| | The independent correlation between high-density lipoprotein cholesterol and subsequent major adverse coronary eventsReceived 5 September 2005; accepted 6 December 2005. BackgroundThere is substantial evidence from clinical trials that lowering low-density lipoprotein cholesterol (LDL-C) reduces cardiovascular risk. There is less evidence for the salutatory effects of raising high-density lipoprotein cholesterol (HDL-C). The predictive strength of an initial HDL-C measurement and its change over time for major adverse coronary events is not well understood. MethodsWe identified a cohort of all 6928 patients in an urban primary care practice who had two or more lipid measurements between January 1985 and December 1997. We used bivariable and multivariable (Cox proportional hazards) techniques to identify independent predictors of subsequent major adverse coronary events (hospitalization for myocardial infarction or acute coronary syndrome) after the second set of lipid measurements. ResultsThe time between first and second lipid measurements averaged 2.6 years. During a mean of 5.1 ± 3.2 years of observation after their second lipid measurements, 2167 (31%) patients had an acute coronary event. Patients having events were significantly older, more often white, male, and smokers and more often had antecedent diabetes, hypertension, coronary heart disease, and myocardial infarctions. Adjusting for covariates, a 10-mg/dL higher initial HDL-C was associated with an 11% (95% CI 7%-14%) lower risk of coronary events. A 10-mg/dL increase in HDL-C between lipid measurements was associated with a 7% (95% CI 3%-10%) lower risk of events. Neither initial or change in triglycerides nor LDL-C predicted subsequent coronary events. ConclusionHigh-density lipoprotein cholesterol measurements and change in HDL-C predicted major adverse coronary events in this urban practice, which provides support studying interventions targeting HDL-C for cardiovascular risk reduction. Coronary heart disease (CHD) is the leading cause of death in the United States.1 Risk factors for CHD were first identified rigorously in the Framingham Study and include age, sex, family history, smoking, diabetes, hypertension, elevated total cholesterol, elevated low-density lipoprotein cholesterol (LDL-C), and reduced high-density lipoprotein cholesterol (HDL-C).2 Randomized controlled trials of bile acid sequestrants3, 4 and HMG-CoA reductase inhibitors (statins)5 have demonstrated conclusively that lowering LDL-C resulted in the primary and secondary prevention of CHD. The National Cholesterol Education Program Adult Treatment Panel III guidelines released in 2001 continued to emphasize lowering of LDL-C as a primary goal of cardiovascular risk reduction.6 However, the guidelines also stated that lowering non–HDL-C (total cholesterol minus HDL-C) should be a secondary goal in persons with elevated triglyceride levels (≥200 mg/dL), based on the assumption that all plasma cholesterol other than HDL-C is atherogenic. Yet, several large trials of drugs to lower triglycerides (the major component of non–HDL-C besides LDL-C) were either negative or raised doubts concerning these drugs' long-term safety.7, 8, 9 Epidemiologic studies have shown a correlation between low levels of HDL-C and risk of adverse coronary events.10 Separating the effects of raising HDL-C from lowering LDL-C, however, is difficult because drugs effective at lowering LDL-C can also increase HDL-C.11 Secondary analyses of the Department of Veterans Affairs' HDL-C Intervention Trial suggested clinical benefit from raising HDL-C levels and lowering triglycerides without any reduction in LDL-C levels in patients with established CHD who had low LDL-C and high triglyceride levels. A 6% increment in HDL-C was associated with a 22% reduction in the composite end point of nonfatal myocardial infarction (MI) and death from coronary artery disease (CAD) with gemfibrozil therapy.12 The AFCAPS/TexCAPS showed that for patients with average LDL-C and low HDL-C levels, baseline HDL-C is an independent risk factor for major adverse coronary events.13 There was an inverse relationship between the change in HDL-C and the subsequent risk of developing a primary end point that did not achieve statistical significance. Although these studies demonstrate that low HDL-C is an independent risk factor and raising it may lower the risk of subsequent major adverse coronary events, these were randomized controlled clinical trials