DOI: 10.1016/j.athoracsur.2008.06.078 2008;86:1431-1437 Ann Thorac Surg

Cooper, Cullen D. Morris, Robert A. Guyton and Vinod H. Thourani Michael E. Halkos, Omar M. Lattouf, John D. Puskas, Patrick Kilgo, William A.

Long-Term Survival After Coronary Artery Bypass SurgeryElevated Preoperative Hemoglobin A1c Level is Associated With Reduced

http://ats.ctsnetjournals.org/cgi/content/full/86/5/1431located on the World Wide Web at:

The online version of this article, along with updated information and services, is

Print ISSN: 0003-4975; eISSN: 1552-6259. Southern Thoracic Surgical Association. Copyright © 2008 by The Society of Thoracic Surgeons.

is the official journal of The Society of Thoracic Surgeons and theThe Annals of Thoracic Surgery

by on June 12, 2013 ats.ctsnetjournals.orgDownloaded from

EACMPR

CE

(b

3tJiwwj(arHw

ltp

CtdirmiDclaww

A

PT

ALA

©P

AD

ULT

CA

RD

IAC

levated Preoperative Hemoglobin A1c Level isssociated With Reduced Long-Term Survival Afteroronary Artery Bypass Surgeryichael E. Halkos, MD, Omar M. Lattouf, MD, PhD, John D. Puskas, MD,

atrick Kilgo, MS, William A. Cooper, MD, Cullen D. Morris, MD,obert A. Guyton, MD, and Vinod H. Thourani, MD

linical Research Unit, Division of Cardiothoracic Surgery, and Department of Biostatistics, Rollins School of Public Health,

mory University School of Medicine, Atlanta, Georgia

5tHcipwbtbicv

mrpp

Background. The predictive role of hemoglobin A1cHbA1c) on long-term outcomes after coronary arteryypass surgery has not been evaluated.Methods. Preoperative HbA1c levels were obtained in

,201 patients undergoing primary, elective coronary ar-ery bypass surgery at Emory Healthcare Hospitals fromanuary 2002 to December 2006 and entered prospectivelynto a computerized database. Long-term survival statusas determined by cross-referencing patient recordsith the Social Security Death Index. Log-rank (unad-

usted) and Cox proportional hazards regression modelsadjusted) were employed to determine whether HbA1cnd diabetes mellitus were independent risk factors foreduced long-term survival, adjusted for 29 covariates.azard ratios for each unit increase in continuous HbA1cere calculated.Results. Patients with HbA1c of 7% or greater had

ower unadjusted 5-year survival compared with pa-ients with HbA1c less than 7% (p � 0.001). Similarly,

atients with diabetes mellitus had lower unadjusted

mtr1g

odtaw[ltm

ong Hospital, 6th Floor, Medical Office Tower, Cardiothoracic Surgery,tlanta, GA 30308; e-mail: vthoura@emory.edu.

2008 by The Society of Thoracic Surgeonsublished by Elsevier Inc

ats.ctsnetjournDownloaded from

-year survival compared with patients without diabe-es (p < 0.001). After multivariable adjustment, higherbA1c (measured as a continuous variable) was asso-

iated with reduced long-term survival for each unitncrease in HbA1c (hazard ratio 1.15, p < 0.001), butreoperative diagnosis of diabetes was not associatedith reduced long-term survival after coronary arteryypass surgery (p � 0.41). Other multivariable predic-

ors of reduced long-term survival included age, cere-rovascular disease, elevated serum creatinine, renal

nsufficiency, congestive heart failure, previous myo-ardial infarction, chronic lung disease, and peripheralascular disease.Conclusions. Poor preoperative glycemic control, aseasured by an elevated HbA1c, is associated with

educed long-term survival after coronary artery by-ass surgery. Optimizing glucose control in theseatients may improve long-term survival.

(Ann Thorac Surg 2008;86:1431–7)

© 2008 by The Society of Thoracic Surgeons

oronary artery bypass surgery (CABG) is the pre-ferred method of revascularization in diabetic pa-

ients with multivessel coronary disease [1]. Althoughiabetes mellitus has historically been associated with

ncreased in-hospital mortality after CABG [2–4], recenteports have documented dramatic reductions in hospitalortality with the use of a continuous perioperative

nsulin infusion to achieve strict glycemic control [5, 6].epending on the severity of diabetes and its associated

omplications and comorbidities, long-term survival isower in diabetic compared with nondiabetic patientsfter CABG [3, 7–10]. Moreover, diabetic patients treatedith insulin appear to have worse outcomes comparedith those managed with diet or oral hypoglycemic

ccepted for publication June 18, 2008.

resented at the Poster Session of the Forty-fourth Annual Meeting ofhe Society of Thoracic Surgeons, Fort Lauderdale, FL, Jan 28–30, 2008.

