Author + information
- Received October 10, 2001
- Revision received January 31, 2002
- Accepted April 30, 2002
- Published online August 7, 2002.
- Jeffrey L. Carson, MD*,* (, )
- Peter M. Scholz, MD†,
- Anita Y. Chen, MS‡,
- Eric D. Peterson, MD, MPH, FACC‡,
- Jeffrey Gold, MD, FACC§∥ and
- Stephen H. Schneider, MD¶
- ↵*Reprint requests and correspondence:
Dr. Jeffrey L. Carson, Robert Wood Johnson Medical School, 125 Paterson Street, New Brunswick, New Jersey 08903, USA.
Objectives The aim of this study was to determine the impact of diabetes mellitus (DM) on short-term mortality and morbidity in patients undergoing coronary artery bypass surgery (CABG).
Background Diabetes mellitus is present in approximately 20% to 30% of patients undergoing CABG, and the impact of diabetes on short-term outcome is unclear.
Methods We performed a retrospective cohort study in 434 hospitals from North America. The study population included 146,786 patients undergoing CABG during 1997: 41,663 patients with DM and 105,123 without DM. The primary outcome was 30-day mortality. Secondary outcomes were in-hospital morbidity, infections and composite outcomes of mortality or morbidity and mortality or infection.
Results The 30-day mortality was 3.7% in patients with DM and 2.7% in those without DM; the unadjusted odds ratio was 1.40 (95% confidence interval [CI], 1.31 to 1.49). After adjusting for other baseline risk factors, the overall adjusted odds ratio for diabetics was 1.23 (95% CI, 1.15 to 1.32). Patients treated with oral hypoglycemic medications had adjusted odds ratio 1.13; 95% CI, 1.04 to 1.23, whereas those on insulin had an adjusted odds ratio 1.39; 95% CI, 1.27 to 1.52. Morbidity, infections and the composite outcomes occurred more commonly in diabetic patients and were associated with an adjusted risk about 35% higher in diabetics than nondiabetics, particularly among insulin-treated diabetics (adjusted risk between 1.5 to 1.61).
Conclusions Diabetes mellitus is an important risk factor for mortality and morbidity among those undergoing CABG. Research is needed to determine if good control of glucose levels during the perioperative time period improves outcome.
Diabetes mellitus (DM) is a major risk factor for cardiovascular disease, and arteriosclerosis is responsible for 80% of deaths in patients with DM. Approximately 20% to 30% of patients undergoing coronary artery bypass surgery (CABG) have DM. With one exception (1)studies are consistent in documenting a 50% to 90% increase in long-term mortality rates in diabetics (2–8). This most likely is a result of accelerated arteriosclerosis associated with diabetes.
The impact of diabetes on short-term mortality and morbidity in patients undergoing CABG is unclear. The best evidence (9)suggests that in-hospital mortality is elevated in DM, although the results are inconsistent (3,6,7,10–17). Only one study evaluated the cause of long-term mortality (3). Most studies were from a single institution, and few studies were large enough to comprehensively evaluated postoperative morbidity.
We performed a large multicenter cohort study using detailed information on preexisting illness, cardiac status, mortality and morbidity collected at 434 institutions under the direction of the Society of Thoracic Surgeons. The aims of our analysis were to determine if DM is independently associated with 30-day mortality and morbidity and to characterize the causes of death.
Study design and population
We performed a retrospective cohort study of patients undergoing coronary artery bypass surgery from the U.S. and Canada that are included in the 1997 STS national database. We included patients undergoing both first and reoperations. We excluded patients who underwent simultaneous valvular heart surgery, patients with DM treated with diet alone, or in whom the treatment was unknown, and those with missing information on the presence of DM.
The primary outcome was 30-day mortality. Secondary outcomes were in-hospital morbidity, infections and composite outcomes of mortality or morbidity and mortality or infection.
We defined in-hospital infection as deep sternum infection, leg infection, septicemia, urinary tract infection or pneumonia. The criteria for a deep sternum infection of the muscle, bone and/or mediastinum included either wound opened with excision of tissue, positive culture or treatment with antibiotics. The criteria for infection of the leg vein harvest site included either wound opened with excision of tissue, positive culture or treatment with antibiotics. The criteria for septicemia required positive blood culture. The criteria for urinary tract infection required a positive urine culture. The criteria for pneumonia were positive cultures of sputum, blood, pleural fluid, empyema fluid, transtracheal fluid or transthoracic fluid, consistent with the diagnosis and clinical findings of pneumonia. Pneumonia is also defined as a chest radiograph diagnostic of pulmonary infiltrates.
