Kidney International, Vol. 63, Supplement 84 (2003), pp. S117–S120

Non–high-density lipoprotein cholesterol (non-HDL-C) as apredictor of cardiovascular mortality in patients withend-stage renal disease

YOSHIKI NISHIZAWA, TETSUO SHOJI, RYUSUKE KAKIYA, YOSHIHIRO TSUJIMOTO,TSUTOMU TABATA, EIJI ISHIMURA, TATSUYA NAKATANI, TAKAMI MIKI, and MASAAKI INABA

Department of Metabolism, Endocrinology, and Molecular Medicine, Department of Nephrology, Department of Urology,Osaka City University Graduate School of Medicine, and Department of Geriatrics and Neurology, Osaka City UniversityMedical School, Osaka, Japan; and Division of Internal Medicine, Inoue Hospital, Suita, Japan

Non–high-density lipoprotein cholesterol (non-HDL-C) as a terol (HDL-C) and increased plasma triglycerides due topredictor of cardiovascular mortality in patients with end-stage accumulation of triglyceride-rich lipoproteins includingrenal disease. very-low-density (VLDL) and intermediate-density lipo-Background. Patients with end-stage renal disease (ESRD)

proteins (IDL). Official recommendations in the generaloften show lipid abnormalities that may promote atherosclero-population [2, 3] employ LDL-C in fasting blood as thesis. Although the standard lipid marker is low-density lipopro-

tein cholesterol (LDL-C) in official recommendations, the need gold standard lipid parameter in lipid-related risk assess-of fasting blood sampling has prevented routine screening for ment. However, there are some problems in the use ofplasma lipids in hemodialysis patients. fasting LDL-C among ESRD patients treated by hemodial-Methods. We therefore evaluated the power of non–high-

ysis. First, the use of only LDL-C ignores the atherogenicdensity lipoprotein cholesterol (non–HDL-C) in predialysispotentials of triglyceride-rich lipoproteins. The degree of(non-fasting) serum as a predictor of cardiovascular mortality

in a cohort of 525 hemodialysis patients. association of IDL and VLDL with aortic sclerosis wasResults. During the mean follow-up of 64 months, 120 deaths, greater than that of LDL in hemodialysis patients [4].

including 44 fatal cardiovascular events, occurred. Patients in Second, it is difficult for many patients to obtain bloodthe highest tertile of non–HDL-C (137 to 285 mg/dL) had asamples after overnight fasting for standard lipid analy-significantly higher risk for cardiovascular mortality (HR, 3.065;sis. Because plasma triglyceride level is greatly affected95% CI, 1.357% to 6.925%; P � 0.007) in a univariate Cox

analysis. The association between non–HDL-C and cardiovas- by eating, the Friedewald formula gives falsely lowercular mortality remained significant in multivariate Cox mod- LDL cholesterol if non-fasting specimens are used [5].els, which included HDL-C, age, gender, duration of hemodial-

Recently, National Cholesterol Education Program Adultysis, blood pressure, presence of diabetes mellitus, serumTreatment Panel III (NCEP ATP-III) [3] listed non–albumin, C-reactive protein, and body mass index.

Conclusion. Non–HDL-C in predialysis serum was a signifi- HDL-C [6], as well as LDL-C, as target lipid parameterscant and independent predictor of cardiovascular mortality in in subjects with hypertriglyceridemia. Non–HDL-C ishemodialysis patients. Non–HDL-C may be a useful marker the sum of LDL-C and cholesterol in triglyceride-richfor risk assessment in routine practice, although predictive

lipoproteins, but can be easily calculated by subtractingpowers of this and the standard fasting LDL-C should be com-HDL-C from total cholesterol. Because HDL-C is hardlypared in future studies.affected by eating [5], non–HDL-C may be a better indexthan LDL-C for routine practice in most of hemodialysispatients. The purpose of this study was to evaluate thePatients with end-stage renal disease (ESRD) oftenpower of non–HDL-C taken at the start of the hemodial-have lipid and lipoprotein abnormalities that may ac-ysis session in predicting cardiovascular outcome in acount for advanced atherosclerosis in this population [1].cohort of hemodialysis patients.Total and low-density lipoprotein cholesterol (LDL-C)

levels are usually within normal range, whereas ESRDpatients show decreased high-density lipoprotein choles- METHODS

