Relationship of Baseline Serum Cholesterol Levels in 3 Large Cohorts of Younger Men to Long-term Coronary, Cardiovascular, and All-Cause Mortality and to Longevity FREE

For middle-aged populations, especially men, serum cholesterol consistently has been shown to be a significant risk factor for coronary heart disease (CHD) and the major cardiovascular diseases (CVDs).^1- 5 The relationship is continuous, graded, strong, independent of other risk factors, predictive, and generally assessed as etiologically significant. This judgment is reinforced by results of many randomized controlled trials in middle-aged and older persons with average, borderline high, and high serum cholesterol levels; these trials demonstrate that effective, sustained reduction of serum cholesterol levels by dietary measures, pharmacologic means, or both reduces CHD and CVD risk.^4- 5

In contrast, for younger adults, the only available long-term prospective data on this relationship are for 3 cohorts of limited size: Johns Hopkins male medical students and men and women aged 31 to 39 years at baseline in the Framingham study.^6- 7 Possibly reflecting this situation, there has been debate on the merits of population-wide measurement of serum cholesterol levels in younger adults to identify and treat those at higher CHD and CVD risk.^8- 14

This article adds extensive information on serum cholesterol levels and risk of CHD and CVD in younger men. For 3 large cohorts of younger men from the Chicago Heart Association Detection Project in Industry (CHA), Chicago Peoples Gas Company (PG), and Multiple Risk Factor Intervention Trial (MRFIT) studies, we provide data on the relationship of serum cholesterol to long-term mortality from CHD, CVD, and all causes.

This report focuses on 3 cohorts of younger men from 3 long-term prospective epidemiologic studies. Baseline methods and follow-up procedures have been described.^3,5,15 A summary of these is provided herein.

Men in the CHA cohort were aged 18 through 39 years at baseline. From late 1967 to early 1973, the CHA study surveyed 39,572 men and women at 84 cooperating Chicago companies and organizations. All employees were invited to participate; the response rate was 55%. Two trained and standardized 4-person field teams collected demographic information, medical history, and medical treatment data; information on smoking status; measurement of height, weight, heart rate, and a single supine blood pressure; resting electrocardiogram (ECG); and performed venipuncture for blood chemistry measurements. Serum cholesterol level was determined by the Levine and Zak method.¹⁶ Criteria of the Pooling Project³ and the Hypertension Detection and Follow-up Program¹⁷ were used to code ECG abnormalities. The cohort in this study is 11,017 men aged 18 through 39 years at baseline, with no history of diabetes mellitus or myocardial infarction (MI).

Methods of follow-up (mean, 25 years) to ascertain vital status include local procedures (eg, inquiry to employers and telephone and letter communications with participants) and use of Social Security Administration and National Death Index records. Vital status is known for more than 99% of the cohort.

Men in the PG cohort were aged 25 through 39 years at baseline. On January 2, 1959, 1609 men aged 25 through 39 years were employed by the Peoples Gas Company. Of these, 1411 (87.7%) underwent standardized examination (December 1959–December 1963) in the PG medical department. The cohort for this study comprises 1266 men who were free of clinical diabetes mellitus and CHD and had complete baseline data. The standardized baseline survey included medical history, physical examination, 3 blood pressure measurements averaged for analyses, and serum cholesterol levels measured using the method described by Abell et al.¹⁸

Long-term follow-up of the PG cohort (mean, 34 years) was done through the PG medical department, which is a recipient of mortality information in relation to employee benefits. Verification included checks with the National Death Index since its inception in 1979. Vital status of every man in the PG cohort is known.

Men in the MRFIT cohort were aged 35 through 39 years at baseline. Altogether, 361,662 men aged 35 through 57 years were screened (1973-1975) at 22 centers in 18 US cities for recruitment for MRFIT. The focus here is on 69,205 men aged 35 through 39 years at baseline with complete data on baseline risk factors and who were free of diabetes mellitus and MI. To assess trial eligibility, the first screening included measurements of blood pressure and serum cholesterol levels; current smoking (by questionnaire), including number of cigarettes per day; and conditions for exclusion, ie, drug treatment for diabetes mellitus and previous hospitalization for MI. Seated blood pressure was measured according to a standardized protocol by trained, certified staff. Three readings per man were taken; the average of the second and third systolic blood pressure readings was used for analyses. Serum total cholesterol level was determined in 15 standardized local laboratories by the Lieberman-Burchard color reaction and use of serum calibrators to yield values equivalent to Abell-Kendall reference values.^5,15,18

The men's vital status (mean follow-up, 16 years) was ascertained through the US National Death Index and, prior to 1979, through the Social Security Administration. Cause of death is known for 99% of decedents.

