Abstract
Objective
To study the association between diabetes status, perioperative hyperglycemia, and adverse events in a statewide surgical cohort.
Background
Perioperative hyperglycemia may increase the risk of adverse events more significantly in patients without diabetes (NDM) than in those with diabetes (DM).
Methods
Using data from the Surgical Care and Outcomes Assessment Program, a cohort study (2010–2012) evaluated diabetes status, perioperative hyperglycemia, and composite adverse events in abdominal, vascular, and spine surgery at 53 hospitals in Washington State.
Results
Among 40,836 patients (mean age, 54 years; 53.6% women), 19% had diabetes; 47% underwent a perioperative blood glucose (BG) test, and of those, 18% had BG ≥180 mg/dL. DM patients had a higher rate of adverse events (12% vs 9%, P < 0.001) than NDM patients. After adjustment, among NDM patients, those with hyperglycemia had an increased risk of adverse events compared with those with normal BG. Among NDM patients, there was a dose-response relationship between the level of BG and composite adverse events [odds ratio (OR), 1.3 for BG 125–180 (95% confidence interval (CI), 1.1–1.5); OR, 1.6 for BG ≥180 (95% CI, 1.3–2.1)]. Conversely, hyperglycemic DM patients did not have an increased risk of adverse events, including those with a BG 180 or more (OR, 0.8; 95% CI, 0.6–1.0). NDM patients were less likely to receive insulin at each BG level.
Conclusions
For NDM patients, but not DM patients, the risk of adverse events was linked to hyperglycemia. Underlying this paradoxical effect may be the underuse of insulin, but also that hyperglycemia indicates higher levels of stress in NDM patients than in DM patients.
Keywords: adverse events, diabetes, hyperglycemia, insulin, perioperative care
Hyperglycemia in patients with diabetes (DM) who undergo surgery is associated with increased rates of surgical site infection (SSI), myocardial infarction, stroke, and death.1–15 Hyperglycemia also occurs in up to two thirds of surgical patients who are not known to have diabetes, and its impact has not been well characterized in surgical patients without diabetes (NDM).16 Recent observational studies have demonstrated an increased risk of complications associated with hyperglycemia in NDM patients when compared with DM patients. Kwon et al17 found that NDM patients who had perioperative hyperglycemia had nearly twice the risk of infections, reoperative interventions, and in-hospital deaths as DM patients and hyperglycemia. Frisch et al18 found an increased risk of 30-day mortality associated with hyperglycemia for NDM patients when compared with those with diabetes.
These recent studies suggest a paradox: that a disease known for complications related to hyperglycemia may have a lower risk of postsurgical complications in the setting of hyperglycemia than in people without diabetes. There are several possible mechanisms to explain this observation. The higher rate of complications among NDM patients may reflect a more extreme inflammatory and stress response that causes a NDM patient to have the same level of hyperglycemia as a DM patient. This increased stress level may be the reason for the increased rate of complications in NDM patients, with hyperglycemia serving as a marker rather than a cause of the problem. Second, the increased risk of complications in NDM patients may be the result of underdiagnosis of diabetes that is revealed in the surgical setting. Third, it may reflect undertreatment of perioperative hyperglycemia with insulin in NDM patients. It is also possible that the lower risk of adverse events among DM patients who have postoperative diabetes is indicative of insulin being poorly tolerated in patients who have not previously been exposed it. Finally, the lower risk may reflect a form of adaptation associated with more chronic exposure to hyperglycemia.
Understanding the mechanism by which NDM patients have an increased risk of complications has important implications for quality improvement activities related to preoperative assessment of diabetes status and management of postoperative hyperglycemia. To evaluate this paradox and explore possible mechanisms, we undertook an observational study of the association between perioperative hyperglycemia, insulin use, and adverse events in a statewide surgical cohort.
METHODS
Study Population and Setting
The Surgical Care and Outcomes Assessment Program (SCOAP) is a physician-led quality improvement collaborative that began in 2006 and now includes 55 hospitals in Washington State. Data are collected from medical records by trained abstractors. This study was a retrospective review of patients undergoing general surgery, bariatric surgery, vascular surgery, and spine operations between 2010 and 2012. These dates were selected as the same covariates were collected in each of the SCOAP modules (general, vascular, and spine) during this time period. Patients were excluded from the analysis if they did not undergo blood glucose (BG) testing in the perioperative period. Data from 53 of 55 Washington State hospitals currently participating in SCOAP were available at the time of this analysis. Research projects using deidentified SCOAP data are exempted from review by the University of Washington institutional review board.
Data Characteristics
Clinical Risk Factors
Demographic information, clinical comorbidities, and operative details were abstracted from the clinical record using standardized definitions. Patient diabetes status, weight, smoking status, history of coronary artery disease, history of hypertension, cancer as indication for surgery, and American Society of Anesthesiologists’ (ASA) class were available. Patients are classified as having diabetes if they have a medical history of diabetes, as identified in the medical record at the time of admission for the index operation, including review of documentation from emergency department visits, previous primary care visits, surgical clinic visits, and anesthesia documentation. DM patients are subsequently divided into 2 categories, based on whether or not they use insulin at home.
