Isolated Low HDL Cholesterol As a Risk Factor for Coronary Heart Disease Mortality
A 21-Year Follow-up of 8000 Men
For the purpose of screening individuals at high risk for coronary heart disease (CHD), serum total cholesterol (TC) of 5.2 mmol/L has been set as a value dividing “desirable” from intermediate high or elevated levels, and HDL cholesterol (HDL-C) <0.9 mmol/L has been labeled as abnormally low, implying high CHD risk. It has been conjectured that low HDL-C poses no risk in the absence of elevated LDL cholesterol or TC. To assess the risk of CHD-free men with “isolated low HDL-C,” ie, abnormally low HDL-C with desirable TC, we examined the CHD and all-cause mortality of some 8000 Israeli men aged 42 years and older during 1965 through 1986. Men with isolated low HDL-C represented one sixth of the cohort. CHD mortality among these men was 36% higher (age adjusted) than in counterparts with desirable TC, of which >0.9 mmol/L was contained in the high-density fraction. In men with TC>5.2 mmol/L, abnormally low HDL-C was associated with a virtually identical CHD mortality risk ratio, 38%. These findings persisted after adjustment for multiple CHD risk factors. The excess CHD risk associated with isolated low HDL-C appeared particularly increased in men with diabetes mellitus, whose death rate was 65% higher than in diabetics with HDL-C >0.9 mmol/L. A second subgroup result was consistent with equal CHD mortality risk among men in the “desirable” TC range, with or without low HDL-C, if systolic blood pressure was >160 mm Hg. These are post hoc findings, and hypotheses arising from these observations would require independent examination. Total mortality was not increased in men with isolated low HDL-C compared with men who had HDL-C <0.9 mmol/L and TC >5.2 mmol/L at baseline. These results indicate that an increased risk of CHD death is associated with abnormally low HDL-C for cholesterol ranges both below and above 5.2 mmol/L. For the individual, therefore, the risk is multiplied by the same amount regardless of TC. Quitting smoking, increasing physical activity, and decreasing body weight would all contribute to raise HDL-C in individuals of most or all age groups. When examined from a community perspective, the results are consistent with a relatively low population-attributable fraction among CHD-free men. This would tend to support the recommended practice of considering a TC level of 5.2 mmol/L (200 mg/dL) as a threshold for further evaluation in screened individuals without manifest CHD.
- Received August 9, 1995.
- Revision received March 28, 1996.
A large body of epidemiological research attests to the role of HDL-C as a marker inversely and independently associated with the risk of developing CHD.1 2 3 High blood levels of HDL-C may reflect genetic and/or environmental influences.4 5 Increasing physical activity,6 decreasing alcohol intake,7 quitting cigarettes,8 and possibly losing weight9 can elevate HDL-C. Drugs such as fibric acid derivatives or niacin have the same effect.10 Ongoing drug trials are underway to test whether raising HDL-C will in fact improve prognosis in CHD patients.11 12
The second report of the Expert Panel on Blood Cholesterol in Adults13 places increased emphasis on CHD risk status as a guide to cholesterol therapy. While the report identifies LDL-C as the primary target, attention is generally recommended for HDL-C concentrations <0.9 mmol/L. However, the management of low HDL-C remains an unresolved issue. In particular, isolated low HDL-C (≤0.9 mmol/L in the absence of TC >5.2 mmol/L) in CHD-free persons is unlikely to obtain attention. In a screening and potential treatment setting conforming to the National Cholesterol Education Program adult treatment panel II recommendations, some of these patients could be subjected to reduction of LDL-C, a therapy of clearly proven efficacy in both primary14 and secondary15 prevention. But since individuals with “desirable total cholesterol” (≤5.2 mmol/L) are usually advised to undergo recurrent reexamination in 5 years, without further action with respect to lipoproteins, isolated low HDL-C would remain partly untraced and without clear indication as to what procedures, if any, are required.