performed in selected populations. In the current study, we used data from one of the largest, most enduring clinical data repositories to assess the independent effect in everyday clinical practice of HDL-C and its change over time on the incidence of major adverse coronary events among patients receiving care in a large urban academic primary care, general internal medicine practice. Methods  Data source and study population This study was approved by the Institutional Review Board of the Indiana University School of Medicine. The study cohort came from patients treated by the outpatient primary care and specialty services of Indiana University Medical Group-Primary Care (IUMG-PC).14 This is the largest primary care practice in Indiana and consists of 11 community health centers affiliated with Wishard Memorial Hospital, an inner-city public teaching hospital. IUMG-PC also provides care to commercially insured patients in 6 offices located in a 50-mile radius of Indianapolis. IUMG-PC was one of the first practice-based research networks funded by the Agency for Healthcare Research and Quality.15 More than 300 physicians (approximately 150 primary care physicians and 150 residents) care for >120 000 patients during >320000 primary care outpatient visits per year. All IUMG-PC and Wishard Hospital data are stored in the Regenstrief Medical Record System16 (RMRS) and include all laboratory data from every outpatient, inpatient, and emergency/urgent patient encounter. Importantly, patients seen at Wishard Hospital and its outpatient practices receive >90% of their care in this system, which is the sole designated provider for indigent patients in Indianapolis. IUMG-PC and Wishard Hospital are also part of the NIH-funded Indianapolis Network for Patient Care and Research, which collects emergency department and selected other electronic data for hospitalized patients from all central Indiana hospitals.17 We identified a retrospective (historical) cohort of subjects that included all IUMG-PC patients who had lipids measured between January 1, 1985 (the first date that HDL-C tests were available in the Wishard clinical laboratory), and December 31, 1997. To assess the effect of change in lipid values on coronary events, we limited our analyses to patients who had two or more lipid measurements. The date of the second set of measurements was considered the cohort inception date for each study patient and ranged from May 1986 through May 21, 2004, the date of initial data extraction from the RMRS. Study variables and analytic methods From each patient's RMRS record, we extracted demographic data (age at the inception date, sex, and race), important cardiovascular risk factors (systolic blood pressure and history of hypertension, diabetes, and smoking), and prior evidence of CHD (by inpatient and outpatient diagnoses of CHD or MI, segmental wall motion abnormalities on a prior echocardiogram, CHD read on a prior cardiac catheterization, or the performance of any prior cardiac revascularization procedure). In addition to CHD, we included a variable indicating patients who had a prior MI, defined as a hospital diagnosis, classic q wave infarction on an electrocardiogram, or elevation in serum concentrations of cardiac enzymes. The outcome variable was the first evidence of an acute coronary event occurring after a patient's inception date (ie, date of the second lipid measurement). An acute coronary event was defined as any evidence of hospitalization for an MI or the acute coronary syndrome (ICD-9-CM codes 411-414.9), an inpatient electrocardiogram or serum cardiac enzyme measurements indicating an acute MI, or a record of any coronary artery intervention (defined as aortocoronary bypass grafting or any percutaneous coronary intervention, ICD-9-CM codes V45.81 or V45.82). The primary independent (predictor) variables were the first recorded (ie, baseline) lipid values (HDL-C, LDL-C, triglyceride, and total cholesterol) and the change in each of these values between the first and second recorded measurements. Before 2000, there was no direct measurement of LDL-C available in the Wishard clinical laboratory. Therefore, the LDL-C was computed from fasting lipid measurements using the standard formula: total cholesterol − HDL-C − triglycerides / 5. Differences between patients who did and did not develop an acute coronary event were assessed using χ2 and t tests for categorical and continuous variables, respectively. We used proportional hazards (Cox) regression to identify independent predictors of incident major adverse coronary events. The dependent variable was the time from each subject's inception date (ie, the date of the second lipid measurements) to the first acute coronary event. We did not include total cholesterol in the analysis because it was highly correlated with LDL-C (r > 0.80). We also adjusted the analysis for history of CHD and MI, demographic data, and other cardiovascular risk factors. Subjects without events were censored as of the last recorded IUMG-PC visit or visit to any Wishard Hospital inpatient or outpatient site before the date of RMRS data extraction (May 21, 2004). All analyses were performed using SAS statistical software, version 8 (SAS Institute, Cary, NC). Results We identified 7506 patients who had at least two HDL-C measurements. Of these, 578 (7.7%) were excluded because of one or more missing lipid measurements: 293 (3.9%) did not have triglycerides and 395 (5.3%) did not have a total cholesterol result or had a triglyceride value of >400 mg/dL at both of the first two HDL-C measurements. Systolic blood pressure on or before the first lipid measurement was not available for 194 (2.6%). The final study cohort consisted of 6928 patients. Table I displays the characteristics of the overall analysis cohort. The mean age was 55 years, two thirds were women, and approximately half were white. One third of the cohort smoked, 38% had diabetes mellitus, and 59% were hypertensive. | ⁎ All differences between those with and without a subsequent acute coronary event are significant (P < .0001 for each comparison). |
During a mean of 5.1 ± 3.2 years of observation, 2167 patients (31% of the cohort) experienced an acute coronary event after their second lipid measurement. Table I compares patients with and without subsequent events. Patients who developed events were significantly older and more often white, male, smoked, and had diabetes and hypertension. Their mean systolic blood pressures were significantly higher. Histories of CHD and MI were much more prevalent among patients with subsequent events. The first and second lipid values are shown in Table II. The mean initial total cholesterol was 240 mg/dL, and the HDL-C was 46 mg/dL. Fifty-nine percent of the cohort had initial HDL-C values >40 mg/dL, the midrange of normal in this laboratory. The mean initial triglyceride value was 203 mg/dL and the mean initial LDL-C was 156 mg/dL. (LDL-C was imputed from total cholesterol, HDL, and triglycerides in 674 [10%] of subjects.) The average time between patients' first and second lipid values was 2.6 ± 2.6 years (median 1.6 years). For the entire cohort between the first and second lipid measurements, total cholesterol declined by a mean of 8.4 mg/dL, HDL-C increased by 1.7 mg/dL, LDL-C declined by 8.9 mg/dL, and triglycerides increased by 0.4 mg/dL. Patients with subsequent major adverse coronary events (after the second lipid measurement) had significantly higher values for total cholesterol at the first measurement and higher triglycerides at the first and second measurements. High-density lipoprotein cholesterol was significantly lower at the first and second measurements for those with a subsequent coronary event compared with those without an event. There was no significant difference in first or second LDL-C values between those with and without subsequent major adverse coronary events. In the univariate analysis, total cholesterol and LDL-C dropped significantly more among those with events, but there were no significant differences between groups in the change in HDL-C or triglycerides. | ⁎ The P value is for comparing patients with versus without subsequent major adverse coronary events. |
Multivariable proportional hazards analysis identified the following significant nonlipid risk factors for coronary events (in decreasing predictive power): history of CHD, age, prior MI, smoking, race, sex, and diabetes (Table III). The strongest and only significant (P < .001) predictor among the initial lipid measurements was HDL-C (hazard ratio [HR] 0.89, 95% CI 0.89-0.93); it was the third strongest predictor behind prior CHD and age. Every 10-mg/dL increase in the baseline HDL-C value was associated with an 11% decrease (95% CI 0.86-0.93) in the risk of an acute coronary event. Change in HDL-C between the first and second measurements was also a significant (P < .001) predictor of events, with an HR of 0.93 (95% CI 0.90%-0.97). A positive