Type/Method of Operation
General surgery, bariatric surgery, vascular surgery, and spine procedures were included in the cohort. General surgery procedures included appendectomies, colon operations, and bariatric operations. Colon operations included right, transverse, and left hemicolectomy, low anterior resection, abdominoperineal resection, total abdominal colectomy, stoma takedown, perineal proctectomy, and abdominal proctectomy. Bariatric operations included laparoscopic and open Rouxen-Y gastric bypass, laparoscopic gastric band placement, sleeve gastrectomy, biliopancreatic bypass with and without duodenal switch, vertical banded gastroplasty, and revision of gastric bypass. Vascular surgery procedures included carotid, aortic, and infrainguinal vascular procedures. Spine operations included cervical and lumbar fusion, corpectomy, discectomy, laminectomy, dural repair, and laminoplasty. Method of operation was identified as laparoscopic or open and as elective or nonelective. The number of hours in the operating room was also recorded.
Hyperglycemia and Insulin Use
SCOAP collects information on the highest BG level at 2 different time periods: perioperatively (including intraoperative time period, and first 24 hours postoperatively), and between 24 and 48 hours postoperatively. Abstractors identify the highest BG within the immediate perioperative period, and between 24 and 48 hours postoperatively, from both finger checks and laboratory tests. In addition, any episodes of hypoglycemia are recorded. On the basis of the highest BG perioperatively, patients were categorized into 3 groups: normal BG (≤125 mg/dL), mildly elevated BG (125–180 mg/dL), and high BG (≥180 mg/dL). In addition, patients with perioperative hyperglycemia are classified on the basis of whether they have persistent hyperglycemia (≥180 mg/dL) on repeat BG check between 24 and 48 hours postoperatively (based on the recommendation of American Diabetes Association that random BG levels be managed to below 180 mg/dL). With regard to insulin use, a binary classification is used to identify individuals as receiving perioperative insulin or not receiving perioperative insulin. Patients are considered to have received insulin if they are given insulin at any time point during the postoperative period.
Outcome Measures
A composite adverse event metric was used as the primary outcome. This overall composite metric was divided into cardiac adverse events, noncardiac adverse events, and death. This framework has previously been used for evaluation of adverse events in a surgical cohort.19 Adverse events were based on in-hospital postoperative events. The cardiac adverse event metric included myocardial infarction, stroke or transient ischemic attack, and atrial arrhythmia requiring treatment. The noncardiac adverse event metric included readmission to the intensive care unit or unplanned intensive care unit stay, fall with injury requiring surgery, infectious complications (Clostridium difficile infection, SSI requiring treatment, pneumonia requiring treatment, urinary tract infection requiring treatment, wound reopening, or debridement), renal insufficiency or initiation of dialysis, reintubation, percutaneous drainage of an abscess, and any reoperative intervention.
Analytic Methods
Demographic and clinical characteristics were compared among patients with and without diabetes both in the entire patient cohort and in the cohort of patients who underwent BG test who make up the study cohort. Characteristics were summarized using frequency distributions for categorical variables and means and standard deviations for continuous variables. Categorical variable comparisons were evaluated for significance using the Pearson χ2 test. Continuous variable comparisons were evaluated for significance using t tests.
The overall adverse event rate and death, cardiac, and noncardiac adverse event rates were compared between patients with and without diabetes. Perioperative insulin administration was compared between patients with and without diabetes, stratified by perioperative BG level. The association between perioperative hyperglycemia and composite adverse events was then evaluated in patients with and without diabetes, adjusting for potential confounders using multivariate logistic regression. Potential confounders included age, sex, race, insurance status, type and method of surgery, operative time, ASA class, comorbidities, smoking status, home medications, and elective or emergent operation classification. Type of surgery was included in regression model in a discrete fashion. Sensitivity analyses were conducted for cardiac adverse events, noncardiac adverse events, and death.
RESULTS
There were 40,836 patients who underwent general, bariatric, vascular, and spine operations between 2010 and 2012. Within this cohort, 19% had a diagnosis of diabetes mellitus, 47% had a perioperative BG test, and of those, 18% had perioperative hyperglycemia (40% of DM patients, 6% of NDM patients; Fig. 1). A total of 6595 DM patients and 12,663 NDM patients underwent BG testing in the perioperative period, and these patients make up the study cohort. Clinical and demographic characteristics of the cohort are described in Table 1.
FIGURE 1.
Flow diagram (ovals indicate patients who comprise the study cohort).
TABLE 1.
