846 Increased Risk of Coronary Heart Disease Death in Men With Low Total and Low-Density Liorotein Cholesterol in the Russian Liid Research Clinics Prevalence Follow-u Study Dmitri B. Shestov, MD; Alexander D. Deev, PhD; Anatoli N. Klimov, MD; Clarence E. Davis, PhD; Herman A. Tyroler, MD Background. A continuously increasing risk of coronary heart disease with increasing levels of cholesterol has been reorted by many observational and exerimental studies. However, this tye of association has not been observed in studies in the Russian Liid Research Clinics. Methods and Results. Twelve-year coronary heart disease mortality among 40- to 59-year-old men was analyzed in the Moscow and St Petersburg examinees in the Russian Liid Research Clinics Program. The baseline survey examined 6431 men fasting and free of revalent coronary heart disease. Liids and lioroteins, blood ressure, body mass, education level, alcohol intake, and smoking history were obtained. Mortality follow-u was based on contacts with articiants or their relatives or neighbors. Coronary heart disease mortality was analyzed based on risk factor levels and was further divided into raid and nonraid deaths. A J-shaed cholesterol-coronary heart disease risk function was resent for both total and low-density liorotein cholesterol. Further examination showed hyocholesterolemic men to have lower low-density and higher high-density liorotein cholesterol, higher alcohol consumtion, leaner body mass, and less education than men with normal or high cholesterol levels. When education level was considered, the J-shaed risk function was resent only among men with less than a high school education. When deaths were classified into raid (less than 4 hours after onset of symtoms) and nonraid, the J-shaed risk function was restricted to raid deaths. Conclusions. The results disclose a sizeable subset of hyocholesterolemics in this oulation at increased risk of cardiac death associated with lifestyle characteristics. (Circulation. 1993;88:846-853.) KEY WORDs * coronary heart disease * cholesterol * lioroteins * Russia Acontinuously increasing risk of coronary heart disease (CHD) with increasing levels of total cholesterol is one of the fundamental tenets of cardiovascular disease eidemiology, based on the results of numerous studies, both observational and exerimental.1 For examle, the monotonically increasing nature of the total cholesterol and CHD mortality relation, extending from total cholesterol levels of 4.68 mmol/l (181 mg/dl) to levels above 6.54 mmol/l (53 mg/dl), was confirmed in the large study of 361 66 men aged 35 to 57 years screened for the Multile Risk Factor Intervention Trial (MRFIT). It was, therefore, unexected to observe a J- or U-shaed relation of total cholesterol and low-density liorotein cholesterol (LDL-C) to CHD mortality among men of similar age Received June 9, 199; revision acceted Aril, 1993. From the Institute of Exerimental Medicine (D.B.S.), Russian Academy of Medical Sciences, St Petersburg, Russia; the Center of Preventive Medicine (A.D.D.), Russian Federation Ministry of Health, Moscow, Russia; the Institute of Exerimental Medicine (A.N.K.), Russian Academy of Medicine, St Petersburg, Russia; the Deartment of Biostatistics (C.E.D.), School of Public Health, University of North Carolina, Chael Hill; and the Deartment of Eidemiology (H.A.T.), School of Public Health, University of North Carolina, Chael Hill. Corresondence to Dr Davis, Collaborative Studies Coordinating Center, Suite 03, 137 E Franklin St, Chael Hill, NC 7514. (40 to 59 years) in the Russian Liid Research Clinics (LRC) Follow-u Study at 4 years3 and 7.5 years4 of follow-u. As reorted herein, there is ersistence of the J-shaed relation at 1 years of follow-u. The urose of this article is to reort the 1-year follow-u exerience and the results of investigation of reasons for the increased risk of CHD deaths among hyocholesterolemic men in the Russian LRC study. Our analyses of the J-shaed risk function exlored four categories of ossible exlanations for the excess risk at low levels of total cholesterol: the effect of the LDL-C and high-density liorotein cholesterol (HDL-C) comonents of total cholesterol; the effect of nonliid biomedical risk factors, eg, blood ressure and body mass; the effect of behavioral and social characteristics such as alcohol consumtion and educational achievement; and the association of hyocholesterolemia with different modes of clinical resentation of CHD, that is, the cholesterol-chd risk function for deaths occurring raidly after onset of symtoms comared with nonraid deaths. Methods The samling, measurements, and quality control rocedures followed the standardized methods for the US-Russian LRC collaborative rogram and have been described in detail reviously.5 A brief summary of
