Reduction of Surgical Mortality and Morbidity in Diabetic Patients Undergoing Cardiac Surgery With a Combined Intravenous and Subcutaneous Insulin Glucose Management Strategy

A retrospective review of 622 consecutive patients who underwent cardiothoracic (CT) surgery between 1 January 2005 and 31 December 2005 at Northwestern Memorial Hospital, a 776-bed tertiary care center in Chicago, Illinois, was performed to evaluate the incidence and treatment of postoperative hyperglycemia and operative morbidity and mortality. Out of 622 patients, 8 cases were excluded: 5 died during the initial procedure, 1 had a sternal wire removed and was not admitted to the hospital postoperatively, 1 had a cardiac laceration with hemopericardium and was not under the primary care of cardiac surgeons postoperatively, and 1 was a duplicate entry into the surgical database. After exclusions, the study population consisted of 614 patients. A predefined subgroup analysis was also performed on 159 consecutive patients who underwent CABG only.

Glucose data were collected from patient medical records. Outcomes data were extracted from the Society of Thoracic Surgeons Adult Cardiac Surgery Database, collected by Northwestern Memorial Hospital CT surgeons. Approval for this study was granted by the Northwestern University Institutional Review Board.

The primary outcome of this study was postoperative mortality, as defined in the Society of Thoracic Surgeons database. Secondary outcomes included the incidence of deep sternal wound infection, other infections, pulmonary complications, cardiac complications, renal complications, neurological complications, and readmission to the hospital within 30 days of the initial procedure (Table 1). If a patient had one or more complication within an outcome category, the category was flagged as a single complication.

Statistical analysis on demographic data were performed with one-way ANOVA (with Tukey test on all statistically significant results) using GraphPad InStat version 3.00 for Windows 95 (GraphPad Software, San Diego, CA). A χ² test or Fisher's exact test (when sample size was small) was used for univariate analysis to compare the incidence between the groups for the primary and secondary outcomes. Subsequently, multivariate analysis was performed using logistic regression to explore the relationship between statistically significant outcomes and the covariates of interest. The multivariate model included the covariates that were significant at the 0.1 level in the univariate analysis. A stepwise model selection technique was used to build the final model. A P value ≤0.05 was considered statistically significant. The statistical analysis was carried out using SAS software, version 9.1 (SAS Institute, Cary, NC).

The study cohort is described with respect to demographic characteristics, cardiac risk profile, and glycemic profile in Table 2, comparing those patients who had a preexisting diagnosis of diabetes (diabetic patients, n = 123, 20%) with those that did not (nondiabetic patients, n = 491, 80%). Significant differences between the diabetic and nondiabetic patients were fewer Caucasians, a higher average age, and a significantly higher BMI in the diabetic group. Traditional cardiac risk factors were significantly more common in the diabetic cohort, specifically hypertension, dyslipidemia, and previous smoking history. Preoperative comorbid illnesses were also present more commonly in the diabetic cohort (specifically chronic lung disease, renal failure, and history of myocardial infarction). Nonelective cardiac surgery was performed more often in the diabetic cohort. Preoperatively, outpatient anti-hyperglycemic regimens for the diabetic cohort included insulin (27%), oral hypoglycemic agents (59%), diet only (6%), and no treatment (8%).

Postoperatively, hyperglycemia (glucose level >6.05 mmol/l) was almost universal (98% diabetic vs. 99% nondiabetic patients) in the ICU. Independent of diabetic status, ICU admission glucose ≥11 mmol/l was predictive of higher rates of postoperative mortality (6.1 vs. 2.3%; P = 0.04) and renal (10.7 vs. 3.9%; P = 0.02), pulmonary (21.4 vs. 11.0%; P = 0.02), and cardiac (32.1 vs. 21.8%; P = 0.02) complications.

In diabetic compared with nondiabetic patients, significantly higher values were found for the mean glucose upon ICU admission, the overall mean glucose in the ICU, and the overall mean glucose after transfer out of the ICU (Table 2). The IV insulin protocol was initiated on 95% of the diabetic and 94% of the nondiabetic patients, and the SC insulin protocol was subsequently initiated on 94% of the diabetic and 88% of the nondiabetic patients. Among all CT surgeries, diabetic patients had higher rates of postoperative mortality (7.3 vs. 3.3%; P = 0.03) and pulmonary complications (19.5 vs. 11.6%; P = 0.02) but had similar rates of deep sternal wound infections, other infections, and cardiac, renal, and neurological complications than nondiabetic patients on univariate analysis (Fig. 1). However, on multivariate analysis, it was found that the presence of diabetes was not a significant factor in either pulmonary complications or mortality. In this multivariate analysis, renal failure, nonelective surgery, and older age were positively associated with pulmonary complications, and renal failure, nonelective surgery, and female sex were positively associated with postoperative mortality independently of the patient's diabetic status.

