Atherosclerosis 154 (2001) 221 – 227
Plasma lipids and cardiovascular risk: a POSCH report
Henry Buchwald
a,*, James R. Boen
b, Phuong A. Nguyen
a, Stanley E. Williams
a,
John P. Matts
baDepartment of Surgery,Uni6ersity of Minnesota,Box290Mayo,420Delaware St SE,Minneapolis,MN55455,USA bDi6ision of Biostatistics in the School of Public Health,Uni6ersity of Minnesota,Minneapolis,MN,USA
Received 20 September 1999; received in revised form 14 February 2000; accepted 2 March 2000
Abstract
Quantifying the relationship between changes in lipid variables and clinical endpoints has been difficult. We studied the predictive value of various lipid variables on three endpoints in the Program on the Surgical Control of the Hyperlipidemias (POSCH): overall mortality, coronary heart disease (CHD) mortality, and CHD mortality and confirmed nonfatal myocardial infarction (MI) combined. We measured lipid variables for the annual visits from baseline to 5 years for actual follow-up values, actual and percentage differences between baseline and follow-up values, as well as the parameters comparing baseline only to 5 years for actual differences, percentage differences, and the ratio of baseline to 5 years. The lipid variables included were total cholesterol, low density lipoprotein (LDL) cholesterol, high density lipoprotein (HDL) cholesterol, very low density lipoprotein (VLDL) cholesterol, triglycerides, and the LDL cholesterol/HDL cholesterol ratio. The analytic method used was that of Cox regression, with age and sex as secondary covariates, and each lipid or ratio of lipids as the primary (univariate) covariate. As a result, 108 univariate Cox regressions were conducted. The combined findings for the control and the intervention groups are presented. The number of events for the combined group were: overall mortality, 190; CHD mortality, 119; and CHD mortality and confirmed nonfatal MI, 262. The highest hazard ratios were found for the lipid variable of the LDL cholesterol/HDL cholesterol ratio (e.g. 1.196 for a 1-unit increase). Only for the combined endpoint of CHD mortality and confirmed nonfatal MI was there a substantial number of statistically significant relationships (PB0.01) of lipid variables and parameters of assessment. © 2001 Elsevier Science Ireland Ltd. All rights reserved.
Keywords:Lipids; Lipoproteins; Cardiovascular diseases; Risk factors
www.elsevier.com/locate/atherosclerosis
1. Introduction
The predictive value of plasma lipids for cardiovascu-lar disease, particucardiovascu-larly coronary heart disease (CHD), has long been established [1 – 4]. The 1990 and 1998 reports of the Program on the Surgical Control of the Hyperlipidemias (POSCH), demonstrated that decreas-ing the total plasma cholesterol and the low density lipoprotein (LDL) cholesterol fraction statistically sig-nificantly decreased multiple cardiovascular endpoints, including by 1995 the trial’s primary endpoint of over-all mortality [5,6]. The statin drug trials, which used an
entirely different mechanism for lipid modification, confirmed the POSCH results for both secondary [7 – 13] and primary [14] intervention.
In spite of these conclusive studies confirming the lipid/atherosclerosis theory, quantifying the relationship between changes in various lipid fractions and clinical endpoint hazard ratios has been difficult. The baseline serum cholesterol and treatment effect [15], and lipo-protein changes and reductions in major CHD events [16], have been analyzed by the Scandinavian Simvas-tatin Survival Study (4S) investigators. Similar analyses were performed in the Cholesterol and Recurrent Events (CARE) study [17,18], and in the major primary intervention trial, the West of Scotland Coronary Pre-vention Study (WOSCOPS) [19]. The implications of these findings have been discussed by Grundy [20]. * Corresponding author. Tel.: +1-612-6258446; fax: +
1-612-6253206.
E-mail address:buchw001@tc.umn.edu (H. Buchwald).
