Contrast-induced acute kidney injury (CI-AKI) is a complication of coronary angiography that occurs commonly in the setting of acute myocardial infarction (AMI) (1- 2) and is associated with severe adverse events, including permanent renal impairment (2- 4), higher in-hospital and long-term mortality (2,5- 10), recurrent ischemic events (10- 11), increased length of stay (2,12), and higher costs (13). Proposed pathophysiologic mechanisms through which contrast administration may potentiate renal injury include oxidative stress, free radical damage, and endothelial dysfunction (3,14- 15). All of these processes also are activated in the setting of hyperglycemia (3,16), which is common in patients with AMIs (17) and has adverse prognostic implications (16,18- 20), particularly among those who do not have established diabetes (19- 21). Thus, it is possible that a combination of pre-procedural hyperglycemia and contrast exposure during coronary angiography could significantly increase the risk for CI-AKI.

However, while diabetes is a well-recognized risk factor for CI-AKI (10,12,14), the association between pre-procedural blood glucose levels and CI-AKI risk (regardless of pre-existing diabetes) is unknown. Since nearly 50% of hyperglycemic AMI patients do not have known diabetes (22), establishing whether elevated glucose levels before coronary angiography are related to subsequent CI-AKI risk would have important clinical implications; primarily by identifying a previously unrecognized high-risk group that might benefit from pre-procedural CI-AKI prophylaxis and also by highlighting a potential target for intervention in the prevention of CI-AKI.

Accordingly, we analyzed data from the Cerner Corporation's (Kansas City, Missouri) Health Facts database, a contemporary registry of patients admitted to 40 hospitals across the U.S., to define the association between pre-procedural hyperglycemia and the risk for CI-AKI during AMI. This database was selected because of its extensive collection of laboratory data, including detailed glucose measurements and assessments of renal function, in a large consecutive cohort of AMI patients. We specifically sought to determine the relationship between pre-procedural glucose levels and the risk for subsequent CI-AKI among patients with and without known diabetes who are referred for coronary angiography in the setting of AMI.

Details of the Health Facts database have been described previously (22). Between January 1, 2000, and December 31, 2005, the Health Facts database collected deidentified information regarding consecutive patients hospitalized at 40 U.S. medical centers. Variables included demographics, medical history, and comorbidities (determined from International Classification of Diseases-Ninth Revision billing codes), laboratory studies (including all venous and fingerstick blood glucose measurements, as well as all serum creatinine levels during hospitalization), medications, in-hospital procedures (including coronary angiography and percutaneous coronary intervention [PCI]), in-hospital mortality, and hospital characteristics (geographic region, number of beds, surgical and procedural capabilities, and teaching vs. nonteaching status).

We identified all patients hospitalized with primary discharge diagnoses of AMI (using International Classification of Diseases-Ninth Revision-Clinical Modification codes 410.xx and excluding 410.x2, which represents readmission after AMI) who had glucose measurements on admission and at least 1 documented abnormal troponin or creatine kinase-myocardial band level. Subsequently, those patients who were transferred from or to other acute care facilities were excluded, as complete laboratory and medication administration details for the entire episode of AMI care were not available. For this analysis, we further restricted the sample to those who underwent coronary angiography during the AMI hospitalization and whose laboratory studies included at least 1 pre-procedural measurement of serum creatinine and at least 1 repeat evaluation within 48 h after the procedure. Of 8,786 AMI patients who underwent coronary angiography during the index hospitalization, 6,358 patients (72%) had complete pre- and post-procedural laboratory data and comprised the study cohort.

The Health Facts database provided access to all of the patients' glucose values, including their measurement times. For clarity, all plasma and capillary glucose measurements are subsequently referred to as “blood glucose.” The glucose level evaluated for each patient was the pre-procedural measurement obtained closest to the time of diagnostic angiography. As in prior studies (22- 24), patients were classified as having recognized diabetes if they had corresponding International Classification of Diseases-Ninth Revision-Clinical Modification codes or were treated with an oral antihyperglycemic agent or any extended-release insulin during the hospitalization.