Cohort Characteristics By Diabetes Status and BG Test
| All Patients (N = 40,836)
|
Patients With BG Test (N =19,258)
|
|||||
|---|---|---|---|---|---|---|
| Diabetes (19.2%) | No Diabetes (80.8%) | P | Diabetes (34.2%) | No Diabetes (65.8%) | P | |
| Number | 7833 | 30,003 | 6595 | 12,663 | ||
| Mean age (SD) | 61.9 (18.4) | 52.1 (18.8) | <0.001 | 62.0 (13.7) | 52.4 (19.5) | <0.001 |
| Sex | 0.86 | <0.001 | ||||
| Male, no. (%) | 3637 (46.5) | 15,286 (46.3) | 3080 (46.7) | 6258 (49.4) | ||
| Female, no. (%) | 4193 (53.6) | 17,704 (53.7) | 3512 (53.3) | 6399 (50.6) | ||
| Insurance, no. (%) | ||||||
| Private | 4850 (62.3) | 22,151 (67.8) | <0.001 | 4148 (63.3) | 8369 (66.5) | <0.001 |
| Medicaid | 1170 (15.0) | 2994 (9.2) | <0.001 | 1009 (15.4) | 1263 (10.1) | <0.001 |
| Weight, mean (SD), kg | 99.5 (27.0) | 85.2 (23.9) | <0.001 | 100.4 (27.2) | 84.0 (23.6) | <0.001 |
| Race | <0.001 | <0.001 | ||||
| White, no. (%) | 6384 (81.7) | 26,774 (81.3) | 5358 (81.4) | 9960 (78.8) | ||
| Black or African American, no. (%) | 353 (4.5) | 938 (2.9) | 304 (4.6) | 407 (3.2) | ||
| Asian, no. (%) | 247 (3.2) | 1165 (3.5) | 198 (3.0) | 513 (4.1) | ||
| American Indian/Alaska Native, no. (%) | 110 (1.4) | 270 (0.8) | 91 (1.4) | 122 (1.0) | ||
| Native Hawaiian or other Pacific Islander, no. (%) | 48 (0.6) | 139 (0.4) | 42 (0.6) | 54 (0.4) | ||
| Unknown/NA, no. (%) | 675 (8.6) | 3654 (11.1) | 589 (9.0) | 1585 (12.5) | ||
| SCOAP module | <0.001 | <0.001 | ||||
| General, no. (%) | 4980 (63.6) | 25,850 (78.3) | 4221 (64.0) | 10,032 (79.2) | ||
| Spine, no. (%) | 655 (8.4) | 3230 (9.8) | 461 (7.0) | 387 (3.1) | ||
| Vascular, no. (%) | 2198 (28.0) | 3923 (11.9) | 1913 (29.0) | 2244 (17.7) | ||
| Type of operation | <0.001 | <0.001 | ||||
| Appendectomy, no. (%) | 660 (8.4) | 11,409 (34.6) | 495 (7.5) | 5383 (42.5) | ||
| Bariatric, no. (%) | 2272 (29.0) | 3750 (11.3) | 2026 (30.7) | 1096 (8.6) | ||
| Colorectal, no. (%) | 1895 (24.2) | 9800 (29.7) | 1569 (23.8) | 3405 (26.9) | ||
| Aortic, no. (%) | 180 (2.3) | 722 (2.2) | 157 (2.4) | 472 (3.7) | ||
| Carotid, no. (%) | 568 (7.2) | 1149 (3.5) | 503 (7.6) | 610 (4.8) | ||
| Infrainguinal, no. (%) | 1450 (18.5) | 2052 (6.2) | 1253 (19.0) | 1162 (9.2) | ||
| Spine, no. (%) | 655 (8.4) | 3230 (9.8) | 461 (7.0) | 387 (3.1) | ||
| Cancer, no. (%) | 153 (2.0) | 891 (2.7) | 131 (2.0) | 148 (1.2) | ||
| Open procedure, no. (%) | 3053 (39.0) | 12,848 (39.0) | 0.928 | 2500 (37.9) | 4446 (35.1) | <0.001 |
| Elective operation no. (%) | 5450 (69.6) | 17,647 (53.5) | <0.001 | 4613 (70.0) | 5168 (40.8) | <0.001 |
| Admission type | ||||||
| Inpatient, no. (%) | 7168 (91.6) | 31,663 (96.0) | <0.001 | 6046 (91.7) | 12,161 (96.1) | <0.001 |
| Hours in OR, mean (SD) | 2.3 (1.5) | 1.9 (1.6) | <0.001 | 2.3 (1.5) | 1.8 (1.5) | <0.001 |
| Rural, no. (%) | 185 (2.4) | 1557 (4.7) | <0.001 | 142 (2.2) | 528 (4.2) | <0.001 |
| ASA class, no. (%) | <0.001 | <0.001 | ||||
| I | 75 (1.0) | 5723 (17.8) | 57 (0.9) | 2451 (20.8) | ||
| II | 1565 (21.1) | 15,308 (47.7) | 1163 (18.6) | 4614 (37.8) | ||
| III | 5168 (69.8) | 9920 (30.9) | 4517 (72.3) | 4303 (35.2) | ||
| IV | 568 (7.7) | 1060 (3.3) | 490 (7.8) | 710 (5.8) | ||
| V | 26 (0.4) | 64 (0.2) | 24 (0.4) | 54 (0.4) | ||
| Risk factors | ||||||
| Cigarette smoking | 1479 (19.0) | 7378 (22.5) | <0.001 | 1219 (18.6) | 2970 (23.6) | <0.001 |
| Hypertension | 6407 (81.8) | 12,674 (38.4) | <0.001 | 5461 (82.8) | 4956 (39.2) | <0.001 |
| Coronary artery disease | 2098 (26.8) | 3046 (9.2) | <0.001 | 1801 (27.3) | 1459 (11.5) | <0.001 |
| Medication use | ||||||
| Statin | 4779 (61.0) | 7127 (21.6) | <0.001 | 4119 (62.5) | 2910 (23.0) | <0.001 |
| β-Blocker | 3161 (40.4) | 5829 (17.7) | <0.001 | 2727 (41.4) | 2459 (19.4) | <0.001 |
| ACE inhibitor | 4722 (60.3) | 6737 (20.4) | <0.001 | 4051 (61.5) | 2673 (21.1) | <0.001 |
| Home oxygen use, no. (%) | 2083 (29.1) | 3792 (12.8) | <0.001 | 1821 (29.7) | 2205 (18.0) | <0.001 |
| Home mobility device, no. (%) | 370 (7.4) | 705 (2.7) | <0.001 | 330 (7.8) | 303 (3.0) | <0.001 |
NA indicates not available.
