Diabetes and Coronary Artery Bypass Surgery: An examination of perioperative glycemic control and outcomes

We conducted a retrospective review of medical records at the University of Alberta Hospital, Edmonton, Canada (a 624-bed tertiary care hospital where all CABG procedures in central and northern Alberta, a catchment area of 1.6 million people, are performed). All patients undergoing CABG at the University of Alberta Hospital in the 12-month period between 1 April 2000 and 31 March 2001 and having a concomitant diagnosis of diabetes were identified (n = 306, 34% of all CABG procedures performed that year). After exclusion of four patients who died intraoperatively or within the first 24 h after CABG, all available charts (n = 291, 96%) were examined and data were abstracted using prestandardized data collection sheets and definitions. Abstraction errors were minimized through the use of detailed definitions and by reabstraction of a random sample of the medical charts by a second abstracter.

We decided a priori to evaluate adequacy of glycemic control by examining: 1) average capillary blood glucose measurements on each of postoperative days 1–5; 2) the proportion of patients with average blood glucose level >6.1 mmol/l, chosen because this was the threshold value for the intensive treatment arm in the Belgian trial (16); and 3) the proportion of time each patient had blood glucose >11.0 mmol/l, chosen because blood glucose above this level correlates with an increased mortality risk in diabetic patients after acute coronary syndromes (5,6).

Adverse outcomes before hospital discharge were defined as death, myocardial infarction, stroke, and septic complications (such as wound infections) and were extracted directly from the medical charts. Myocardial infarction was defined as any one of the following: new Q wave(s) or disappearance of R wave(s) on postoperative electrocardiograms, any new and persistent regional wall motion abnormalities on postoperative echocardiography (performed in only a minority of these patients), CK-MB isoenzyme levels ≥100 IU/l within 48 h of the operation, or troponin I level ≥3.9 μg/l within 24 h of the operation (18). Stroke was defined as a new focal or global disturbance of cerebral function lasting longer than 24 h with no apparent nonvascular cause and proven by computed tomography or magnetic resonance imaging. Although silent myocardial infarction and stroke were not routinely sought in these patients, specially trained infection control nurses do perform daily postoperative wound surveillance in CABG patients at our institution and any wound infections at either the sternal or saphenous graft sites are recorded (using the Centers for Disease Control and Prevention definitions) (19). The accuracy of end point assessment was confirmed in a randomly chosen subset (n = 30, 93% agreement, κ = 0.83) by the senior investigator, independent of the initial abstraction and blinded to perioperative glycemic control.

In addition to reporting summary statistics for the overall sample, we compared the frequency (for categorical data) and means (for continuous variables) between those patients who did and did not experience an adverse outcome (death, myocardial infarction, stroke, or septic complication). We also examined the perioperative glucose management and degree of glycemic control in these two groups. We examined differences between the two groups using χ² tests for categorical variables and Student’s t test for continuous variables. Due to the multiple comparisons being made, we used the Bonferroni correction and a priori established statistical significance at P < 0.005 for the comparisons (shown in Table 1). Multiple linear regression models were used to examine the relationship between a variety of baseline characteristics and glycemic control (the backward stepwise technique was used with P to remove set at 0.15; statistical significance was set at P < 0.05). We examined outcome rates in quartiles of glycemic control and used multiple logistic regression to examine the association between clinical and operative factors, perioperative glycemic control, and postoperative outcomes (the forward stepwise technique was used with P to remove set at 0.15 and statistical significance at P < 0.05). All analyses were performed using SPSS statistical software (version 11; SPSS, Chicago, IL).

Clinical features and medication usage at baseline for all 291 patients are outlined in Table 1. Virtually all of our patients had type 2 diabetes; median duration of diabetes was 10 years; and 40% of patients had known diabetic microvascular complications at the time of surgery. Median duration of hospital stay was 7 days (interquartile range 6–13 days); median time to complication was 6 days (interquartile range 4–11 days). After exclusion of the four patients who died intraoperatively or within the first 24 h, adverse outcomes during the initial hospitalization occurred in 78 (27%) of the 291 remaining patients: 8 (3%) died, 7 (2%) suffered nonfatal stroke, and 63 (22%) had a septic complication (57% of which were wound infections). As can be seen in Table 1, adverse outcomes after surgery were more common if the surgery was emergent (P < 0.001), and patients suffering adverse outcomes had higher BMI (P = 0.003) and had undergone longer operations (P = 0.003).