To evaluate strength of the relationship between serum cholesterol level and CHD risk for younger vs middle-aged men, summary data are reported for cohorts of men aged 40 through 59, 40 through 59, and 40 through 57 years at baseline in these 3 studies (n = 8955, 1416, and 258,570 for the CHA, PG, and MRFIT studies, respectively).

For CHA and PG cohort decedents, underlying cause of death was coded by a trained staff professional, using the International Classification of Diseases, Eighth Revision (ICD-8). For the MRFIT cohort, this was done by a nosologist using International Classification of Diseases, Ninth Revision (ICD-9). Coders had access only to death certificates; they were blinded to baseline data.

Death from all CHD was defined for CHA and PG cohorts as ICD-8 codes 410 to 414 and for the MRFIT cohort as ICD-9 codes 410 to 414 and 429.9; MI, code 410 (all cohorts); all CVD deaths, ICD-8 codes 400 to 445.9 for the CHA and PG cohorts and ICD-9 codes 390 to 459 for the MRFIT cohort; all cancers, codes 140 to 209 (all cohorts); violence, for the CHA and PG cohorts, ICD-8 codes E800 to E999 exclusive of codes E930-E936 and for the MRFIT cohort, ICD-9 codes 800 to 999.

Mortality rates were age-adjusted (direct method) to age distribution of all men in an age stratum. Cox multivariate proportional hazards regression was used to calculate relative risks (RRs) and 95% confidence intervals across baseline serum cholesterol level strata and to obtain multivariate-adjusted coefficients for the relation of serum cholesterol level to mortality end points.

Cox multivariate proportional hazards regression coefficients for the relation of serum cholesterol level to all-cause mortality were used to estimate years of greater life expectancy for men with baseline serum cholesterol levels less than 200 mg/dL (<5.17 mmol/L) vs men with serum cholesterol levels of at least 240 mg/dL (≥6.21 mmol/L).^5,15 Thus, the coefficient for this relationship for men in the CHA cohort aged 18 through 39 years is 0.0051. Average serum cholesterol level for the 6888 men with values less than 200 mg/dL (<5.17 mmol/L) was 167.8 mg/dL (4.35 mmol/L); for the 974 men with values of at least 240 mg/dL (≥6.21 mmol/L), it was 262.1 mg/dL (6.79 mmol/L), ie, 94.3 mg/dL (2.44 mmol/L) higher. By exponentiation, estimated RR of death for the former vs the latter subcohort is e^{−0.0051 × 94.3} = e^−0.4809 = 0.618.

To estimate impact of this lower serum cholesterol level on life expectancy, we used the Cox coefficient for the relationship of age to all causes mortality, 0.0718. The product for serum cholesterol level exponentiation, e^{0.0051 × 94.3} = e^0.4809, is also obtained with multiplication of the age coefficient, 0.0718, by 6.70, indicating that a serum cholesterol level of 167.8 mg/dL (4.35 mmol/L) vs 262.1 mg/dL (6.79 mmol/L) is equivalent to these men being, on average, 6.7 years younger, eg, age 23.0 years rather than age 29.7 years. From US life tables,¹⁹ male life expectancy at age 29.7 years is 43.1 years and at age 23.0 years, 50.5 years; ie, 7.4 years estimated greater longevity is attributable to serum cholesterol levels of 167.8 mg/dL (4.35 mmol/L) vs 262.1 mg/dL (6.79 mmol/L).

In the earliest baseline survey, mean serum cholesterol levels of the PG cohort were higher than that of the other 2 cohorts, as was prevalence of cigarette smoking (Table 1). Results for baseline serum cholesterol levels and long-term mortality by cause were consistent for the 3 younger adult male cohorts (Table 2, Table 3, and Table 4).

