Do Total and High Density Lipoprotein Cholesterol and Triglycerides Act Independently in the Prediction of Ischemic Heart Disease?
Ten-Year Follow-Up of Caerphilly and Speedwell Cohorts
Several studies have suggested that men with raised plasma triglycerides (TGs) in combination with adverse levels of other lipids may be at special risk of subsequent ischemic heart disease (IHD). We examined the independent and combined effects of plasma lipids at 10 years of follow-up. We measured fasting TGs, total cholesterol (TC), and high density lipoprotein cholesterol (HDLC) in 4362 men (aged 45 to 63 years) from 2 study populations and reexamined them at intervals during a 10-year follow-up. Major IHD events (death from IHD, clinical myocardial infarction, or ECG-defined myocardial infarction) were recorded. Five hundred thirty-three major IHD events occurred. All 3 lipids were strongly and independently predictive of IHD after 10 years of follow-up. Subjects were then divided into 27 groups (ie, 33) by the tertiles of TGs, TC, and HDLC. The number of events observed in each group was compared with that predicted by a logistic regression model, which included terms for the 3 lipids (without interactions) and potential confounding variables. The incidence of IHD was 22.6% in the group with the lipid risk factor combination with the highest expected risk (high TGs, high TC, and low HDLC) and 4.7% in the group with the lowest expected risk (P<0.01). A comparison of the predicted number of events in the 27 groups with the number of events observed showed that a logistic regression provided an adequate fit without the need to incorporate interactions between lipids in the model. Conclusions are as follows: (1) Serum TGs, TC, and HDLC are independently predictive of IHD at 10 years of follow-up. (2) Combinations of adverse levels of the 3 major lipid risk factors have no greater impact on IHD than that expected from their individual contributions in a logistic regression model. There was no evidence that men with low HDL/raised TGs were at significantly greater risk than that predicted from the independent effects of the 2 lipids considered individually.
Plasma total cholesterol concentration is well established as a major risk factor for ischemic heart disease (IHD).1,2 Within the past 20 years, HDL cholesterol has also been added as a major lipid risk factor independent of total cholesterol.3,4
The role of triglycerides remains uncertain.5,6 We have previously reported that at the first follow-up (5 years) of the Caerphilly and Speedwell Studies, triglyceride was an important and statistically significant predictor of IHD, even after adjustment for the major conventional nonlipid risk factors and for total and HDL cholesterol.7 In the Copenhagen Male Study,8 fasting triglyceride was an independent risk factor for subsequent IHD at the 8-year follow-up.
In a report based only on the Caerphilly cohort at 10 years of follow-up, serum total cholesterol, HDL cholesterol, and triglyceride concentrations were independently predictive of IHD.9
The effects of combinations of lipid factors on IHD have been examined in several studies, notably the Framingham,10 Munster,11 Honolulu,12 Physicians Health,13 and Helsinki Heart14 Studies and the Copenhagen Male Study.15 These studies have been reviewed in detail elsewhere9,11,16 and suggest that risk of IHD is significantly increased in subjects with high triglyceride levels and either elevated total or LDL cholesterol or low HDL cholesterol. In the present report, we examine the effect of combinations of lipid levels on the 10-year risk of subsequent IHD in a British cohort of middle-aged men. Since triglyceride measurements have been reported to show a greater degree of measurement error (because of laboratory and biological variability), 17,18 the effect of measurement error was also estimated in a previous report from the Caerphilly study.9
In Caerphilly, all men were selected from within a defined area. They were aged 45 to 59 years when first examined. A total of 2512 men were seen, ie, 89% of the 2818 that were found to be eligible. In Speedwell, all men were selected from the age-sex registers of 16 general practitioners working from 2 neighboring health centers. A total of 2348 men (92% of those eligible) were seen during the following 3 years. The men were aged 45 to 59 years when chosen, immediately before the study began in 1978; thus, they were aged between 45 and 63 years when first examined. The combined cohort consisted of 4860 men.