Most of these patients (92%) were managed with intravenous insulin infusion and a standardized protocol whereby the infusion rate (or insulin concentration) was adjusted based on serial glucose measurements (Table 2). This standardized protocol had been developed several years earlier by the Division of Endocrinology and the Department of Cardiovascular Anesthesia and Intensive Care at the University of Alberta and has been used routinely in our cardiovascular ICU and wards since the late 1990s. In addition, almost four-fifths of these patients were seen by an endocrinologist during their admission (9% were managed by an endocrinologist throughout their admission, including preoperatively). Patients started on insulin preoperatively were less likely (P = 0.01) to experience an adverse outcome, but there were no other statistically significant differences in management between those patients who did and those who did not experience an adverse outcome (Table 2). The association between preoperative insulin and outcomes disappeared after controlling for the lower use of preoperative insulin in emergent operations.

Glycemic control in this cohort of patients was suboptimal using any of our predefined criteria, particularly in the first 24 h postoperatively (Table 3). Although control gradually improved over the course of the hospital stay, these patients had blood glucose levels >11.0 mmol/l for 42% (95% CI 40–44) of their hospitalization. Multiple linear regression (examining all factors in Table 1) showed that only the urgency of the operation was associated with glycemic control: patients undergoing emergent CABG had poorer postoperative glycemic control than those undergoing elective CABG operations (P = 0.03). Other factors, such as age, sex, type of diabetes, history of diabetic control, presence/absence of microvascular complications, duration of diabetes, and endocrinologist involvement were not associated with glycemic control.

The average capillary glucose in the first postoperative day was significantly associated with subsequent adverse outcomes (P = 0.03, Table 3). The rates of in-hospital adverse outcomes by quartiles of average glucose on postoperative day 1, from lowest quartile to highest, were 18, 21, 30, and 36%, respectively. Compared with patients in the lowest quartile (blood glucose 4.8–10.1 mmol/l on postoperative day 1), patients in the highest quartile (blood glucose ≥12.5 mmol/l) exhibited a significant increase in nonfatal stroke, septic complications, or death after 24 h: unadjusted odds ratio ([OR] 2.5, 95% CI 1.1–5.3), the adjusted OR (after controlling for differences in baseline clinical features and operative factors) remained significant at 2.7 (1.2–6.1). Patients in the second quartile (glucose 10.2–11.2 mmol/l) and third quartile (11.3–12.4 mmol/l) exhibited trends toward higher outcomes: OR 1.2 (0.5–2.7) and 1.9 (0.9–4.2), respectively. Multiple logistic regression analyses examining the association between the factors in Table 1, glycemic control, and outcomes showed that the following factors were independently associated with adverse outcomes after the first postoperative day: emergent operation (OR 3.25, P < 0.001), average glucose on first postoperative day (1.17 for each 1-mmol/l increase in glucose, P = 0.03), duration of operation (1.01 for every extra minute the operation lasted over the median of 225 min, P < 0.001), and age (0.97 per year, P = 0.02). Results were virtually identical in a preplanned subanalysis examining the risk of septic complications alone (data available from corresponding author on request). Glycemic control on postoperative days 2–5 was not associated with adverse outcomes in this cohort.

In summary, individuals with diabetes who undergo CABG are at increased risk for adverse cardiovascular and septic complications. Our data provide further support for arguments that the complication rates derived from administrative data (2) are underestimates in CABG patients. We found that 1.4% of diabetic individuals undergoing CABG in a large tertiary care center died intraoperatively or within the first 24 h and another 27% suffered nonfatal stroke or myocardial infarction, developed a septic complication (most frequently wound infection), or died during their hospitalization.