ddress correspondence to Dr Thourani, 550 Peachtree St, Crawford

edications [3, 9, 10]. For these patients, aggressivereatment to achieve glycemic control is associated with aeduced incidence of diabetes-related complications [11,2]. Therefore, treatment aimed at achieving long-termlycemic control may improve long-term survival.Hemoglobin A1c (HbA1c) is the recommended method

f monitoring long-term glycemic control in patients withiabetes mellitus [13]. The American Diabetes Associa-

ion currently recommends that patients with diabeteschieve HbA1c levels less than 7%, which is associatedith a lower risk of diabetes-associated complications

11, 12]. Although patients with diabetes appear to haveower long-term survival after CABG, there is no objec-ive risk factor that quantifies the risk of long-term

ortality among these patients. We previously reported

Drs Puskas and Guyton disclose that they have afinancial relationship with Medtronic and Maquet

Cardiovascular; Dr Lattouf with Medtronic.

0003-4975/08/$34.00doi:10.1016/j.athoracsur.2008.06.078

by on June 12, 2013 als.org

ttfpv

M

SIaopTq

iCpTTdofadCpntb

T

R

P

P

Bm

1432 HALKOS ET AL Ann Thorac SurgHEMOGLOBIN A1C AND SURVIVAL AFTER CABG 2008;86:1431–7A

DU

LTC

AR

DIA

C

hat elevated preoperative HbA1c was a powerful predic-or of adverse in-hospital events after CABG [14]. There-ore, the purpose of this study was to determine theredictive role of preoperative HbA1c on long-term sur-ival after CABG.

aterial and Methods

tudy Populationn compliance with HIPAA (Health Insurance Portabilitynd Accountability Act) regulations and the Declarationf Helsinki, and after Institutional Review Board ap-roval was granted by Emory University, The Society ofhoracic Surgeons (STS) Adult Cardiac Database wasueried for all patients who underwent primary, elective,

able 1. Preoperative and Intraoperative Variables

Hisk Factor

reoperative risk factorsDiabetes mellitus (%)Diabetes control method

None (%)Diet (%)Oral (%)Insulin (%)

Age (mean �SD)Female (%)Caucasian (%)Renal insufficiency (%)Renal failure dialysis-dependent (%)Cerebrovascular disease (%)Cerebrovascular accident (%)NYHA class IV (%)CCS class V (%)Current smoker (%)Congestive heart failure (%)Myocardial infarction (%)Ejection fraction (mean � SD)Last creatinine value (mean � SD)Hypertension (%)Intraoperative glucose POD 0 (mean � SD)Postoperative glucose POD 1–3 (mean � SD)Left main disease � 50% stenosis (%)No. diseased vessels (mean � SD)Chronic lung diseasePeripheral vascular disease (%)

erioperative factorsArterial grafts (mean � SD)Vein grafts (mean � SD)Total grafts (mean � SD)Cardiopulmonary bypass time (mean � SD)OPCAB (%)LIMA or BIMA used (%)

IMA � bilateral internal mammary arteries; CCS � Canadian Cardiovaammary artery; NHYA � New York Heart Association; OPCAB � off

ats.ctsnetjournDownloaded from

solated CABG at Emory University Hospital, Emoryrawford Long Hospital, and Wellstar Kennestone Hos-ital between January 1, 2002, and December 31, 2006.he study cohort consisted of 3,201 consecutive patients.he time frame was chosen to include the entire perioduring which preoperative HbA1c was collected as partf our routine preoperative laboratory analysis. All dataor consecutive patients were prospectively entered into

computerized cardiac surgical database, utilizing theata fields and definitions of the STS National Adultardiac Database (available at: www.sts.org/documents/df/AdultCVDataSpecifications2.61.pdf). Cardiopulmo-ary bypass was utilized according to the discretion of

he attending surgeon. Conventional coronary arteryypass with cardiopulmonary bypass was performed

c � 7.0% HbA1c � 7.0%2,360 n � 841 p Value

8 (22.8) 747 (88.8) �0.001

0 (1.27) 36 (4.28) �0.0014 (3.56) 37 (4.40)3 (13.26) 386 (45.9)1 (4.70) 288 (34.24)� 11.1 61.7 � 10.1 �0.001

5 (26.5) 284 (33.8) �0.0010 (83.3) 607 (75.0) �0.0012 (4.8) 77 (9.2) �0.0012 (1.8) 32 (3.8) �0.0016 (15.5) 163 (19.4) 0.0094 (8.2) 96 (11.4) 0.0063 (20.7) 188 (22.7) 0.243 (18.6) 146 (17.6) 0.535 (28.2) 220 (26.2) 0.262 (12.8) 174 (20.7) �0.0011 (45.4) 401 (47.7) 0.25� 11.6 51.0 � 12.1 0.14� 1.03 1.32 � 1.32 0.015

5 (79.0) 738 (87.8) �0.001� 19.3 142.8 � 28.8 �0.001� 23.4 154.7 � 45.8 �0.001

7 (21.9) 164 (19.5) 0.14� 0.67 3.62 � 0.60 �0.001

5 (26.1) 100 (11.9) 0.503 (12.4) 150 (17.8) �0.001

� 0.74 1.42 � 0.72 0.90� 1.12 1.87 � 1.10 0.002� 1.13 3.29 � 1.08 0.002� 28.2 100.6 � 27.6 0.62