Shestov et al Increased CHD in Russian Men With Low Cholesterol 847 TABLE 1. Number of Particiants and Age-Adjusted Death Rate for Moscow and St Petersburg Samles St Petersburg Moscow Both clinics N Rate N Rate N Rate Total 3165 366 6431 Survivors 538 67 5165 Mortality CVD 7 7.59 73 6.94 545 7. CHD 156 4.37 178 4.47 334 4.41 Stroke 57 1.60 41 1.09 98 1.3 Other CVD 59 1.6 54 1.38 113 1.48 Non-CVD 355 9.84 366 9.8 71 9.51 Cancer 195 5.54 1 5.6 416 5.54 Other causes 160 4.30 145 3.66 305 3.97 Total mortality 67 17.43 639 16. 166 16.73 Follow-u time (y) 1.10 1.10 1.13 Peole with revalent coronary heart disease (CHD) are excluded. Rates are er 1000 erson-years. CVD, cardiovascular disease. rocedures follows: Men born between 1916 and 1935 were samled from voter registration lists of selected residential districts in Moscow and St Petersburg from 1975 to 1977; articiation rates were 77.7% in Moscow and 78.1% in St Petersburg. Of the 7815 men screened, 6431 men aged 40 to 59 years were fasting, free of revalent CHD, and eligible for these analyses. Baseline measurements of liids and lioroteins were determined by standardized LRC methods.6 Blood ressure was measured four times, first using a standard mercury shygmomanometer and then a random-zero device, with reeat of both. The blood ressures used for this study consisted of the average of the two random-zero blood ressures. The body mass index (BMI), calculated as body weight (in kilograms) divided by height (in meters) squared, was used as an index of obesity. Level of education was determined by interview and classified by Russian eidemiologists into categories comarable to those used in US studies: high school graduate and more or less than that level of education. Measurements of total cholesterol, HDL-C, and triglycerides and derived LDL-C were obtained from all examinees at the first contact, as were blood ressure, height, and weight. Information regarding cigarettes smoked and alcohol intake was obtained by interview on a subsamle recalled to a second clinic visit. Alcohol consumtion was calculated based on 7-day recall and exressed as average daily gram intake. Follow-u, which averaged 1.1 years at the time of the analyses, began in January 1979, based on contacts with articiants or their relatives or neighbors. Address bureaus were used for those who moved recently. For articiants who had died, death certificates were obtained; when cardiovascular disease was listed as a direct cause of death or as a related disease, information from the hosital, hysician, relative, neighbor, or witness was collected. Final event information was evaluated by a classification anel. A samle of 58 Russian decedent cause of death classifications was indeendently classified by two US cardiologists exerienced in similar activities for the US LRC studies; comarison indicated good agreement between the two classifications.7 All deaths classified as due to CHD were further identified as raid if death occurred within 4 hours of onset of the final clinical eisode and as nonraid if survival exceeded 4 hours of onset. The exeriences of the Moscow and St Petersburg cohorts were aggregated for this reort. From the original cohort of 7815 men, we have excluded 44 who were fasting less than 10 hours at the time of blood drawing, 148 who were either younger than 40 years or older than 59, and 99 who had either clinically manifest CHD or a history of CHD at the time of the screening. This leaves a total of 6431 men aged 40 to 59 years who were fasting and free of CHD for this reort. Statistical Methods As a first ste, the distribution of each imortant otential risk factor was stratified by quintiles, and age-adjusted death rates were comuted for each quintile. The age adjustment was accomlished by direct standardization using the World Health Organization standard oulation. Further analyses centered on the association of CHD mortality with total cholesterol, LDL-C, and HDL-C. For multivariable analyses, the roortional hazards models was used. In the roortional hazards model, all deaths from causes other than CHD are treated as losses in the analysis. A restricted cubic sline of LDL-C was fit to test for nonlinearity of the risk ratio function.9 If this cubic sline model indicated statistically significant deartures from linearity, a model was fit using a quadratic term for the risk factors. The risk ratios recorded are comuted from the roortional hazards model adjusted for other covariates, and the multivariable models were reeated searately for raid and nonraid CHD deaths. Variables included in these models were restricted to those measurements obtained on the first clinic visit; data on cigarettes smoked and alcohol consumed (obtained on the subsamle that was called to the second visit) were analyzed only in aggregate comarisons of the quintiles of liids or liorotein estimates.