Survey Methods and Follow-Up Procedure
The 2 studies had a common core protocol and common procedures. These have been described in detail elsewhere.7,19 The studies were each approved by the appropriate ethics committee, and each subject gave informed consent. Briefly, at recruitment, the men attended an afternoon or evening clinic. At the clinic, a standard medical and smoking history was obtained; the London School of Hygiene and Tropical Medicine Chest Pain questionnaire was administered; height, weight, and blood pressure were measured; and a 12-lead ECG was recorded. The men then returned after an overnight fast to an early morning clinic, during which a blood sample was taken with minimal venous stasis. Fasting samples were obtained from 4641 men.
The 10-year follow-up in Caerphilly was at a nearly constant interval of 120 months (SD 6 months) and was the second follow-up of the cohort. In Speedwell, results related to the third follow-up, and the mean interval was 112 months (SD 3 months).
At each follow-up, the chest pain questionnaire was again administered, and a further ECG was recorded. All ECGs were coded by 2 experienced coders using the Minnesota scheme. The chest pain questionnaire was extended to include questions about hospitalization for severe chest pain. These, together with Hospital Activity Analysis notifications of admissions coded as 410 to 414 (IHD) in the 9th revision of the International Classification of Diseases, were used as the basis for a search of hospital notes, which were checked by an epidemiologist for events that satisfied the World Health Organization criteria for definite acute myocardial infarction. For men who had died before the end of the follow-up, a copy of the death certificate was automatically received from the National Health Service Central Register. From this information, 3 categories of incident IHD events were defined: IHD death, clinical nonfatal (definite acute) myocardial infarction, and ECG myocardial infarction, as previously described.7,19 A major IHD event was defined as any of the 3 possible outcomes described above.
Because of the heavy workload, separate laboratories had to be used for the lipid analyses for the 2 areas. Plasma samples were transported at 4°C by rail to the laboratories on the day of venipuncture. Cholesterol and triglyceride concentrations were measured by enzymatic procedures.20,21 The HDL fraction was isolated by precipitation of LDLs and VLDLs with sodium phosphotungstate and magnesium chloride (Caerphilly)22 or with heparin and manganese chloride (Speedwell).23 All results presented are based on plasma samples obtained after an overnight fast.
Over the course of the recruitment phase, 91 of the Speedwell blood samples were split, with 1 aliquot going to the Speedwell laboratory and 1 aliquot going to the Caerphilly laboratory. Mean differences in concentration between laboratories were 0.50 mmol/L for cholesterol, 0.01 mmol/L for HDL cholesterol, and 0.11 mmol/L for triglycerides. The Speedwell laboratory found higher values for cholesterol and HDL cholesterol and lower values for triglycerides. These between-laboratory differences were consistent from sample to sample, and the variation between laboratories was indistinguishable from the variation when 2 aliquots were sent to the same laboratory. In all analyses, the concentrations of cholesterol, HDL cholesterol, and triglycerides were adjusted to those of the Speedwell laboratory by using the mean differences cited.
Mean differences in lipids between various groups (Table 1) were adjusted for age by ANCOVA. The rest of the analysis was performed by multiple logistic regression, with the occurrence or not of a major IHD event as the dependent variable. The slight difference in the duration of follow-up between the 2 studies was taken into account by the inclusion of study area as a covariate in the model. Survival analysis was not used because no time to event was available for ECG-defined myocardial infarctions. The lipid distributions were divided into equal thirds (defined separately for each study area), and the results were presented as the odds of major IHD in each third relative to the baseline third; 95% CIs for these relative odds were also estimated.
Preexisting IHD was adjusted for by including 3 factors in the logistic regression models: (1) angina (none, grade I, or grade II) from the chest pain questionnaire, (2) history of severe chest pain lasting half an hour or more (yes or no) from the chest pain questionnaire, and (3) evidence of ischemia (none, possible, or probable) from the Minnesota-coded ECG.
Combinations of lipid risk factors were derived from the division of the distributions of each of the lipids into thirds denoting high, medium, and low risk. This yielded 33 (ie, 27) lipid risk factor combinations. Logistic regression models were used to calculate the predicted risk of major IHD events for each of the 27 groups of men. Interaction terms between pairs of lipids were added as appropriate.