0 (73.3) 563 (66.9) �0.0012 (95.4) 810 (96.3) 0.28

bA1n �

53

38

3111

63.162

1,92114

361948436630

1,0751.71.20

1,86125.0125.7

513.513129

1.421.733.1599.6

1,732,25

scular Society; HbA1c � hemoglobin A1c; LIMA � left internal-pump coronary artery bypass surgery; POD � postoperative day.

by on June 12, 2013 als.org

wmtpOoa

PAii2gmiosftmdrtaUibgsttwutc

OTHtcHv�rwSwnd((e

tm(ou

mdwdrdofbpcrdspu

SdSEspiStpwdpbwcmnsd

F(7t

1433Ann Thorac Surg HALKOS ET AL2008;86:1431–7 HEMOGLOBIN A1C AND SURVIVAL AFTER CABG

AD

ULT

CA

RD

IAC

ith standard techniques, utilizing roller head pumps,embrane oxygenators, cardiotomy suction, arterial fil-

ers, cold antegrade or retrograde, or both, blood cardio-legia, and moderate systemic hypothermia (32° to 34°C).ff-pump coronary artery bypass was performed with

ne of several commercially available cardiac positioningnd coronary artery stabilizing devices.

erioperative Glucose Controlll patients were treated with a uniform perioperative

ntravenous insulin protocol. In the operating room, annsulin infusion was premixed with 125 units of insulin in50 cc 0.9% normal saline. Routine measurement of bloodlucose was obtained from serial arterial blood gaseseasured every 30 minutes. In the intensive care unit,

nsulin infusion was continued and glucose levels werebtained from arterial blood gas samples or finger-stickamples every 2 hours. The insulin infusion was initiatedor blood glucose greater than 120 mg/dL and adjusted toarget intraoperative blood glucose between 80 and 110

g/dL according to the discretion of the attending car-iac anesthesiologist. In the intensive care unit, patientseceived a continuous insulin infusion that was adjustedo maintain blood glucose between 80 and 110 mg/dLccording to a sliding scale (blood glucose – 60 � 0.04 �nits of insulin per hour). Upon transfer out of the

ntensive care unit, most patients were monitored withlood glucose monitoring every 4 to 6 hours (goal bloodlucose � 150 mg/dL) and maintained according to theliding scale with subcutaneous insulin in addition toheir preoperative subcutaneous regimens. If necessaryo achieve glucose control, continuous insulin infusionas continued after transfer out of the intensive carenit. Newly diagnosed or poorly controlled diabetic pa-

ients also received endocrinology consultation for betterontrol in the postoperative period.

utcomeshe primary aim of this study was to determine whetherbA1c was an independent risk factor for all-cause long-

erm mortality. To this end, a total of 29 covariates wereollected to use as risk adjustors to ensure that the effect ofbA1c was not confounded by their influence. These co-

ariates are listed in Table 1 by HbA1c groupings (�7% and7%). The HbA1c classifications were chosen according to

ecommendations of the American Diabetes Association,hich recommends a target HbA1c of less than 7% [11, 12].tandard STS definitions of each risk factor and outcomeere used. Diabetes control method (diet, oral, insulin,one) was a nominal variable that was analyzed as separateichotomous variables. Canadian Cardiovascular Society

CCS) classification and New York Heart AssociationNYHA) classification are ordinal variables that were mod-led as discrete numerical variables.

To adjust for the effect of perioperative glucose control,wo different mean glucose values were computed and

odeled: the mean glucose value for the day of surgerypostoperative 0), and the mean glucose value for post-perative days 1 through 3. These two variables were

tilized in the Cox proportional hazards regression o

ats.ctsnetjournDownloaded from

odel. In this study, patients were classified as having aiagnosis of diabetes if they presented to the hospitalith a history of diabetes or if an established diagnosis ofiabetes was made preoperatively. Because this was aetrospective study, confirmatory testing to establish theiagnosis of diabetes using fasting blood glucose levelsr an oral glucose tolerance test was not routinely per-ormed. Patients who were diagnosed postoperativelyecause of persistent or refractory hyperglycemia oratients with undiagnosed diabetes preoperatively werelassified as nondiabetic. These definitions were made toeflect actual practice patterns where a preoperativeiagnosis of diabetes may potentially alter managementtrategies (preoperative insulin drip, more aggressiveostoperative management of hyperglycemia, decision totilize bilateral internal mammary arteries, and so forth).Long-term survival data was obtained using the Social

ecurity Death Index, which is a public use nationalatabase of death records extracted from the Unitedtates Social Security Administration’s Death Master Filextract. Persons who have died since 1963 who had aocial security number and whose death has been re-orted to the Social Security Administration will be listed

n the Social Security Death Index. The sensitivity of theocial Security Death Index (92.2%) is comparable with

hat of the National Death Index among American-bornersons (87% to 98%) [15]. Arrangements were madeith the Social Security Death Index whereby individualeath records for all patients in the study cohort wereurchased in bulk. Thus, for each patient who diedefore the cutoff date of March 31, 2007, a mortality dateas provided, allowing construction of Kaplan-Meier

urves and product-limit estimates of survival time toeasure 1-, 3-, and 5-year survival. Cause of death was

either considered nor available; the purpose of thistudy was to compare all-cause mortality according toiabetes and HbA1c classification.