Study design and patientsKey words: dyslipidemia, atherosclerosis, outcome, hemodialysis. We registered 790 patients with ESRD on maintenance

hemodialysis at Inoue Hospital, Suita, Japan, during the 2003 by the International Society of Nephrology

S-117

Non-HDL-C predicts CV mortality in ESRDS-118

period of 1993 through 2000. In the present study, weexcluded 265 patients because of the lack of measure-ments of non–HDL-cholesterol, and analyzed data from525 out of the 790 patients in whom baseline non–HDL-Cwas available. Mean (� SD) age was 55 � 11 years, andthe male:female ratio was 309:216. The underlying renaldiseases were chronic glomerulonephritis (N � 311), dia-betic nephropathy (N � 96), polycystic disease (N � 27),nephrosclerosis (N � 18), toxemia of pregnancy (N � 10),gout (N � 9), lupus nephritis (N � 7), chronic pyelone-phritis (N � 7), renal allograft failure (N � 19), others(N � 13), and unknown (N � 8). The follow-up wasperformed until the end of 2000, and was 64 � 28 (mean �SD) months. This study was approved by the institutional Fig. 1. Kaplan-Meier curves of causes of death from cardiovascular

and non-cardiovascular death in relation to non–HDL cholesterol levelsethical committee (Inoue Hospital Approval No. 109).in a cohort of 525 hemodialysis patients. The ranges of non–HDLcholesterol in the lowest, middle, and highest tertiles were 1.09 to 2.74Serum lipids and other measurementsmmol/L (42 to 106 mg/dL), 2.76 to 3.51 mmol/L (107 to 136 mg/dL),and 3.54 to 7.36 mmol/L (137 to 285 mg/dL), respectively. P valuesBlood samples were taken at the start of hemodialysiswere by log-rank test.session. Patients were not asked to stay fasted. Total

cholesterol was assayed by an enzymatic method, andHDL-C by a precipitation method with dextran sulfateand magnesium chloride. Other parameters were mea- 4.21 � 0.98 (163 � 38 mg/dL), 0.96 � 0.31 (37 � 12 mg/dL),

and 3.26 � 0.98 (126 � 38 mg/dL), respectively. The prev-sured by routine biochemical methods.alence of total cholesterol �4.65 mmol/L (180 mg/dL),

Outcome data collection HDL-C �1.03 mmol/L (40 mg/dL), and non–HDL-C�3.36 mmol/L (130 mg/dL) was 28%, 63%, 41%, respec-Outcome data were obtained for all subjects by re-

viewing the hospital record forms. During the follow tively.up, 120 deaths, including 44 fatal cardiovascular events

Univariate predictors of cardiovascular mortalityoccurred. The 44 cardiovascular deaths were attributableto ischemic heart disease (N � 12), cerebrovascular dis- In univariate Cox analysis, HDL-C was a negative

predictor (HR, 0.326 per 1 mmol/L; 95% CI, 0.108% toease (N � 6), congestive heart failure (N � 19), andsudden death (N � 7), as defined elsewhere [7]. The 0.985%; P � 0.047), whereas non–HDL-C was a positive

predictor (HR, 1.553 per 1 mmol/L; 95% CI, 1.196% toother causes included infectious disease (N � 26), cancer(N � 14), hepatic cirrhosis (N � 5), cachexia (N � 4), 2.016%; P � 0.0009). The HR of those in the highest

tertile of non–HDL-C was 3.07 (95% CI, 1.36% to 6.23%,ileus (N � 2), pulmonary embolism (N � 1), acute pan-creatitis (N � 1), suicide (N � 1), and unspecified (N � P � 0.007) as compared with those in the lowest tertile

(Fig. 1). When the subjects were stratified into three22). These 76 cases were combined and analyzed as non-cardiovascular deaths. groups using cut-off levels for non–HDL-C of 100 and

130 mg/dL, the HR of the group with the highest non–Statistical analysis HDL-C was 3.62 (95% CI, 1.24% to 10.56%; P � 0.018)

as compared with the group with the lowest non–HDL-C.Survival curves were estimated by the Kaplan-Meiermethod, followed by log rank test. Prognostic variables Other significant univariate non-lipid predictors were

age (HR, 1.061 per year; 95% CI, 1.031% to 1.092%; P �were evaluated by univariate and multivariate Cox pro-portional hazards models, and summarized as hazards 0.0001), systolic blood pressure (HR, 1.025 per 1 mm Hg;

