term effect of diabetes mellitus, glu-cose tolerance and other risk factors. AmJ Epidemiol 121:541-47, 1985
10. Pan WE, Cedres LB, Liu K, Dyer A,Schoenberger JA, Shekelle RB, Stamler R,Smith D, Collette P, Stamler J: Relation-ship of clinical diabetes and asymptom-atic hyperglycemia to risk of coronaryheart disease mortality in men andwomen. Am J Epidemiol 123:504-16,1986
11. Stamler J: Lectures on Preventive Cardiol-ogy. New York, Grune & Stratton, 1967
12. Stamler J: Epidemiology, established riskfactors, and the primary prevention ofcoronary heart disease. In Cardiology. Vol2. Parmley WW, Chatterjee K, Eds. Phil-adelphia, PA, Lippincott, 1987, p. 1-41
13. Kannel WB, McGee DL: Diabetes and car-diovascular risk factors in the Framingham
442 DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993
Stamler and Associates
Study. Circulation 59:8-13, 197914. Dupree EA, Meyer MB: Role of risk fac-
tors in complications of diabetes melli-tus. Am] Epidemiol 112:100-12, 1980
15. Suarez L, Barrett-Connor E: Interactionbetween cigarette smoking and diabetesmellitus in the prediction of death attrib-uted to cardiovascular disease. Am J Ep-idemiol 120:670-75, 1984
16. Fuller JH, Shipley MJ, Rose J, Jarrett RJ,Keen H: Mortality from coronary heartdisease and stroke in relation to degree ofhyperglycemia: the Whitehall study. BrMedJ 27:867-70, 1983
17. Fitzgerald AP, Jarrett RJ: Are conven-tional risk factors for mortality relevantin type 2 diabetes? Diabetic Med 8:475-80, 1991
18. Morrish NJ, Stevens LK, HeadJ, Fuller JH,Jarrett RJ, Keen H: A prospective study ofmortality among middle-aged diabetic pa-tients (the London cohort of the WHOMultinational Study of Vascular Disease inDiabetics) 11: associated risk factors. Diabe-tologia 33:542-48, 1990
19. Rosengren A, Welin L, Tsipigianni A,Wilhelmsen L: Impact of cardiovascularrisk factors on coronary heart disease andmortality among middle-aged diabeticmen: a general population study. Br MedJ 299:1127-31, 1989
21. Sherwin R, Kaelber CT, Kezdi P, Kjels-berg MO, Thomas HE: The Multiple RiskFactor Intervention Trial (MRFIT) 11:The development of the protocol. PrevMed 10:402-25, 1981
22. Neaton JD, Grimm RH, Cutler JA: Re-cruitment of participants for the MultipleRisk Factor Intervention Trial. ControlledClin Trials 8:42-53S, 1988
23. Dischinger P, DuChene AG: Quality con-trol aspects of blood pressure measure-ments in the Multiple Risk Factor Inter-vention Trial. Controlled Clin Trials6:137-57S, 1986
24. Department of Health, Education andWelfare, Lipid Research Clinics Program:Manual of Operations, Vol. 1: Lipid andLipoprotein Analysis. Washington, DC,
U.S. Govt. Printing Office, 197425. Wentworth DN, Neaton JD, Rasmussen
WL: An evaluation of the Social SecurityAdministration MRB file and the Na-tional Death Index in the ascertainmentof vital status. Am J Public Health 73:1270, 1983
26. International Classification of Diseases.Clinical Modification. 9th rev. Vol. 1. AnnArbor, MI, Edwards Bros., 1981
27. Fleiss JL: Statistical Methods for Rates andProportions. New York, Wiley, 1973
28. Cox DR: Regression models and life-tables. J R Statist Soc B 24:187-220,1972
29. National Center for Health Statistics: To-tal Serum Cholesterol Levels of Adults20-74 Years of Age, United States, 1976-80. Vital and Health Statistics, Series II.Washington, DC, U.S. Govt. Printing Of-fice, 1986 (NCHS publ. no. 236)
30. National Center for Health Statistics:Blood Pressure Levels and Hypertension inPersons Ages 6-74 Years: United States,1976-80. Advance data. WashingtonDC, U.S. Govt. Printing Office, 1982(NCHS publ. no. 84)
31. National Center for Health Statistics: Se-rum Cholesterol Levels of Persons 4-74Years of Age by Socioeconomic Character-istics. Vital and Health Statistics, Series II.Washington, DC, U.S. Govt. Printing Of-fice, 1980 (NHANES I; NCHS publ. no.217)
32. National Center for Health Statistics:Blood Pressure Levels of Persons 6-74Years, United States, 1971-74. Vital andHealth Statistics, Series II. Washington,DC, U.S. Govt. Printing Office, 1977(NCHS publ. no. 203)
33. National Center for Health Statistics:Health United States 1985. WashingtonDC, U.S. Govt. Printing Office, 1985,(DHHS publ. no. PHS 86-1232)
34. Harris MI, Hadden WC, Knowler WC,Bennett PH: Prevalence of diabetes andimpaired glucose tolerance and plasmaglucose levels in the U.S. populationaged 20-74 years. Diabetes 36:523-34,1987
35. Col well J A, Franz MJ, Ginsberg H, Gw-ynne JT, Landsberg L, Lockwood D, Ser-vice FJ, Stern MP, Vinik AI, Weir GC:Role of cardiovascular risk factors in pre-
vention and treatment of macrovasculardisease in diabetes: Consensus statementof the American Diabetes Association.Diabetes Care 12:573-79, 1989.