848 Circulation Vol 88, No 3 Setember 1993 A T a U 0 0 0 v lq 6 409 4.30 4.4 6.1 1 a 4 6 4 5.6 8.46 1 5 4 6 SBP OUINTILES 9.19 t0 10 A T a IE 0 00 v 4 4.10..1 em a 8 * 0 1 a 4 LDL-C QUINTILES 6.15 4 6.15 4.80 C~~~~~~~~ 1 5 4 5 BMI OUINTILES A T U a U 4) 0 0 Vo v s ~~~~~~~.68 ~~~65...0 1 a a 4 HDL'C OUINTILES EDUCATION LEVEL FIG 1. Grahs show age-adjusted coronary heart disease death rate by selected riskfactors (articiants with revalent coronary heart disease at screening are excluded). SBP indicates systolic blood ressure; LDL-C, low-density liorotein cholesterol; BMI, body mass index; HDL-C, high-density liorotein cholesterol; PlY, erson-years; and HS, high school. Results Table 1 sets out the number of deaths and death rates er 1000 erson-years for the articiants in the study. A total of 166 articiants died during the 1 years of follow-u, and 334 of the deaths were attributed to CHD. Death rates for CHD and most other major causes of death were similar for articiants residing in St Petersburg and Moscow, and the aggregate of articiants residing in both cities was used for all subsequent analyses. Fig 1 dislays the age-adjusted CHD mortality by quintiles of candidate risk factors, for which cutoints shown in Table. It is noticeable that the death rate in the lowest quintile for both total cholesterol and LDL-C is higher than the death rates in the second quintile, with the death rate highest in the fifth quintile, 448 me leading to a J-shaed curve. This result reviously has been reorted for shorter follow-u eriods of these samled oulations. There is a negative association between HDL-C and CHD mortality, although the gradient is of lesser magnitude and less consistent than for total cholesterol and LDL-C. There is a shallow U-shaed relation between BMI and CHD mortality. In contrast, there is a strong ositive association between systolic blood ressure and CHD mortality and a strong inverse relation between level of attained education and CHD mortality. Age-adjusted death rates by quintiles of total cholesterol, LDL-C, and HDL-C within the three educational achievement strata are resented in Fig. For total cholesterol and LDL-C, the mortality rate varies in a *48
Shestov et al Increased CHD in Russian Men With Low Cholesterol 849 TABLE. Quintile Cut-Points of Selected Risk Factors Quintile 1 3 4 5 Cholesterol < 188 188-07 08-6 7-50 51+ LDL-C <11 11-13 133-150 151-17 173+ HDL-C <40 41-47 48-54 55-64 65+ SBP < 117 117-16 17-135 136-150 151+ BMI <.73.73-4.74 4.75-6.51 6.5-8.5 8.53+ LDL-C indicates low-density liorotein cholesterol; HDL-C, high-density liorotein; SBP, systolic blood ressure; and BMI, body mass index. J-shaed manner among individuals with less than a high school education. In contrast, among those who had a high school education or more, a ositive association between CHD mortality and quintiles of the cholesterol levels is observed. It is to be noted, however, that desite the aarent difference of the shae of the risk function by educational achievement, there is a stewise increase in the mortality rate within each quintile level of total cholesterol and LDL-C extending from those men with greater than a high school education to high school education only to those men with less than a high school education. The relative risk of CHD mortality among men with less than a high school education comared with men with more than a high school education is greatest in the lowest total cholesterol quintile (.9) and decreases to 1.9 and 1.6 in the two highest quintiles, resectively. The relation of CHD mortality to HDL-C quintiles also varies among educational strata. There is a suggestive U-shaed relation among those with less than a high school education (although these differences are not statistically significant); in contrast, the relation aears negative or inverse among individuals with more than a high school education. At each level of HDL-C, the mortality is greatest among those with less than a high school education, intermediate among those with a high school education, and lowest among those with more than a high school education. The mortality rate ratio increases with increasing HDL levels, from 1.8 at the lowest HDL-C level to 3.1 at the highest HDL-C level. The relation of CHD death rates to liid and liorotein estimates for those deaths occurring raidly and contrasted with nonraid deaths is shown in Fig 3. For both total cholesterol and LDL-C, the mortality rate for raid deaths resented a U-shaed relation with highest rates in the lowest and highest quintiles; in shar contrast, there was a generally ositive increasing death rate for nonraid CHD