ig 1. Kaplan-Meier 5-year overall survival by hemoglobin A1cHbA1c) stratification: 7.0% or greater (light line) versus less than.0% (heavy line). At 5 years, 87.6% of patients with HbA1c lesshan 7.0% survived, compared with 82.3% of patients with HbA1c

f 7.0% or greater (p � 0.001).

by on June 12, 2013 als.org

SCtAmcmaccicmucsatlbmt

me0mMdndaNSa

R

Ab7itT

Fssl

FdlA5

T

S

A

D

1434 HALKOS ET AL Ann Thorac SurgHEMOGLOBIN A1C AND SURVIVAL AFTER CABG 2008;86:1431–7A

DU

LTC

AR

DIA

C

tatistical Analysishecks for data quality are employed both at the insti-

utional level and before final entry into the STS Nationaldult Cardiac Database. To statistically evaluate theain effects of HbA1c in the presence of 29 potential

onfounders, a Cox proportional hazards regressionodel was constructed that modeled the mortality haz-

rd as a function of HbA1c and 29 other risk factorovariates. Of interest was whether HbA1c, modeled as aontinuous factor, was predictive of long-term mortalityn the presence of diabetes status and other potentialonfounders. Adjusted odds ratios were reported andeasured the additional increase in odds of outcome per

nit increase in HbA1c. Because of the inherent ineffi-iencies of such a large model (eg, artificially inflatedtandard errors, collinearity), a backward eliminationlgorithm was performed to determine which risk fac-ors, considered together, were significant predictors ofong-term survival. The backwards elimination approachegins with the full (saturated) model estimates (afterultiple imputation and parameter estimate combina-

ion), removes the least significant risk factor and fits the

ig 2. Kaplan-Meier 5-year overall survival by diabetes mellitustatus. At 5 years, 89.8% of patients without diabetes (light line)urvived, compared with 81.1% of patients with diabetes (heavy

able 2. Unadjusted Kaplan-Meier Survival Estimates: Effect

ubgroup 1-Year Survival

ll patientsHbA1c � 7.0 (n � 2,360) 0.968HbA1c � 7.0 (n � 841) 0.948Yes, diabetes (n � 1,285) 0.953No, diabetes (n � 1,916) 0.969iabetic patientsHbA1c � 7.0 (n � 538) 0.954HbA1c � 7.0 (n � 747) 0.952Insulin control (n � 399) 0.930Noninsulin control (n � 886) 0.963

ine; p � 0.001). g

ats.ctsnetjournDownloaded from

odel again. This process is sequentially repeated untilvery model term left in the model is significant at the � �.05 level. Hazard ratios were generated for each significantodel term along with 95% confidence intervals. Kaplan-eier curves were generated to determine observed

ifferences in HbA1c groups (�7.0, �7.0), diabetic (yes/o), HbA1c among diabetic patients, and insulin depen-ence among diabetic patients. The data were managednd analyzed using SAS version 9.1 (SAS Institute, Cary,orth Carolina) and STATA 9.0 (Stata Corp, Collegetation, Texas). All statistical tests were two-sided usingn � � 0.05 level of significance.

esults

total of 3,201 patients from January 1, 2002, to Decem-er 30, 2006, was included in this analysis (2,360 HbA1c �%; 841 HbA1c � 7%). Preoperative demographics, clin-cal variables, and perioperative factors comparing pa-ients with HbA1c greater than or less than 7 are listed inable 1. Forty-two percent of patients (538 of 1,285) with

ig 3. Kaplan-Meier 5-year overall survival among patients withiagnosis of diabetes mellitus according to hemoglobin A1c (HbA1c)ess than 7.0% (heavy line) versus 7.0% or greater (light line).mong diabetic patients, there was no significant difference in-year survival according to HbA1c less than 7.0% versus 7.0% or

emoglobin A1c (HbA1c) and Diabetes Mellitus on Survival

ar Survival 5-Year Survival Log-Rank p Value

0.922 0.876 0.0010.888 0.8230.877 0.811 �0.0010.937 0.898

0.859 0.801 0.370.890 0.8190.838 0.783 0.0060.894 0.824

of H

3-Ye

reater (p � 0.37).