We defined in-hospital morbidity as infections (as defined above), myocardial infarction (MI), renal failure, stroke or multisystem failure. Postoperative MI required two of the following four criteria: prolonged (>20 min) typical chest pain not relived by rest and/or nitrates; enzyme level elevation with either creatine kinase (CK)-MB >5% of total CK, CK greater than twice normal, lactase dehydrogenase (LDH) subtype 1 > LDH subtype 2, or troponin >0.2 μg/ml; new wall motion abnormalities; or serial electrocardiogram (at least two) showing changes from baseline or serially in ST-T and/or Q waves that are 0.03 s in width and/or plus one-third of the total QRS complex in two or more contiguous leads. Renal failure required either an increase of serum creatinine >2.0 mg/dl, a 50% or greater increase in creatinine over the baseline preoperative value or a new requirement for dialysis. Stroke was defined as a persistent central neurological deficit lasting >72 h. Multisystem failure was defined as two or more major organ systems suffering compromised function.
Definition of DM
We defined DM as a history of diabetes currently receiving treatment with either oral medications or insulin. We further subclassified DM based on method of glucose control, including oral medication or insulin, at time of surgery.
The data were collected as part of the National Database of the Society of Thoracic Surgeons, including approximately 65% of all cardiac surgery programs across North America. Data were collected using standardized data collection instruments and prospectively defined explicit data definitions. The data are harvested semiannually and electronically transmitted to the Duke Clinical Research Institute. It is then cleaned for predetermined standards and analyzed. Data managers receive biannual training and updates on STS data definitions and procedures. Sites also receive biannual reports on data quality and completeness. The data collection instruments included questions regarding demographic characteristics, preoperative risk factors, previous interventions, preoperative cardiac status, cardiac catheterization results, medications, intraoperative management and postoperative complications.
All statistical analyses were performed using SAS version 8.0. First, we used simple 2 × 2 tables to examine the univariate relation between diabetes status and each outcome. Then we used multivariate analyses to examine the independent effect of DM after controlling for other known baseline predictors of surgical risk for potential confounding. We performed separate multiple logistic regression models for the primary outcome and each secondary outcome. For the first set of multivariate analyses, we examined the diabetes dichotomous variable to determine its significance after adjusting for covariates. For the second set of multivariate analyses, we dichotomized diabetes patients into two treatment groups—patients treated with medications or patients treated with insulin. We then examined the differential effect of each diabetes treatment group to determine its significance after adjusting for covariates. We did not examine confounding using propensity scores because developing a model predicting the presence or absence of DM had little face validity.
All models included independent patient variables found to be risk factors for mortality in previous modeling efforts in this database (17). These variables included: demographic (age, gender, race), preoperative risk factors (renal failure, renal failure-dialysis, cerebrovascular accident—when, chronic lung disease, peripheral vascular disease, cerebrovascular disease, last creatinine preop, body surface area), previous interventions (number of prior cardiac operations requiring bypass, prior percutaneous catheter angioplasty/including balloon, atherosclerosis, and/or stent interval), preoperative cardiac status (congestive heart failure, MI, cardiogenic shock, arrhythmia, New York Heart Association classification), preoperative medications (diuretics, corticosteroids, digitalis, intravenous nitrates), preoperative hemodynamics and catheterization hemodynamic data (ejection fraction, number of diseased coronary vessels, left main disease >50%), operative (status of the procedure), cardiopulmonary bypass and support (intra-aortic balloon pump).
The data were collected as part of the National Database of the Society of Thoracic Surgeons after approval by the Institutional Review Board of Duke University.
There were 156,046 patients who underwent coronary artery bypass surgery. We excluded 4,564 patients with DM treated only with diet, 2,143 patients in whom the treatment was unknown and 2,553 patients who were missing information on DM status. The final study population included 146,786 patients, 105,123 without DM and 41,663 (28.4%) patients with DM. Of the patients with DM, 16,660 (40.0%) used insulin, and 25,003 (60.0%) used oral medications. The mean age of the study population was 65.0 (SD ± 10.7). The clinical characteristics of the study population are described in Table 1.
