Preprocedural white blood cell count and deathafter percutaneous coronary interventionHitinder S. Gurm, MBBS, MRCP, Deepak L. Bhatt, MD, Ritesh Gupta, MBBS, Stephen G. Ellis, MD, FACC,Eric J. Topol, MD, FACC, and Michael S. Lauer, MD, FACC Cleveland, Ohio

Background Elevated inflammatory markers are associated with worse outcome after percutaneous coronary ar-tery interventions (PCI). An elevation in the white blood cell (WBC) count is a nonspecific response to inflammation. Wehypothesized that an elevated WBC count would be a predictor of death in patients undergoing PCI.

Methods A total of 4450 patients undergoing percutaneous coronary artery intervention were divided into quintiles,based on their preprocedural WBC count (mean WBC count: quintile 1, 5.08 � 103/�L; quintile 2, 6.58 � 103/�L;quintile 3, 7.70 � 103/�L; quintile 4, 9.14 � 103/�L; and quintile 5, 13.4 � 103/�L). Vital status was assessedthrough the use of the Social Security Death Index.

Results There were a total of 504 deaths over a follow-up period of 48 months. The best survival was seen in quintile2, with an increase in long-term mortality rates seen with both a higher or a lower WBC count (P � .001). This J-shapedcurve was preserved after multivariate adjustment, with the adjusted hazard ratio of mortality relative to quintile 2 being1.95 (95% CI, 1.40 to 2.73) in quintile 1, 1.66 (95% CI, 1.18 to 2.33) in quintile 3, 2.31 (95% CI, 1.67 to 3.17) inquintile 4, and 2.42 (95% CI, 1.76 to 3.34) in quintile 5.

Conclusions A low or an elevated preprocedural WBC count in patients undergoing PCI is associated with an in-creased risk of long-term death. Our result provides further evidence to support the important role of inflammation in coro-nary artery disease. (Am Heart J 2003;146:692–8.)

Inflammation plays a key role in coronary artery dis-ease.1 Elevated markers of inflammation have beenconsistently associated with worse outcome in largecohorts of apparently healthy subjects,2 patients withstable and unstable angina, and in those undergoingcoronary revascularization.3–8 Evidence from basic andclinical studies support a central role for inflammationin human coronary artery disease.1

The white blood cell (WBC) count is a readily avail-able marker of systemic inflammation. Previous workover the past few decades has demonstrated an ele-vated WBC count to be a marker of poor survival inpatients with myocardial infarction,9–11 in those withstable coronary artery disease,4 and in apparentlyhealthy subjects.12–17 Recent data suggest that C-reac-tive protein (CRP) is a strong and independent predic-tor of poor outcome in patients undergoing percutane-

ous coronary intervention (PCI).18 The prognosticvalue of preprocedural WBC count in predicting out-come in patients undergoing PCI has not been deter-mined. Since inflammatory processes are intimatelyinvolved in coronary atherosclerosis and the vascularresponse to PCI,19,20 we hypothesized that an elevatedWBC count would be a predictor of poor outcome inpatients undergoing PCI. We accordingly studied theassociation of preprocedural WBC count and death inpatients undergoing PCI at a large tertiary care center.

MethodsAll patients undergoing PCI at the Cleveland Clinic are fol-

lowed in a prospective PCI registry. Clinical, demographic,procedural, and angiographic characteristics and long-termfollow-up data are collected on all patients.21 Baseline datainclude patient demographics, acuity of presentation, conven-tional risk factors, comorbidities, medications, number of dis-eased vessels, left ventricular ejection fraction, and AmericanCollege of Cardiology/American Heart Association (ACC/AHA) lesion classification.22,23 The physician-operator deter-mined the type of device used and adjunct use of plateletglycoprotein IIb/IIIa inhibitor. Our study population com-prised of all patients who underwent PCI between October28, 1996, and August 31, 2000, and had a complete bloodcount measured as a part of routine pre-procedural protocolwithin 7 days before the procedure and had a social security

