The impact of diabetes on postoperative outcomes following lower-extremity bypass surgery

Abstract

Objective

The effect of diabetes type (noninsulin dependent vs insulin dependent) on outcomes after lower-extremity bypass (LEB) has not been clearly defined. Therefore, we analyzed associations between diabetes type and outcomes after LEB in patients with critical limb ischemia.

Methods

We performed a retrospective analysis of 1977 infrainguinal LEB operations done for critical limb ischemia between 2003 and 2010 within the Vascular Study Group of New England. Patients were categorized as nondiabetic (ND), noninsulin-dependent diabetic (NIDD), or insulin-dependent diabetic (IDD) based on their preoperative medication regimen. Our main outcome measures were in-hospital mortality and major adverse events (MAEs) – a composite outcome, including myocardial infarction, dysrhythmia, congestive heart failure, wound infection, renal insufficiency, and major amputation. We compared crude and adjusted rates of mortality and MAEs using logistic regression across diabetes categories.

Results

Overall, 41% of patients were ND, 28% were NIDD, and 31% were IDD. Crude rates of in-hospital mortality were similar across these groups (1.7% vs 3.1% vs 2.1%; P = .211). Adjusted analyses accounting for differences in patient characteristics showed that diabetes is not associated with increased risk of in-hospital mortality. However, type of diabetes was associated with a higher risk of MAEs in both crude (15.1% for ND; 21.1% for NIDD; and 25.2% for IDD; P <.001) and adjusted analyses (odds ratio for NIDD, 1.41; 95% confidence interval, 1.2–1.7; odds ratio for IDD, 1.53; 95% confidence interval, 1.3–1.8).

Conclusions

Diabetes is a significant contributor to the risk of postoperative complications after LEB surgery, and insulin dependence is associated with higher risk. Quality measures aimed at limiting complications after LEB may have the most impact if these initiatives are focused on patients who are IDD.

Several studies have shown that diabetes is an independent risk factor for adverse postoperative outcomes, such as mortality and surgical site infection, in patients undergoing a wide range of surgical interventions.^1–3 However, the effect of diabetes on surgical risk among patients undergoing lower-extremity bypass (LEB) surgery remains controversial. Although some have found that patients with diabetes undergoing LEB experience increased rates of amputation, cardiac complications, and infection,⁴ other studies have demonstrated that patients with diabetes are not at increased risk for worse outcomes compared to patients who are nondiabetic.^5,6

We hypothesize that this discrepancy may result from analytical methods that did not account for the extent or type of diabetes within the study population. The impact of diabetes on postoperative outcomes after LEB may vary according to the type of diabetes, if the extent of diabetes (noninsulin dependent vs insulin dependent) manifests its effect differently. Although several prior authors have investigated the impact of type of disease among diabetics on outcomes after vascular surgery, these evaluations were limited in scope⁷ or reported a narrow range of clinical outcomes.⁸ Accordingly, the aims of our study were to, first, examine the influence of the type of diabetes (as defined by insulin dependence) on a wide range of clinical outcomes after LEB in patients with critical limb ischemia (CLI). Second, we sought to define which patient and operative characteristics were most closely associated with these outcomes among noninsulin-dependent patients with diabetes alone and patients with insulin-dependent diabetes.

METHODS

Subjects and databases

For this report, we analyzed data collected by the Vascular Study Group of New England, a regional cooperative quality improvement initiative developed in 2002. Further details on the registry used by Vascular Study Group of New England have been published previously⁹ and are available at http://www.vsgne.org.

Data were examined from 1977 patients who underwent primary LEB for CLI. We studied procedures performed by 69 participating surgeons across 14 study hospitals between January 1, 2003, and June 1, 2010. Because we were unable to differentiate patients with disabling claudication in our database from those with less severe claudication, patients whose indication for bypass surgery did not include rest pain or tissue loss were excluded from our analysis.