by on June 12, 2013 als.org

d7dgmHdccatc

PPmtdoiamcfastcmp

FKvtDc

FHsHIdHAsti

ttm2n1cmtdisidepsfatcbds

C

T

Fdtyp0

TM

R

H

ACRCMSCP

1435Ann Thorac Surg HALKOS ET AL2008;86:1431–7 HEMOGLOBIN A1C AND SURVIVAL AFTER CABG

AD

ULT

CA

RD

IAC

iabetes were well controlled with HbA1c levels less than%. Of patients without an established diagnosis ofiabetes, 4.9% (94 of 1,916) had HbA1c levels of 7% orreater, suggesting undiagnosed and untreated diabetesellitus. Similarly, 11.2% of patients (94 of 841) withbA1c of 7% or greater had no previous history ofiabetes. Most patients in this study had multivesseloronary artery disease, more than 95% of patients re-eived arterial grafting with the left internal mammaryrtery or bilateral internal mammary artery grafting, andhe majority of cases were performed without the use ofardiopulmonary bypass.

revalence of Comorbiditiesatients with a preoperative HbA1c of 7% or greater wereore likely to have comorbid conditions and complica-

ions associated with diabetes, even if a prior diagnosis ofiabetes had not been made. Patients with HbA1c of 7%r greater were more likely to have preoperative renal

nsufficiency, renal failure requiring dialysis support, and higher baseline creatinine level. In addition, they wereore likely to have a history of cerebrovascular disease,

erebrovascular accident, hypertension, congestive heartailure, and peripheral vascular disease. These patientslso had a significantly higher number of diseased ves-els and received more bypass grafts. In addition, despitehe same protocols in place for perioperative glucoseontrol, patients with HbA1c of 7% or greater had higherean blood glucose levels on postoperative day 0 and

ostoperative days 1 through 3.

ive-Year Survival Outcomesaplan-Meier survival curves were constructed to pro-ide estimates of 5-year survival. The mean follow-upime for patients in this study was 2.81 � 1.40 years.iabetic patients had a significantly lower 5-year survival

ompared with nondiabetic patients (81.1% versus 89.8%;

ig 4. Kaplan-Meier 5-year overall survival among patients withiagnosis of diabetes mellitus according to treatment method: insulinherapy (light line) versus no insulin therapy (heavy line). At 5ears, 78.3% of diabetic patients treated with insulin survived, com-ared with 82.4% of diabetic patients treated without insulin (p �

a.006).

ats.ctsnetjournDownloaded from

ig 1, Table 2). Similarly, patients with a preoperativebA1c of 7% or greater had a significantly lower 5-year

urvival compared with patients with a preoperativebA1c less than 7% (82.3% versus 87.6%; Fig 2, Table 2).

n this unadjusted analysis, there was no 5-year mortalityifference observed among diabetic patients according tobA1c level of 7% or greater versus less than 7% (Fig 3).mong diabetic patients treated with insulin, 5-year

urvival was significantly worse compared with patientsreated with diet, oral hypoglycemic medication, or noth-ng (78.3% versus 82.4%; Fig 4).

To statistically evaluate the main effects of HbA1c inhe presence of 29 potential confounders, a Cox propor-ional hazards regression model was constructed that

odeled the mortality hazard as a function of HbA1c and9 other risk factor covariates. Although HbA1c is orga-ized into risk groups in Figures 1 through 4 and Tables and 2 for descriptive purposes, it was analyzed as aontinuous factor in the proportional hazards regressionodel (Table 3). Higher HbA1c percentage, adjusted for

he 29 covariates, was associated with an increased inci-ence of death (odds ratio � 1.15, p � 0.001) for each unit

ncrease in HbA1c. This corresponds with a statisticallyignificant 15% reduction in 5-year survival for each unitncrease in HbA1c. Furthermore, age, cerebrovascularisease, chronic lung disease, peripheral vascular dis-ase, renal insufficiency, congestive heart failure, andrevious myocardial infarction were all associated with aignificant reduction in 5-year survival. After adjustingor confounding variables in this Cox proportional haz-rds regression analysis, neither insulin treatment norhe diagnosis of diabetes was associated with a signifi-ant increase in long-term mortality. Furthermore, meanlood glucose on postoperative day 0 and postoperativeays 1 through 3 (Table 1) did not affect long-termurvival in the regression analysis.

omment

he prevalence of diabetes mellitus is increasing at an

able 3. Adjusted Hazard Ratios for Significantultivariable Predictors of Long-Term Survival

isk Factor

Adjusted HazardRatio (95%Confidence

Interval) p Value

emoglobin A1c (ascontinuous variable)