deaths in relation to increasing total cholesterol and LDL-C. The relative risk, contrasting the highest with the minimum death rates, was aroximately.1 for the U-shaed raid deaths and aroximately 3 to 1 for the linearly increasing nonraid deaths. The relation of HDL-C quintiles was different for raid and nonraid deaths. Among raid deaths, a U-shaed relation was resent; in contrast, there was an inverse association of nonraid CHD death rates with increasing levels of HDL-C. For the raid deaths, the relative risk contrasting the maximum with the minimum was 1.; for the nonraid deaths, the mortality rate was 1.5 times higher in the lowest than in the highest HDL-C quintile. The relation of raid and nonraid CHD death rates to education level is also resented in Fig 3. The major difference was in a markedly higher death rate for raid CHD deaths among men with less than a high school education. We next set out the relation of raid and nonraid CHD death rates to liid and liorotein levels within educational strata. As shown in Fig 4, there is a general reservation of the educational level gradient at all quintile levels of total cholesterol, LDL-C, and HDL-C, inverse in nature such that raid and nonraid death rates at each liid and liorotein level were higher, the lower the educational achievement. Furthermore, there is an emergent attern of a U-shaed relation for both total cholesterol and LDL-C and raid deaths, the lower the education stratum, in shar contrast to the generally ositive increasing level of nonraid death rates with increasing levels of both total cholesterol and LDL-C for all educational strata. The risk function also was different for raid and nonraid deaths in relation to education level and HDL-C level. For raid deaths there was no association in the lowest education stratum and a slight U-shaed relation otherwise; in marked contrast, for nonraid deaths there was generally an inverse relation with decreasing nonraid death rates with increasing levels of HDL-C. Statistical tests of significance were erformed to evaluate the shaes of the risk functions derived from the stratified analyses. The J- or U-shaed relation of CHD death rates modeled by quadratic liid terms was statistically significant for raid deaths among men with less than a high school education: for LDL-C, P<.05, with minimum risk at LDL-C, of 143 mg/dl. For raid deaths among men with a high school education there was no significant association with LDL-C, and for raid deaths among men with more than a high school education, the quadratic term was not statistically significant, whereas the linear term was at P<.05. In contrast to the risk functions for raid deaths, the quadratic term for LDL-C in relation to nonraid deaths was not statistically significant within any of the education strata. However, the linear term for LDL-C in relation to nonraid deaths was statistically significant (P<.05) for each education achievement grou. Thus, the nature of the risk function for LDL-C in relation to nonraid CHD deaths was similar to the ositive, monotonically increasing risk observed in most other settings. There was neither a significant linear nor quadratic relation of HDL-C to raid deaths. In contrast, among men with less than a high school education there was a statistically significant inverse association of HDL-C
850 Circulation Vol 88, No 3 Setember 1993 10. 4 a 4 6 1 a 4 6 LDL-C OUINTILES 1 a a 4 a HDL-C OUINTILES FIG. Grahs show age-adjusted coronary heart disease death rate by education and quintiles of lids (articiants with revalent coronary heart disease at screening are excluded). LDL-C indicates low-density liorotein cholesterol; HDL-C, high-density liorotein cholesterol; P1Y, ersonyears; and HS, high school. Rates are shown as er 1000 Ply with nonraid CHD deaths (P<.05). For each of the other two education strata, ie, men with a high school education and those with more than a high school education, the linear term coefficient for HDL-C in relation to nonraid CHD deaths was negative but did not reach statistical significance. We next fit multivariable models adjusting for age, BMI, systolic blood ressure, HDL-C, and education, categorizing the CHD deaths into raid and nonraid deaths (not shown). The U-shaed relation between LDL-C and CHD mortality remained strong when restricted to deaths occurring within 4 hours after the onset of symtoms. However, for deaths occurring more than 4 hours after the onset of symtoms, the U-shaed relation disaeared. Finally, we fit multivariable models within education categories for raid CHD deaths and nonraid CHD deaths searately. An inverse relation