Our exposure variable was the degree to which the patients were affected by diabetes. Subjects were divided into three cohorts based on diabetic status: nondiabetic (ND), noninsulin-dependent diabetic (NIDD), and insulindependent diabetic (IDD). Insulin dependence was defined as reliance on insulin administration at baseline to control diabetes. Patients who were diabetic, but did not rely on insulin, were categorized as noninsulin dependent. In our study, insulin dependence is not synonymous with type I diabetes, but rather defines the patient-level pattern of insulin utilization at the time of LEB.

Outcome measures

Our study had two main outcome measures: in-hospital mortality and major adverse events (MAEs) – a composite outcome, including myocardial infarction (MI), congestive heart failure (CHF), dysrhythmia, wound infection, renal insufficiency, and major amputation. Patients experiencing one or more of the above adverse events were included in the composite outcome. Only MAEs that occurred during the surgical admission were included in this analysis. MI included both ST-segment elevation MIs, diagnosed by a change from baseline electrocardiogram, and non-ST elevation MIs, diagnosed by elevated cardiac enzymes in the setting of an electrocardiogram unchanged from baseline. Dysrhythmia included the presence of any new cardiac rhythm that was not normal sinus. The diagnosis of postoperative CHF was left to the discretion of the operating surgeon and was based primarily on clinical signs and symptoms. The designation of wound infection was also made on clinical suspicion or data indicative of wound infection (culture positive or requiring antibiotics). Renal insufficiency was defined as an increase in plasma creatinine of ≥0.5 from baseline or the need for dialysis. Last, major amputation was defined as amputation occurring below the knee or more proximal. Outcomes for each strata of patients with diabetes were compared to the patients who were nondiabetic.

Statistical analysis

First, using data from all patients, we examined univariate associations between individual patient characteristics and each of our main outcome measures. Rates were compared across strata (ND, NIDD, and IDD) using χ². Differences with P < .05 were considered to be significant.

Next, multivariable analyses were performed to adjust for confounding due to baseline risk factors and then to identify patient-level risk factors associated with in-hospital death or MAEs after LEB surgery. All variables with univariate significance P < .20 were used to develop hierarchical regression models. Hierarchical models were used to account for the nonrandom assignment of patients within hospitals. Backward stepwise logistic regression was used to create a hierarchical model for death, as well as a separate model for MAEs. We used nested likelihood ratio tests while clustering for center-effect. Discrimination for final models was assessed using the area under the receiver operating curve (AUC), and model calibration was assessed using goodness of fit (GOF) testing across strata of patient risk. Finally, to examine the effect of patient risk within each strata of diabetes, we developed models for mortality and MAEs within each strata of diabetes (ND, NIDD, and IDD). Statistical significance was defined as P ≤ .05, and data were analyzed using Stata 11.0 (StataCorp, College Station, Tex).

RESULTS

Patient demographics

Baseline patient characteristics, stratified by type of diabetes (ND, NIDD, IDD), are represented in Table I.¹⁰ Overall, 59% of patients (1161) undergoing LEB surgery for CLI were diabetic. Of patients with diabetes, 47% were categorized as NIDD. Among patients who are NIDD, 122 (23%) were described as diet controlled, whereas 419 (77%) relied on oral hypoglycemics. Overall, 620 patients (53% of all patients with diabetes, and 31% of the entire cohort) were dependent on insulin preoperatively for glucose control.

Table I.