Perioperative Insulin Use
Perioperative insulin administration was compared between patients with and without diabetes, stratified by level of perioperative hyperglycemia (Fig. 2). At each level of BG, DM patients were significantly more likely to receive insulin than NDM patients (P < 0.001). We evaluated the effectiveness of correction of hyperglycemia (to ≤180 mg/dL) after initial hyperglycemia among patients with and without diabetes. Among all patients who had perioperative hyperglycemia and received insulin, those who had persistent hyperglycemia had significantly higher adverse event rates than those whose BG was corrected (40.3% vs 25.6% for NDM patients; 17.5% vs 8.3% for DM patients).
FIGURE 2.
Receipt of insulin, by glucose level. Differences were statistically significant between patients at different levels of hyperglycemia for both patients with and without diabetes.
Overall Rate of Adverse Events
In our complete patient cohort of more than 40,000 patients, DM patients had a higher composite adverse event rate than NDM patients (12.0% vs 8.9%, P < 0.001). In our study cohort of patients with BG testing, we found no significant difference in the rate of adverse events between DM patients with diabetes and NDM patients (12.6% vs 12.1%, P = 0.26) in an adjusted analysis. This difference between the overall cohort and the study cohort suggests that NDM patients are selected for perioperative BG testing in a nonrandom way. Differences in risk characteristics are addressed in the multivariate analysis.
Risk Adverse Events Associated With Hyperglycemia
The rate of adverse events was compared by level of hyperglycemia among patients with and without diabetes (Fig. 3). Among DM patients, those with hyperglycemia did not have increased odds of adverse events {odds ratio [OR] = 0.76 [95% confidence interval (CI), 0.57–1.0, P = 0.06], for BG 125–180 mg/dL; and OR = 0.94 (95% CI 0.72–1.2; P = 0.65) for BG ≥180 mg/dL}in an unadjusted analysis. After controlling for confounders, DM patients with a BG level between 125 and 180 mg/dL were found to have decreased odds of an adverse event [OR = 0.66 (95% CI = 0.49–0.91)] when compared with the reference group (BG ≤125 mg/dL; Table 2). Patients with a BG level more than 180 mg/dL, however, had no significant difference in odds of an adverse event [OR = 0.78 (95% CI = 0.58–1.04)] when compared with the reference group.
FIGURE 3.
Adverse events, by diabetes status and postoperative glucose level.
TABLE 2.
Multivariate Model for Risk of Composite Adverse Events Among Patients With and Without Diabetes
| Univariate | Multivariate | |||
|---|---|---|---|---|
| OR | 95% CI | OR | 95% CI | |
| Patients with diabetes | ||||
| Age | 1.03 | 1.02–1.04 | 1.02 | 1.01–1.02 |
| Male sex | 1.23 | 1.06–1.43 | 1.12 | 0.92–1.35 |
| Perioperative insulin use | 1.33 | 1.03–1.71 | 1.19 | 0.97–1.47 |
| Perioperative BG | ||||
| BG ≤125 | Ref | Ref | ||
| 125 <BG <180 | 0.76 | 0.57–1.01 | 0.66 | 0.49–0.91 |
| BG ≥180 | 0.94 | 0.72–1.23 | 0.78 | 0.58–1.04 |
| Private insurance | 0.62 | 0.50–0.76 | 0.84 | 0.63–1.12 |
| Medicaid insurance | 1.54 | 1.22–1.93 | 1.38 | 1.03–1.83 |
| Open procedure | 3.24 | 2.53–4.15 | 1.30 | 1.15–1.29 |
| Type of operation (%) | ||||
| Appendectomy | Ref | Ref | ||
| Bariatric | 0.61 | 0.35–1.06 | 0.93 | 0.55–1.58 |
| Colorectal | 4.47 | 3.01–6.63 | 3.39 | 2.07–5.57 |
| Aortic | 1.90 | 1.24–2.91 | 1.03 | 0.55–1.92 |
| Carotid | 0.85 | 0.45–1.63 | 0.47 | 0.21–1.08 |
| Infrainguinal | 1.18 | 0.78–1.79 | 0.94 | 0.51–1.71 |
| Spine | 0.83 | 0.41–1.65 | 0.85 | 0.41–1.75 |
| Cancer | 2.48 | 1.28–4.80 | 2.50 | 1.51–4.43 |
| ASA class | ||||
| I | Ref | Ref | ||
| II | 0.89 | 0.46–1.72 | 0.58 | 0.24–13.8 |
| III | 1.13 | 0.62–2.06 | 0.79 | 0.35–1.79 |
| IV | 3.79 | 2.06–6.94 | 1.48 | 0.67–3.23 |
| V | 14.17 | 3.93–51.03 | 3.21 | 0.87–11.8 |
| Elective operation | 0.48 | 0.38–0.62 | 0.46 | 0.36–0.59 |
| Mean hours in operative room | 1.26 | 1.17–1.36 | 1.22 | 1.15–1.29 |
| Patients without diabetes | ||||
| Age | 1.04 | 1.03–1.05 | 1.02 | 1.01–1.02 |