between LDL-C and CHD death was resent only for raid deaths of men with less than a high school education. The U-shaed relation initially observed for the total cohort is rimarily a result of the strong inverse association between LDL-C and raid CHD deaths among the less-educated articiants. Table 3 lists characteristics of the men in the lowest LDL-C quintile comared with those in the third and fifth quintiles. Since the categorization is based on LDL-C, it is not surrising that there is a sizeable difference in total cholesterol between the grous. However, men in the lowest quintile of LDL-C also have lower BMI, higher HDL-C, and higher alcohol intake; in this quintile there is a larger roortion of smokers and a larger roortion with education less than high school than in the four higher LDL-C quintiles. The age distribution and blood ressure means for the grous are essentially the same. Discussion The total cholesterol-chd mortality risk function observed in the Russian LRC study differs markedly from that observed in numerous other eidemiological studies. The incremental risk of deaths from CHD observed among men with the lowest levels of total cholesterol now has been observed to ersist through 1 years of follow-u. Methodological errors in measurement of total cholesterol and certification of cause of death differences between the Russian and the US LRC rograms have been virtually eliminated as ossible causes.5,6,10,11 The association with mortality is similar for LDL-C as for total cholesterol. The increased risk among hyocholesterolemic men cannot be attributed to low levels of HDL-C; in fact, in this oulation, men with lowest total cholesterol and LDL-C levels had the highest HDL-C levels. The aarent U-shaed total cholesterol-chd mortality rate relation observed in men from this oulation-based study has been found to be restricted to a subset of the examinees characterized by educational achievement level. It was only among men with the lowest educational achievement, identified in this study as equivalent to less than a high school education, that the U-shaed relation was manifest. In contrast, among men with the highest educational achievement (those with more than a high school education), the more usually observed ositive association between total cholesterol and LDL-C and CHD mortality was resent. Not only was there modification of the risk function by educational achievement level but in addition, there was a major association of education level er se with CHD mortality levels. In this samle of middle-aged male residents of neighborhoods in Moscow and St Petersburg, there was a two- to threefold gradient of increasing CHD mortality with decreasing educational achievement level at all levels of total cholesterol and LDL-C. The finding of an inverse association of educational level with CHD in the Russian cohort study is not unique. A similar strong inverse relation of CHD with educational level has emerged in the United States, transforming a reviously ositive or nonexistent asso-
Shestov et al Increased CHD in Russian Men With Low Cholesterol 851 *r RATE PER 1000 P/Y As RATE PER 100 PlY 3 I 01 +..._-- 11 13 1. 16 %~RAPID -. 1.48.7 3.01 1.49 1.34 1* 1.6J 3.11 3 A a 1h a 4. 1 131 4 1 RAID - 5.63..17.31 I 7 NON-RAID - -.S.07.09 1AS 13 HDLC OUINTLES...... -AM RATE PER 10 P/y 4 3 4 /,,+/ --~ 3 RATE PER 1000 Ps 4% 1 ; 3 4 6 44 148s *08 RAPID 3.03 1.71 1.61 S6.9 RAPID - 3.67.17 1.69 NON RAID - +-1.07 1.7 1.47 1.91 3.16 NON-RAPID - +-.16.41 1.34 LDL-C OUINTILES EDUCATION LEVEL FIG 3. Plots show age-adjusted raid and nonraid coronary heart disease death rate by selected risk factors (articiants with revalent coronary heart disease at screening are excluded). HDL-C indicates high-density liorotein cholesterol; LDL-C, low-density liorotein cholesterol; PIY, erson-years; and HS, high school. ciation before 1960 to a strong inverse association during the decline of CHD mortality after the mid 1960s in the United States; this was reorted on comarison of mortality findings in 1960 with the NHANES Follow-u Study findings of 1971 to 1984.1 In addition, an inverse association of CHD with education level or other indexes of socioeconomic status has been reorted in eidemiological studies from other countries: Great Britain,13 Sweden,14 Norway,16 Finland,16 and Scotland.17 Among the exlanations advanced for these findings are the increasing divergence of risk factor distributions among members of different social classes. This is articularly true for smoking atterns, weight, and blood ressure distributions, each increasing with decreasing social status. In contrast, total