Consistent with long-term observations in other studies, a previous study of Israeli male civil servants has demonstrated that CHD-free survival is directly associated with TC and inversely with HDL-C levels.16 Given these findings, we undertook the present investigation of the same cohort to address the following issues. First, is isolated low HDL-C associated with an increased risk of CHD mortality? Second, does the detection of isolated low HDL-C carry public health significance, such that individuals with no manifest CHD and TC <5.2 mmol/L should be screened for HDL-C?
In the present study we used the results of a 21-year mortality follow-up of 8586 Israeli men aged 42 years and older at onset. Our purpose was to estimate the frequency of isolated low HDL-C in a working male population, compare their characteristics to counterparts with other HDL-C/TC combinations, and comparatively assess the prognosis of these groups in terms of CHD and all-cause mortality. Results are not available for women because in the 1960s there were too few middle-aged women working in tenured government and municipal positions in Israel for an eventual assessment of risk factor role in CHD development among them.
Subjects and Measures
This study is based on an investigation of 10 059 men, civil servants and municipal employees in Israel, who underwent extensive clinical, anthropometric, biochemical, and psychosocial evaluations in 1963, 1965, and 1968. Details of the methodology for these investigations and the mortality follow-up and analysis are available.16 17
Venous blood from nonfasting subjects was drawn into 20-mL dry evacuated tubes (no anticoagulant) after electrocardiography but before a psychosocial and dietary survey. Drawing venous blood from nonfasting individuals, a common denominator of all the early large epidemiological studies, primarily affects the levels of TGs and VLDL cholesterol, which were not assessed in this study; this practice would have little effect on HDL-C levels. Blood samples were kept refrigerated and were shipped daily in ice-cooled containers to a central laboratory, where tubes were centrifuged until the serum separated; cells were discarded. Cholesterol levels were determined by using Anderson and Keys'18 modification of the Abell method. HDL-C (“alpha lipoprotein”) was measured by using the method of Burstein and Samaille,19 which involved precipitating the “beta lipoproteins” (the LDL and VLDL fractions, in contemporary terminology) by adding heparin (0.04 mL) in the presence of manganese chloride (0.05 mL) to 1 mL serum. Centrifugation was done for 15 minutes. The precipitant, essentially beta lipoprotein, included no alpha lipoproteins, and the supernatant, upon electrophoresis, remained free of beta and exhibited the band for alpha lipoprotein. Taking the dilution into account, the cholesterol level in the supernatant was obtained by multiplying the supernatant concentration by 1.09.
The serum cholesterol values were determined on the basis of two replications. The SDs were 7.1 mg/dL for known duplicates and 13.3 mg/dL for blind replicates. The “technical error” would therefore lie between 4.9 and 9.4 mg/dL (the above figures divided by 2). Although HDL-C was also determined in 1963, albeit for only the last two thirds of the examined men, in that examination some HDL apparently precipitated with the lipoproteins of lower densities, so that on average the 1963 levels were 4.2 mg/dL lower.
Design and Statistical Methods
An upper limit of 5.2 mmol/L for desirable TC and 0.9 mmol/L for undesirable HDL-C were chosen, in conformation with National Cholesterol Education Program adult treatment panel policy,13 to permit comparison with other studies. Direct age-adjusted rates of CHD and all-cause mortality have been computed for HDL-C by TC groups per person-year according to 10-year age groups (40 to 49, 50 to 59, 60 to 69, 70 to 79, and 80 and above), taking into account the shifting age of the study population over the long follow-up period.20
The pooled risk of CHD and all-cause mortalities associated with HDL-C ≤0.9 and TC ≤5.2 mmol/L relative to that incurred by having HDL-C >0.9 mmol/L among men with cholesterol ≤5.2 mmol/L was estimated by using the Cochran-Mantel-Haenszel method. Stratification was done by age (<50 and ≥50 years), diabetes mellitus, current smoking versus nonsmoking at the beginning of the 21-year follow-up (1965), SBP ≤160 and >160 mm Hg, and QI (weight in kilograms divided by height in meters squared) ≤25 or >25. The above RRs were compared with the respective RRs associated with HDL-C ≤0.9 mmol/L risk in men whose 1965 serum TC was >5.2 mmol/L.