Overall, the 30-day mortality was 3.74% in patients with DM and 2.70% in those without DM (Table 2); the unadjusted odds ratio was 1.40 (95% confidence interval [CI], 1.31 to 1.49). After adjusting for other baseline risk factors, the adjusted odds ratio was 1.23 (95% CI, 1.15 to 1.32). The impact of DM on mortality is similar to gender (adjusted odds ratio for female gender = 1.31), race (adjusted odds ratio for Caucasian = 0.89), peripheral vascular disease (adjusted odds ratio = 1.36) and chronic lung disease (adjusted odds ratio = 1.33) but is less important than preexisting renal failure or dialysis (adjusted odds ratio = 1.73), preoperative cardiogenic shock (adjusted odds ratio = 2.68) and reoperation (adjusted odds ratio = 2.87).
When the analysis was stratified by diabetes treatment and adjusted for confounders, patients treated with insulin (adjusted odds ratio 1.39; 95% CI, 1.27 to 1.52) had a significant elevation in risk associated with 30-day mortality (Table 3). Those on oral medications also had a significant, but small, increase in 30-day mortality (adjusted odds ratio 1.13; 95% CI, 1.04 to 1.23).
The distributions of causes of death are described in Table 4. The majority of patients died from cardiac causes. The other most common causes of death were neurologic, pulmonary disease and infection. Compared with patients without DM, infections were the cause of death more often in patients with DM treated with insulin (p = 0.033). However, infections were not a more common cause of death in patients with DM (treated with oral medication and insulin) than patients without DM (p = 0.22). Neurologic causes of death occurred more frequently in DM patients overall (p = 0.01) and in DM patients treated with oral medications (p < 0.005).
Morbidity, infection and composite outcomes
Morbidity (13.9% vs. 9.1%), infections (7.9% vs. 5.2%) and the composite outcomes of death or morbidity (15.5% vs. 10.4%) and death or infection (10.6% vs. 7.2%) occurred more commonly in diabetic patients (Table 2). For each of these outcomes, the adjusted risk was about 35% higher in diabetics than nondiabetics (Table 3, A and B). Diabetic patients using insulin had the highest risk of mortality and morbidity compared with diabetic patients who were treated with oral medications (Table 3, A and B). The adjusted risk for each of these outcomes was 50% to 61% higher for insulin-treated diabetics compared with nondiabetics for each outcome.
Length of hospital stay
The median length of hospital stay was 7.0 days (interquartile range; 5, 10) in patients without DM and 8.0 (interquartile range; 6, 11) in patients with DM (p < 0.001). The length of hospital stay was 7.0 days (interquartile range; 6, 11) in patients with DM treated with oral medication and 8.0 days (interquartile range; 6, 12) in patients with DM treated with insulin. The length of stay was significantly longer in patients treated with insulin compared with patients treated with oral medication and patients without DM (p < 0.001).
Summary of findings
This study of 146,786 patients undergoing coronary artery bypass surgery found that patients with DM had a 23% to 37% increase in 30-day mortality and in-hospital morbidity compared with patients without DM. The increased risk of death (39%) or postoperative complications (50% to 61%) occurred most commonly in diabetic patients treated with insulin. Cardiac disease was the most common mode of death in DM patients, although neurological, pulmonary and infectious disorders were relatively common modes of death.
Previous studies had conflicting results, although most did not identify a significantly elevated risk of death in patients with DM (3,6,7,10–12). Two studies have found associations between short-term mortality and DM. One study based on data from 25 years ago found higher mortality in DM with preserved left ventricular function, although no differences were found in patients with poor ventricular function, and confounding was not adjusted for (13). A recent large study included 2,278 patients with DM and 9,920 patients without DM (9). Short-term mortality was significantly higher in DM (3.9% vs. 1.6%) after controlling for other risk factors for death. This study was limited by the fact that data were not contemporary (1978 to 1993), year of surgery was not controlled for, infections were not reported and cause of death was not evaluated. Nearly all of the studies that developed predictive indexes for mortality after CABG surgery included DM in the regression model (14–17).
This is the first study to contrast the causes of death in patients with diabetes and patients without diabetes. Most patients died from cardiac disease, although neurological causes were very common. Infection was a more common cause of death only in insulin-treated diabetics compared with patients without DM.