H.Buchwald et al./Atherosclerosis154 (2001) 221 – 227 222
2. Methods
2.1. POSCH o6er6iew
POSCH was a multiclinic, randomized, prospective, secondary intervention trial, utilizing the partial ileal bypass operation as the intervention modality. Between 1975 and 1983, 838 survivors of a single myocardial infarction (MI), documented by electrocardiograms and changes in cardiac enzyme values, were entered into the study at four widely geographically spaced institutions: 417 patients were treated with diet instruction only and 421 patients with diet instruction plus partial ileal by-pass surgery. Lipid values were determined at baseline, and annually for 5 years after randomization [21,22]. All analyses were based on randomization assignment (intention-to-treat). Throughout the study, no patient was lost to follow-up. The formal POSCH trial ended July 20, 1990, with a mean patient follow-up of 9.7 years and a highly statistically significant difference in CHD mortality and confirmed nonfatal MI (PB0.001) [5]. Five-year post-trial follow-up ended September 30, 1995 with the differences in all of the endpoints statisti-cally significant, including overall mortality (P=0.049) and CHD mortality (P=0.03) [6].
2.2. E6ents and lipid analyses
We analyzed the relationship between changes in plasma lipids for the three POSCH endpoints of: over-all mortality, CHD mortality, and CHD mortality and confirmed nonfatal MI. The findings for the control and intervention groups combined (n=838) are pre-sented. The number of events for the combined group were: overall mortality, 190; CHD mortality, 119; and CHD mortality and confirmed nonfatal MI, 262.
We evaluated lipid variables for all visits (baseline and annually after randomization to 5 years) for actual follow-up values, actual differences between baseline and follow-up values, and percentage differences be-tween baseline and follow-up values, as well as for baseline compared only to 5 years for actual differ-ences, percentage differdiffer-ences, and the ratio of baseline to 5 years. The lipid variables included were total cholesterol, LDL cholesterol, high dnesity lipoprotein (HDL) cholesterol, very low density lipoprotein (VLDL) cholesterol, triglycerides, and the LDL choles-terol/HDL cholesterol ratio.
2.3. Statistical analysis
The relationships between endpoints and lipid values were analyzed using Cox regression, with the single lipid or ratio of lipids as the primary (univariate) time-dependent covariates (proportional hazards model) [23]. All analyses contained baseline (secondary)
covariates for age and gender. Possible combinations numbered 106. The hazard ratios represent the increase or decrease in relative risk for a 1-unit increase in the lipid variable. To obtain the hazard ratio for a 2-unit increase the hazard ratio must be taken to the 2nd power, for a 10-unit increase, to the 10th power, and so on. For example, 1.1 to the 10th power yields a 2.59 relative risk for a 10-unit increase. The P-values are 2-sided. Because of the multiple assessments performed, a P-value of 50.01 is considered statistically significant.
3. Results
To facilitate comparisons, the data analyses are sum-marized as a single table (Table 1). The table can be read vertically to assess the relationship between the lipid variables and each of the six parameters of analy-sis for each of the three endpoints; or it can be read horizontally for these relationships by specific endpoint. Of the 108 possible combinations, 20 are statistically significant with a P50.01.
3.1. Lipid 6ariables
Total cholesterol is statistically significantly related only to the endpoint of CHD mortality and confirmed nonfatal MI for actual values, actual differences, per-centage differences for all visits, and perper-centage differ-ences between baseline and 5 years. LDL cholesterol is statistically significantly related only to CHD mortality for actual values and to CHD mortality and confirmed nonfatal MI for actual values, actual differences, and percentage differences for all visits, and actual differ-ences and percentage differdiffer-ences between baseline and 5 years. HDL cholesterol is not statistically significantly related to any of the parameters of analysis. VLDL cholesterol is statistically significantly related only to the endpoint of overall mortality for actual values all visits; the same is true for triglycerides. The LDL cholesterol/HDL cholesterol ratio has the most statisti-cally significant relationships of the lipid variables. It is statistically significantly related to CHD mortality for actual values and percentage differences for all visits, and percentage differences between baseline and 5 years, as well as to CHD mortality and confirmed nonfatal MI for actual values, actual differences, and percentage differences for all visits, and for actual differences and percentage differences between baseline and 5 years.
3.2. O6erall mortality
H
.
Buchwald
et
al
.