The outcome for this study was the occurrence of CI-AKI, defined as an absolute serum creatinine increase ≥0.3 mg/dl or a relative increase in serum creatinine ≥50% that occurred within 48 h after coronary angiography, in accordance with the Acute Kidney Injury Network criteria (25). The rates of CI-AKI were calculated using pre- and post-procedural serum creatinine measurements. Pre-procedural serum creatinine level was defined as the measurement obtained during hospitalization that occurred before and closest to the time of coronary angiography. Post-procedural creatinine level was defined as the highest serum creatinine measurement within 48 h after coronary angiography. If a patient underwent more than 1 coronary angiography procedure, pre- and post-procedural creatinine levels for the first procedure were considered for this analysis.

Patients were stratified into 5 groups according to pre-procedural glucose levels: <110 mg/dl, 110 to <140 mg/dl, 140 to <170 mg/dl, 170 to <200 mg/dl, and ≥200 mg/dl. Baseline characteristics were compared among groups using chi-square analysis for categorical variables and analysis of variance for continuous variables.

Rates of CI-AKI were compared between subgroups of pre-procedural glucose levels using chi-square analysis, first in the entire sample and then within the a priori defined subgroups of patients with and without known diabetes. Multivariable logistic regression models were subsequently developed to evaluate whether the association between pre-procedural glucose values and CI-AKI persisted after adjustment for other patient characteristics and potential confounders. In these models, pre-procedural glucose level was included as a categorical variable, using the 5 groups described previously. Patient characteristics clinically considered to be prognostically important, and covariates identified in bivariate analyses as predictors of CI-AKI (at a level of significance of p ≤ 0.05), were entered into the models. Covariates included demographic factors (age, gender, and race), comorbidities (diabetes, heart failure, hypertension, and cerebrovascular disease), laboratory values (admission hematocrit and peak troponin value), mechanical reperfusion during hospitalization (coronary artery bypass grafting [CABG] or PCI), medications during hospitalization (aspirin, clopidogrel, ticlopidine, beta-blockers, calcium-channel blockers, angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, diuretics, 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors, oral antihyperglycemic agents, and insulin), hospital site, and hospital length of stay. A pre-procedural glucose × diabetes interaction term was also introduced into the models to formally test whether the glucose-associated risk for CI-AKI differed in patients with and without known diabetes.

Importantly, the models also adjusted for the baseline estimated glomerular filtration rate (GFR), calculated according to the Modification of Diet in Renal Disease study equation (26). Because contrast volume is a significant factor in the development of CI-AKI (1), models adjusted for the performance of PCI (categorized as none, single vessel, or multivessel), as well as the number of coronary angiography and PCI procedures during hospitalization (all surrogates of contrast volume). In addition, to account for pre-procedural CI-AKI prophylaxis measures, the final models also adjusted for the administration of sodium bicarbonate. Models were then repeated within the subgroups of patients with and without diabetes.

Several sensitivity analyses were performed. First, multivariable models were repeated after excluding all patients who underwent CABG within 48 h of coronary angiography, to ensure that our results were not confounded by the potential impact of CABG on renal function. Second, cardiogenic shock was incorporated as a covariate into the models, given its associations with both hyperglycemia and CI-AKI (2,19). Third, a different definition of CI-AKI was used (increase in serum creatinine >0.5 mg/dl or >25%).

Additional analyses were performed to examine whether the relationship between pre-procedural glucose levels and CI-AKI differed according to baseline renal function. To accomplish this, patients were divided into 3 groups according to baseline GFR (<30 ml/min, 30 to 59 ml/min, and ≥60 ml/min). The rates of CI-AKI were then compared across the 5 pre-procedural glucose categories (<110 mg/dl, 110 to <140 mg/dl, 140 to <170 mg/dl, 170 to <200 mg/dl, and ≥200 mg/dl) within each of the GFR subclasses, using chi-square analysis. These analyses were then performed separately in patients with and without diabetes.