Death due to CHD accounted for 26%, 34%, and 28% of all deaths in the CHA, PG, and MRFIT cohorts, respectively. Multivariate-adjusted risk of CHD death was higher by 3.46 (CHA cohort), 2.15 (PG cohort), and 3.63 (MRFIT cohort) times for each subcohort with elevated serum cholesterol levels according to the criterion of the US National Cholesterol Education Program (≥240 mg/dL [≥6.21 mmol/L]) compared with the subcohort with desirable levels (<200 mg/dL [<5.17 mmol/L]). Results were similar for fatal acute MI (RRs: 3.99, 3.22, and 3.47 for the CHA, PG, and MRFIT cohorts, respectively) and for other CHD death (CHD death excluding MI) (RRs: 2.88, 1.64, and 3.86 for the 3 cohorts, respectively) (data not shown).

Absolute risk of CHD death and absolute excess risk—the difference in absolute risk with high baseline serum cholesterol levels vs risk with favorable levels—were related to duration of follow-up (Table 2, Table 3, and Table 4). Thus, with serum cholesterol levels of 240 mg/dL or higher (≥6.21 mmol/L), absolute risk was 15.5 per 1000 men in 16 years (from average age 37 to age 53 years) for the MRFIT cohort; 54.2 per 1000 men in 25 years (from age 30 to age 55 years) for the CHA cohort; and 154.0 per 1000 men in 34 years (from age 32 to age 66 years) for the PG cohort. Absolute excess risk for these subcohorts, compared with those with serum cholesterol levels less than 200 mg/dL (<5.17 mmol/L), was 12.1 per 1000 men (MRFIT), 43.6 per 1000 men (CHA), and 81.4 per 1000 men (PG).

Relative risks were particularly large for subcohorts with severe hypercholesterolemia (≥280 mg/dL [≥7.24 mmol/L]) vs those with low serum cholesterol levels (<160 mg/dL [<4.14 mmol/L]); for all CHD deaths, the RRs were 11.93, 8.06, and 8.09 for the CHA, PG, and MRFIT cohorts, respectively (Table 2, Table 3, and Table 4). The relation of serum cholesterol levels to risk was continuous and graded across the entire distribution; rates were lowest with serum cholesterol levels lower than 160 mg/dL (<4.14 mmol/L).

For all CHD death, Cox multivariate-adjusted coefficients for the relation of baseline serum cholesterol level to risk were 0.0147, 0.0088, and 0.0099 for the CHA, PG, and MRFIT cohorts, respectively (Table 2, Table 3, and Table 4), yielding RRs of 1.80, 1.42, and 1.49 for the CHA, PG, and MRFIT cohorts, respectively, for serum cholesterol levels higher by 40 mg/dL (1.03 mmol/L). These are much higher RRs than for the cohorts of middle-aged men (aged 40-59, 40-59, and 40-57 years) in these studies—1.20, 1.10, and 1.29 for the CHA, PG, and MRFIT cohorts, respectively (Cox multivariate coefficients, 0.0046, 0.0023, and 0.0064, respectively). Results were similar for other CHD mortality end points.

Because the CHA and PG cohorts had longer follow-up (25 and 34 years, respectively), analyses were done for 2 periods to assess whether the strong serum cholesterol level–CHD death relationship prevailed during both earlier and later years of follow-up. For the CHA cohort, during the first 15 years of follow-up, multivariate Cox coefficients for this relationship were 0.0183 (t = 6.508; 57 CHD deaths) and for the next 10 years, 0.0130 (t = 6.293; 146 CHD deaths) (RRs, 2.08 and 1.68 for serum cholesterol levels higher by 40 mg/dL (1.03 mmol/L); the difference between these coefficients is not significant (t = −1.528). For the PG cohort, Cox coefficients for the first 23 years of follow-up were 0.0109 (t = 3.241; 49 CHD deaths) and for the next 11 years, 0.0072 (t = 2.325; 66 CHD deaths) (RRs, 1.55 and 1.33, respectively); the difference between these 2 coefficients is not significant (t = −0.430).