Glycemic control in these patients was suboptimal relative to the standards achieved in recent clinical trials (16,17). Indeed, the mean blood glucose level on the first postoperative day was 11.4 mmol/l and 98% of these patients had average blood glucose levels >6.1 mmol/l (the threshold triggering intensification of insulin therapy in one of the trials). This may be a particularly important care gap to address because the degree of hyperglycemia in the first postoperative day was significantly associated with adverse outcomes; after adjustment for potential confounders, risk increased by 17% for every 1-mmol/l increase in average blood glucose level >6.1 mmol/l. The observed association between glycemic control and outcomes in diabetic patients undergoing CABG is very similar to what would be extrapolated from the trial evidence (where a 2.8-mmol/l decrease in blood glucose level in surgical ICU patients translated into a 34% reduction in in-hospital mortality) (16). Interestingly, the poor glycemic control in our patients occurred despite the frequent involvement of endocrinologists and the use of a standard insulin infusion protocol at our institution since the late 1990s.

Our study confirms earlier studies suggesting a relationship between glycemic control and cardiovascular outcomes or septic complications in individuals with diabetes undergoing CABG (6–15,20–25). Furthermore, our data suggest that the occurrence of these complications is correlated with blood glucose level on the first postoperative day, even after adjusting for other confounding factors such as presence/absence of diabetic microvascular complications and known prognostic factors in CABG. Patients with diabetes exhibit impaired platelet, coagulation, and fibrinolytic function, all of which are more pronounced with higher glucose levels (26–29). Furthermore, hyperglycemia can lead to dehydration, electrolyte disorders, and potential arrhythmias and has also been shown to adversely affect endothelial-dependent vasodilation (30). Finally, hyperglycemia leads to various in vitro abnormalities that predispose nosocomial infections, including delayed chemotaxis, diminished granulocyte adherence, impaired phagocytosis, and reduced microbiocidal capacity (31). These abnormalities develop when blood glucose levels exceed 11 mmol/l and do improve in vitro with glycemic control (31). We believe our findings support the conclusions from a before/after case series (32) and the Belgian ICU trial (16) that tight glycemic control perioperatively reduces septic complications. However, an observational study can never be considered definitive and alternative explanations could account for our findings: for example, 1) blood glucose may not be a causal risk factor but merely a marker for another unmeasured prognostic factor, or 2) the treatment used to control glucose (i.e., insulin) may be the beneficial factor rather than the lower circulating glucose levels. Although secondary analysis of the Belgian ICU trial suggests that the benefits of lower blood glucose levels are independent of insulin levels (indeed high doses of insulin were associated with worse outcomes in that trial) (29), resolution of this debate awaits the results of the ongoing Bypass Angioplasty Revacularization Investigation-2D (BARI-2D) study (33).

As with any observational study, there are other limitations to our study. First, the retrospective extraction of data may raise concern. However, we used all possible means to minimize any potential bias: we enrolled consecutive patients, looked at objective end points with predefined rigorous definitions, and for a subset of patients, double-extracted the data (independently and, for outcomes, blinded to the perioperative glycemic control). Second, we relied on hospital discharge abstracts to identify diabetic patients; however, previous studies have confirmed the accuracy of our data sources and coding for diabetes (with reported sensitivity of 91% and specificity of 92%) (34). Third, due to incomplete angiographic data in the medical charts, we were unable to examine the relationship between severity of coronary artery disease and glycemic control or outcomes. However, we believe the duration of cardiopulmonary bypass time is a reasonable proxy for coronary lesion scores (only 2 of these 291 operations were performed “off-pump”), and we did control for duration of cardiopulmonary bypass in all analyses. Fourth, due to the ad-hoc nature of cognitive assessments after CABG in our institution, we were unable to examine the association between glycemic control and other end points such as neurocognitive function. Finally, as discussed more fully in the preceding paragraph, in an observational study, one can only examine associations and cannot attribute causation.

Although there are no nationally endorsed clinical practice guidelines establishing standards for glycemic control in surgical patients, there is emerging evidence that stringent glucose control benefits critically ill patients. As such, we believe the question is no longer whether glucose control is required in diabetic patients undergoing CABG, but rather, how much does current glycemic control need to be improved? We found high glucose levels in our diabetic patients undergoing CABG, and hyperglycemia was associated with poor outcomes (in-hospital complication rates increased by 17% for each 1-mmol/l increase in postoperative day 1 glucose >6.1 mmol/l). Because most patients were seen by endocrinologists and managed with standardized insulin infusion protocols, we believe there is an urgent need for development and assessment of novel system-based interventions to improve perioperative glycemic control.