/
Atherosclerosis
154
(2001)
221
–
227
223
Table 1
POSCH data analysis
Total cholesterol LDL cholesterol HDL cholesterol VLDL cholesterol Triglycerides LDL Chol/HDL Chol
P-value Hazard ratio P-value Hazard ratio P-value Hazard ratio P-value Hazard ratio P-value Hazard ratio P-value Hazard ratio
O6erall mortality,control and surgery groups combined
All6isits
0.6102 1.001 0.0861 1.003 0.4116 1.006
Actual values 0.0143 0.991 0.0082 0.998 0.1492 1.070
1.033
0.8596 1.000 0.9209 1.000 0.6047 0.996 0.4730 0.998 0.5482
Actual 1.000 0.4760
differences
1.002
0.7164 0.998 0.8806 1.000 0.6291
Percentage 0.998 0.0966 0.998 0.2447 0.998 0.3666
differences
Baseline to5
years
1.025
0.6302 1.001 0.7341 1.001 0.9043 1.001
Actual 0.7432 0.999 0.6869 1.000 0.6241
differences
0.7090 1.002 0.5355 1.002 0.8267 1.001 0.2121 0.998 0.2024 0.998 0.4024 1.002
Percentage differences
0.982
0.9291 0.973 0.6084
Ratio of 0.924 0.7313 1.126 0.2331 1.035 0.4042 1.115 0.8803
baseline to 5 years
CHD mortality,control and surgery groups combined All6isits
0.0951 1.003 0.0108 1.005 0.7601 0.997
Actual values 0.1253 0.993 0.0715 0.998 0.0044 1.175
Actual 0.3411 1.002 0.2312 1.002 0.1697 0.985 0.6285 0.998 0.6792 1.000 0.0268 1.140
differences
1.006
0.3759 1.005 0.1355 1.005 0.2188 0.995
Percentage 0.4261 0.999 0.5134 0.999 0.0123
differences
Baseline to5
years
1.164
0.0919 1.004 0.1097 1.004 0.3859
Actual 0.990 0.8185 1.001 0.9546 1.000 0.0257
differences
1.007
0.0937 1.011 0.0619 1.007 0.3831 0.996
Percentage 0.7667 1.000 0.6526 0.999 0.0130
differences
0.1885 0.794
0.1115 0.505 0.0914 0.684
Ratio of 0.1676 1.823 0.7606 0.973 0.9113 0.977
baseline to 5 years
CHD mortality and confirmed nonfatal myocardial infarction,control and surgery groups combined All6isits
0.0001 1.005 0.0001 1.006
Actual values 0.0658 0.987 0.8567 1.000 0.5246 1.000 0.0001 1.196
1.161
0.0048 1.004 0.0013 1.005 0.0504
Actual 0.985 0.8921 1.000 0.4496 1.000 0.0003
differences
Percentage 0.0030 1.011 0.0026 1.007 0.0872 0.995 0.7221 1.000 0.1648 0.998 0.0002 1.006
H
.
Buchwald
et
al
.
/
Atherosclerosis
154
(2001)
221
–
227
224
Table 1 (Continued)
VLDL cholesterol Triglycerides LDL Chol/HDL Chol
Total cholesterol LDL cholesterol HDL cholesterol
Hazard ratio P-value Hazard ratio P-value Hazard ratio P-value
P-value Hazard ratio
P-value Hazard ratio P-value Hazard ratio
Baseline to5
years
0.0475 0.983 0.9329 1.000 0.7201 1.000
0.0043 0.0006
1.004 1.005 1.192
Actual 0.0157
differences
0.0558 0.993 0.9129 1.000 0.5777 0.999 0.0003
Percentage 0.0090 1.012 0.0054 1.008 1.008
differences
0.704 0.0486 1.941 0.7286 0.973 0.3800 1.111
0.452 0.0393 0.0269 0.740
Ratio of 0.0156
H.Buchwald et al./Atherosclerosis154 (2001) 221 – 227 225
to overall mortality, with an increase in these values associated with a lower risk.