Patient and operative characteristics by diabetes status

	ND n = 816	NIDD n = 541	IDD n = 620	P value (ND to NIDD)	P value (ND to IDD)	P value (ND to IDD)
Patient characteristics
Male, %	67.2	67.5	66.8	.905	.879	.802
Age, %				.281	<.001	<.001
<60 years	19.5	16.6	23.9
60–80 years	53.7	57.7	61.1
≥80 years	26.8	25.7	15.0
Obese, % (BMI ≥30)a	11.8	20.3	24.4	<.001	<.001	.101
Hypertension, %	81.7	92.2	93.9	<.001	<.001	.313
CAD, %	32.6	40.3	50.0	.004	<.001	.001
Chronic obstructive pulmonary disease, %	32.5	25.3	30.5	.005	.421	.051
Smoking, %	83.8	81.1	77.8	.196	.004	.169
CHF, %	14.7	19.8	31.3	.014	<.001	<.001
Chronic renal insufficiency, % (Cr ≥1.8)	11.2	15.2	31.4	.036	<.001	<.001
Dialysis, %	4.4	7.6	15.7	.014	<.001	<.001
Prior percutaneous intervention, %	10.3	10.5	12.1	.886	.281	.403
Ankle brachial index, mean	0.46	0.51	0.58	.046	<.001	.018
Indication for surgery, %				<.001	<.001	<.001
Rest pain	51.4	30.1	19.5
Tissue loss	48.6	69.9	80.5
Operative characteristics
General anesthesia, %	77.0	75.0	79.7	.417	.217	.059
Common femoral artery inflow, %	66.9	64.9	62.7	.439	.101	.450
Conduit, %
Reversed GSV	21.7	23.8	19.0	.353	.217	.046
Nonreversed GSV	48.0	44.4	44.4	.184	.166	.998
Arm vein	1.8	2.6	4.0	.350	.012	.173
Prosthetic	26.1	24.8	27.4	.581	.576	.306

BMI, Body mass index; CAD, coronary artery disease; CHF, congestive heart failure; Cr, creatinine; GSV, great saphenous vein; IDD, insulin-dependent diabetic; ND, nondiabetic; NIDD, noninsulin-dependent diabetic.

^aOf note, BMI was not recorded for 468 of 1977 patients (24%) in our analysis. Therefore, we used imputation methods¹⁰ and assigned these patients the mean BMI in our cohort of 27.2 kg/m².

We found that patients who are IDD had an increased number and magnitude of surgical risk factors compared to both patients who are ND and NIDD. For example, when compared with patients who are ND and NIDD, patients who are IDD were significantly more likely than the other two groups to have coronary artery disease (CAD; 32.6% ND; 40.3% NIDD; and 50.0% IDD; P < .001), CHF (14.7% ND; 19.8% NIDD; and 31.3% IDD; P < .001), and chronic renal insufficiency (11.2% ND; 15.2% NIDD; and 31.4% IDD; P < .001), pointing out the need for risk adjustment when comparing outcomes across these groups.

Main outcome measure: In-hospital mortality

In univariate analyses, overall in-hospital mortality was 2.2% (n = 44 of 1977) after LEB surgery. Crude mortality rates did not differ for patients who are ND (1.7%), NIDD (3.1%), and IDD (2.1%; P = .211).

In multivariable analysis, studying the effect of type of diabetes across all patients, after adjustment for differences in baseline patient characteristics, diabetes, independent of type, was not a predictor of mortality. Characteristics that were most strongly associated with in-hospital mortality included age greater than 80 (odds ratio [OR], 2.2; 95% confidence interval [CI], 1.2–4.1; P= .01), CHF (OR, 3.5; 95% CI, 1.9–6.4; P< .001), and tissue loss as the indication for surgery (OR, 3.3; 95% CI, 1.6–6.9; P = .002; Table II).

Table II.

Multivariate models for predicting in-hospital mortality after LEB surgery in all patients with CLI

Covariates	OR	95% CI	P value
Age >80 years	2.18	1.17–4.07	.01
CHF	3.53	1.94–6.41	<.001
Dialysis dependence	1.79	1.07–2.96	.024
Obesity	0.12	0.04–0.40	<.001
Aspirin use	0.57	0.33–0.97	.039
Tissue loss as indication for surgery	3.28	1.56–6.91	.002
Nonuse of GSV	2.19	1.23–3.90	<.001
AUC, 0.8130; GOF, P = .98

AUC, Area under curve; CHF, congestive heart failure; CI, confidence interval; CLI, critical limb ischemia; GOF, goodness of fit; GSV, great saphenous vein; LEB, lower-extremity bypass; OR, odds ratio.