From the aDepartment of Cardiovascular Medicine, Cleveland Clinic Foundation,Cleveland, Ohio.Dr Lauer is partially supported by NIH grant HL 66004-01.Submitted October 4, 2002; accepted February 5, 2003.Reprint requests: Michael S. Lauer, MD, FACC, F25, Cleveland Clinic Foundation,9500 Euclid Ave, Cleveland, OH 44195.E-mail: lauerm@ccf.org© 2003, Mosby, Inc. All rights reserved.0002-8703/2003/$30.00 � 0doi:10.1016/S0002-8703(03)00230-8

number on record. The WBC count obtained closest to theprocedure was used in patients with more than one value onrecord. In patients having multiple procedures during thestudy period, only the first was included. Routine 30-day fol-low-up data included vital status and myocardial infarction.Creatinine kinase data was recorded at 8 hours and on themorning after the PCI.

The primary outcome of our study was all-cause death. All-cause death is an objective and unbiased end point unlikecardiac or noncardiac death, which may be significantly af-fected by ascertainment bias.24 Vital status was assessedthrough the use of the Social Security Death Index, whichhas been previously shown to be highly specific and unbi-ased.25,26 In a previous study, the sensitivity of the Social Se-curity Death Index in the Cleveland Clinic population hasbeen confirmed to be 97%.27 Mean follow-up was 3 years,with a median follow-up among patients alive at the end ofthe study period being 33 months.

Statistical analysisContinuous variables are expressed as mean � SD; discrete

variables are expressed as frequency counts and percentages.Difference in discrete variables between groups was deter-mined by means of the �2 test; the t test was used for contin-uous variables. The association of WBC count with death wasassessed by using proportional hazards regression. Propor-tional hazards assumption was continued by testing time-de-pendent covariates. A multivariable model was used to assessthe predictive value of WBC count after adjusting for knownpredictors of survival after PCI; age, sex, left ventricular ejec-tion fraction, renal function, ACC/AHA worst lesion class, thevessel being intervened upon, medication use at time of PCI,acute myocardial infarction, and chronic obstructive pulmo-nary disease (COPD). Mean or mode imputation was used tocorrect for missing variables, thus allowing all patients to beincluded in the model. None of the variables imputed had�5% missing values. Since time-censored data limited calcu-lating goodness of fit for the model, Akaike information crite-ria (AIC) were calculated for a model without WBC quintilesand compared with the above model. The �i or the differ-ence between AIC of the two models and the Akaikeweight,28,29 an estimate of the likelihood that a given modelis the best model of those studies, were calculated to deter-mine if addition of WBC improved goodness of fit. Further-more, the association between WBC count and death wasassessed across tertiles of age and certain prespecified sub-groups known to affect WBC count: patients with history ofmyocardial infarction,16 diabetes,30 or smoking31; andsex.16,32 The relation between 30-day death and WBC quin-tiles was assessed by means of logistic regression.

Statistical comparisons were performed with the use ofSPSS; logistic regression and proportional hazards regressionwere performed with the use of SAS. A P value of � .05 wasconsidered significant. The institutional review board ap-proved the study.

ResultsA total of 8586 PCIs were performed in 7211 pa-

tients during the study period. We excluded 2761 pa-

tients because they either did not have social securitynumbers on record (1042) or did not have a WBCcount performed at the Cleveland Clinic within 7 daysor less of the PCI.17,19 Table I describes the baselinedemographic, angiographic, and procedural differencesbetween the patients included and excluded from theanalysis. No marked differences were noted betweenthe groups.