| Male sex | 1.02 | 0.87–1.19 | 1.07 | 0.91–1.27 |
| Perioperative insulin use | 3.62 | 2.28–5.74 | 1.28 | 0.99–1.64 |
| Perioperative BG | ||||
| BG ≤125 | Ref | Ref | ||
| 125 < BG < 180 | 2.39 | 1.89–3.03 | 1.26 | 1.08–1.47 |
| BG ≥180 | 5.12 | 3.80–6.90 | 1.63 | 1.27–2.10 |
| Private insurance | 0.60 | 0.61–0.72 | 0.77 | 0.67–0.90 |
| Medicaid insurance | 1.77 | 1.46–2.15 | 1.16 | 0.93–1.45 |
| Open procedure | 5.45 | 4.53–6.55 | 1.73 | 1.44–2.08 |
| Type of operation (%) | ||||
| Appendectomy | Ref | Ref | ||
| Bariatric | 2.06 | 0.87–4.94 | 1.17 | 0.61–2.26 |
| Colorectal | 12.39 | 9.36–16.40 | 3.34 | 2.35–4.76 |
| Aortic | 6.63 | 5.33–11.21 | 1.15 | 0.69–1.93 |
| Carotid | 2.21 | 1.54–3.16 | 0.42 | 0.29–0.61 |
| Infrainguinal | 3.24 | 2.27–4.61 | 0.92 | 0.62–1.37 |
| Spine | 2.86 | 1.71–4.78 | 0.81 | 0.46–1.43 |
| Cancer | 8.12 | 5.51–11.99 | 2.43 | 1.35–4.38 |
| ASA class | ||||
| I | Ref | Ref | ||
| II | 3.62 | 2.50–5.23 | 1.38 | 0.96–2.00 |
| III | 9.45 | 6.42–13.90 | 2.76 | 1.85–4.12 |
| IV | 44.87 | 28.66–70.27 | 7.22 | 4.66–11.2 |
| V | 206.55 | 86.37–493.97 | 32.8 | 11.7–91.6 |
| Elective operation | 1.21 | 0.88–1.67 | 0.53 | 0.42–0.66 |
| Mean hours in operative room | 1.41 | 1.25–1.59 | 1.19 | 1.12–1.27 |
Bold values indicate statistical significance (odds ratios that do not include one).
In contrast, among NDM patients, those with hyperglycemia had significantly higher odds of a composite adverse event [OR = 2.4 (95% CI, 1.9–3.0) for BG 125–180 mg/dL; and OR = 5.1 (95% CI, 3.8–6.9) for BG ≥180 mg/dL; P < 0.001 for both] in an unadjusted analysis. After controlling for confounders, this increased risk of complications associated with hyperglycemia for NDM patients persisted, with a dose-response relationship between the level of hyperglycemia and the odds of a composite adverse event [OR = 1.26 (95% CI, 1.08–1.47) for BG 125–180 mg/dL; OR = 1.63 (95% CI, 1.27–2.10) for BG ≥180 mg/dL; Table 2]. When patients with and without diabetes were compared, the overall rates of adverse events increased in a dose-response fashion for each increasing level of hyperglycemia among NDM patients, but not for DM patients.
For both those with diabetes and those without diabetes, advanced age, patients undergoing open operations, patients with colorectal disease and patients with cancer, and patients with an increased operative time had higher odds of an adverse event. In both cohorts, patients undergoing elective operations had decreased odds of an adverse event. Patients with Medicaid had increased odds of an adverse event among DM patients but not among those without. Patients who were considered to be ASA class III, IV, and V had increased odds of an adverse event among NDM patients but not among those with diabetes. Patients with private insurance and patients undergoing carotid operations had decreased odds of an adverse event among NDM patients but not among those with diabetes.
Sensitivity Analyses
Sensitivity analyses were conducted for cardiac adverse events, noncardiac adverse events, and death. The findings from these analyses mirrored the main analysis of composite adverse events, with the exception that for cardiac adverse events and death, NDM patients did not have increased odds of adverse events for BG 125 to 180 mg/dL when compared with the reference group (≤125 mg/dL).
DISCUSSION
The results from this statewide evaluation of surgical patients demonstrate that although DM patients have a higher rate of adverse events than NDM patients overall, among those with any given level of hyperglycemia, NDM patients have higher rates of adverse events than those with diabetes. In addition, among NDM patients, there is a dose-response relationship between the level of hyperglycemia and the risk of adverse events, with higher rates at higher glucose levels. For patients—with and without diabetes alike—who have perioperative hyperglycemia, receive insulin, and still have persistent hyperglycemia, there is an even higher rate of adverse events.