36. Harris MI: Hypercholesterolemia in dia-betes and glucose intolerance in the U.S.population. Diabetes Care 14:366-74,1991
37. Pyorala K, Laakso M, Uusitupa M: Dia-betes and atherosclerosis: an epidemio-logic view. Diabetes Metab Rev 3:463-•542, 1987
38. Hanefeld M, Fischer S, Schmechel H,Rothe G, Schulze J, Dude H, Schawne-beck U, Julius U: Diabetes InterventionStudy: Multiple Intervention Trial in anewly diagnosed NIDDM. Diabetes Care14:308-17, 1991
39. National Cholesterol Education ProgramExpert Panel: Report of the NationalCholesterol Education Program ExpertPanel on Detection, Evaluation, andTreatment of High Blood Cholesterol inAdults. Arch Intern Med 148:36-69,1988
40. National Institutes of Health, NationalCholesterol Education Program Popula-tion Panel: Repori of the Expert Panel onPopulation Strategies for Blood CholesterolReduction. Bethesda, MD, U.S. Govt.Printing Office, 1990 (NIH publ. no. 90- •3046)
41. Joint National Committee on Detection,Evaluation, and Treatment of High BloodPressure: The 1992 Report of the JointNational Committee on Detection, Eval-uation and Treatment of High BloodPressure. Arch Intern Med. In press
42. Department of Health and Human Ser-vices, Public Health Service, Centers forDisease Control, Center for Chronic Dis-ease Prevention and Health Promotion,Office on Smoking and Health: TheHealth Benefits of Smoking Cessation: AReport of the Surgeon General. Wash-ington, DC, U.S. Govt. Printing Office,1990 (DHHS publ. no. CDC-90--8416)
43. Department of Health and Human Ser-vices, Public Health Service: Healthy Peo-ple 2000: National Health Promotion andDisease Prevention. Washington, DC, U.S.Govt. Printing Office, 1991 (DHHS publ.no. PHS-91-50212)
DIABETES CARE, VOLUME 16, NUMBER 2, FEBRUARY 1993 443
Diabetes and cardiovascular mortality in MRFIT
44. European Atherosclerosis Society StudyGroup: Strategies for the prevention ofcoronary heart disease: A policy state-ment of the European AtherosclerosisSociety. Eur Heart] 8:77-88, 1987
45. Katz LN, Stamler J: Experimental Athero-sclerosis. Springfield, IL: Charles C. Tho-mas, 1953
46. Snapper I: Chinese Lessons to WesternMedicine. New York, Interscience, 1941
51. Berkson DM, Stamler J: Epidemiology ofthe killer chronic diseases. In Nutritionand the Killer Diseases. Winick M, Ed.New York, Wiley, 1981, p. 17-55
52. Stamler J: Epidemiologic findings onbody mass and blood pressure in adults.Ann Epidemiol 1:347-62, 1991
53. American Diabetes Association, Ameri-can Dietetic Association Committee onFood and Nutrition: Principles of nutri-tion and dietary recommendations forpatients with diabetes mellitus: 1971. A
special report. Diabetes 20:633-34,1971
54. American Diabetes Association, Commit-tee on Food and Nutrition: Principles ofnutrition and dietary recommendationsfor individuals with diabetes mellitus:1979. Diabetes Care 2:520-23, 1979.
55. Fontbonne A, Charles MA, Thibult N,Richard JL, Clande JR, Warnet JM,Rosselin GE, Eschwege E: Hyperinsu-linemia as a predictor of coronary heartdisease mortality in a healthy population:the Paris Prospective study, 15 years fol-low-up. Diabetologia 34:356-61, 1991
56. Fontbonne AM, Eschwege E: Insulin andcardiovascular disease: the Paris Prospec-tive study. Diabetes Care 14:461-69,1991
57. Jarrett RJ, Shipley MJ: Type 2 (non-insulin-dependent) diabetes mellitusand cardiovascular diseaseputative asso-ciation via common antecedents: Furtherevidence from the Whitehall study. Dia-betologia 31:737-40, 1988
59. Brownlee M, Cerami A, Vlassara H: Ad-vanced glycosylation end products in tis-sues and the biochemical basis of dia-betic complications. N Engl J Med 318:1315-21, 1988
60. Makita Z, Radoff S, Rayfield EJ, Yang Z,
Skolnick E, Delaney V, Friedman EA,Cerami A, Vlassara H: Advanced glyco-sylation end products in patients withdiabetic nephropathy. N Engl J Med 325:836-42, 1991
61. Steffes MW, Mauer SM: Toward a basicunderstanding of diabetic complications.N EnglJ Med 325:883-84, 1991
62. Moller DE, Flier JS: Mechanisms of dis-ease: insulin resistance mechanisms, syn-dromes, and implications. N Engl J Med325:938-48, 1991
63. Stern MP: Kelly West Lecture: Primaryprevention of type II diabetes mellitus.Diabetes Care 14:399-410, 1991
64. Farinaro E, Stamler J, Upton M, Mojon-nier L, Hall Y, Moss D, Berkson DM:Plasma glucose levels: Long-term effectof diet in the Chicago Coronary Preven-tion Evaluation Program. Ann Intern Med86:147-54, 1977
65. Rivellese A, Riccardi G, Giacco A, PacioniD, Genovese S, Mattioli PL, Mancini M:Effect of dietary fibre on glucose controland serum lipoproteins in diabetic pa-tients. Lancet 2:447-50, 1980