change of 10 mg/dL was associated with a 7% lower risk of subsequent major adverse coronary events. Neither the initial values of LDL-C or triglycerides nor their changes predicted events (HR of 0.99 and 1.0, respectively). To investigate the lack of prediction for LDL-C and triglycerides, we repeated the multivariable analysis after excluding the initial HDL-C and change in HDL-C, with similar results (ie, neither LDL-C nor triglycerides, nor their changes, were significant predictors). We also repeated the analysis to check for interactions between HDL-C and change in HDL-C and age, sex, race, and history of diabetes. None of these interactions was significant. Finally, because change in HDL-C was only significant multivariably and not univariably, we repeated the analysis while, one by one, eliminating the other variables in the final multivariable model. Only when the baseline HDL-C was removed from the model did the change in HDL-C lose its statistical significance (P = .5). | ⁎ All indicator variables were coded 1 if present and 0 if not present. †Hazard ratios are for a 10-unit change in the variable. |
We repeated the multivariable proportional hazards analysis in subgroups of patients with and without evidence of prior CAD. There was no significant difference in the effect of baseline HDL-C (HR 0.93, 95% CI 0.89-0.97, and HR 0.85, 95% CI 0.80-0.91, respectively) or change in HDL-C (HR 0.93, 95% CI 0.89-0.97 and HR 0.94, 95% CI 0.88-1.01, respectively). We performed an additional multivariable proportional hazards analysis after substituting baseline non–HDL-cholesterol (total cholesterol minus HDL-C) for LDL-C and triglycerides and change in non–HDL-cholesterol for change in LDL-C and change in triglycerides. The resulting model had similar results for baseline HDL-C (HR 0.90, 95% CI 0.87-0.93) and change in HDL-C (HR 0.93, 95% CI 0.90-0.97). Only 2.5% and 0.6% of patients, respectively, were taking a statin or a fibrate at the time of their first lipid measurement. Between the first and second measurements, 17% of patients were treated with a statin and 3% with a fibrate. Between the second lipid measurement and the last recorded visit, 49% were treated with a statin and 8% with a fibrate. We did not include indicators of treatment with these drugs in the multivariable analysis because of significant confounding by indication18 (ie, physicians treating the highest-risk patients). However, we included change in lipid values in the model to account for the direct effects of these drugs on lipids. Yet because there may be indirect effects of these classes of drugs on vascular outcomes, we repeated the multivariable analysis on subgroups of patients taking a statin or fibrate (HR for change in HDL-C 0.95, 95% CI 0.89-1.02) and those not taking a statin or fibrate (HR for change in HDL-C 0.94, 95% CI 0.90-0.98) on or before the second set of lipid measurements. For both models, the resulting model was practically identical to the full analysis, except that due to the reduced number of subjects on one of these two classes of drugs, the change in HDL-C was no longer statistically significant (P = .15). Discussion In this cohort of urban primary care patients, considering demographic information and other known cardiovascular risk factors, patients with higher initial HDL-C concentrations and those whose HDL-C increased from the first to second lipid measurements had significantly lower risk of incurring subsequent major adverse coronary events. The other components of IUMG-PC's standard lipid profile (LDL-C and triglycerides) did not provide important prognostic information. The change in HDL-C only became significant when the initial HDL-C was included in the analysis, indicating an interaction between these two terms. For example, a patient with a lower initial HDL-C and a large increase in HDL-C may have had a significant reduction in risk compared with a patient with a low initial HDL-C and no change over time. The level of risk reduction was nontrivial: every 1 mg/dL higher baseline HDL-C was associated with about 1% lower in cardiac risk. Comparably, for every 1-mg/dL increase in HDL-C change over the average 2.6-year measurement interval, cardiac risk was 0.7% lower. These results support a focused treatment on HDL-C in lipid management when needed and are consistent with the recent position paper by the European Consensus Panel on HDL-C.19 We were surprised that, contrary to prior research,2, 3, 4, 5, 6, 20 LDL-C and its