The findings from this study prompt exploration of a number of theories regarding the mechanism for this diabetes paradox—that hyperglycemic NDM patients have higher odds of adverse events than those with diabetes. The first theory is that hyperglycemia in NDM patients is really a marker for increased surgical stress or severity of illness. Although we endeavored to correct for this in our multivariate model, including operation type, method (open vs minimally invasive), duration, ASA class, and other proxies for surgical stress, it is possible that some residual confounding exists. Studies have demonstrated that surgical stress results in decreased insulin sensitivity and subsequent hyperglycemia and that this relationship is relative to the magnitude of the surgical procedure.20 Given that DM patients have a baseline level of insulin resistance and hyperglycemia, it is possible that it requires lower levels of surgical stress for such patients to reach each level of hyperglycemia, or, conversely, that it requires higher levels of surgical stress for NDM patients to reach a given level of BG. In this way, hyperglycemia in those without diabetes may be an indicator of patients with greater surgical stress or severity of illness—patients who are known to be at higher risk for complications.
A second theory is that, in the perioperative setting in association with surgical stress, previously undiagnosed diabetes is revealed. Studies indicate that nearly 40% of surgical patients with perioperative hyperglycemia more than 200 mg/dL have no previous diagnosis of diabetes21 and that up to 60% of NDM patients with postoperative hyperglycemia and a myocardial infarction were ultimately diagnosed with diabetes.22 Preoperative HbA1c or fasting BG is not currently the standard for patients without a known diagnosis of diabetes, and as such, it is not clear how many of the patients in this study had previously undiagnosed diabetes. However, if undiagnosed diabetes were the major driver of increased risk for NDM patients, it would not explain why their observed risk was higher than that of patients with a known diagnosis of diabetes.
Third, insulin treatment in the setting of perioperative hyperglycemia may be better delivered in patients with a diagnosis of diabetes. Our findings indicate that this may be true, as only 40% of NDM patients with a BG level of 180 to 250 mg/dL and 55% of those with a BG level of 250 mg/dL or more received insulin therapy in the postoperative period. This rate of insulin use by level of hyperglycemia was statistically different than rates of insulin use in DM patients (60% for BG 180–250 mg/dL and 80% for BG ≥250 mg/dL; P < 0.001). In DM patients, providers may be more likely to suspect and test for hyperglycemia and system protocols may enhance that detection as well.
A fourth theory is that insulin treatment in patients previously unexposed to insulin—NDM patients and DM patients not on home insulin—may not be well tolerated. Insulin administration, especially in the setting of acute hyperglycemia, may increase inflammation and oxidative stress through a pathway involving the potentiation of NFκB21 and through an increase in IL-6 and TNF-α,22 may worsen atherosclerosis by inducing smooth muscle cell proliferation and sterol synthesis23–26 and create a hypercoagulable state by increasing procoagulant factors (plasma activator inhibitor-1, tPA antigen, von Willenbrand factor, factor VII, and fibrinogen).27 However, in a post hoc analysis among DM patients, we did not find that those starting new insulin during the hospitalization had increased odds of an adverse event. In addition, we observed hypoglycemia in less than 1% of the patients in our cohort.
Finally, in keeping with the idea of a paradox, whereas diabetes is a disease characterized by complications related to chronic hyperglycemia, diabetes may be “protective” for patients with hyperglycemia, as the very result of chronic exposure to elevated BG. There is experimental evidence to suggest that patients tolerate chronic hyperglycemia better than acute hyperglycemia because of conditioning. In the setting of chronic hyperglycemia, myocytes become resistant to hypoxia-induced apoptosis and necrosis through upregulation of Bcl-2 (anti-apoptotic) and reduction of intracellular calcium.28 With acute hyperglycemia, the sensitivity of myocytes to ischemia increases and this negates the beneficial effects of ischemia preconditioning.29 As a result, DM patients, in the setting of chronic hyperglycemia, may have better outcomes associated with hypoxia and lower rates of myocardial ischemia. Inflammatory or oxidative preconditioning may exist as well. This form of conditioning has yet to be described in relation to immune function or tolerance of surgical stress.
In summary, there are 5 theories we posit as to why NDM patients have higher observed odds of adverse events associated with perioperative hyperglycemia: (1) hyperglycemia is a marker for greater surgical stress or severity of illness; (2) previously undiagnosed diabetes is revealed in the setting of surgical stress; (3) insulin therapy is better delivered in patients with a known diagnosis of diabetes; (4) insulin administration is associated with potential harm and inflammation in previously insulin-naive patients; and (5) DM patients may have an adaptation to hyperglycemia. Our group is currently engaged in ongoing studies to explore the plausibility of the first, second, and fifth theories. Our current research suggests that—at the very least—the third theory is true and that administration of insulin for NDM patients is a target for quality improvement. We do not find any evidence in our results to suggest that the fourth theory—positing potential harm from insulin in insulin-naive patients—is true.
This study must be interpreted in the context of its limitations. First, the major limitation is the retrospective nature of the analysis and missing information in our data set. Given that NDM patients do not routinely undergo fasting BG, preoperative HbA1c, or even post-operative BG testing, it is difficult to make clear conclusions about the mechanism of increased risk. It is possible that our findings may just reveal previously undiagnosed diabetes, although this does not explain increased odds of adverse events compared with DM patients. Only 38% of NDM patients underwent a perioperative BG test, and as a result, confounding by indication may play a role in our findings. We attempted to adjust for this as best we could in multivariate models, including both demographic and clinical factors, such as age, sex, weight, comorbidities, smoking status, and home medications in our models. However, some residual confounding may exist. In addition, among DM patients, limited information on diabetes status was available, including the duration since diagnosis and current home medication regimens. We also did not have information on the type of insulin (dose or continuous infusion vs basal bolus vs sliding scale) used in-hospital.