Subgroup analysis was also performed on patients who underwent CABG surgery (n = 159; 25.9%) without any other simultaneous surgical procedures. The subgroup consisted of 48 diabetic patients (30.2%) and 111 nondiabetic patients (69.8%). Table 3 describes the demographics, cardiac risk profile, and comorbid illnesses of the CABG-only subgroup. Similar to the entire cohort, diabetic patients undergoing CABG surgery had higher rates of hypertension, dyslipidemia, preexisting renal failure, and chronic lung disease. The nondiabetic cohort had a higher percentage of previous smokers.

In the CABG-only subgroup, use of both IV and SC insulin protocols was similar among diabetic and nondiabetic patients. Mean glucose levels on ICU admission, in the ICU, and outside the ICU were higher in diabetic patients (Table 3). In the CABG-only cohort, no statistically significant differences were seen in primary or secondary outcomes between diabetic and nondiabetic patients using univariate analysis (Fig. 2). No postoperative deaths occurred in this cohort, and there was only one deep sternal wound infection.

The major complication of intensive insulin therapy in the perioperative period is hypoglycemia. Out of 37,626 glucose values reviewed in our study, only 465 (1.2%) were <3.3 mmol/l. Rates were higher in individuals with a prior history of diabetes compared with individuals without such a history in the total group (1.7 vs. 1.1%, P < 0.0001) and those undergoing CABG only (1.2 vs. 0.6%, P = 0.012). None of these episodes were considered severe, and none resulted in loss of consciousness or seizure.

Hyperglycemia has been shown to be an independent risk factor for poor clinical outcomes in multiple patient populations (8–10). Similarly, we found that a glucose level on ICU admission ≥11 mmol/l was predictive of higher rates of mortality and postoperative complications (renal, pulmonary, and cardiac).

Diabetic patients undergoing CT surgery have been shown to have higher rates of mortality and postoperative complications than nondiabetic patients (4,11,12). Our study demonstrates that glucose management with a combined IV and SC insulin regimen eliminates this known increased postoperative morbidity and mortality in patients with preexisting diabetes not only in those undergoing CABG, similar to what has been shown with 3 days of continuous IV insulin infusion (3,6,13,14) but in all CT surgeries. It should be noted that the diabetic patients in our study had postoperative morbidity and mortality rates similar to those without a prior diagnosis of diabetes despite having higher age, BMI, traditional cardiac risk factors (hypertension, dyslipidemia, and smoking history) and more preexisting comorbid illnesses (chronic lung disease, chronic renal insufficiency, previous myocardial infarction, and previous stroke). Although postoperative pulmonary complications and mortality were statistically significant in the diabetic cohort for all cardiac surgery on univariate analysis, they were found to be unrelated to diabetes status on multivariate analysis. The difference in pulmonary complications between diabetic and nondiabetic patients was attributable to higher rates of prolonged ventilation and need for reintubation, and these were positively associated with preexisting renal failure, nonelective surgery, and older age independently of the presence of diabetes. Postoperative mortality was associated with preexisting renal failure, nonelective surgery, and female sex. Mean glucose levels in the immediate postoperative period were slightly higher in the diabetic cohort, but this calculated mean is influenced by the higher mean glucose on ICU admission.

Of those individuals not known to have diabetes, 99% were found to be hyperglycemic postoperatively in our study, with 94% requiring an IV insulin drip for control. Critical illness causes an impairment of insulin secretion and insulin action, resulting in hyperglycemia, even in normal individuals, and a worsening of the hyperglycemia in patients with diabetes (10,13–17). Biochemical studies suggest that mortality is reduced with improved glycemic control due to favorable alterations in myocardial and skeletal muscle metabolism, improved cell membrane stability, myocardial contractility, and endothelial cell function, as well as increased nitric oxide and decreased inflammatory mediators, platelet aggregation, and thrombosis (18–20).