The WBC count ranged from 1.86 � 103/�L to 93.2� 103/�L. The highest WBC count was seen in a pa-tient with chronic lymphocytic leukemia who was stillalive at the end of the study. The median WBC countwas 7.69 � 10 3 /�L; values for the 1st, 10th, 25th,75th, 90th, and 99th percentiles were 3.8 � 103/�L,5.2 � 103/�L, 6.31 � 103/�L, 9.62 � 103/�L, 12.25 �103/�L, and 20.38 � 103/�L, respectively. Table II de-scribes the range for the different WBC quintiles andthe distribution of baseline demographic and clinicalcharacteristics. Patients in the highest quintile weremore likely to be women, younger, and more likely tobe smokers or have COPD. They had a higher acuityof illness with more likelihood of having recent oracute myocardial infarction, a lower ejection fraction,and worse lesion morphology. The incidence of 3-ves-

Table I. Comparison of patients included and excluded fromthe analysis

Baseline characteristics

Patientsexcluded

(n � 2761)

Patientsincluded

(n � 4450)

Age (mean) (y) 63.20 64.13Male (%) 1938 (70) 3107 (70)History of hypertension (%) 1782 (65) 3029 (68)History of hyperlipidemia (%) 1303 (47) 1848 (42)Family history of premature coronary

artery disease (%)986 (36) 1761 (39)

History of myocardial infarction (%) 1273 (47) 1873 (44)Current smoker (%) 607 (22) 968 (22)COPD (%) 190 (7) 348 (8)History of peripheral vascular disease

(%)244 (9) 475 (11)

History of creatinine �1.5 mg/dL (%) 86 (3) 209 (5)History of transient ischemic attack/

stroke (%)249 (9) 443 (10)

Diabetes—on insulin (%) 289 (11) 496 (11)Diabetes—not on insulin (%) 480 (17) 772 (17)Acute myocardial infarction (%) 148 (5) 432 (9)Unstable angina (%) 1620 (59) 2677 (61)Left ventricular ejection fraction

(mean)53.27% 52.01%

3-Vessel disease (%) 831 (30) 1476 (34)Stent used (%) 2032 (74) 3430 (77)Worst ACC lesion class B2 or C (%) 1305 (48) 2140 (48)Left anterior descending artery

intervention (%)1181 (43) 1870 (42)

Procedural success (%) 2490 (90) 4085 (92)Post procedural non-Q–MI (%) 112 (4) 180 (4)

American Heart JournalVolume 146, Number 4

Gurm et al 693

sel disease or of unstable angina preceding the PCIwas, however, not any higher. Although no differencewas noted in the postprocedural myocardial infarction,patients in the fifth quintile were more likely to need

intra-aortic balloon pump, to have postprocedural re-nal failure, or to need blood products.

There were a total of 504 deaths over 48 months.Figure 1 depicts the mortality rates of the cohort perWBC quintiles. Although the highest mortality rate wasseen in quintile 5, the best survival was seen in quin-tile 2. There was a J-shaped relation between mortalityrate and pre procedural WBC count, with mortalityrates increasing with either a rise or a fall in the totalWBC count. (Figures 1 and 2).

The relation between mortality rate and WBC countremained significant after adjusting for age, sex, acutemyocardial infarction, serum creatinine, left ventricularejection fraction, ACC/AHA worse lesion score, ad-junct medications, diabetes, and other comorbidities,site of lesion, and maximal residual stenosis in the in-tervened vessel. The adjusted hazard of death relativeto quintile 2 was 1.95 (95% CI, 1.40 to 2.73) in quin-tile 1, 1.66 (95% CI, 1.18 to 2.33) in quintile 3, 2.31(95% CI, 1.67 to 3.17) in quintile 4, and 2.42 (95% CI,1.76 to 3.34) in quintile 5 (Figure 3). The J-shapedcurve was preserved and the increased hazard was ob-served with rising WBC count even within the rangeconsidered normal. Other significant predictors of out-come were age, low ejection fraction, acute myocar-dial infarction, renal function, diabetes, COPD, use of

Table II. Patient characteristics per WBC quintile

Baseline characteristics

Quintile

1 (n � 892) 2 (n � 889) 3 (n � 886) 4 (n � 891) 5 (n � 892)