cholesterol values in the United States and Great Britain vary less or directly with social class. In the Russian LRC Follow-u Study cohort reort herein, men with the lowest total cholesterol and LDL-C lasma levels in addition to having less education were also observed to have a history of greater alcohol consumtion, higher HDL-C levels, a slightly higher history of current smoking, and lower mean BMI. This constellation of characteristics suggests a nutritional syndrome in association with alcohol consumtion that might redisose to low total cholesterol and acute liver failure, with deaths misclassified as sudden coronary deaths.18 Cardiomyoathies secondary to alcohol excess have also been reorted as risk factors for sudden cardiac death in a Russian oulation.19 The use of alcohol may be a marker of an underlying genetic disorder. 0. The lack of alcohol intake and cigarette smoking data on a large ortion of the samled oulations is a weakness of the study. Were such measures available on all articiants, the statistical models could assess directly the joint effects of smoking, alcohol, and LDL-C. Lacking this direct assessment, it can only be inferred indirectly that excess alcohol intake and smoking may be contributing to the excess CHD mortality in men with low LDL-C, articularly in the lower-educated men. Selenium was not measured in this study, but it is ossible that a low dietary intake of selenium and fat-soluble vitamins could be associated with low cholesterol levels. Low selenium levels have been associated with increased risk of coronary disease in Finland.0 The men in this study were free of clinically diagnosed revalent CHD at time of entry into the study. However, this does not rule out the ossibility of reexistent underlying coronary atherosclerosis in a subset of men vulnerable to sudden arrhythmic deaths. It is noteworthy in the US and British studies that the comosite effect of the major risk factors of smoking, elevated blood ressure, and total cholesterol statistically exlained only a relatively small fraction of the excess CHD mortality associated with low social status and/or educational achievement. The remainder of the excess may be attributable to random measurement errors and intraindividual variation of risk factors with resultant regression bias,1 to other biological risk factors not measured in these studies such as hemostatic factors, or to sychosocial and behavioral risk factors associated with lower socioeconomic status.3 Further
85 Circulation Vol 88, No 3 Setember 1993 ms P. too, RATH en 10A ty 6-0 -. 1 0- a 4 6 ant an too0 Py a 4 6 1 a a 4 LDL-C OUINTILES LDL-C OUINTILES I a a 4 6 1 a a 4 6 HDL-C QUINTILES HDL-C QUINTILES FIG 4. Grahs show age-adjusted raid and nonraid coronary heart disease (CHD) death rate by education and quintiles of liids (articiants with revalent CHD at screening are excluded). LDL-C indicates low-density liorotein; HDL-C, high-density liorotein; P/Y, erson-years; and HS, high school. Rates are er 1000 PlY investigation of the contribution of the other factors measured in the Russian LRC Study are in rogress. This study of CHD mortality cannot distinguish between determinants of disease incidence and casefatality rates. The strong association of excess mortality from noncardiovascular causes with hyocholesterolemia and low educational achievement oints to the need for study of rocesses of host suscetibility more general than that assessed by cardiovascular risk factors.4 In summary, the U-shaed relation of total cholesterol and LDL-C with all-cause mortality and the J-shaed relation with CHD mortality observed in the Russian LRC Follow-u Study was resent only among the men in the lowest educational achievement stratum. Behavioral (alcohol consumtion and smoking), anthroometric (BMI), and other liid and liorotein characteristics (HDL-C) further characterized the J-shaed henomenon for deaths occurring within 4 hours of onset of the final eisode. In contrast, a ositive association of total cholesterol and LDL-C to CHD was observed among middle-aged men with high educational achievement in the Russian LRC Study; these were men who drank less, had lower levels of HDL-C and higher BMI, smoked less, and whose risk of dying of CHD during 1 years of follow-u was one third to one half lower than that of men with the lowest educational achievement. The Russian LRC Follow-u Study was carried out during a eriod of increasing CHD mortality rates in the Soviet Union.5 The results of the study reorted herein suggest the otential for CHD revention by combined high-risk subgrou and community intervention strategies.