Within our diabetes type-specific models, we identified several patient-level risk factors associated with postoperative mortality. For example, within patients who are IDD, independent risk factors associated with in-hospital mortality were CHF (OR, 9.1; 95% CI, 2.6–32.7; P = .001), bypass origin distal to the common femoral artery (CFA; OR, 3.9; 95% CI, 1.2–13.0; P = .025), and lack of great saphenous vein (GSV) conduit (OR, 2.9; 95% CI, 1.3–6.7; P = .010). Indication for surgery (rest pain vs tissue loss) was not included in this model because no patients who were IDD who were operated on for an indication of rest pain died during their surgical admission. This predictive model had relatively good discrimination (AUC, 0.83) and calibration (GOF, P= .532). Similarly, within patients who were NIDD, independent predictors of in-hospital mortality were age >80 (OR, 3.1; 95% CI, 1.1–8.8; P = .028), CAD (OR, 3.1; 95% CI, 1.5–6.4; P = .002), and tissue loss as the indication for surgery (OR, 3.1; 95% CI, 1.1–8.8; P= .028). Hypertension was protective against mortality in this cohort of patients (OR, 0.21; 95% CI, 0.1–0.9; P = .030). This predictive model had modest discrimination (AUC, 0.75) and good calibration (GOF, P = .994).

Main outcome measure: Major adverse events

Overall, 19.9% (n = 393) of patients experienced an MAE during the admission after surgery. In univariate analysis, crude rates of wound infection before discharge (5.2% ND; 4.8% NIDD; and 5.8% IDD) and dysrhythmia (4.4% ND; 5.6% NIDD; and 6.5% IDD) did not differ across categories of diabetes. However, rates of major amputation, MI, CHF, and renal insufficiency were all directly related to the type of diabetes (Fig). Finally, overall MAE rates were also directly related to the type of diabetes (15.1% ND; 21.1% NIDD; and 25.2% IDD; P < .001).

Fig.

Crude rates of major adverse events (MAEs) after lower-extremity bypass surgery, stratified by patient diabetes status: nondiabetic (ND), noninsulin-dependent diabetic (NIDD), and insulin-dependent diabetic (IDD).

In multivariable analyses considering the entire cohort, MAE rates revealed a similar diabetic treatment-type relationship. Patients with NIDD had a higher likelihood of experiencing MAEs than did patients who were ND (OR, 1.41; 95% CI, 1.1–1.7; P < .001), and those with IDD also experienced increased risk (OR, 1.53; 95% CI, 1.3–1.8; P < .001). Other independent predictors of MAEs included female gender, CHF, CAD, estimated glomerular filtration rate (eGFR) <60, and bypass origin distal to the CFA (Table III). However, this model’s ability to predict MAEs had only modest discrimination (AUC, 0.65), likely due to the use of a composite end point and good calibration (GOF, P < .981).

Table III.

Multivariate models for predicting MAEs after LEB surgery in all patients with CLI

Covariates	OR	95% CI	P value
Diabetes
ND	Referent	—	—
NIDD	1.41	1.15–1.74	.001
IDD	1.53	1.32–1.78	<.001
CHF	2.08	1.47–2.95	<.001
CAD	1.36	1.07–1.73	.012
eGFR <60	1.40	1.10–1.77	.006
Female gender	1.22	1.08–1.39	.001
Inflow distal to the CFA	1.32	1.09–1.61	.005
AUC, 0.6462

AUC, Area under curve; CAD, coronary artery disease; CFA, common femoral artery; CHF, congestive heart failure; CI, confidence interval; CLI, critical limb ischemia; eGFR, estimated glomerular filtration rate; IDD, insulin-dependent diabetic; LEB, lower-extremity bypass; MAEs, major adverse events; ND, nondiabetic; NIDD, noninsulin-dependent diabetic; OR, odds ratio.