The adjusted population-attributable risk of mortality was calculated as the weighted average of the age-specific rate difference where ai and bi are the number of cases (CHD deaths) among the “exposed” (abnormally low HDL-C) and unexposed subjects in age group i. Ai and Bi are the number of person-years of follow-up in exposed and unexposed men in age group i. The weights Wi were The PAF of mortality was calculated from the RR as where E is a weighted mean WiEi of the Ei, the proportion of men exposed to the factor (say, HDL-C ≤0.9 mmol/L) in age group i. In this formula, the weights Wi are the proportion that men in age group i, regardless of HDL-C level, constitute in the total sample. Calculations were done by using RATES2 of Computer Programs for Epidemiologic Analysis (CPEA).21 Relative hazards of all-cause and CHD mortality associated with HDL-C ≤0.9 mmol/L were also estimated by using the Cox proportional hazards model.22 The Kaplan-Meier curves,23 which depict CHD mortality over 21 years in men without MI or definite angina in 1965, were derived from SAS LIFETEST,24 and the estimated CHD mortality risk ratios were derived from cumulative probabilities by using the SURVIVAL procedure of CPEA.21
Multiple Comparison Issue
The subdivision of TC to “desirable” and higher blood levels as well as that of HDL-C to “abnormally low” and above are arbitrary, as they are based on approximate trends of lipoprotein-associated risks. Statistical treatment of mortality rates within cholesterol ranges, as if the latter constituted distinct biological-clinical entities, is open to criticism on grounds of “data dredging.” An examination and comparison of rates in a certain combination of TC/HDL-C ranges should be considered as only one in a large number of possible contrasts. We therefore qualify the statistical comparisons accordingly. Without quantitatively calculating an “experimentwise” rather than “comparisonwise” statistical error, we focus on the similarity or dissimilarity of risks associated with abnormally low HDL-C in the “desirable” and increased TC ranges but refrain from formal statistical evaluation of risk ratios as to whether they are >1.
Of the 10 059 men under study, 8586 (85%) participants from the second round of examinations (1965; see “Methods”) had a valid measurement of serum HDL-C. Of these 8586 men, 4051 (47.2%) had TC ≤5.2 mmol/L in 1965. Of the latter, 1494 (36.9%) exhibited HDL-C ≤0.9 mmol/L. Table 1⇓ shows demographic, clinical, anthropometric, and biochemical findings for four groups in 1965: “isolated low” (HDL-C ≤0.9 mmol/L and TC ≤5.2 mmol/L, n=1494 [17.4%]); “high-low” (HDL-C >0.9 mmol/L but TC ≤5.2 mmol/L, n=2557 [29.8%], representing those with the “best” lipoprotein profile); “high-high” (HDL-C >0.9 mmol/L and TC >5.2 mmol/L, n=3357 [39.1%]); and “low-high” (with both unfavorable HDL-C ≤0.9 mmol/L and TC >5.2 mmol/L, n=1178 [13.7%]). This nomenclature is not meant to imply that all men with HDL-C >0.9 mmol/L have high HDL-C.
Arising from the definition of the groups by cholesterol and its fractions, TC in the high-high and low-high groups was approximately 1.3 mmol/L higher than in the isolated low and high-low ones, while HDL-C was 0.34 to 0.38 mmol/L lower in the low HDL-C groups than in the “>0.9 mmol/L” ones. Several expected differences between TC categories (Table 1⇑) included statistically higher ages and mean systolic and diastolic pressures in the high-high and low-high groups compared with isolated low and high-low. The four groups differed significantly (P<.01 by ANOVA) in means of QI, the frequencies of current smoking habits, occupational physical activity, and history of heart attack. High-low men were the leanest (QI=25.0 kg/m2) among the four groups, had the smallest percent of current smokers (45.0%) when the study began, and included the largest proportion of men who were physically active on the job. There was almost no difference in the reported frequency of nonsporadic leisure time activity between the groups.