Univariate analyses were based on 4362 men who provided a fasting blood sample at recruitment and who had a complete set of data for all variables used in the analysis. By 10 years, 533 (12.2%) had had a major IHD event.
Lipid levels in subjects experiencing a major IHD event by 10 years and in those who did not are shown in Table 1. Age-adjusted mean levels of total cholesterol and triglycerides were significantly (P<0.001) elevated among men who developed IHD over the 10-year follow-up period, whereas levels of HDL cholesterol were significantly (P<0.001) lower.
In the remaining analyses, subjects were divided into approximately equal thirds of the distribution of each of the lipid variables. Initially, relative odds were adjusted for age and area only, and then additional potential confounders were added. Nonlipid factors included in the analyses in addition to age and area of residence were as follows: preexisting IHD, smoking habit, diastolic blood pressure, and body mass index. These results are shown in Table 2.
Relative odds of IHD increase steadily as total cholesterol and triglyceride levels increase and as the HDL cholesterol level decreases. With adjustment for all nonlipid risk factors and for triglycerides and HDL cholesterol, the relative odds of IHD in the third of men with the highest levels of total cholesterol were 1.59 (95% CI 1.25 to 2.03) compared with the third of men with the lowest total cholesterol. The corresponding relative odds, with adjustment for all nonlipid and the other lipid factors, for triglycerides and HDL cholesterol were 1.53 (95% CI 1.18 to 2.00) and 1.43 (95% CI 1.12 to 1.82), respectively; low levels of HDL cholesterol are presented as high risk. Therefore, all 3 lipid factors contribute independently to the risk of IHD.
We next examined the 10-year incidence of IHD in relation to combinations of lipid risk factors. All lipid distributions were divided, as previously, into approximate thirds of their distributions. This gave 33 (ie, 27) groups of men with different lipid risk factor combinations, which were then individually examined for 10-year risk of IHD. Table 3 shows these results displaying combinations of lipid risk factor in order of predicted risk.
The lowest observed 10-year incidence of major IHD (4.7%) was found among the group of men who were in the lowest risk category for each of the 3 lipids. Similarly, the highest observed incidence (22.6%) was found among the men who were in the highest risk category for all 3 lipids. The risk of a major IHD event was estimated for each man from a logistic regression model containing terms for each of the 3 lipid risk factors (considered by thirds, but without interaction terms) and incorporating nonlipid risk factors. These risks were then summed within each of the 27 (33) lipid risk factor combinations to give a predicted number of major IHD events. The number of major IHD events observed within each of the 27 subgroups closely matched the numbers predicted. The test of goodness of fit of the model comparing the predicted number of events in the 27 groups with the observed number of events was not significant. (χ2 22.69, df 20, P=0.30). Furthermore, addition of terms to the model to represent the 3 pairwise interactions between lipids did not lead to a significant improvement in fit (χ2 16.42, df 12, P=0.17).
Observed and expected event rates are shown in full in the Figure, with the 27 subgroups arranged on the horizontal axis in ascending order of predicted risk, taking into account differences in the distribution of other (nonlipid) risk factors.
The figure confirms and extends the findings shown in Table 3. The observed event rates, calculated from the number of men experiencing major IHD, shown for each of the 27 combinations of lipids, arranged in the order of their predicted risk, demonstrate a progression from the lowest to the highest level of predicted risk. The continuous line shows the predicted event rate when other major risk factors in addition to lipids are included in the model. This shows that observed and predicted risks are in good agreement, and the 95% CIs for the observed rates overlap the predicted IHD risk for all lipid combinations. In comparison, the event rate predicted for the model containing nonlipid risk factors only is also shown (broken line). This model showed much less ability to discriminate between the high and low risk associated with the extreme of the lipid distributions, with the overall predicted risk varying from 11% to 15% across the range of men (compare 6% to 23% for the model incorporating lipids also).
To address the hypothesis that men with low HDL cholesterol and high triglyceride levels were at excess risk, we specifically examined the subgroup of men whose HDL cholesterol levels were in the lowest third and whose triglyceride levels were in the highest third irrespective of their total cholesterol level. We found this subgroup of 653 men to be at no greater risk than would be predicted from a logistic regression model without interaction terms (119 observed events, 120.8 predicted events).