CONCLUSIONS
This study confirms and expands upon previously reported findings regarding the impact of perioperative hyperglycemia in NDM patients. We demonstrate a dose-response relationship between hyperglycemia and adverse events, suggesting that NDM patients may fare worse with postoperative hyperglycemia. Further prospective studies are needed to help clarify the mechanism of increased risk and to determine whether or not hyperglycemia among NDM patients is, in fact, a marker for increased surgical stress or undiagnosed diabetes. A clearer understanding of the risk of hyperglycemia for surgical NDM patients has profound implications for future management of hyperglycemia in a perioperative setting.
Footnotes
The administrative home for the Surgical Care and Outcomes Assessment Program (SCOAP) is the Foundation for Healthcare Quality.
Disclosure: The Comparative Effectiveness Research Translation Network (CERTAIN) is supported by the Life Sciences Discovery Fund (LSDF) of Washington State and the Agency for Healthcare Research and Quality (AHRQ). Dr Kotagal is supported by a University of Washington Department of Surgery T32 training fellowship grant from the National Institute of Diabetes & Digestive & Kidney Diseases (grant no. 5T32DK070555-03).
Disclosure: Dr Hirsch has received research support from Sanofi USA and Halozyme; Dr Umpierrez has received research support from Merck and Sanofi. None of the other authors have any conflicts of interests related to this article.
References
- 1.Zerr KJ, Furnary AP, Grunkemeier GL, et al. Glucose control lowers the risk of wound infections in diabetics after open heart operations. Ann Thorac Surg. 1997;63:356–361. doi: 10.1016/s0003-4975(96)01044-2. [DOI] [PubMed] [Google Scholar]
- 2.Latham R, Lancaster AD, Covington JF, et al. The association of diabetes and glucose control with surgical-site infections among cardiothoracic surgery patients. Infect Control Hosp Epidemiol. 2001;22:607–612. doi: 10.1086/501830. [DOI] [PubMed] [Google Scholar]
- 3.Zacharias A, Habib RH. Factors predisposing to median sternotomy complications: deep vs superficial infection. Chest. 1996;110:1173–1178. doi: 10.1378/chest.110.5.1173. [DOI] [PubMed] [Google Scholar]
- 4.Vuorisalo S, Haukipuro K, Pokela R, et al. Risk features for surgical-site infections in coronary artery bypass surgery. Infect Control Hosp Epidemiol. 1998;19:240–247. doi: 10.1086/647802. [DOI] [PubMed] [Google Scholar]
- 5.Slaughter MS, Olson MM, Lee JT, Jr, et al. A fifteen-year wound surveillance study after coronary artery bypass. Ann Thorac Surg. 1993;56:1063–1068. doi: 10.1016/0003-4975(95)90014-4. [DOI] [PubMed] [Google Scholar]
- 6.Nagachinta T, Stephens M, Reitz B, et al. Risk factors for surgical-wound infection following cardiac surgery. J Infect Dis. 1987;156:967–973. doi: 10.1093/infdis/156.6.967. [DOI] [PubMed] [Google Scholar]
- 7.Lillenfeld D, Engin M, Vlahov D, et al. Obesity and diabetes as risk factors for postoperative wound infections after cardiac surgery. Am J Infect Control. 1988;16:3–6. doi: 10.1016/0196-6553(88)90003-x. [DOI] [PubMed] [Google Scholar]
- 8.Borger MA, Rao V, Weisel RD, et al. Deep sternal wound infection: risk factors and outcomes. Ann Thorac Surg. 1998;65:1050–1056. doi: 10.1016/s0003-4975(98)00063-0. [DOI] [PubMed] [Google Scholar]
- 9.McAlister FA, Man J, Bistritz L, et al. Diabetes and coronary artery bypass surgery: an examination of perioperative glycemic control and outcomes. Diabetes Care. 2003;26:1518–1524. doi: 10.2337/diacare.26.5.1518. [DOI] [PubMed] [Google Scholar]
- 10.McConnel YJ, Johnson PM, Porter GA. Surgical site infections following colorectal surgery in patients with diabetes: association with postoperative hyperglycemia. J Gastrointest Surg. 209(13):508–515. doi: 10.1007/s11605-008-0734-1. [DOI] [PubMed] [Google Scholar]
- 11.Ata A, Lee J, Bestle SL, et al. Postoperative hyperglycemia and surgical site infection in general surgery patients. Arch Surg. 2010;145:858–864. doi: 10.1001/archsurg.2010.179. [DOI] [PubMed] [Google Scholar]
- 12.Pomposelli JJ, Baxter JK, III, Babineau TJ, et al. Early postoperative glucose control predicts nosocomial infection rate in diabetic patients. J Parenter Enteral Nutr. 1998;22:77–81. doi: 10.1177/014860719802200277. [DOI] [PubMed] [Google Scholar]