1.15 (1.06, 1.24) �0.001

ge 1.07 (1.06, 1.08) �0.001erebrovascular disease 1.62 (1.25, 2.09) �0.001enal insufficiency 1.84 (1.20, 2.83) 0.005ongestive heart failure 1.40 (1.07, 1.83) 0.015yocardial infarction 1.43 (1.12, 1.83) 0.005

erum creatinine 1.18 (1.10, 1.27) �0.001hronic lung disease 1.39 (1.21, 1.60) �0.001eripheral vascular disease 1.44 (1.10, 1.89) 0.009

larming rate and is a significant risk factor for cardio-

by on June 12, 2013 als.org

vlpohugircuw

ttbihtssTmHrc

HAgtAstiaUHc15w5vcws

dardpmHtaahl

wbsptcchsidShvc

sesabloirwtttcdttrC

taatketlotioamwtitcarTc

1436 HALKOS ET AL Ann Thorac SurgHEMOGLOBIN A1C AND SURVIVAL AFTER CABG 2008;86:1431–7A

DU

LTC

AR

DIA

C

ascular disease. The overall mortality from cardiovascu-ar disease is two to five times greater among diabeticatients than among nondiabetic patients [16]. Numer-us studies have documented the increased risk of in-ospital morbidity and mortality among diabetic patientsndergoing CABG [2, 3, 17]. However, perioperativelucose control using a continuous insulin infusion dur-

ng and after surgery has been shown to significantlyeduce the incidence of in-hospital mortality and majoromplications [5, 6, 18]. These results have been attrib-ted to the improved glycometabolic state associatedith strict glycemic control.The American Diabetes Association has recommended

he use of blood HbA1c as a method of assessing long-erm glycemic control in diabetic patients [3]. Hemoglo-in A1c is formed when glucose in the blood binds

rreversibly to hemoglobin to form a stable glycatedemoglobin complex. Because red cell turnover is con-

inuous (life span 90 to 120 days), HbA1c is not affected byhort-term glycemic lability, and thus allows better as-essment of glucose control over a 3 to 4 month period.he American Diabetes Association currently recom-ends that patients with diabetes aim for a targetbA1c of less than 7% [13], which is associated with a

educed incidence of macrovascular and microvascularomplications [11, 12].

In this study, we examined the impact of preoperativebA1c as a risk factor for 5-year mortality after CABG.mong patients with diabetes and with HbA1c of 7% orreater, 5-year survival was significantly lower comparedo patients without diabetes or with HbA1c less than 7%.mong diabetic patients, those treated with insulin had a

ignificantly lower 5-year survival compared to patientsreated without insulin. However, there was no differencen unadjusted 5-year survival among diabetic patientsccording to HbA1c stratification (�7% versus �7%).sing a Cox proportional hazard regression model,bA1c (as a continuous variable) emerged as a signifi-

ant predictor of reduced long-term survival (odds ratio.15, p � 0.001). This corresponds with a 15% reduction in-year survival for each unit increase in HbA1c. Otherell-known risk factors were also significant for reduced

-year survival including renal insufficiency, peripheralascular disease, age, congestive heart failure, andhronic lung disease. Diabetes and insulin treatmentere not significant predictors of reduced long-term

urvival in the logistic regression model.These findings have several implications. First, these

ata suggest that it may be the associated comorbiditiesnd complications of diabetes that are associated witheduced long-term survival, not simply the diagnosis ofiabetes or treatment methods. As seen in Table 1,atients with HbA1c of 7% or greater had significantlyore preoperative comorbidities than did patients withbA1c less than 7%. Although Figure 3 suggests that

here is no difference in survival among diabetic patientsccording to HbA1c level, this unadjusted Kaplan-Meiernalysis does not take into account the impact of theigher incidence of comorbidities in patients with HbA1c

evels of 7% or greater nor does it account for patients r

ats.ctsnetjournDownloaded from

ith undiagnosed diabetes. Therefore, these results maye misleading, as 11.2% of patients (94 of 841) in thistudy with preoperative HbA1c of 7% or greater had norevious or preoperative history of diabetes and were

herefore not included in this subgroup analysis. Weonclude that patients with good preoperative glucoseontrol, reflected by HbA1c less than 7%, may have aigher 5-year survival compared with patients havinguboptimal or poor glucose control (HbA1c � 7%). Themprovement in survival may be due to a reduction iniabetes-associated complications and comorbidities.pecifically, the adverse metabolic effects of prolongedyperglycemia as well as the associated arteriopathy andasculopathy may be attenuated in patients with well-ontrolled diabetes [5, 11].

Long-term survival after CABG has been the subject ofeveral investigations. Mohammadi and colleagues [9]xamined long-term cardiac-specific mortality on 9,125urvivors of CABG and found that cardiac survival wasdversely affected by the need for insulin therapy. Dia-etic patients not treated with insulin before surgery had

ong-term cardiac-related survival comparable with thatf nondiabetic patients. In a previous report from our

nstitution [3], patients with diabetes had a significantlyeduced 5- and 10-year survival compared with patientsithout diabetes. Among the diabetic cohort, insulin-

reated diabetes had a significantly higher 10-year mor-ality compared with patients treated with oral medica-ion or diet [3]. Similarly, in a report from Leavitt andolleagues [10], diabetic patients with peripheral vascularisease and renal insufficiency had markedly worse long-

erm survival compared with diabetic patients withouthese associated comorbidities. Diabetic patients withetinopathy also have reduced long-term survival afterABG [19].The results from our study are similar to the aforemen-