Mean WBC (� 103/�L) 5.08 6.58 7.70 9.14 13.4WBC range (� 103/�L) 1.86–5.97 5.98–7.16 7.17–8.31 8.32–10.16 10.17–93.20Age (y, mean) 64.32 65.08 64.26 63.41 62.64Male (%) 674 (75.6) 636 (71.5) 627 (70.8) 610 (68.5) 567 (63.6)Diabetes—no insulin (%) 136 (15.2) 151 (17.0) 164 (18.5) 168 (18.9) 147 (16.5)Diabetes—on insulin (%) 79 (8.9) 80 (9.0) 116 (13.1) 117 (13.1) 111 (12.4)History of hypercholesterolemia (%) 372 (41.7) 380 (42.7) 399 (45.0) 359 (40.3) 335 (37.6)Prior myocardial infarction (%) 343 (39.6) 383 (44.3) 376 (43.7) 414 (48.3) 360 (42.6)Family history of premature coronary artery disease (%) 371 (41.6) 354 (40.0) 359 (40.6) 354 (40.0) 319 (36.1)Current smoker (%) 112 (16.0) 142 (16.0) 163 (18.4) 236 (26.5) 306 (34.3)History of hypertension (%) 617 (69.2) 611 (68.7) 621 (70.2) 588 (66.0) 597 (67.2)History of stroke or transient ischemic attack (%) 74 (8.3) 102 (11.5) 99 (11.2) 85 (9.6) 82 (9.2)History of serum creatinine �1.5mg/dL (%) 48 (5.4) 35 (3.9) 46 (5.2) 37 (4.2) 47 (5.3)History of peripheral vascular disease (%) 84 (9.4) 89 (10.0) 87 (9.8) 107 (12.0) 106 (11.9)History of COPD (%) 48 (5.4) 63 (7.1) 55 (6.2) 89 (10.0) 98 (11.0)Unstable angina (%) 524 (58.9) 567 (63.9) 544 (61.6) 567 (64.4) 471 (55.1)Recent myocardial infarction (%) 70 (7.8) 109 (12.3) 145 (16.4) 201 (22.6) 246 (27.6)Left ventricular systolic function (% mean) 52.9% 53.26% 52.94% 51.66% 48.73%ACC class C lesion (%) 204 (22.9) 166 (18.7) 208 (23.6) 198 (22.3) 240 (26.9)3-Vessel disease (%) 305 (34.6) 303 (34.5) 324 (37.0) 289 (32.7) 249 (28.2)Stent use (%) 657 (73.7) 687 (77.3) 666 (75.2) 689 (77.4) 700 (78.6)Procedure success (%) 814 (91.6) 834 (94.0) 813 (92.0) 809 (91.0) 817 (91.8)IABP use (%) 23 (2.6) 15 (1.7) 22 (2.5) 33 (3.7) 116 (13.0)Postprocedural non-Q–MI (%) 37 (4.2) 38 (4.3) 39 (4.4) 34 (3.8) 29 (3.3)Postprocedural renal failure (%) 14 (1.6) 12 (1.4) 13 (1.5) 21 (2.4) 48 (5.4)Postprocedural blood product use (%) 46 (5.2) 32 (3.6) 35 (4.0) 50 (5.7) 115 (13.1)

Figure 1

Kaplan-Meier curve of survival by WBC quintiles.

American Heart JournalOctober 2003

694 Gurm et al

aspirin, statins, and diuretics, and left circumflex arteryintervention. (Table III). The AIC for the model withWBC quintiles was 7506.949; the AIC for the modelwithout WBC was 7546.697, leading to a �i of 39.745.The Akaike weight for the model without WBC quin-tiles was �0.001, compared with a weight of �0.999for the final model, suggesting substantial improve-ment in the goodness of fit of the model with incorpo-ration of WBC quintiles.

Figure 4 demonstrates the relation between deathand WBC count in certain prespecified subgroups. Nostatistically significant interactions were noted. Further-more, no interaction with glycoprotein IIb/IIIa inhibi-tor use was noted. The J-shaped relation was pre-served in all subgroups, and a strong association wasseen between death and WBC count in all patients.