All lipid variables show strong univariate associations with incident IHD at 10 years of follow-up in these cohorts (Table 1). Multivariate analysis also shows that each lipid is independently predictive of IHD (Table 2), which confirms and extends the findings based only on the Caerphilly cohort.9 In the Caerphilly cohort, regression dilution bias was estimated by using repeated lipid measures, but comparable data were not available from the Speedwell cohort; therefore, no estimates of regression dilution bias have been made in the present report. The possible effects of measurement errors are discussed in detail in a previous report.9
Maturity-onset diabetic patients were not excluded from the present analysis, and cardiovascular risk factors have been shown to be particularly common in this group.24 Glucose levels >5.0 mmol/L were found to be significantly associated with increased risk of subsequent IHD independent of lipid and nonlipid cardiovascular risk factors.25 The inclusion of glucose in the present analysis did not affect the findings (data not shown). Among nondiabetic subjects, however, there is little evidence of the specific clustering of cardiovascular risk factors that is sometimes labeled the “metabolic syndrome.”19 Exclusion of 162 subjects with ECG evidence of IHD at baseline (major or moderate Q waves on the baseline ECG) did not materially affect these results. Among the thirds of men with highest risk shown on Table 2, relative odds for total cholesterol and HDL cholesterol were reduced to 1.50 (95% CI 1.17 to 1.94) and 1.36 (95% CI 1.06 to 1.76) (adjusted for all nonlipid and the other lipid risk factors), but those for triglyceride were increased to 1.68 (95% CI 1.27 to 2.22).
Details of the intercorrelations between the lipids have been published previously.7 Briefly, total cholesterol and HDL cholesterol are not associated statistically (r=0.04), HDL cholesterol and triglycerides are negatively associated with each other (r=−0.26), and triglycerides are positively associated with total cholesterol (r=0.35). LDL cholesterol, calculated from the Friedewald formula as in the Munster Study,11 was not considered an independent risk factor in the present report because it is correlated highly with total cholesterol (r=0.92).
In these samples of British men, we found no evidence of interaction between the 3 lipids and the risk of subsequent IHD. This is in contrast to studies conducted in Germany11 and among Japanese Americans in Hawaii,12 both of which have suggested that certain lipid combinations are associated with a greater risk than would have been predicted from their individual effects. By examining the atherosclerotic coronary artery disease risk in numerous subgroups of patients in the Munster study,11 the authors suggested that hypertriglyceridemia is a powerful additional coronary risk factor when it coincides with a high ratio of LDL to HDL cholesterol. When we used our data to address this issue, we found no evidence of interaction between elevated triglyceride level (≥2.3 mmol/L) and high LDL/HDL ratio (>5.0). We also reproduced the analysis of the Honolulu Heart Program12 by using their cut points for defining high triglyceride, low HDL cholesterol, and elevated, borderline-elevated, and desirable cholesterol levels; contrary to their findings, we observed that the odds of IHD in the high-risk (high triglyceride/low HDL cholesterol) group relative to the low-risk (low triglyceride/high HDL cholesterol) group were similar in subgroups of men with desirable, borderline-elevated, and elevated cholesterol levels. These apparent inconsistencies may be attributable to spurious findings arising as an inevitable consequence of examining numerous subgroups of the data, and a preferable statistical approach is to fit interaction terms in a statistical model. Alternatively, the inconsistencies might reflect differences in the mixture of cardiovascular risk factors present in the different populations studied. For example, high triglyceride levels have been reported in a Chinese population26 known to be at low risk for IHD but at high risk for hypertension and stroke.27 Such differences between populations remain to be fully explored.
In conclusion, our results indicate that total and HDL cholesterol and triglyceride levels are each independently predictive of subsequent major IHD at 10 years of follow-up in our cohorts of British men. The notion that the combination of a low HDL cholesterol level with a raised triglyceride level or other combinations of lipid risk factors confers unexpected levels of risk is unsubstantiated by our data.