Shestov et al Increased CHD in Russian Men With Low Cholesterol 853 TABLE 3. Characteristics of Russian Men by LDL-C Quintiles LDL-C quintile Characteristic 1 3 5 N 1438 1447 1436 Age (y) 48.4 (5.1) 48.6 (5.3) 48.8 (5.3) LDL-C (mg/dl) 93.8 (16.3) 14.0 (5.3) 198.5 (3.4) TC (mg/dl) 175.7 (3.6) 16.0 (15.4) 74.0 (8.1) HDL-C (mg/dl) 6.4 (.8) 51.3 (14.1) 48.4 (1.) SBP (mm Hg) 135.9 (1.9) 134.8 (1.6) 135.8 (.3) BMI (kg/m) 4.7 (3.6) 6.0 (3.6) 6.7 (3.) Alcohol* (g/d) 18.9 (5.0) 11.5 (15.9) 11. (0.1) Current smoker* (%) 66.3 (1.)t 5.0 (1.3)t 57.1 (1.3)t Education <High school (%) 53.8 (1.3)t 39.3 (1.3)t 3.7 (1.)t High school (%). (1.1)t 3.8 (1.1)t 4.8 (1.1)t >High school (%) 4.0 (1.1)t 36.8 (1.3)t 4.5 (1.3)t LDL-C indicates low-density liorotein cholesterol; TC, total cholesterol; HDL-C, high-density liorotein cholesterol; SBP, systolic blood ressure; and BMI, body mass index. Values are mean (standard deviation in arentheses). *Estimates based on second visit subsamle. tstandard error. Acknowledgments This work was art of the Joint US-USSR Program in Research in Cardiovascular Disease and was suorted in art by the National Heart, Lung, and Blood Institute, Bethesda, Md, and under contracts NO-1HV-143 and NO-1HV-0811 with the University of North Carolina, Chael Hill. Comuting was done by Jeffrey Abolafia, Haesook O'Grady, and Diane Williams. Tying was done by Lesa McPherson and Pat Coley. References 1. Tyroler HA. Review of liid-lowering clinical trials in relation to observational eidemiologic studies. Circulation. 1987;76:515-5.. Martin MJ, Browner WS, Hulley SB, Kuller LH, Wentworth D. Serum cholesterol, blood ressure, and mortality: imlications from a cohort of 361 6 men. Lancet. 1986;:933-936. 3. Shestov DB. Results of four years of follow-u of USSR oulations. In: USA-USSR First Liorotein Symosium: Laboratory Research, Joint Poulation Studies, Leningrad, USSR, May 6 to 7, 1981. Bethesda, Md: US Deartment of Health and Human Services; 198. National Institutes of Health ublication No. 83-1966. 4. Shestov DB, Deev AD, Zhukovsky GS, Thorn MD, Davis CE, Klimov AN. Coronary heart disease risk factors and mortality in the USSR. Liid Research Clinics Follow-u Study. In: Levy RI, ed. Atherosclerosis Reviews. New York: Raven Press, Publishers; 1988;17:61-75. 5. Poulation Descritions and Methodology for US-USSR Collaboration in Problem Area 1 and Organization of Research in Soviet Liid Centers in Problem Area 1 of the USSR-USA Collaboration: Descrition of Poulations and Methodology. In: USA-USSR First Liorotein Symosium: Laboratory Research, Joint Poulation Studies, Leningrad, USSR, May 6 to 7, 1981. Bethesda, Md: US Deartment of Health and Human Services; 198. National Institutes of Health ublication No. 83-1966. 6. Liid Research Clinics Program. Manual oflaboratory Oerations. Vol 1: Liid and Liorotein Analysis. Washington, DC: US Government Printing Office; 1974, revised 198. DHEW ublication No. (NIH) 75-68. 7. Davis CE, Trobaugh GB, Zhukovsky GS, Liovetsky BM, Shalnova SA, Plavinskaya SI, Prineas R. Methods for comarative mortality studies in the USSR and US. In: Levy RI, ed. Atherosclerosis Reviews. New York: Raven Press, Publishers; 1988;17:57-60. 