This study further clarifies the risk of postoperative complications in patients with diabetes undergoing coronary artery bypass surgery. Prior studies that evaluated postoperative complications largely focused on individual diseases and were much too small to detect differences in morbidity outcomes (6,7,9–12,18–20). We chose to evaluate complications grouped together because most individual diseases were very uncommon. We demonstrate that the risk of infection and other serious life-threatening complications is 36% to 38% higher in diabetics after adjusting for differences in risk factors. Similar to the analysis of death, insulin-treated diabetics had the highest risk of serious complications.
Possible explanations for poor outcome in DM
There are a number of possible explanations for the relation between DM and increased mortality and morbidity after CABG. The most obvious explanation is that patients with DM have more comorbidity or more advanced cardiac disease at the time of surgery. While we controlled for many known risk factors previously demonstrated to be associated with mortality (demographic characteristics, preoperative risk factors including most common comorbidities, previous interventions, preoperative cardiac status, preoperative medications, preoperative hemodynamics and catheterization data, and operative information), it is still possible there is residual confounding. Other potential limitations include incomplete 30-day follow-up for mortality, although a validation study suggests that this information is accurate (21), and difficulty in distinguishing pneumonia from congestive heart failure in postoperative coronary artery bypass surgery patients. Neither of these potential problems should have biased the results of the study because it is unlikely that there are differential misclassifications between patients with and without DM.
It is also interesting to consider the possibility that the metabolic abnormalities associated with DM are responsible for some of the increased mortality and morbidity. Dehydration and electrolyte disturbances as a result of uncontrolled hyperglycemia could contribute. Free fatty acids levels are elevated after major surgery and could suppress cardiac function, increase myocardial oxygen demand and may be arrhythmogenic (22–25).
Hyperglycemia per se could impact on perioperative mortality and morbidity by a number of mechanisms. Hyperglycemia interferes with the function of polymorphonuclear leukocytes predisposing to infection and may impair wound healing (10,11,26–29). Some studies suggest a relation between improved glucose control in the perioperative period and lower rates of wound infection and dehiscence (30–32). Hyperglycemia could contribute to increased platelet activity and disordered coagulation and fibrinolytic function (33)as well as abnormalities in lipid metabolism. Hyperglycemia may also adversely affect endothelial function (34). The few clinical trials of more intensive insulin therapy of hyperglycemia during and after a MI (35–38)and in intensive care unit patients (39)suggest improved outcomes.
Our data suggest that patients with DM are at significantly greater risk of death or suffering a serious postoperative complication when compared with nondiabetics. Diabetic patients represent 28% of all patients undergoing CABG. While the absolute difference in mortality and morbidity between patients with and without DM is modest, the absolute difference is substantial when comparing patients with insulin-treated DM to those without DM; the difference in mortality is 1.9%, mortality or morbidity is 8.2%, and mortality or infection is 5.4%. There are also significant healthcare costs associated with these poor outcomes (40).
☆ Support was provided internally by the Society of Thoracic Surgery National Cardiac Database.
- coronary artery bypass grafting
- confidence interval
- creatine kinase
- diabetes mellitus
- lactase dehydrogenase
- myocardial infarction
- Received October 10, 2001.
- Revision received January 31, 2002.
- Accepted April 30, 2002.
- American College of Cardiology Foundation
- Herlitz J.,
- Wognsen G.B.,
- Emanuelsson H.,
- et al.
- Barsness G.W.,
- Peterson E.D.,
- Ohman E.M.,
- et al.
- Kuan P.,
- Bernstein S.B.,
- Ellestad M.H.
- Peterson E.D.,
- Alexander K.P.,
- O’Connor G.T.,
- et al.
- Henderson A.H.,
- Most A.S.,
- Parmley W.W.,
- Gorlin R.,
- Sonnenblick E.H.
- Pomposelli J.J.,
- Baxter J.K. III.,
- Babineau T.J.,
- et al.
- Williams S.B.,
- Goldfine A.B.,
- Timimi F.K.,
- et al.
- Fath-Ordoubadi F.,
- Beatt K.J.
- Malmberg K.
- Gwilt D.J.,
- Petri M.,
- Lamb P.,
- Nattrass M.,
- Pentecost B.L.
- Van den Berghe G.,
- Wouters P.,
- Weekers F.,
- et al.