Finally, we sought to define predictors of postoperative MAEs among patients within each subclass of diabetes. Among patients who are NIDD, predictors of postoperative MAEs included CAD (OR, 1.8; 95% CI, 1.3–2.5; P = .001), CHF (OR, 2.5; 95% CI, 1.3–4.6; P = .005), and the use of GSVconduit (OR, 1.7; 95% CI, 1.2–2.5; P = .004; Table IV). This model demonstrated modest discrimination (AUC, 0.65) and good calibration (GOF, P = .284). Within multivariable analysis studying only patients who were IDD, independent risk factors associated with MAEs were eGFR <60 (OR, 1.4; 95% CI, 1.1–1.8; P = .002) and CHF (OR, 2.3; 95% CI, 1.5–3.6; P < .001; Table IV). As with previous models for MAEs, this model demonstrated modest discrimination (AUC, 0.61) and good calibration (GOF, P = .321).

Table IV.

Multivariate models for predicting MAEs after LEB surgery in NIDDs and IDDs with CLI

	NIDDs			IDDs
Covariates	OR	95% CI	P value	OR	95% CI	P value
CAD	1.76	1.26–2.46	.001	—	—	—
CHF	2.46	1.31–4.61	.005	2.31	1.48–3.62	<.001
eGFR <60	—	—	—	1.40	1.13–1.75	.002
Use of GSV conduit	1.70	1.18–2.45	.004	—	—	—
AUC		0.6482			0.6124

AUC, Area under curve; CAD, coronary artery disease; CHF, congestive heart failure; CI, confidence interval; CLI, critical limb ischemia; eGFR, estimated glomerular filtration rate; GSV, great saphenous vein; IDDs, insulin-dependent diabetics; LEB, lower-extremity bypass; MAEs, major adverse events; NIDDs, noninsulin-dependent diabetics; OR, odds ratio.

DISCUSSION

Impact of diabetes on outcomes after vascular surgery

The relative impact of diabetes on postoperative outcomes after vascular surgery remains controversial,^4–7 with several studies demonstrating increased risk of perioperative morbidity⁴ and mortality¹¹ among diabetics, whereas others report no added risk in this population of patients.^6,12 In 2004, Virkkunen et al⁴ studied 5709 bypass surgeries performed for an indication of CLI and found that, after adjustment for differences in baseline comorbidities, patients with diabetes had increased risk of below-knee amputation (OR, 1.7), cardiac complications (OR, 1.5), and wound infection (OR, 1.3) compared to patients who were ND. Similarly, Luther et al¹¹ reported higher rates of perioperative mortality (14% vs 1%) and worse 5-year survival (42% vs 71%) among patients with diabetes compared to those without diabetes undergoing femorotibial reconstruction for CLI. Further, Roghi et al¹³ found that diabetes was an independent predictor for major cardiac events (OR, 2.5) among intermediate-risk patients undergoing vascular surgery, including LEB.

However, others have reported evidence to the contrary, including Akbari et al,⁶ who demonstrated reduced in-hospital mortality in patients with diabetes compared to nondiabetics (0.9% vs 4.2%) and reported no difference between the groups in 5-year survival or limb salvage. Likewise, Hamdan et al¹² reported that diabetes is protective against perioperative mortality (OR, 0.55) and is not an independent predictor of MI after lower-extremity revascularization.

Critical review of previous studies suggests that there may be room to better define the impact of diabetes on outcomes after LEB. First, previous studies made little attempts to control for differences in baseline patient and operative characteristics.⁶ Second, some of the studies were limited in sample size and had low event rates,¹³ and most studies aggregated multiple types of vascular surgery as a combined entity.¹³

Our study, derived from a large sample of generalizable data from real-world vascular practice, limits the analysis to a cohort of patients undergoing LEB for CLI, and provides information pertaining to a wide range of clinically important outcomes. Further, although we found that diabetics on the whole suffer from increased rates of several postoperative complications, we also examined the impact of the type of diabetes on postoperative outcomes after LEB and found that this stratification offers additional important information.