The calculation and comparison of 21-year mortality have been restricted to men free of old MI and definite angina pectoris (n=7686). Of these, 2432 (31.6%) died through 1986. Of the latter, 697 died of CHD. Table 2⇓ shows the crude rates for 21-year CHD and all-cause mortality as well as the crude and age-adjusted rates per 10 000 man-years in the four groups. The highest age-adjusted rate for CHD mortality was observed in low-high men (6.6) and the lowest one in their high-low counterparts (2.8). CHD mortality for high-high (4.9) and isolated low (3.8) were intermediate. For all-cause mortality, the age-adjusted rates in the low-high, high-high, high-low, and isolated low groups were 18.0, 15.3, 13.9, and 14.1, respectively. The adjusted isolated low/high-low RR for CHD mortality, namely the risk among men with TC ≤5.2 mmol/L associated with isolated low HDL-C vis-à-vis men with higher HDL-C, was 1.38 (95% CI=1.06 to 1.78), the attributable risk (risk difference, see “Methods”), 4.9 per 10 000 man-years (95% CI=−3.1 to 13.0), and the PAF, 12.2% (95% CI=2.3 to 22.3%). For total mortality, the adjusted isolated low/high-low RR was 1.02, and the PAF attributable to isolated HDL-C was estimated at 0.
For men with TC>5.2 mmol/L, the respective ratio (low-high/high-high) was 1.35 (95% CI=1.10 to 1.64), so that the risk ratio was similar to the value observed for the isolated low/high-low group. The attributable risk associated with HDL-C <0.9 mmol/L, however, was larger in men with TC>5.2 mmol/L (8.3 per 10 000 man-years [95% CI=−3.9 to 20.5]; the PAF, 8.1% [95% CI=2.6 to 14.1%]). The attributable risk reflects the higher CHD mortality in men with TC >5.2 mmol/L than in counterparts with “desirable TC.” The smaller PAF reflects the somewhat smaller proportion of men with abnormally low HDL-C among all men with TC >5.2 mmol/L.
Calculation of life-table cumulative estimates permits a consideration of the timing of mortality. Kaplan-Meier curves depicting CHD mortality over 21 years in men without MI or definite angina in 1965 within HDL-C×TC categories were generated (the Figure⇓). The cumulative probabilities of CHD death were 8.0%, 6.4%, 12.0%, and 14.6% for the isolated low, high-low, high-high, and low-high men, respectively. This yielded an estimated risk ratio of 1.24 associated with isolated low HDL-C (cumulative CHD probability of isolated low divided by high-low). By comparison, the estimated cumulative risk ratio (low-high/high-high) associated with HDL-C ≤0.9 mmol/L among men with cholesterol >5.2 mmol/L was 1.21.
Subgroup and Multivariate Analysis
Since diabetes mellitus and smoking strongly affect both HDL-C and survival, we analyzed mortality in categories of diabetes and current smoking. The results indicate that the increased risk of CHD death associated with isolated low HDL-C versus high-low was estimated at 1.65 for diabetics versus 1.18 in nondiabetics and 1.21 for current (in 1965) cigarette smokers and 1.26 for nonsmokers (Table 3⇓). We also calculated the respective risks for men with SBP >160 mm Hg (RR=1.00) and SBP ≤160 mm Hg (RR=1.37). The corresponding relative risks for men above and below QI=25 kg/m2 (not tabulated) were 1.12 and 1.38, respectively. These subgroup analyses were not subjected to formal statistical testing as the multiplicity of post hoc testing for data-generated hypotheses is notorious for yielding spurious associations.
The next step in the analysis was simultaneous control for variables that may independently predict mortality and potentially confound the association between isolated low HDL-C and mortality. We evaluated the risk ratios for CHD and all-cause death associated with the isolated low and high-low groups (Table 4⇓). For CHD mortality, HDL-C ≤0.9 mmol/L (isolated low) resulted in a risk ratio of 1.25 vis-à-vis high-low, ie, counterparts in the same desirable cholesterol range but without the frankly reduced HDL-C. For men with TC >5.2 mmol/L, the respective risk ratio was similar, 1.22.