Death due to CVD accounted for 34%, 42%, and 39% of all deaths in the CHA, PG, and MRFIT cohorts, respectively. Results for the relationship of serum cholesterol level to risk of CVD death were similar to those for CHD deaths (Table 2, Table 3, and Table 4), with RRs of 2.18, 2.10, and 2.87 for the CHA, PG, and MRFIT cohorts, respectively, for serum cholesterol levels of at least 240 mg/dL (≥6.21 mmol/L).

For all 3 cohorts, baseline serum cholesterol level was significantly related to risk of mortality from all causes (Table 2, Table 3, and Table 4). The relationship was generally continuous, graded, strong, and independent of other risk factors. With serum cholesterol level higher by 40 mg/dL (1.03 mmol/L), multivariate-adjusted Cox coefficients (0.0051, 0.0029, and 0.0043) yielded RR estimates of 1.23, 1.12, and 1.19 for the CHA, PG, and MRFIT cohorts, respectively.

For men with favorable baseline serum cholesterol levels (<200 mg/dL [<5.17 mmol/L]) vs men with hypercholesterolemia (≥240 mg/dL [≥6.21 mmol/L]), baseline average serum cholesterol levels were lower by 94.3, 91.0, and 92.1 mg/dL (2.44, 2.36, and 2.39 mmol/L) for the CHA, PG, and MRFIT cohorts, respectively (Table 2, Table 3, and Table 4). These lower serum cholesterol levels translate into estimated greater life expectancy of 6.1, 8.7, and 3.8 years for the CHA, PG, and MRFIT cohorts, respectively.

All Cancer Deaths. Cancer deaths numbered 225, 132, and 753 for the CHA, PG, and MRFIT cohorts, respectively; there was no significant relation of baseline serum cholesterol levels to cancer mortality (multivariate Cox coefficients, 0.0028, −0.0027, and −0.0005; P>.10) (Table 5, Table 6, and Table 7).

All Violent Deaths. There were 130 (CHA cohort), 11 (PG cohort), and 414 (MRFIT cohort) deaths from violent causes (injury, suicide, or homicide); no significant association of serum cholesterol levels with risk of violent death was seen (Cox multivariate coefficients, −0.0023, −0.0027, and 0.0020; P>.10) (Table 5, Table 6, and Table 7).

All Other Deaths. For the CHA and PG cohorts, there was no significant relation of serum cholesterol level to risk of death from other causes (ie, non-CVD, noncancer, nonviolent causes) (157 and 54 deaths, respectively; coefficients: −0.0012 and 0.0011, respectively; P>.10); for the MRFIT cohort, with 412 such deaths, this association was inverse and statistically significant (Cox multivariate coefficient, −0.0037; P<.01) (Table 5, Table 6, and Table 7).

All Non-CVD Deaths. There was no significant relation of serum cholesterol level to all non-CVD mortality (Cox multivariate coefficients, 0.0004, −0.0016, −0.0006; P>.10) (Table 5, Table 6, and Table 7). For men in the CHA and PG cohorts (aged 18-39 years and 25-39 years, respectively, at baseline), those with the lowest baseline serum cholesterol levels (<160 mg/dL [<4.14 mmol/L]) generally had the lowest mortality rate for each of the 4 listed non-CVD causes of death; for men in the MRFIT cohort (aged 35-39 years at baseline), this subcohort had death rates higher than for most other strata, significantly so only for other non-CVD causes of death (Table 5, Table 6, and Table 7).

The main results of this study, consistent for these cohorts of younger men, were first, there is a continuous, graded, strong, independent relationship of serum cholesterol level to long-term risk of CHD and CVD death, stronger than for men with like serum cholesterol levels measured in middle age; second, with high serum cholesterol levels in young adulthood, there is substantial absolute risk and absolute excess risk of CHD and CVD death over the decades from young adulthood through middle age; third, there is almost no countervailing evidence of any greater non-CVD mortality for younger adult men with favorable serum cholesterol levels compared with others; fourth, consequently, for all 3 cohorts, there is a significant, positive, continuous graded relationship of serum cholesterol level to long-term mortality from all causes; and finally, substantially longer estimated life expectancy—3.8 to 8.7 years longer—is observed for younger adult men with favorable serum cholesterol levels (<200 mg/dL [<5.17 mmol/L]) vs those with unfavorable levels (≥240 mg/dL [≥6.21 mmol/L]). Almost certainly, these strong relationships of serum cholesterol level to long-term mortality are underestimates, because they are based on only 1 cholesterol measurement per person, thus tending to produce misclassification (regression dilution bias).²⁰