change were not associated with subsequent cardiovascular risk. We can only speculate on the reasons for this finding. Because only 17% of the subjects were taking statins on or before their second lipid measurements, initiating treatment with statins after the second lipid measurements could have limited the negative impact of any LDL-C change or baseline values. Analyzing subsets of patients taking and not taking statins before the second lipid measurement did not influence the results. Our observational study was not a clinical trial where patients were randomized to receive a lipid-lowering drug or placebo. Patients were prescribed drugs at the discretion of the physician and adherence was up to the patient. Yet unlike prior controlled trials that focused on white patients at academic medical centers who had few comorbid conditions, our cohort was more representative of what physicians see in everyday practice in the inner city who have a higher prevalence of minority race and comorbid conditions (Table I). All these factors may have contributed, at least in part, to the deviation of our LDC-C findings from those in the lipid trials. The study has several limitations. One is selection bias. Our results may not apply to the general population. Clinical data repositories, such as the one used in our study, contain information collected from routine medical care. Patient data in clinical data repositories18 do not reflect data collected from population-based studies.2 Individuals in clinical databases tend to be sicker compared with population-based samples.18 Unlike studies such as the Framingham Heart Study,2 lipid values were not measured routinely in a general population of people. Rather, they were measured at the discretion of their primary care physicians. Also, fewer than half of the adult patients had lipid measurements despite computer reminders to do so.21 It is therefore possible that those patients whose physicians complied with the reminder had higher cardiovascular risk than those whose physicians did not comply. Moreover, only adults who had their lipids monitored at least twice were eligible for our study. All these factors may have biased the study cohort. Clinical practice has changed over the study period, including changes in the criteria for diagnosis of MI. There is no reason to believe that the diagnosis of MI is systematically different for different levels of change in HDL-C, and therefore, the nondifferential misclassification is not likely to bias our results but may have diluted our estimates toward observing no effect of change in HDL-C on CAD. In addition, differences in clinicians' responses to elevated lipids over time would have resulted in more variability in the predictive value of baseline lipid measurements and hence less predictive power. We therefore believe that our estimates of the predictive power of HDL and change in HDL-C are likely the lower bound of their actual predictive power in an environment where clinician response is more stable and predictable. We did not have information on family history of CHD, and information on height (and thus body mass index) was missing on half of the patients. Therefore, we could not control for these variables as confounding cardiovascular risk factors in the multivariate models. Although we used data from a citywide network, it is likely that we did not capture all major adverse coronary events, including events for which patients did not seek medical attention or that resulted in sudden death. This may have biased our cohort toward patients with lower socioeconomic status who more often receive care from Wishard Hospital. In this urban primary care, general internal medicine practice, initial HDL-C measurements and change in HDL-C predicted major adverse coronary events. High-density lipoprotein cholesterol was the only lipid fraction that affected cardiovascular risk. Our data suggest that physicians practicing in similar urban settings and with a similar patient mix should pay more attention to HDL-C as a predictor of cardiovascular disease and possibly as a target for lipid-lowering interventions. The actual benefit of such an approach generally will require further clinical trial study. Our results also suggest that antilipemic drugs and nonpharmaceutical interventions specially targeting HDL deserve attention. References 1. 1Mokdad AH, Marks JS, Stroup DF, et al. Actual causes of death in the United States, 2000. JAMA. 2004;291:1238–1245.