The analysis, based on the Cochran-Mantel-Haenszel pooled risk ratio for person-year data, controlled for age (below or above 50 years), SBP, current cigarette smoking, QI, and diabetes mellitus (see “Methods”). Adjusted all-cause mortality was equal in isolated low and high-low men. Calculation of the presumed “net effect” of isolated low HDL-C assuming proportional hazards, controlling for the same variables as well as serum glucose, and using a continuous scale for age and all the continuous variables resulted in somewhat higher estimates. The adjusted risk ratio associated with isolated low HDL-C was 1.37 (95% CI=1.06 to 1.77) for 21-year CHD mortality, and HDL-C ≤0.9 mmol/L in men with TC >5.2 mmol/L (low-high) was associated with an adjusted risk ratio of 1.36 (95% CI=1.10 to 1.66) versus the high-high group.
It has been conjectured that if TC and LDL blood cholesterol levels are low enough then the risk of coronary atherosclerosis is small, and other risk factors may be inconsequential. This has already been disproved as far as the contribution of hypertension25 and smoking to the risk of CHD in normocholesterolemic men is concerned. For abnormally low HDL-C unaccompanied by elevated LDL-C, the issue deserves special attention. We found only one presentation, so far unpublished, specifically dealing with the significance of isolated low HDL-C in CHD-free individuals. Luepker et al26 studied 6192 men and 7565 women aged 25 to 74 years in six communities for an average of 5 years. Men with HDL-C ≤0.9 mmol/L and TC ≤5.2 mmol/L constituted 11% of all men in the cohort compared with 17% in the Israeli cohort. Similar concentrations were found in 12% of the women. CHD incidence was 1.32% and 1.41%, respectively, among men with TC <5.2 mmol/L in the groups with HDL-C ≤0.9 and >0.9 mmol/L, respectively. This is comparable to the small differences in a limited incidence phase of our study. Comparing the 3-year (1965 through 1968) CHD incidence among older men in the respective groups in the Israeli Ischemic Heart Disease Study,16 the incidence of acute MI was 2.83% and 3.21% and the incidence of definite angina 2.78% and 3.17%, respectively, in the low and normal HDL-C groups among men with desirable TC. Perhaps a short follow-up cannot fully reveal the extent of risk associated with isolated low HDL-C. Alternatively, low HDL-C may bear a yet unclear relevance to case-fatalities.
In women, the corresponding rates in the six-community study were 0.95% and 1.27%.26 For reasons that remain to be clarified, men and women in these six communities revealed excess CHD risk associated with HDL-C ≤0.9 mmol/L in the intermediate cholesterol range of 5.2 to 6.2 mmol/L but not at TC >6.2 mmol/L. The authors of the above study suggest that isolated low HDL-C is of little consequence. In our study, men with isolated low HDL-C numbered slightly over one sixth of the healthy population under study. Their risk of long-term CHD mortality was elevated by approximately one third, whereas their risk of all-cause mortality was only slightly elevated. The attributable risk associated with isolated low HDL-C in our study was small. A tentative conclusion is that for the total population a finding of TC >5.2 mmol/L implies that information potentially added by measuring HDL-C in the young and perhaps middle-aged man is limited. This may not be the case among the elderly or CHD patients.27 The major sex differences in the distributions of blood cholesterol fractions, notably HDL-C, and the age-dependent sex differences in the incidence of CHD underscore the need for a separate investigation in women.
Our subgroup analyses are consistent with a particularly increased risk of long-term fatal CHD risk associated with isolated low HDL-C in diabetics compared with nondiabetics. Because it is apparent that the diabetic state may influence lipid and lipoprotein metabolism, emphasis should be placed on managing dyslipoproteinemia in diabetic patients. Low HDL-C may substantially increase the hazard of CHD in diabetics.28 Hypertriglyceridemia, which was not investigated in the present study, often appears in conjunction with abnormally low HDL-C and is probably of major importance in increasing the risk of CHD death in diabetics.29 However, the suggested coeffect of diabetes and HDL-C on CHD mortality in men with a desirable cholesterol level is an impression generated post hoc and should be independently investigated in a large group of normocholesterolemic diabetics. The absence of any association between isolated low HDL-C and CHD mortality in men with SBP >160 mm Hg in our cohort lacks biological plausibility and may represent a chance finding typical of post hoc subgroup analysis. At best, it should be considered as a hypothesis to be tested in another cohort.