ioned studies with the exception of examining preoper-tive HbA1c as a risk factor in an adjusted multivariablenalysis. In the present study, we have provided a quan-ifiable risk factor, HbA1c, after adjusting for other well-nown risk factors, which can provide 5-year survivalstimates. Using the method of treatment or complica-ions associated with diabetes to determine the risk ofong-term mortality has limitations. First, the treatmentf diabetes has evolved substantially, so that many pa-ients are treated with different and combined regimensn an attempt to reach euglycemia [20]. The combinationf oral hypoglycemics with short- and long-term insulinnalogues as well as diet and lifestyle modification isore common in the current era. Thus, many patientsithout diabetes-related complications may still be

reated aggressively with insulin and insulin analoguesn addition to oral medication and lifestyle modifica-ion in an effort to prevent these well-known compli-ations. Finally, patients who achieve euglycemia havereduced incidence of diabetes-related complications,

egardless of their treatment modality [11–13, 16, 20].herefore, using a marker for long-term glycemicontrol such as HbA1c may provide a more accurate

eflection of long-term risk after CABG.

by on June 12, 2013 als.org

LTcvbopioptdTpduosc1adbCdwOtsiospriaagkt

rvcfidopt

TUSd

R

1

1

1

1

1

1

1

1

1

1

2

2

1437Ann Thorac Surg HALKOS ET AL2008;86:1431–7 HEMOGLOBIN A1C AND SURVIVAL AFTER CABG

AD

ULT

CA

RD

IAC

imitationshe retrospective nature of this study does not permitomplete accounting for all sources of bias, despite ad-anced statistical methodology designed to correct foroth treatment selection bias and potential confoundersf outcomes in the preplanned analyses. In this study,atients were classified as having a diagnosis of diabetes

f they presented to the hospital with a history of diabetesr if an established diagnosis of diabetes was madereoperatively. Patients with no known history of diabe-

es were classified as “nondiabetic” even if they wereetermined to be diabetic in the postoperative period.he definitions in this study were made to reflect actualractice patterns where the preoperative diagnosis ofiabetes was made according to the patient’s historypon presentation. Importantly, in this study as well asther retrospective analyses, how the presence or ab-ence of diabetes is determined has far-reaching impli-ations with regard to outcomes. This is relevant because1.2% of patients (94 of 841) in this study with preoper-tive HbA1c of 7% or greater had no previous history ofiabetes. With these patients, their diagnosis would haveeen overlooked without a preoperative HbA1c level.urrently, HbA1c is not recommended as a modality toiagnose diabetes, although it is apparent that patientsith an elevated HbA1c may in fact have diabetes [21].ther than preoperative HbA1c, routine confirmatory

esting in all preoperative patients to establish a diagno-is of diabetes was not performed. Therefore, the methodn which diabetics were defined may have influenced theutcomes in this study. An additional limitation of thistudy is that more than 70% patients underwent off-ump coronary artery bypass surgery, which does noteflect national practice patterns for surgical revascular-zation, although this variable was accounted for in thedjusted analysis. Finally, although we had preoperativend perioperative glucose control data, we do not havelucose control data after hospital discharge nor do wenow if there were treatment changes (ie, oral medica-ion to insulin) or changes in comorbidity status.

In conclusion, patients with diabetes and diabetes-elated complications appear to have worse long-term sur-ival after CABG. These outcomes may be more related toomplications associated with diabetes such as renal insuf-ciency and peripheral vascular disease than to simply aiagnosis of diabetes or insulin-treatment. Using a markerf glucose control such as a preoperative HbA1c mayrovide more accurate risk stratification to predict long-

erm outcomes of patients with diabetes mellitus.

he authors wish to express their gratitude to Clinical Researchnit Director Kim Baio for project oversight, to Jean Walker andusan Joyce for data abstraction, and to Deborah Canup foratabase management.

eferences

1. The BARI Investigators. Seven-year outcome in the Bypass

Angioplasty Revascularization Investigation (BARI) by treat-ment and diabetic status. J Am Coll Cardiol 2000;35:1122–9.

ats.ctsnetjournDownloaded from

2. Carson JL, Scholz PM, Chen AY, Petersen ED, Gold J,Schneider SH. Diabetes mellitus increases short-term mor-tality and morbidity in patients undergoing coronary arterybypass graft surgery. J Am Coll Cardiol 2002;40:418–23.

3. Thourani VH, Weintraub WS, Stein B, et al. Influence ofdiabetes mellitus on early and late outcomes after coronaryartery bypass grafting. Ann Thorac Surg 1999;67:1045–52.

4. Calafiore AM, Di Mauro M, Di Giammarco G, et al. Effect ofdiabetes on early and late survival after isolated first coro-nary bypass surgery in multivessel disease. J Thorac Cardio-vasc Surg 2003;125:144–54.

5. Furnary AP, Wu Y. Eliminating the diabetic disadvantage:the Portland Diabetic Project. Semin Thorac CardiovascSurg 2006;18:302–8.