3.3. CHD mortality
The LDL cholesterol for all visits actual values, and the LDL cholesterol/HDL cholesterol ratios for all visits actual values, all visits percentage differences, and baseline to 5 years percentage differences, are statisti-cally significantly related to CHD mortality, with an increase in these values associated with a higher risk.
3.4. CHD mortality and confirmed nonfatal MI
Essentially all of the parameters for all visits and for baseline to 5 years (14 of 18) for the lipid variables of total cholesterol, LDL cholesterol, and the LDL choles-terol/HDL cholesterol ratio, are statistically signifi-cantly related to CHD mortality and confirmed nonfatal MI, with an increase in these values associated with a higher risk.
4. Discussion
In the 4S [15], CARE [17], and WOSCOPS [19] studies, baseline lipid values are correlated with cardio-vascular risk: total plasma cholesterol, LDL choles-terol, and triglyceride levels are positively correlated and HDL cholesterol level is inversely correlated. These expected trends are in concordance with prospective epidemiologic data [1 – 4]. According to quartile or quintile analyses within the total plasma cholesterol, LDL cholesterol, and HDL cholesterol ranges reported in the 4S [15] and WOSCOPS [19] studies, simvastatin and pravastatin each provided relatively uniform benefits of therapy. In the CARE study’s lowest choles-terol quintile (B125 mg/dl), which is lower than the lowest LDL cholesterol values in the 4S and WO-SCOPS studies, no beneficial trend was observed with pravastatin therapy [17]. Analysis of therapy in the 4S study by the Cox proportional hazards model (for lipid levels at baseline and at 1 year, and for percentage change between baseline and 1 year) shows a highly correlated reduction in major coronary events with changes in total and LDL cholesterol, less so with HDL cholesterol, and no clear relationship with triglycerides [16]. In the CARE study, similar analyses show that the coronary event rate is significantly correlated with the LDL cholesterol level achieved (P=0.007) but not with the extent of the LDL cholesterol reduction [18]. Also, a threshold level of effect was found at a LDL choles-terol level of 125 mg/dl. Triglycerides but not HDL cholesterol levels are weakly but statistically signifi-cantly associated with the coronary event rate. In con-trast, in the WOSCOPS study, again using Cox
regression analyses, the percentage decrease in LDL cholesterol in the pravastatin-treated group does not correlate with CHD risk reduction and a decrease in the range of 24% was sufficient to produce the full benefits of therapy, with no further incremental benefits demonstrated for LDL cholesterol reductions up to 39% [19].
Grundy points out that the 4S data best fit a curvilin-ear model for the effects of reducing plasma cholesterol levels on relative risk for CHD; the CARE data, a threshold model; and the WOSCOPS data, either a curvilinear or threshold relationship [20]. What remains unclear from the available evidence is the impact on clinical practice. We do not know if there is a level below which further lipid reductions, primarily LDL cholesterol reductions, confer no additional clinical benefits (threshold model), or if there is a continuum in the derivable risk reduction as the LDL cholesterol level is lowered (curvilinear or even linear models). Grundy appreciates the impossibility of deducing from the statin trials a simple predictive relationship between plasma lipid changes and reductions in CHD risks; nevertheless, he uses these data to substantiate his National Institutes of Health [24] and American Heart Association [25] panels’ recommendations on detection, evaluation, and treatment of lipid levels.
Do the POSCH analyses make these relationships any clearer? Do they further substantiate the difficulty of deriving any formula for changes in lipid variables that can predict changes in clinical endpoints? Our analyses of lipid variables for all visits to 5 years (actual values, actual differences, percentage differences), and between baseline and 5 years only (actual differences, percentage differences, ratio of baseline to 5 years), demonstrates certain statistically significant (PB0.01) correlations. A firm formula, however, for predicting clinical events based on lipid changes did not emerge. The 4S, CARE, and WOSCOPS analyses used Cox regressions quite similar to our statistical analyses and they had equally large numbers of assessments. Yet, they chose to define statistical significance as P50.05. If we use P50.05 for statistical significance for our data, we obtain 29, instead of 20, statistically significant relationships out of a possible 108, with no appreciable change in the distribution or relevance of affirmative findings.