Evidence for the effect on CHD incidence of correcting low levels of HDL-C in CHD-free individuals is available only from the controlled Helsinki Heart Trial.30 For each 1 mg/dL (0.025 mmol/L) rise in HDL-C among the study subjects, the average response was an ≈3% decrease in the risk of new CHD. None of the 4081 study participants in the trial fulfilled the definition of isolated low HDL-C.
In large epidemiological studies it has been repeatedly observed that the risk curves of new CHD, as related to high LDL-C or low HDL-C, closely resemble each other. It is tempting to think that departures from high to average HDL-C levels would justify the treatment already established for individuals with above-average LDL-C levels. Currently, US National Cholesterol Education Program guidelines advocate the lowering of already “normal” LDL-C to as low as 2.6 mmol/L in patients with CHD and high-risk CHD-free individuals. No official recommendations to intervene with drugs against abnormally low HDL-C levels, even in CHD patients, have been formulated. For men and women with a history of clinically manifest CHD, the scope of the problem associated with abnormally low HDL-C may be larger, and the prognostic implications different, than in their CHD-free counterparts. Although the issue of CHD patients with isolated low HDL-C has been addressed,27 31 32 large-scale controlled clinical trials have not provided a clue for intervention. Two such studies are currently examining treatment of isolated low HDL-C. These trials, incorporating over 5500 patients, examine the efficacy of bezafibrate11 or gemfibrozil12 versus placebo in reducing the incidence of nonfatal MI and CHD death in CHD patients by elevating their HDL-C and reducing TG levels. Similar studies are neither undertaken nor envisaged in persons free of CHD, given the size and length of a trial that would be needed to establish the effect of LDL-C–raising therapy in persons without CHD.
Two main limitations apply to the current study. First, because the metabolism of HDL-C is intimately linked to that of serum TGs, an adjustment for serum TGs would have been appropriate. It was found impractical to collect fasting blood on 10 000 subjects examined at all times during the day, so TGs were not investigated. A large number of studies have found that an independent contribution of TGs to CHD risk has been diluted or has disappeared upon adjustment for HDL-C, but not vice versa. Presumably, the associations reported here would be weakly affected or unaffected by adjusting for TGs. Second, information on the changes in cholesterol fractions during the long follow-up period is not available.
In summary, the current results indicate that abnormally low HDL-C levels carry a similar excess risk in CHD-free men with TC <5.2 mmol/L as they do for counterparts with higher TC levels. In addition to the significance of isolated low HDL-C for the CHD-free individual, our results should be evaluated in terms of their preventive cardiology/public health significance. Isolated low HDL-C was associated with a small increase in attributable risk of dying from CHD. Despite all the evidence from our study sample and others that supports the unquestionable role of HDL-C as a protective factor in CHD, we hesitate to recommend its unqualified measurement in the absence of CHD with the exception of diabetics. For individuals with an extremely unfavorable risk due to family history of premature CHD and/or severe hypertension, individual clinical consideration will apply. Nonpharmacological public health advice, such as abstinence from smoking and lifelong commitment to physical activity, is prudent on established grounds. It is part and parcel of general recommendations to the public, including individuals with isolated low HDL-C.
Selected Abbreviations and Acronyms
|CHD||=||coronary heart disease|
|SBP||=||systolic blood pressure|
Gordon DJ, Probstfield JL, Garrison RJ, Neaton JD, Castelli WP, Knoke JD, Jacobs DR Jr, Bangdiwala S, Tyroler HA. High-density lipoprotein cholesterol and cardiovascular disease: four prospective American studies. Circulation. 1989;79:8-15.
Rader DJ, Brewer HB Jr. Lipids, apolipoproteins and lipoproteins. In: Goldbourt U, de Faire U, Berg K, eds. Genetic Factors in Coronary Heart Disease. Dordrecht, Netherlands: Kluwer; 1994:83-103.