Impact of type of diabetes on outcomes after vascular surgery

Although few prior studies have examined the impact of type of diabetes (determined by the need for home insulin) on outcomes in patients undergoing all types of vascular surgery, those studies which have been published on this topic have differed in their conclusions. For example, in 2002, Axelrod et al⁸ published a study performed in a veteran population showing that IDD is a significant predictor of increased risk for cardiovascular events after vascular surgery (adjusted OR, 1.5; 95% CI, 1.5–1.9). However, this study did not find significant associations between diabetes and postoperative mortality or NIDD and other adverse postoperative outcomes. These results differ from ours which suggest that patients who were NIDD fair significantly worse than their nondiabetic counterparts, both with regard to rates of inhospital mortality and other MAEs.

Other studies have reached similar conclusions to ours. In 2007, Hertzer et al⁷ also examined outcomes stratified by type of diabetes after LEB. In this descriptive analysis of over 600 LEB procedures performed by a single surgeon over a 16-year period, postoperative complications among patients with diabetes are similar to those presented in our series. However, Hertzer et al⁷ also examined long-term outcomes and demonstrated significantly higher rates of 1- and 5-year mortality among both patients who were NIDD (OR, 1.4; 95% CI, 1.1–1.8) and patients who were IDD (OR, 1.5; 95% CI, 1.2–1.8). This study also showed that IDD is a significant predictor of both short-term and long-term amputation.

The impact of severity of diabetes has been examined more extensively in cardiac surgery patients, with findings similar to ours. For example, Carson et al¹⁴ examined the impact of type of diabetes on outcomes after coronary artery bypass graft (CABG) in over 140,000 patients and demonstrated a clear dose-response curve with risk of infection (deep sternal, leg, septicemia, urinary tract infection, or pneumonia) significantly higher among patients who were NIDD than those who were ND (adjusted OR, 1.24) and even higher among patients who were IDD (OR, 1.55). Another study, by Zacharias et al,¹⁵ reports IDD as an independent predictor of sternal wound infection (OR, 5.9) after CABG or combined CABG/valve surgery. Similar studies have shown insulin dependence portrays increased risk of perioperative mortality¹⁶ and longer length of stay¹⁷ compared to NIDD.

Our findings add important context to this literature in two ways. First, although prior studies reporting outcomes among diabetics after LEB offer a single surgeon’s experience, ⁷ the data presented here provide the perspective of a large multicenter experience, representing 14 academic and community hospitals and nearly 70 surgeons’ experiences, which may make the findings more generalizable. Second, within our regional quality improvement collaborative, this study served to provoke interest and focus direction toward improving the care of patients with diabetes undergoing LEB. As described in previous work,⁹ our regional collaborative collects data across a broad spectrum of clinical detail in a prospective fashion and reviews this data at biannual meetings. These real-time findings regarding the risks involved for patients with diabetes were of interest to several in our region who participate in quality improvement activities, and served to motivate several initiatives toward improving outcomes for patients with diabetes. Therefore, this work demonstrates the utility of our dataset not only for descriptive analyses of the current circumstances surrounding clinically important measures, but also helps to direct quality improvement efforts toward the largest opportunities for meaningful change.

One additional finding that warrants discussion is the effect of using GSV in those patients with NIDD. In nearly every major article, the use of conduit other than GSV is associated with poorer outcomes. In our study, although this relationship holds true for patients who are IDD, we found among patients who are NIDD, the direction of the effect is reversed (OR, 1.6), with a very wide CI (1.04–2.47). We believe the explanation behind these competing associations is likely multifactorial—a relatively small number of diverse MAEs in the subgroups by insulin type, selection for use of GSV in patients with tissue loss, the need for crural bypass, and the fact that these data represent only short-term outcomes are three of the most likely explanations.