CrossRef
2. 2D'Agostino RB, Grundy S, Sullivan LS, et al.for the CHD Risk Prediction Group Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA. 2001;286:180–197. MEDLINE |
CrossRef
3. 3Lipid Research Clinics Program . The Lipid Research Clinics Coronary Primary Prevention Trial Results: I, reduction in the incidence of coronary heart disease. JAMA. 1984;251:351–374. MEDLINE 4. 4Lipid Research Clinics Program . The Lipid Research Clinics Coronary Primary Prevention Trial Results: II, the relationship of reduction in incidence of coronary heart disease to cholesterol lowering. JAMA. 1984;251:365–374. MEDLINE 5. 5Rossouw JE, Lewis B, Rifkind BM. The value of lowering cholesterol after myocardial infarction. N Engl J Med. 1990;323:1112–1119. MEDLINE |
CrossRef
6. 6Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults . Executive summary of the third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel III). JAMA. 2001;285:2508–2509. MEDLINE |
CrossRef
7. 7Secondary prevention by raising HDL cholesterol and reducing triglycerides in patients with coronary artery disease: the Bezafibrate Infarction Prevention (BIP) study. Circulation. 2000;102:21–27. 8. 8WHO cooperative trial on primary prevention of ischaemic heart disease with clofibrate to lower serum cholesterol . Final mortality follow-up. Report of the Committee of Principal Investigators. Lancet. 1984;2:600–604. MEDLINE 9. 9Frick MH, Elo O, Haapa K, et al. Helsinki Heart Study: primary prevention trial with gemfibrozil in middle-aged men with dyslipidemia: safety of treatment, changes in risk factors, and incidence of coronary heart disease. N Engl J Med. 1987;317:1237–1245. MEDLINE |
CrossRef
10. 10Gordon DL, Rifkind BM. High-density lipoprotein—the clinical implications of recent studies. N Engl J Med. 1989;321:1311–1316. MEDLINE |
CrossRef
11. 11Gotto AM. High-density lipoprotein cholesterol and triglycerides as therapeutic targets for preventing and treating coronary artery disease. Am Heart J. 2002;144(Suppl):33–42. 12. 12Boden WE. High-density lipoprotein cholesterol as an independent risk factor in cardiovascular disease: assessing the data from Framingham to the Veterans Affairs High-Density Lipoprotein Intervention Trial. Am J Cardiol. 2000;86(Suppl):19L–22L. MEDLINE 13. 13Gotto AM, Whitney E, Stein EA, et al. Relation between baseline and on-treatment lipid parameters and first acute major coronary vents in the Air Force/Texas Coronary Atherosclerosis Prevention Study. Circulation. 2000;101:477–484. 14. 14Tierney WM, Miller ME, Hui SL, et al. Practice randomization and clinical research: the Indiana experience. Med Care. 1991;29:JS57–JS64. MEDLINE 15. 15Agency for Healthcare Research and Quality . Fact sheet: primary care practice–based research networks. http://www.ahrq.gov/research/pbrnfact.htm. 16. 16McDonald CJ, Overhage JM, Tierney WM, et al. The Regenstrief Medical Record System: a quarter century experience. Int J Med Inf. 1999;54:225–253. 17. 17Overhage JM, McDonald CJ, Tierney WM. Design and implementation of the Indianapolis Network for Patient Care and Research. MLA Bull. 1995;83:48–56. 18. 18Tierney WM, McDonald CJ. Practice databases and their uses in clinical research. Stat Med. 1991;10:541–557. MEDLINE |
CrossRef
19. 19Chapman MJ, Assmann G, Fruchart JC, et al. European Consensus Panel on HDL-C. Raising high-density lipoprotein cholesterol with reduction of cardiovascular risk: the role of nicotinic acid—a position paper developed by the European Consensus Panel on HDL-C. Curr Med Res Opin. 2004;20:1253–1268. MEDLINE |
CrossRef
20. 20Robins SJ, Collins D, Wittes JT, et al. Relation of gemfibrozil treatment and lipid levels with major coronary events. VA-HIT: a randomized controlled trial. JAMA. 2001;285:1585–1591. MEDLINE |
CrossRef
21. 21McDonald CJ, Hui SL, Smith DM, et al. Reminders to physicians from an introspective computer medical record. A two-year randomized trial.. Ann Intern Med. 1984;100:130–138. MEDLINE a GlaxoSmithKline, Inc., Collegeville, PA b Regenstrief Institute, Inc., Indianapolis, IN c Indiana University School of Medicine, Indianapolis, IN  Reprint requests: William M. Tierney, MD, Room M200-OPW, Wishard Memorial Hospital, 1001 West Tenth Street, Indianapolis, IN 46202.
The authors are solely responsible for the content of this article that represents their opinions and not necessarily their respective institutions. This study was funded by a research grant from GlaxoSmithKline to Dr Tierney. PII: S0002-8703(05)01067-7 doi:10.1016/j.ahj.2005.12.007 © 2006 Mosby, Inc. All rights reserved. | |
|
FULL TEXT