There was a total of 72 deaths at 30 days. The mor-tality rate was 0.7% (6) in quintile 1, 0.7% (6) in quin-tile 2, 0.8% (7) in quintile 3, 1.5% (13) in quintile 4,and 4.5% (40) in quintile 5. The unadjusted hazard ra-tio (HR) for 30-day death for quintile 5 was 6.91 (95%CI, 2.92 to 16.38, P � .001). Quintile 5 remained anindependent predictor of 30-day death after adjustingfor age, sex, acute myocardial infarction, unstable an-gina, renal insufficiency, left ventricular ejection frac-tion, maximal stenosis in the target vessel at end ofPCI, ACC/AHA class, and left anterior descending ar-tery disease (HR, 3.16; 95% CI, 1.85 to 5.39, P �.001).

DiscussionThe relation between death and WBC count has

been the focus of numerous studies in the past.11,15,16

In our current analysis, we demonstrate a strong andindependent association of WBC count with death inpatients undergoing PCI. Compared with the secondquintile, the patients with highest WBC count demon-strated a 2.5-times excess mortality rate, whereas thosewith the lowest WBC count had almost 2-fold excessin mortality rate, even after adjusting for major predic-tors of death after PCI. Furthermore, the addition ofWBC quintiles substantially improved the goodness offit of the model. Furthermore, the increased hazardwith rising WBC count is seen within the range con-sidered normal. Our findings corroborate and extendthe previous work on the association of WBC countand long-term outcome in a non-PCI population.

With the use of the Second National Health and Nu-tritional Examination Survey (NHANES II), Brown andcolleagues described a 1.4-times relative risk of deathafter adjusting for multiple covariates among healthyadults in the highest tertile of WBC compared withthose in the lowest tertile.16 Weijenberg et al13 de-scribed a 65% increase in risk of death with each stan-dard deviation increase in WBC count in a cohort of

elderly Dutch subjects. A meta-analysis of 19 studiesdescribed a coronary heart disease risk of 1.5 amongpatients in the top third of WBC count compared withthose in the bottom third.4 Multiple studies have de-scribed an excess mortality rate with elevated WBCcount in patients with myocardial infarction.10,11,33

Our study was the first work exploring the relationbetween WBC count at the time of PCI and 30-day andlong-term death. Furthermore, we looked at prespeci-fied subgroups to minimize the influence of factorsknown to influence WBC count. As has been previ-ously reported, WBC count was higher in smokers, theyoung, and patients with myocardial infarction anddiabetes.16,31,32 By dividing the cohort into subgroupsbased on these factors, we were able to confirm thevalidity of our findings. Furthermore, to rule out a dif-

Figure 2

Three-year Kaplan-Meier mortality rates according to median pre-procedural WBC count.

Figure 3

Hazard for death for WBC quintiles with multivariate adjustment.

American Heart JournalVolume 146, Number 4

Gurm et al 695

fering impact of WBC count on mortality rates at dif-ferent times of follow-up, we confirmed the propor-tional hazards assumption by testing time-dependentcovariates. The hazard associated with a higher orlower WBC count was maintained throughout thestudy period.

Our study is the first one to describe a J-shaped rela-tion between baseline WBC count and long-termdeath. A low WBC count may be a marker of general

bone marrow depression or may reflect an inability tomount an inflammatory response and hence may re-flect a sicker population. The lack of a previous de-scription of such an effect may be secondary to theuse of tertiles in most previous analysis. This wouldhave the effect of combining the quintile with the bestsurvival with one with a high mortality rate and hencenegate the difference in survival. Further sicker pa-tients are excluded from clinical trials and such pa-tients are more likely to be followed in registry popu-lations such as ours. Furthermore, since this relationwas preserved in multiple subgroups in our study, it islikely to genuine. However, a chance finding cannotbe excluded. Recent work has identified leukocyteATP-binding cassette transporter 1 (ABCA1), a key reg-ulator of HDL metabolism, as a leukocyte factor thatcontrols the recruitment of inflammatory cells and pro-tects against atherosclerosis.34 It is likely that differentleukocyte subtypes have differing roles in the athero-sclerotic process, with some having protective and theother negative effects. Patients with a low WBC countmay thus possibly be at a greater risk secondary to therelative lack of this protective function of WBCs.