Previously, in a prospective randomized controlled study, Van den Berghe et al. (10) showed that normalization of elevated glucose levels by intensive insulin therapy in patients in the surgical ICU decreased in-hospital mortality by 34%. Post-CABG mortality in diabetic patients undergoing CABG only has been reported to be between 2 and 4% (21–23), but continuous insulin infusion for 72 h postoperatively has been shown to reduce this higher risk (6). In a study of 3,554 diabetic patients undergoing isolated CABG surgery, Furnary et al. (6) reported a mortality rate of 2.5% in patients treated with a 72-h postoperative insulin infusion (mean glucose 9.73 ± 1.65 mmol/l) compared with 5.3% of patients treated with SC insulin therapy (mean glucose 11.71 ± 2.25 mmol/l). Although Furnary et al. did not study the nondiabetic population, they reported an overall CABG mortality rate of 2.8% (n = 14,051) at their institution. In comparison, in our study of 159 CABG patients, postoperative mortality was nonexistent in both the diabetic and nondiabetic groups.

It is difficult to perform inter-institutional comparison because of differences in surgical experience and techniques. However, we were able to compare our observed mortality results of 0% for both patient groups undergoing CABG surgery with those from the national Society of Thoracic Surgeons 2005 database. According to this database, which adjusts for the various risk factors in our patients, the expected mortality for the group without a prior diagnosis of diabetes was 1.30% (95% CI 0–3.75%), for those with a prior diagnosis of diabetes it was 2.1% (0–8.24%), and for the entire group of 159 patients undergoing CABG it was 1.50% (0–2.65%).

Complications of intensive insulin therapy in our patients were very low. Despite over 90% of patients being treated intensively with insulin therapy, hypoglycemia (glucose <3.3 mmol/l) rates remained low but were higher (1.7 vs. 1.1%, P < 0.0001) in individuals with a prior diagnosis of diabetes. None of the hypoglycemic values were deemed severe, and none were associated with loss of consciousness or seizures. Nondiabetic patients were often titrated off SC insulin therapy on postoperative days 4 or 5 because of improvements in postoperative insulin resistance and normalization of blood glucose. The difference in mean glucose levels outside of the ICU can also be explained by the need for continuing to monitor euglycemic nondiabetic patients after cessation of insulin therapy.

There are limitations to our study. Although it was a prospectively designed protocol, it was not a randomized trial comparing different treatment regimens, and there was no control group. We also could not compare surgical complication rates before and after institution of the glycemic protocols because of concomitant changes in the surgeons and changes in a number of aspects of the surgery and perioperative care. Another limitation is the relatively small numbers of patients reported for this single year period. Finally, some of the patients designated as not having diabetes could possibly have had diabetes; it is generally estimated that ∼30% of people with diabetes have not been diagnosed (24). Based on the number of patients in our series with diagnosed diabetes, we can estimate that perhaps an additional 54 (11.0%) patients of the “nondiabetic” group for all CT surgery and an additional 15 (13.5%) patients of the “nondiabetic” group undergoing CABG may actually have had diabetes.

In conclusion, in this study, diabetic patients had similar rates of postoperative mortality, deep sternal wound infections, other infections, and pulmonary, cardiac, renal, and neurological complications compared with nondiabetic patients after CT surgery. Thus, for the first time, we have shown that a combination of IV insulin (in the ICU) and SC insulin (outside the ICU), a less costly and less nursing-intensive therapy than 3 days of IV insulin postoperatively, results in a reduction of the increased surgical morbidity and mortality in diabetic patients after CT surgery that has been previously described in the literature.

This work was supported in part by an investigator-initiated grant from Sanofi-Aventis.

The authors wish to thank the Endocrinology Fellows who participated in the care of the patients, Drs. Herman Blomeier, Allison Hahr, and Rehka Ramamurthy, MD; the nursing staffs of the Cardiovascular Intensive Care Unit; and the general surgical nursing staff at Northwestern Memorial Hospital. Furthermore, the authors would like to acknowledge Dr. Jie (Jenny) Huang and Nancy Jin of the Northwestern University Biostatistics Collaboration Center for their assistance with statistical analysis and Stacie Landon, RN, MS, Society of Thoracic Surgeons and ICD Database Manager, Clinical Trials Unit, Bluhm Cardiovascular Institute, Northwestern University Feinberg School of Medicine.

This work was presented in part at the 66th Annual Meeting of the American Diabetes Association, Washington, DC, June, 2006.