These findings for younger men are in agreement with those for Johns Hopkins medical students,⁶ (1017 men; average age, 22 years) with serum cholesterol levels measured 1 to 11 times (median, 3) and averaged for analyses. During follow-up of the Hopkins cohort for 27 to 42 years (median, 30.5 years), serum cholesterol level was strongly related to CHD and CVD incidence and mortality.⁶ After adjustment for possible confounders, for serum cholesterol levels higher by 36 mg/dL (0.9 mmol/L) (75th vs 25th percentile), CHD and CVD RRs were 1.72 to 2.02. The non-CVD mortality rate was also higher across quartiles of serum cholesterol levels, and risk of death before age 50 years was significantly greater (RR, 1.64 with serum cholesterol levels higher by 36 mg/dL [0.9 mmol/L]). Based on only 1 measurement, RRs were somewhat lower, eg, 1.50 for CVD incidence (vs 1.72 with multiple measurements), reflecting regression dilution bias.⁶

For 597 men and 677 women aged 31 through 39 years at baseline in the Framingham study, with 65 and 23 CVD deaths during 30 years of follow-up,⁷ findings are again similar to those of our study: univariate Cox coefficient for the relationship of serum cholesterol to CVD mortality 0.0096 for men and 0.0089 for women (RR with serum cholesterol higher by 40 mg/dL [1.03 mmol/L], 1.47 and 1.43, respectively) (multivariate coefficients not given). In the CHA study, of the 7676 women aged 18 through 39 years at baseline (mean, 26.8 years), only 17 died from CHD during 25 years of follow-up, too few to permit valid multivariate analyses of risk factor relations. In Cox models like those for men, the coefficient for the serum cholesterol level–CHD relationship was 0.0059 (t = 0.915),⁷ ie, qualitatively similar to the CHA men, the Framingham women, and the Johns Hopkins and Framingham men.

All these data are consistent in showing high RRs of premature CHD and CVD death for younger adults with high serum cholesterol levels vs those with favorable levels. The CHA and Johns Hopkins findings further indicate that for those with young adult hypercholesterolemia, both absolute risks and absolute excess risks of CVD disease and death are high during the decades of middle age, to age 65 years.

In sum, together with previously published findings, data from this study provide powerful additional support to current public policy. First, the results underline the strategic importance of population-wide primary prevention of unfavorable serum cholesterol levels (and other major risk factors), by improvement of lifestyles—particularly primary eating habits—from conception and weaning on, so that a progressively higher proportion of adults of all ages are at low risk.^5,21- 22 Second, they support population-wide efforts to identify children, teenagers, and young adults—as well as others—with unfavorable serum cholesterol levels (and other major risk factors), so that early therapeutic efforts can be instituted, first and foremost, to improve nutrition.^{2,4,14,21- 22} This second component of public policy is further buttressed by data from autopsy studies showing the significant coronary atherosclerosis in young American men and the relationship of serum lipids to such lesions.^23- 24

The feasibility of these objectives—and the fact that significant progress has been made toward their achievement—is indicated by the extensive data on population-wide reductions in intake of cholesterol, saturated fat, and total fat over recent decades, and the concomitant declines in adult prevalence rates of hypercholesterolemia and in population-wide average serum cholesterol level from about 235 to 240 mg/dL (6.08-6.21 mmol/L) in the 1950s to about 200 mg/dL (5.17 mmol/L) currently.⁵ The national health goal of an adult average serum cholesterol level no greater than 200 mg/dL (5.17 mmol/L) has been achieved.²² The value of implementing these goals is supported by extensive data from many randomized controlled trials. These data demonstrate efficacy for primary and secondary (dietary and pharmacologic) interventions for CHD and CVD prevention in achieving sustained reductions of unfavorable serum cholesterol levels.^4- 5,14 While there are no such trials in young adults (such studies would need more than 20 years to assess effects on "hard" clinical end points), it is a reasonable inference that improved nutrition to reduce serum cholesterol levels is efficacious for the whole population, including all young adults and particularly those who already have unfavorable levels.