This study was supported by the Medical Research Council, UK. P.M.S. is currently supported by the British Heart Foundation.
Received March 13, 2001; revision accepted May 18, 2001.
Verschuren WMM, Jacobs DR, Bloemberg BPM, Kromhout D, Menotti A, Aravanis C, Blackburn H, Buzina R, Dontas A, Fidanza F, et al. Serum total cholesterol and long-term coronary heart disease mortality in different cultures: twenty-five year follow-up of the Seven Countries Study. JAMA. 1995; 274: 131–136.
Law MR, Wald NJ, Wu T, Hackshaw A, Bailey A. Systematic underestimation of association between serum cholesterol concentration and ischaemic heart disease in observational studies: data from the BUPA study. BMJ. 1994; 308: 363–366.
Pocock SJ, Shaper AG, Phillips AN. Concentrations of high density lipoprotein cholesterol, triglycerides and total cholesterol in ischaemic heart disease. BMJ. 1989; 298: 998–1002.
Austin MA. Review: plasma triglyceride and coronary heart disease. Arterioscler Thromb. 1991; 11: 2–14.
Bainton D, Miller NE, Bolton CH, Yarnell JWG, Sweetnam PM, Baker IA, Lewis B, Elwood PC. Plasma triglyceride and high density lipoprotein cholesterol as predictors of ischaemic heart disease in British men: the Caerphilly and Speedwell Collaborative Heart Disease Studies. Br Heart J. 1992; 68: 60–66.
Jeppesen J, Hein HO, Suadicani P, Gyntelberg F. Triglyceride concentration and ischaemic heart disease: an eight-year follow-up in the Copenhagen Male Study. Circulation. 1998; 97: 1029–1036.
Burchfiel CM, Laws A, Benfante R, Goldberg RJ, Hwang LJ, Chin D, Rodriguez BL, Curb JD, Sharp DS. Combined effects of HDL cholesterol, triglyceride, and total cholesterol concentrations on 18-year risk of atherosclerotic disease. Circulation. 1995; 92: 1430–1436.
Manninen V, Tenkanen L, Koskinen P, Huttunen JK, Manttari M, Heinonen OP, Frick MH. Joint effects of serum triglyceride and LDL cholesterol and HDL cholesterol concentrations on coronary heart disease risk in the Helsinki Heart Study: implications for treatment. Circulation. 1992; 85: 37–45.
Jeppesen J, Hein HO, Suadicani P, Gyntelberg F. Relation of high TG-low HDL cholesterol and LDL cholesterol to the incidence of ischaemic heart disease: an 8-year follow-up in the Copenhagen Male Study. Arterioscler Thromb Vasc Biol. 1997; 17: 1114–1120.
Austin MA. Plasma triglycerides as a risk-factor for coronary heart disease: the epidemiological evidence and beyond. Am J Epidemiol. 1989; 129: 249–259.
Yarnell JWG, Patterson CC, Bainton D, Sweetnam PM. Is metabolic syndrome a discrete entity in the general population?: evidence from the Caerphilly and Speedwell population studies. Heart. 1998; 79: 248–252.
Buccolo G, David H. Quantitative determination of serum triglycerides by the use of enzymes. Clin Chem. 1973; 19: 476–482.
Lopes-Virella MF, Stone P, Ellis S, Colwell JA. Cholesterol determination in high-density lipoproteins separated by three different methods. Clin Chem. 1977; 23: 822–824.
Warnick GR, Albers JJ. A comprehensive evaluation of the heparin-manganese precipitation procedure for estimating high density lipoprotein cholesterol. J Lipid Res. 1978; 19: 65–76.
Yarnell JWG, Patterson CC, Thomas HF, Sweetnam PM. Fasting plasma glucose and subsequent macrovascular disease after 10 years follow-up: a collaborative study on two populations. QJM. 1999; 92: 207–210.
Chen ZM, Peto R, Collins R, MacMahon S, Lu JR, Li WX. Serum-cholesterol concentration and coronary heart disease in a population with low cholesterol concentrations. BMJ. 1991; 303: 276–282.