95% CI, 1.009% to 1.041%; P � 0.002), and presenceratio (HR) and 95% confidence interval (CI). Signifi-cance was defined as P � 0.05. All of these calculations of diabetes mellitus (HR, 5.054 vs. non-diabetics; 95%

CI, 2.673% to 9.555%; P � 0.0001). Gender, durationwere performed with statistical software (StatView 5,SAS Institute Inc., Cary, NC, USA). of hemodialysis, serum albumin, C-reactive protein, and

BMI were not statistically prognostic for cardiovascularmortality. Non–HDL-C did not predict non-cardiovascu-

RESULTSlar mortality.

Total cholesterol, HDL-C, and non–HDL-Clevels at baseline Independent predictors of cardiovascular mortality

To evaluate whether the association of HDL-C andTotal cholesterol, HDL-C, and non–HDL-C showedalmost normal distributions, and mean � SD levels were non-HDL-C with cardiovascular deaths are independent

Non-HDL-C predicts CV mortality in ESRD S-119

Table 1. Cox proportional hazards models showing the value of in hypertriglyceridemia. Third, previous cross-sectionalnon-HDL cholesterol in predicting cardiovascular mortality in a

studies showed that non–HDL-C was an independentcohort of ESRDfactor associated with carotid artery intima-media thick-

Parameter Model 1 Model 2 ness [8] and aortic sclerosis [4, 7] in ESRD patients.Non-HDL cholesterol 1.479 (1.109–1.973) 1.509 (1.104–2.063) And fourth, non–HDL-C predicts future cardiovascular

per 1 mmol/L P � 0.008 P � 0.009mortality in a cohort of hemodialysis patients, as shownHDL cholesterol 0.531 (0.182–1.546) 0.565 (0.172–1.856)

per 1 mmol/L P � 0.246 P � 0.347 in the present study.Age at entry 1.044 (1.014–1.075) 1.049 (1.015–1.084) It is convenient if overnight fasting is not required forper 1 year P � 0.004 P � 0.004

assessment of lipid-related risk. Because eating affectsDiabetes mellitus 3.314 (1.697–6.544) 4.005 (1.853–8.658)presence vs. absence P � 0.0006 P � 0.0004 LDL-C, but not HDL-C, in non-uremic subjects [5], ca-

Systolic blood pressure 1.012 (0.997–1.028) 1.017 (1.001–1.034) sual sampling may be acceptable for non–HDL-C. Fur-per 1 mm Hg P � 0.121 P � 0.043ther studies are needed to clarify postprandial changesGender — 1.049 (0.513–2.146)

female vs. male P � 0.896 of individual lipid parameters in patients with ESRD.Serum albumin — 0.987 (0.891–1.092) In the present study, we could not obtain sufficientper 1 g/L P � 0.794

information on preexisting cardiovascular morbidity atC-reactive protein 1.003 (0.967–1.040)per 1 g/L — P � 0.884 baseline. Preexisting cardiovascular disease is one of the

Duration of hemodialysis — 1.005 (0.999–1.012)important predictors of cardiovascular mortality [9].per 1 month P � 0.106

Body mass index — 0.993 (0.875–1.127) Apolipoprotein B is also an independent factor associ-per 1 unit P � 0.912 ated with morbidity of myocardial infarction in hemodi-

Global model significance P � 0.0001 P � 0.0001alysis patients [10]. Therefore, it is possible that the pre-

The table gives hazards ratios (95% confidence intervals) with statistical sig-existing cardiovascular disease might have confoundednificance. Model 1 includes five covariates that are significant in univariate Cox

analysis. Model 2 includes five additional covariates that may be relevant. the association between non–HDL-C and cardiovasculardeath during the follow up.