- 13.Ambiru S, Kato A, Kimura F, et al. Poor postoperative blood glucose control increases surgical site infections after surgery for hepatobiliary-pancreatic cancer: a prospective study in a high-volume institute in Japan. J Hosp Infect. 2008;68:230–233. doi: 10.1016/j.jhin.2007.12.002. [DOI] [PubMed] [Google Scholar]
- 14.Malmstedt J, Wahlberg E, Jorneskog G, et al. Influence of perioperative blood glucose levels on outcome after infrainguinal bypass surgery in patients with diabetes. Br J Surg. 2006;93:1360–1367. doi: 10.1002/bjs.5466. [DOI] [PubMed] [Google Scholar]
- 15.Olsen MA, Nepple JJ, Riew KD, et al. Risk factors for surgical site infection following orthopaedic spinal operations. J Bone Joint Surg Am. 2008;90:62–69. doi: 10.2106/JBJS.F.01515. [DOI] [PubMed] [Google Scholar]
- 16.Kiran RP, Turina M, Hammel J, et al. The clinical significance of an elevated postoperative glucose value in nondiabetic patients after colorectal surgery: evidence for he need for tight glucose control? Ann Surg. 2013;258:599–605. doi: 10.1097/SLA.0b013e3182a501e3. [DOI] [PubMed] [Google Scholar]
- 17.Kwon S, Thompson R, Dellinger P, et al. Importance of perioperative glycemic control in general surgery: a report from the surgical care and outcomes assessment program. Ann Surg. 2013;257:8–14. doi: 10.1097/SLA.0b013e31827b6bbc. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Frisch A, Chandra P, Smiley D, et al. Prevalence and clinical outcome of hyperglycemia in the perioperative period in noncardiac surgery. Diabetes Care. 2010;33:1783–1788. doi: 10.2337/dc10-0304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kulier A, Levin J, Moser R, et al. Impact of preoperative anemia on outcome in patients undergoing coronary artery bypass graft surgery. Circulation. 2007;116:471–479. doi: 10.1161/CIRCULATIONAHA.106.653501. [DOI] [PubMed] [Google Scholar]
- 20.Thorell A, Nygren J, Ljungqvist O. Insulin resistance: a marker of surgical stress. Curr Opin Clin Nutr Metab Care. 1999;2:68–78. doi: 10.1097/00075197-199901000-00012. [DOI] [PubMed] [Google Scholar]
- 21.Levetan CS, Passaro M, Jablonski K, et al. Unrecognized diabetes among hospitalized patients. Diabetes Care. 1993;21:246–249. doi: 10.2337/diacare.21.2.246. [DOI] [PubMed] [Google Scholar]
- 22.Golovchenko I, Goalstone ML, Watson P, et al. Hyperinsulinemia enhances transcriptional activity of nuclear factor-κb induced by angiotensin ii, hyperglycemia, and advanced glycosylation end products in vascular smooth muscle cells. Circ Res. 2000;87:746–752. doi: 10.1161/01.res.87.9.746. [DOI] [PubMed] [Google Scholar]
- 23.Ruge T, Lockton JA, Renstrom F, et al. Acute hyperinsulinemia raises plasma interleukin-6 in both nondiabetic and type 2 diabetes mellitus subjects, and this effect is inversely associated with body mass index. Metabolism. 2009;58:860–866. doi: 10.1016/j.metabol.2009.02.010. [DOI] [PubMed] [Google Scholar]
- 24.Stout RW, Bierman EL, Ross R. Effect of insulin on the proliferation of cultured primate arterial smooth muscle cells. Circ Res. 1975;36:319–327. doi: 10.1161/01.res.36.2.319. [DOI] [PubMed] [Google Scholar]
- 25.Stout RW. The effect of insulin and glucose on sterol synthesis in cultured rat arterial smooth muscle cells. Atherosclerosis. 1977;27:271–278. doi: 10.1016/0021-9150(77)90036-3. [DOI] [PubMed] [Google Scholar]
- 26.Chait A, Bierman EL, Albers JJ. Regulatory role of insulin in the degradation of low density lipoprotein by cultured human skin fibroblasts. Biochim Biophys Acta. 1978;529:292–299. doi: 10.1016/0005-2760(78)90072-3. [DOI] [PubMed] [Google Scholar]
- 27.Meigs JB, Mittleman MA, Nathan DM, et al. Hyperinsulinemia, hyperglycemia, and impaired hemostasis: the Framingham offspring study. JAMA. 2000;283:221–228. doi: 10.1001/jama.283.2.221. [DOI] [PubMed] [Google Scholar]
- 28.Schaffer SW, Croft CB, Solodushko V. Cardioprotective effect of chronic hyperglycemia: effect on hypoxia-induced apoptosis and necrosis. Am J Physiol Heart Circ Physiol. 2000;278:H1948–H1954. doi: 10.1152/ajpheart.2000.278.6.H1948. [DOI] [PubMed] [Google Scholar]
- 29.Kersten JR, Schmeling TJ, Orth KG, et al. Acute hyperglycemia abolishes ischemic preconditioning in vivo. Am J Physiol Heart Circ Physiol. 1998;275:H721–H725. doi: 10.1152/ajpheart.1998.275.2.H721. [DOI] [PubMed] [Google Scholar]