Inflammation has acquired increasing importance asa marker and medium of cardiovascular mortality andmorbidity. Multiple studies have previously reported astrong relation between CRP and outcome afterPCI.7,18 Leukocytes are the major modulators of inflam-matory processes and play a key role in plaque devel-opment and instability. PCI is associated with leuko-cyte activation, increased expression of adhesionmolecules, and formation of platelet-leukocyte com-plexes.19,20,35 The relation between leukocyte activa-tion and long-term outcome has not been well defined.Although a causal interpretation cannot be ascribed tothe relation between WBC count and death, it is likely

Table III. Significant predictors of long-term mortality after PCI in the order of selection in a multivariable model

Variable Hazard ratio 95% CI �2 P

Age 1.060 1.050–1.070 201.36 �.0001Creatinine �1.5 mg/dL 1.301 1.242–1.361 224.50 �.0001Ejection fraction �40% 2.007 1.670–2.411 117.20 �.0001Diuretic use 1.995 1.644–2.421 71.20 �.0001COPD 2.061 1.640–2.590 45.75 �.0001Diabetic (on insulin) 1.745 1.385–2.197 21.75 �.0001Acute MI 1.486 1.125–1.963 20.53 �.0001Circumflex PCI 1.473 1.231–1.762 14.52 .0001WBC Quintile 5 2.425 1.760–3.341 11.74 .0006WBC Quintile 4 2.305 1.671–3.179 14.32 .0002Maximal stenosis in the target vessel at end of PCI 1.005 1.001–1.008 9.12 .0025WBC Quintile 1 1.952 1.395–2.732 8.33 .0039WBC Quintile 3 1.661 1.182–2.333 9.41 .0021Diabetic (not on insulin) 1.395 1.119–1.740 8.05 .0045Statin use (at time of PCI) 0.747 0.603–0.926 6.93 .0085Aspirin use (at time of PCI) 0.720 0.543–0.956 5.19 .0227

Figure 4

Rate of death in various subgroups stratified by WBC quintiles.

American Heart JournalOctober 2003

696 Gurm et al

that an elevation in WBC count is a reflection of thesame inflammatory state associated with cardiovasculardisease that is also manifest as elevation of other in-flammatory markers. Indeed, such an association be-tween leucocyte count, serum amyloid A (SAA) pro-tein, and CRP has been previously described.3

LimitationsWe used a single value of preprocedural WBC count.

No data were available with respect to changes inWBC count after PCI or during subsequent follow-up.Previous studies suggest that the neutrophils are re-sponsible for most of the increased risk associatedwith an elevated WBC count.36 We did not have ac-cess to differential WBC count in most of our patientsand were hence unable to study the association of var-ious cell types with death. Furthermore, no other mea-sures of inflammation such as CRP, SAA protein, orinterleukin-6 were available. WBC count is known tobe affected by catecholamine and cortisol levels. Dataon these variables were not available. It is noteworthythat there was no association noted between unstableangina and WBC count. This could argue against thisbeing a marker of inflammation. However, we did findan association between WBC count and acute and re-cent myocardial infarction in which the existence of aproinflammatory state has been well documented.33

ConclusionsWBC count is an inexpensive and simple test that is

routinely obtained in patients undergoing PCI. Wewere able to demonstrate a strong, independent associ-ation between WBC count and death in patients under-going PCI. Our findings provide additional evidence tosupport the implications of inflammation in cardiovas-cular pathophysiology. Further studies are warrantedto confirm our findings and to define the role of WBCcount in clinical decision-making in patients undergo-ing PCI.

References1. Libby P, Ridker PM, Maseri A. Inflammation and atherosclerosis.