H.Buchwald et al./Atherosclerosis154 (2001) 221 – 227 226
In conclusion, the lipid/atherosclerosis theory has been proven: both causatively, by epidemiologic studies [1 – 4], and interventionally, by POSCH [5,6] and the statin trials [7 – 14]. A simple relationship between changes in certain lipid variables and patient prognosis, however, remains elusive.
Acknowledgements
This study was supported by grants RO1-HL-15265 and RO1-HL-49522 from the National Heart, Lung, and Blood Institute, Bethesda, Maryland and a grant from the Office of the Dean and the Department of Surgery, University of Minnesota Medical School, Min-neapolis, Minnesota.
References
[1] Kannal WB, Castelli WP, Gordon T, McNamara PM. Serum cholesterol, lipoproteins, and the risk of coronary heart disease: the Framingham Study. Ann Intern Med 1971;74:11 – 2. [2] Pooling Project Research Group. Relationship of blood pressure,
serum cholesterol, smoking habit, relative weight and ECG abnormalities to incidence of major coronary events: final report of the Pooling Project Research Group. J Chronic Dis 1978;31:201 – 306.
[3] Goldbourt V, Holtzman E, Neufeld NH. Total and high density lipoprotein cholesterol in the serum and risk of mortality: evi-dence of a threshold effect. Br Med J 1985;290:1239 – 43. [4] Stamler J, Wentworth D, Neaton JD. Is the relationship between
serum cholesterol and risk of premature death from coronary heart disease continuous and graded? Findings in 356,222 pri-mary screenees of the Multiple Risk Factor Intervention Trial (MRFIT). J Am Med Assoc 1986;256:2823 – 8.
[5] POSCH Group, Buchwald H, Varco RL, Matts JP, Long JM, Fitch LL, Campbell GS, Pearce MB, Yellin AE, Edmiston WA, Smink RD, Jr, Sawin HS, Jr, Campos CT, Hansen BJ, Tuna N, Karnegis JN, Sanmarco ME, Amplatz K, Castaneda-Zuniga WR, Hunter DW, Bissett JK, Weber FJ, Stevenson JW, Leon AS, Chalmers TC. Effect of partial ileal bypass surgery on mortality and morbidity from coronary heart disease in patients with hypercholesterolemia. Report of the Program on the Surgi-cal Control of the Hyperlipidemias (POSCH). New Engl J Med 1990;323:946 – 55.
[6] POSCH Group, Buchwald H, Campos CT, Varco RL, Boen JR, Williams SE, Hansen BJ, Campbell GS, Pearce MB, Yellin AE, Edmiston WA, Smink RD, Jr, Sawin HS, Jr. Effective lipid modification by partial ileal bypass reduced long-term coronary heart disease mortality and morbidity: five-year posttrial follow-up report from the POSCH. Arch Intern Med 1998;158:1253 – 61. [7] Scandinavian Simvastatin Survival Study Group. Randomized trial of cholesterol lowering in 4,444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet 1994;344:1383 – 9.
[8] MAAS Investigators. Effect of simvastatin on coronary atheroma: a multicentre antiatheroma study. Lancet 1994;344:633 – 8.
[9] Waters D, Higginson L, Gladstone P, Kimball B, Le May M, Boccuzzi SJ, Lesperance J. Effects of monotherapy with an HMG-CoA reductase inhibitor on the progression of coronary atherosclerosis as assessed by serial quantitative arteriography:
the Canadian Coronary Atherosclerosis Intervention Trial. Cir-culation 1994;89:959 – 68.
[10] Furberg CD, Adams HP, Jr, Applegate WB, Byington RP, Espeland MA, Hartwell T, Hunninghake DB, Lefkowitz DS, Probstfield J, Riley WA, Young B. Effect of lovastatin on early carotid atherosclerosis and cardiovascular events. Circulation 1994;90:1679 – 87.