8. Cox DR. Regression models and life tables. JR Stat Soc B. 197; 34:187-0. 9. Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationshis between redictors and resonse. J Natl Cancer Inst. 1988;80:198-0. 10. Liel K, Ahmed S, Albers JJ, Bachorik, PS, Cooer C, Helms R, Williams J. Analytical erformance and comarability of the determination of cholesterol by 1 liid research clinics. Clin Chem. 1977;3:1744-175. 11. US-USSR Steering Committee for Problem Area 1. The athogenesis of atherosclerosis: collaborative US-USSR study on the revalence of dysliioroteinemias and ischemic heart disease in American and Soviet oulations. Am J Cardiol. 1977;40:60-68. 1. Feldman JJ, Makuc DM, Kleinman JC, Cornoni-Huntley J. National trends in educational differentials in mortality. Am J Eidemiol. 1989;19:919-933. 13. Rose G, Marmot MG. Social class and coronary heart disease. Br Heart J. 1981;45:13-19. 14. Rosengren A, Wedel H, Wilhelmsen L. Coronary heart disease and mortality in middle aged men from different occuational classes in Sweden. BMJ. 1988;97:1497-1500. 15. Holme I, Helgeland A, Hjermann I, Levern P, Lund-Larsen PG. Four-year mortality by some socioeconomic indicators: the Oslo Study. J Eidemiol Community Health. 1980;34:48-5. 16. Salonen JT. Socioeconomic status and risk of cancer, cerebral stroke, and death due to coronary heart disease and any disease: a longitudinal study in eastern Finland. J Eidemiol Community Health. 198;36:94-97. 17. Woodward M, Shewry MC, Smith WCS, Tunstall-Pedoe H. Social status and coronary heart disease: results from the Scottish Heart Health Study. Prev Med. 199;1:136-148. 18. Regan TJ, Haider B. Ethanol abuse and heart disease. Circulation. 1981;64(sul III):III-14-III-19. 19. Vikhert AM, Tsilenkova VG, Cherachenko NM. Alcoholic cardiomyoathy and sudden cardiac death. JAm Coll Cardiol. 1986; 8:3A-1 1A. 0. Salonen JT, Alfthan G, Huttunen JK, Pikkarainen J, Puska P. Association between cardiovascular death and myocardial infarction and serum selenium in a matched-air longitudinal study. Lancet. 198;891:175-179. 1. MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, Abbott R, Godwin J, Dyer A, Stamler J. Blood ressure, stroke, and coronary heart disease, I: rolonged differences in blood ressure: rosective observational studies corrected for the regression dilution bias. Lancet. 1990;335:765-774.. Meade TW, Chakrabarti R, Hanes AP. Hemostatic function and ischemic heart disease rincial results of the Northwick Park Heart Study. Lancet. 1986;:533-537. 3. Weinblatt E, Ruberman W, Goldberg JD, Frank CW, Shaior S, Chaudhary BS. Relation of education to sudden death after myocardial infarction. N Engl J Med. 1978;99:60-65. 4. Cassel JC. The contribution of the social environment to host resistance: the Fourth Wade Hamton Frost Lecture. Am J Eidemiol. 1976;104:107-13. 5. Deev AD, Oganov RG. Trends and determinants of cardiovascular mortality in the Soviet Union. Int J Eidemiol. 1989;18(sul 1): S137-S144.