In conclusion, non–HDL-C in predialysis blood wasa significant and independent predictor of cardiovascularof other confounding variables, multivariate Cox modelsmortality in hemodialysis patients. Therefore, use of ca-were used (Table 1). The first model, into which HDL-C,sual non–HDL-C will provide a useful tool for risk as-non–HDL-C and the three significant non-lipid univari-sessment among hemodialysis patients, although the pre-ate predictors were forced, showed that the positive asso-dictive power of this and the standard fasting LDL-Cciation between non–HDL-C and cardiovascular mortal-should be compared in future studies.ity remained significant, whereas the impact of HDL-C

was no longer significant. The close association betweenACKNOWLEDGMENTSnon–HDL-cholesterol and cardiovascular mortality was

We gratefully acknowledge the special effort by Mr. Narutoshiagain significant in the second model that included gen-Okada and Ms. Sugako Muro at Inoue Hospital in accumulating clinicalder, serum albumin, C-reactive protein, duration of he-data. Part of this study was presented in a Symposium at the Interna-

modialysis, and body mass index as additional covariates. tional Congress on Mechanisms of Uremic Toxicity, Uremia, and theCardiovascular System, Wurzburg, Germany, September 13-16, 2002.

Reprint requests to Yoshiki Nishizawa, M.D., Ph.D., DepartmentDISCUSSIONof Metabolism, Endocrinology, and Molecular Medicine, Osaka City

We showed for the first time that non–HDL-C in the University Graduate School of Medicine, 1-4-3, Asahi-machi, Abeno-predialysis samples is an independent predictor of car- ku, Osaka 545-8585, Japan.

E-mail: yynisizawa@med.osaka-cu.ac.jpdiovascular mortality in a cohort of ESRD patients onmaintenance hemodialysis.

REFERENCESThere have been at least four reasons for proposingnon–HDL-C as the marker of lipid abnormality in clini- 1. Kawagishi T, Nishizawa Y, Konishi T, et al: High-resolution

B-mode ultrasonography in evaluation of atherosclerosis in uremia.cal practice. First, non–HDL-C is a marker that inte-Kidney Int 48:820–826, 1995grates atherogenic potentials carried by VLDL, IDL, 2. Wood D, De Backer G, Faergeman O, et al: Prevention of coro-

and LDL. In our previous study [4], although IDL was nary heart disease in clinical practice: Recommendations of theSecond Joint Task Force of European and other Societies on Coro-the lipoprotein that was most closely associated withnary Prevention. Atherosclerosis 140:199–270, 1998aortic sclerosis, VLDL and LDL were also significantly

3. Executive Summary of The Third Report of The National Choles-associated with aortic sclerosis independent of other non- terol Education Program (NCEP) Expert Panel on Detection,

Evaluation, And Treatment of High Blood Cholesterol In Adultslipid variables. Therefore, use of only one lipoprotein(Adult Treatment Panel III). JAMA 285:2486–2497, 2001level may miss the significant contribution of other lipo-

4. Shoji T, Nishizawa Y, Kawagishi T, et al: Intermediate-densityproteins. Second, non–HDL-C can be easily calculated lipoprotein as an independent risk factor for aortic atherosclerosis

in hemodialysis patients. J Am Soc Nephrol 9:1277–1284, 1998from total cholesterol and HDL-C without limitation

Non-HDL-C predicts CV mortality in ESRDS-120

5. Wilder LB, Bachorik PS, Finney CA, et al: The effect of fasting 8. Shoji T, Kawagishi T, Emoto M, et al: Additive impacts of diabetesand renal failure on carotid atherosclerosis. Atherosclerosisstatus on the determination of low-density and high-density lipo-

protein cholesterol. Am J Med 99:374–377, 1995 153:257–258, 20009. Mailloux LU, Napolitano B, Bellucci AG, et al: The impact of6. Frost PH, Havel RJ: Rationale for use of non-high-density lipo-

protein cholesterol rather than low-density lipoprotein cholesterol co-morbid risk factors at the start of dialysis upon the survival ofESRD patients. ASAIO J 42:164–169, 1996as a tool for lipoprotein cholesterol screening and assessment of

risk and therapy. Am J Cardiol 81:26B–31B, 1998 10. Koch M, Kutkuhn B, Trenkwalder E, et al: Apolipoprotein B,fibrinogen, HDL cholesterol, and apolipoprotein(a) phenotypes7. Shoji T, Emoto M, Shinohara K, et al: Diabetes mellitus, aortic

stiffness, and cardiovascular mortality in end-stage renal disease. predict coronary artery disease in hemodialysis patients. J Am SocNephrol 8:1889–1898, 1997J Am Soc Nephrol 12:2117–2124, 2001