[11] REGRESS Study Group, Jukema JW, Bruschke AVG, van Boven AJ, Reiber JH, Bal ET, Zwinderman AH, Jansen H, Boerma GJ, van Rappard FM, Lie KI. Effects of lipid lowering by pravastatin on progression and regression of coronary artery disease in symptomatic men with normal to moderately elevated serum cholesterol levels: the Regression Growth Evaluation Statin Study (REGRESS). Circulation 1995;91:2528 – 40. [12] Cholesterol and Recurrent Events Trial Investigators, Sacks FM,
Pfeffer MA, Moye LA, Rouleau JL, Rutherford JD, Cole TG, Brown L, Warnica JW, Arnold JMO, Wun CC, Davis BR, Braunwald E. The effect of pravastatin on coronary events after MI in patients with average cholesterol levels. New Engl J Med 1996;335:1001 – 9.
[13] The Long-Term Intervention with Pravastatin in Ischaemic Dis-ease (LIPID) Study Group. Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels. New Engl J Med 1998;339:1349 – 57.
[14] West of Scotland Coronary Prevention Study Group, Sheperd J, Cobbe SM, Ford I, Isles CG, Lorimer AR, MacFarlane PW, McKillop JH, Packard CJ. Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia. New Engl J Med 1995;333:1301 – 7.
[15] Scandinavian Simvastatin Survival Study Group. Baseline serum cholesterol and treatment effect in the Scandinavian Simvastatin Survival Study (4S). Lancet 1995;345:1274 – 5.
[16] Scandinavian Sismvastatin Survival Study Group, Pedersen TR, Olsson AG, Faergeman O, Færgeman O, Kjekshus J, Wedel H, Berg K, Wilhelmsen L, Haghfelt T, Thorgeirsson G, Pyo¨ra¨la¨ K, Miettinen T, Christophersen B, Tobert JA, Musliner TA, Cook TJ. Lipoprotein changes and reduction in the incidence of major coronary heart disease events in the Scandinavian Simvastatin Survival Study (4S). Circulation 1998;97:1453 – 60.
[17] Pfeffer MA, Sacks FM, Moye LA, East C, Goldman S, Nash DT, Rouleau JR, Rouleau JL, Sussex BA, Theroux P, Vanden Belt RJ, Braunwald E. Influence of baseline lipids on effective-ness of pravastatin in the CARE Trial. J Am Coll Cardiol 1999;33:125 – 30.
[18] Sacks FM, Moye LA, Davis BR, Cole TG, Rouleau JL, Nash DT, Pfeffer MA, Braunwald E. Relationship between plasma LDL concentrations during treatment with pravastatin and re-current coronary events in the Cholesterol and Rere-current Events trial. Circulation 1998;97:1446 – 52.
[19] West of Scotland Coronary Prevention Study Group. Influence of pravastatin and plasma lipids on clinical events in the West of Scotland Coronary Prevention Study (WOSCOPS). Circulation 1998;97:1440 – 5.
[20] Grundy SM. Statin trials and goals of cholesterol-lowering ther-apy. Circulation 1998;97:1436 – 9.
[21] POSCH Group, Campos CT, Matts JP, Fitch LL, Speech JC, Long JM, Buchwald HB. Lipoprotein modification achieved by partial ileal bypass: five-year results of the POSCH trial. Surgery 1987;102:424 – 32.
[22] Campos CT, Matts JP, Santilli SM, Fitch LL, Long JM, Speech JC, Buchwald H. Predictors of total and LDL cholesterol change following partial ileal bypass. Am J Surg 1988;155:138 – 46. [23] Cox DR. Regression models and life-tables. J R Stat Soc (B)
1972;34:1887.
H.Buchwald et al./Atherosclerosis154 (2001) 221 – 227 227
program: second report of the Expert Panel on Detection, Eval-uation, and Treatment of High Blood Cholesterol (Adult Treat-ment Panel II). Circulation 1994;89:1329 – 45.
[25] Grundy SM, Balady GJ, Criqui MH, Fletcher G, Greenland P, Hiratzka LF, Houston-Miller N, Kris-Etherton P, Krumholz
HM, LaRosa J, Ockene IS, Pearson TA, Reed J, Smith SC, Jr, Washington R. When to start cholesterol-lowering therapy in patients with coronary heart disease: a statement for healthcare professionals from the American Heart Association Task Force on Risk Reduction. Circulation 1997;95:1683 – 5.