Goldbourt U, Medalie JH. Characteristics of smokers, non-smokers and ex-smokers among 10,000 adult males in Israel, II: physiological, biochemical and genetic characteristics. Am J Epidemiol. 1977;105:75-86.
Wood PD, Stefanick MI, Dreon DM, Frey-Hewitt B, Garay SC, Williams PT, Superko HR, Fortmann SP, Albers JJ, Vranizan KM, Elsworth NM, Terry RB, Haskell WL. Changes in plasma lipids and lipoproteins in overweight men during weight loss through dieting as compared with exercise. N Engl J Med. 1988;319:1173-1179.
Goldbourt U, Behar S, Reicher-Reiss H, Agmon J, Kaplinsky E, Graff E, Kishon Y, Caspi A, Weisbort J, Mandelzweig L, et al. Rationale and design of a secondary prevention trial of elevating serum high-density lipoprotein cholesterol and reducing triglyceride in patients with clinically manifest atherosclerotic heart disease: the Bezafibrate Infarction Prevention Trial. Am J Cardiol. 1993;71:909-915.
Rubins HB, Robins SJ, Iwane MK, Boden WE, Elam MB, Fye CL, Gordon DJ, Schaefer EJ, Schectman G, Wittes JT. Rationale and design of the Department of Veterans Affairs High-Density Lipoprotein Cholesterol Intervention Trial (HIT) for the secondary prevention of coronary artery disease in men with low high-density lipoprotein cholesterol and desirable low density lipoprotein cholesterol. Am J Cardiol. 1993;71:45-52.
Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. Summary of the Second Report of the National Cholesterol Education Program (NCEP) Expert Panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel II). JAMA. 1993;269:3015-3023.
Anderson JT, Keys A. Cholesterol in serum lipoprotein fractions. Clin Chem. 1962;21:145-159.
Gahlinger PM, Abramson JH. RATES2: comparison of person-time incidence rates. Version 1.4. Computer Programs for Epidemiologic Analysis. Honolulu, Hawaii: Makapuu Medical Press; 1993:119-125.
Cox DR. Regression models and life tables. J R Stat Soc (Ser B). 1972;34:187-202.
The LIFETEST Procedure. In: SAS/STAT User's Guide, Vol 2, GLM-VARCOMP. Version 6, 4th ed. Cary, NC: SAS Institute; 1990:1028-1069.
Goldbourt U, Yaari S. Cholesterol and CHD mortality: a 23-year follow-up study of 9902 men in Israel. Arteriosclerosis. 1990;10:512-519.
Luepker RV, Sprafka JM, McGovern PG, Hannan PJ, Blackburn H. Isolated low HDL cholesterol as a risk factor for coronary heart disease. Circulation. 1993;88(suppl I, part 2):I-511. Abstract.
Sacks FM. Desirable serum total cholesterol with low HDL cholesterol levels: an undesirable situation in coronary heart disease. Circulation. 1992;86:1341-1344.
Goldschmid MG, Barret-Connor E, Edelstein SL, Wingard DL, Cohn BA, Herman WH. Dyslipidemia and ischemic heart disease mortality among men and women with diabetes. Circulation. 1994;89:991-997.
Fontbonne A, Eschwege E, Cambien F, Frick MH, Elo O, Haapa K, Heinonen OP, Heinsalmi P, Helo P, Huttunen J. Hypertriglyceridemia as a risk factor of coronary heart disease mortality in subjects with impaired glucose tolerance or diabetes: results from 11-year follow-up of the Paris Prospective Study. Diabetologia. 1989;32:300-304.
Miller M, Seidler A, Kwiterovich PO, Pearson TA. Long-term predictors of subsequent cardiovascular events with coronary artery disease and ‘desirable’ levels of total cholesterol. Circulation. 1992;86:1165-1170.
Goldbourt U, Benderly M, Graff E, Mandelzweig L, for the BIP Study Group. Lipids and lipoproteins in symptomatic coronary heart disease: distribution, intercorrelation, and significance for risk classification in 6,700 men and 1,500 women. Circulation. 1992;86:839-848.