Circulation 2002;105:1135–43.2. Ridker PM, Buring JE, Shih J, et al. Prospective study of C-reactive

protein and the risk of future cardiovascular events among appar-ently healthy women. Circulation 1998;98:731–3.

3. Danesh J, Whincup P, Walker M, et al. Low grade inflammationand coronary heart disease: prospective study and updated meta-analyses. BMJ 2000;321:199–204.

4. Danesh J, Collins R, Appleby P, et al. Association of fibrinogen,C-reactive protein, albumin, or leukocyte count with coronary heartdisease: meta-analyses of prospective studies. JAMA 1998;279:1477–82.

5. Liuzzo G, Biasucci LM, Gallimore JR, et al. The prognostic value ofC-reactive protein and serum amyloid a protein in severe unstableangina. N Engl J Med 1994;331:417–24.

6. Haverkate F, Thompson SG, Pyke SD, et al. Production of C-reac-tive protein and risk of coronary events in stable and unstable an-gina: European Concerted Action on Thrombosis and DisabilitiesAngina Pectoris Study Group. Lancet 1997;349:462–6.

7. Buffon A, Liuzzo G, Biasucci LM, et al. Preprocedural serum levelsof C-reactive protein predict early complications and late resteno-sis after coronary angioplasty. J Am Coll Cardiol 1999;34:1512–21.

8. Biasucci LM, Liuzzo G, Grillo RL, et al. Elevated levels of C-reactiveprotein at discharge in patients with unstable angina predict recur-rent instability. Circulation 1999;99:855–60.

9. Modan B, Schor S, Shani M. Acute myocardial infarction: prog-nostic value of white blood cell count and blood glucose level.JAMA 1975;233:266–7.

10. Furman MI, Becker RC, Yarzebski J, et al. Effect of elevated leuko-cyte count on in-hospital mortality following acute myocardial in-farction. Am J Cardiol 1996;78:945–8.

11. Barron HV, Cannon CP, Murphy SA, et al. Association betweenwhite blood cell count, epicardial blood flow, myocardial perfu-sion, and clinical outcomes in the setting of acute myocardial in-farction: a thrombolysis in myocardial infarction 10 substudy. Cir-culation 2000;102:2329–34.

12. Folsom AR, Wu KK, Rosamond WD, et al. Prospective study ofhemostatic factors and incidence of coronary heart disease: theAtherosclerosis Risk in Communities (ARIC) Study. Circulation1997;96:1102–8.

13. Weijenberg MP, Feskens EJ, Kromhout D. White blood cell countand the risk of coronary heart disease and all-cause mortality inelderly men. Arterioscler Thromb Vasc Biol 1996;16:499–503.

14. Yarnell JW, Baker IA, Sweetnam PM, et al. Fibrinogen, viscosity,and white blood cell count are major risk factors for ischemicheart disease: the Caerphilly and Speedwell collaborative heartdisease studies. Circulation 1991;83:836–44.

15. de Labry LO, Campion EW, Glynn RJ, et al. White blood cellcount as a predictor of mortality: results over 18 years from theNormative Aging Study. J Clin Epidemiol 1990;43:153–7.

16. Brown DW, Giles WH, Croft JB. White blood cell count: an inde-pendent predictor of coronary heart disease mortality among anational cohort. J Clin Epidemiol 2001;54:316–22.

17. Kannel WB, Anderson K, Wilson PW. White blood cell count andcardiovascular disease: insights from the Framingham Study.JAMA 1992;267:1253–6.

18. Chew DP, Bhatt DL, Robbins MA, et al. Incremental prognosticvalue of elevated baseline C-reactive protein among establishedmarkers of risk in percutaneous coronary intervention. Circulation2001;104:992–7.

19. Serrano CV Jr, Ramires JA, Venturinelli M, et al. Coronary angio-plasty results in leukocyte and platelet activation with adhesionmolecule expression: evidence of inflammatory responses in coro-nary angioplasty. J Am Coll Cardiol 1997;29:1276–83.