Improving outcomes among diabetics undergoing LEB

We and others have found that patients who are IDD have significantly higher rates of postoperative complications than their ND counterparts,^7,8 and our study suggests that these risks are present, to a lesser degree, in patients who are NIDD as well. This relationship provides the framework for designing quality improvement initiatives aimed at improving the care that we provide for patients undergoing LEB, indicating that initial efforts should focus largely on patients who are IDD. Therefore, our future work will aim to examine the impact of various measures for improving and standardizing the management of patients with diabetes postoperatively in an effort to reduce postoperative complications. Specifically, at our center and others in our region, we are examining the impact of insulin administration protocols (both intermittent and continuous) and the utility of a glucose management service on glycemic control and outcomes after LEB. Preliminary results indicate that both interventions result in significant improvements in glucose management compared to previous standards of care¹⁸ and will serve for a larger, multicenter trial aimed at improving the quality of care we provide diabetics undergoing vascular surgery.

Study limitations

Our study has several limitations. First, the MAEs reported only account for events occurring during a patient’s surgical admission and do not reflect 30-day or longer-term outcomes. This limitation is likely the reason our study failed to identify statistically significant differences in some of the independent adverse events that are more likely to occur after a patient’s discharge from the hospital, such as wound infection and amputation. However, we have established that patients with diabetes are at increased risk for many complications in the immediate postoperative period and suspect these trends would persist at follow-up as well. Second, because additional measures of diabetes disease severity, such as HA1C and actual glucose values, were not available for this study, the use of insulin was used as a surrogate for disease severity. However, we recognize that not all patients who are IDD have more severe disease than patients who are NIDD, and there is an associated potential for confounding by extent of medical therapy. Further, data pertaining to other potentially important variables that may impact the risk of adverse postoperative outcomes, such as the presence of a preoperative infection or anemia, and details regarding outflow were not available in the database. Third, the use of a composite outcome (MAEs) limits the predictive ability of our multivariable model because each included adverse event is likely to have different independent predictors. However, we chose to use a combined end point to increase the power of the analysis and to include all clinically important outcomes in our study, as this end point incorporates a variety of outcomes valuable to clinicians who care for diabetics after LEB. Fourth, the impact of diabetes on outcomes after LEB may be attenuated in our analysis due to the fact that our multivariable models adjusted for conditions that may be the result of poorly controlled diabetes, and therefore mediate the relationship between diabetes and outcomes, rather than being independent of these relationships. Likewise, there are undeniably unmeasured confounders that are unaccounted for in this analysis. Fifth, this analysis does not account for multiple comparisons for two reasons. First, whereas this is a single study evaluating more than one hypothesis, each hypothesis is tested using a single test and many logistic regression models have previously been published without accounting for multiple comparisons. Further, given the relatively low event rates in our analysis, the risk of a type II error for some outcomes is not negligible and therefore further increasing this risk by lowering the alpha is not ideal. Finally, while the advantages of reporting prospective data from multiple centers are obvious, the disadvantage to this approach lies in the inherent variation in both the management of patients with diabetes and the rates of complications across individual centers. While our models used hierarchical methods to account for this center-specific effect (patients “nested” within centers), our future efforts will aim to limit variation and better standardize both diabetic care and diabetic care measurement (such as the time and frequency of blood glucose measurement) across centers in our collaborative.

CONCLUSIONS

Although only NIDD seems to have a significant impact on short-term mortality after LEB surgery, all patients with diabetes are at increased risk for other major postoperative complications such as cardiovascular events, amputation, renal insufficiency, and wound infection. The impact of diabetes on many of these outcomes is greatest among IDDs. Future work is needed to identify best practices for managing diabetics in the postoperative period, and these efforts should begin by focusing on IDDs.