6. Furnary AP, Gao G, Grunkemeier GL, et al. Continuousinsulin infusion reduces mortality in patients with diabetesundergoing coronary artery bypass grafting. J Thorac Car-diovasc Surg 2003;125:1007–21.

7. Rajakaruna C, Rogers CA, Suranimala C, Angelini GD,Ascione R. The effect of diabetes mellitus on patients under-going coronary surgery: a risk adjusted analysis. J ThoracCardiovasc Surg 2006;132:802–10.

8. Barsness GW, Peterson ED, Ohman EM, et al. Relationshipbetween diabetes mellitus and long-term survival after cor-onary bypass and angioplasty. Circulation 1997;96:2551–6.

9. Mohammadi S, Dagenais F, Mathieu P, et al. Long-termimpact of diabetes and its comorbidities in patients under-going isolated primary coronary artery bypass graft surgery.Circulation 2007;116(Suppl 1):220–5.

0. Leavitt BJ, Sheppard L, Maloney C, et al. Effect of diabetesand associated conditions on long-term survival after coro-nary artery bypass graft surgery. Circulation 2004;110:41–4.

1. The Diabetes Control and Complications Trial ResearchGroup. The effect of intensive treatment of diabetes on thedevelopment and progression of long-term complications ininsulin-dependent diabetes mellitus. N Engl J Med. 1993;329:977–86.

2. UK Prospective Diabetes Study (UKPDS) group. Intensiveblood-glucose control with sulphonylureas or insulin com-pared with conventional treatment and risk of complicationsin patients with type 2 diabetes (UKPDS 33). Lancet 1998;352:837–53.

3. American Diabetes Association. Standards of medical carein diabetes. Diabetes Care 2005;28(Suppl):4–36.

4. Halkos ME, Puskas JD, Lattouf OM, et al. Elevated preoper-ative hemoglobin A1c level is predictive of adverse eventsafter coronary artery bypass surgery. J Thorac CardiovascSurg 2008;136:631–40.

5. Schisterman EF, Whitcomb BW. Use of the Social SecurityAdministration Death Master File for ascertainment of mor-tality status. Popul Health Metr 2004;2:2.

6. Hammoud T, Tanguay JF, Bourassa MG. Management ofcoronary artery disease: therapeutic options in patients withdiabetes. J Am Coll Cardiol 2000;36:355–65.

7. Woods SE, Smith JM, Sohail S, Sarah A, Engle A. The influenceof type 2 diabetes mellitus in patients undergoing coronaryartery bypass graft surgery. Chest 2004;126:1789–95.

8. Furnary AP, Zerr KJ, Grunkemeier GL, Starr A. Continuousintravenous insulin infusion reduces the incidence of deepsternal wound infection in diabetic patients after cardiacsurgical procedures. Ann Thorac Surg 1999;67:352–62.

9. Ono T, Kobayashi J, Sasako Y, et al. The impact of diabeticretinopathy on long-term outcome following coronary arterybypass graft surgery. J Am Coll Cardiol 2002;40:428–36.

0. Nathan DM, Buse JB, Davidson MB, et al. Management ofhyperglycemia in type 2 diabetes: a consensus algorithm forthe initiation and adjustment of therapy. Diabetes Care2006;29:1963–72.

1. Saudek CD, Derr RL, Kalyani RR. Assessing glycemia indiabetes using self-monitoring blood glucose and hemoglo-

bin A1c. JAMA 2006;295:1688–97.

by on June 12, 2013 als.org

DOI: 10.1016/j.athoracsur.2008.06.078 2008;86:1431-1437 Ann Thorac Surg

Cooper, Cullen D. Morris, Robert A. Guyton and Vinod H. Thourani Michael E. Halkos, Omar M. Lattouf, John D. Puskas, Patrick Kilgo, William A.

Long-Term Survival After Coronary Artery Bypass SurgeryElevated Preoperative Hemoglobin A1c Level is Associated With Reduced

& ServicesUpdated Information

http://ats.ctsnetjournals.org/cgi/content/full/86/5/1431including high-resolution figures, can be found at:

References http://ats.ctsnetjournals.org/cgi/content/full/86/5/1431#BIBL

This article cites 21 articles, 8 of which you can access for free at:

Citations

shttp://ats.ctsnetjournals.org/cgi/content/full/86/5/1431#otherarticleThis article has been cited by 7 HighWire-hosted articles:

Subspecialty Collections

http://ats.ctsnetjournals.org/cgi/collection/coronary_disease disease

Coronary http://ats.ctsnetjournals.org/cgi/collection/cardiac_other Cardiac - other

following collection(s): This article, along with others on similar topics, appears in the

Permissions & Licensing

.healthpermissions@elsevier.comemail: orhttp://www.us.elsevierhealth.com/Licensing/permissions.jsp

in its entirety should be submitted to: Requests about reproducing this article in parts (figures, tables) or

Reprints reprints@elsevier.com

For information about ordering reprints, please email:

by on June 12, 2013 ats.ctsnetjournals.orgDownloaded from