20. Mickelson JK, Lakkis NM, Villarreal-Levy G, et al. Leukocyte acti-vation with platelet adhesion after coronary angioplasty: a mecha-nism for recurrent disease? J Am Coll Cardiol 1996;28:345–53.

21. Ellis SG, Guetta V, Miller D, et al. Relation between lesion charac-teristics and risk with percutaneous intervention in the stent andglycoprotein IIb/IIIa era: an analysis of results from 10,907 lesionsand proposal for new classification scheme. Circulation 1999;100:1971–6.

22. Ryan TJ, Faxon DP, Gunnar RM, et al. Guidelines for percutane-ous transluminal coronary angioplasty: a report of the AmericanCollege of Cardiology/American Heart Association Task Force on

American Heart JournalVolume 146, Number 4

Gurm et al 697

Assessment of Diagnostic and Therapeutic Cardiovascular Proce-dures (Subcommittee on Percutaneous Transluminal Coronary An-gioplasty). Circulation 1988;78:486–502.

23. Ellis SG, Vandormael MG, Cowley MJ, et al. Coronary morpho-logic and clinical determinants of procedural outcome with angio-plasty for multivessel coronary disease: implications for patientselection: Multivessel Angioplasty Prognosis Study Group. Circula-tion 1990;82:1193–202.

24. Lauer MS, Blackstone EH, Young JB, et al. Cause of death in clini-cal research: time for a reassessment? J Am Coll Cardiol 1999;34:618–20.

25. Boyle CA, Decoufle P. National sources of vital status information:extent of coverage and possible selectivity in reporting. Am J Epi-demiol 1990;131:160–8.

26. Newman TB, Brown AN. Use of commercial record linkage soft-ware and vital statistics to identify patient deaths. J Am Med In-form Assoc 1997;4:233–7.

27. Nishime EO, Cole CR, Blackstone EH, et al. Heart rate recoveryand treadmill exercise score as predictors of mortality in patientsreferred for exercise ECG. JAMA 2000;284:1392–8.

28. Akaike H. A Bayesian analysis of the minimum AIC procedure.Ann Inst Stat Math 1978;30:9–14.

29. Burnham KP, Anderson DR. Model selection and inference: apractical information theoretic approach. New York: Springer;1998.

30. Ford ES. Leukocyte count, erythrocyte sedimentation rate, and dia-betes incidence in a national sample of US adults. Am J Epidemiol2002;155:57–64.

31. Yarnell JW, Sweetnam PM, Rogers S, et al. Some long-term effectsof smoking on the haemostatic system: a report from the Caer-philly and Speedwell Collaborative Surveys. J Clin Pathol 1987;40:909–13.

32. Held C, Hjemdahl P, Rehnqvist N, et al. Haemostatic markers, in-flammatory parameters and lipids in male and female patients inthe Angina Prognosis Study in Stockholm (APSIS): a comparisonwith healthy controls. J Intern Med 1997;241:59–69.

33. Cannon CP, McCabe CH, Wilcox RG, et al. Association of whiteblood cell count with increased mortality in acute myocardial in-farction and unstable angina pectoris: OPUS-TIMI 16 Investigators.Am J Cardiol 2001;87:636–9, A10.

34. van Eck M, Bos IS, Kaminski WE, et al. Leukocyte ABCA1 controlssusceptibility to atherosclerosis and macrophage recruitment intotissues. Proc Natl Acad Sci U S A 2002;99:6298–303.

35. Ikeda H, Nakayama H, Oda T, et al. Neutrophil activation afterpercutaneous transluminal coronary angioplasty. Am Heart J1994;128:1091–8.

36. Sweetnam PM, Thomas HF, Yarnell JW, et al. Total and differen-tial leukocyte counts as predictors of ischemic heart disease: theCaerphilly and Speedwell studies. Am J Epidemiol 1997;145:416–21.

American Heart JournalOctober 2003

698 Gurm et al