A 2-sided p value of ≤0.05 was considered statistically significant. Analyses were conducted with SAS version 8.02 (SAS Institute Inc., Cary, North Carolina). Use of the Health Facts database was approved by the Saint Luke's Mid America Heart Institute Institutional Review Board.

Baseline characteristics are detailed in (Table 1), stratified by pre-procedural glucose levels. Pre-procedural hyperglycemia (glucose ≥140 mg/dl) occurred in 42% of the patient sample, and 48% of patients with elevated glucose before coronary angiography did not have known diabetes. The median time from pre-procedural glucose assessment to coronary angiography was 8.3 h (interquartile range: 5.3 to 16.3 h). Diabetes was present in 1,929 patients (30%). Overall, 823 patients (13%) developed CI-AKI after coronary angiography. Mean creatinine in these patients increased from 1.73 ± 1.5 mg/dl at baseline to a peak of 2.54 ± 1.8 mg/dl, whereas patients without CI-AKI experienced no significant change in serum creatinine (baseline: 1.16 ± 0.8 mg/dl; post-procedure: 1.14 ± 0.7 mg/dl). The median number of post-procedural creatinine measurements was 1 for patients with glucose <110 mg/dl and 2 in all other glucose groups. The median number of coronary angiograms performed during AMI hospitalization was 1, and the median number of PCI procedures during hospitalization was 1 across all 5 of the glucose groups. Overall, 43% of all patients underwent diagnostic coronary angiography only, 50% underwent single-vessel PCI, and 7% underwent multivessel PCI. Median hospital length of stay was 113 h (interquartile range: 74 to 193 h) and is stratified by glucose level in (Table 1).

Compared with patients who had lower pre-procedural glucose, greater proportions of those with higher glucose were female and had prior heart failure, AMIs, and diabetes. Hyperglycemic patients also had lower GFRs, had higher peak troponin levels and lower hematocrit on admission, and were treated more frequently with diuretics, angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, oral antihyperglycemic agents, and insulin. Periprocedural infusions of sodium bicarbonate, in-hospital PCI, and in-hospital CABG were used more frequently in patients with higher glucose levels.

In unadjusted analyses, higher pre-procedural glucose levels were strongly associated with higher rates of CI-AKI (Table 2). However, the nature of the relationship between glucose levels and CI-AKI was significantly different in patients with and without known diabetes (pre-procedural glucose × diabetes interaction p < 0.001). There was a strong association between glucose and CI-AKI risk in patients without diabetes (CI-AKI rates across the 5 glucose groups from lowest to highest: 8.2%, 9.9%, 12.4%, 14.9%, and 24.3%; p < 0.001). In contrast, no significant relationship was observed between glucose levels and CI-AKI in patients with established diabetes (CI-AKI rates across the 5 glucose groups from lowest to highest: 20.9%, 16.1%, 16.3%, 14.8%, and 19.2%; p = 0.24).

After adjusting for confounders (including baseline GFR), the nature of these relationships between glucose levels and CI-AKI risk persisted (Figure 1). In patients without known diabetes, there was a gradual, incremental increase in the risk for CI-AKI associated with higher pre-procedural glucose levels (odds ratios [95% confidence intervals] for glucose groups of 110 to <140 mg/dl, 140 to <170 mg/dl, 170 to <200 mg/dl, and ≥200 mg/dl: 1.31 [1.00 to 1.71], 1.51 [1.11 to 2.10], 1.58 [1.03 to 2.43], and 2.14 [1.46 to 3.14], respectively, vs. glucose <110 mg/dl). However, no significant increase in glucose-associated risk for CI-AKI was seen among patients with established diabetes (corresponding odds ratios [95% confidence intervals]: 0.71 [0.44 to 1.14], 0.82 [0.52 to 1.30], 0.73 [0.45 to 1.12], and 0.94 [0.63 to 1.40]; adjusted p value for glucose × diabetes interaction = 0.04). In 3 sensitivity analyses—first excluding patients who underwent CABG within 48 h of coronary angiography, second adjusting for the rates of cardiogenic shock, and third using a different definition of CI-AKI—the results were similar (data not shown).

Among patients without known diabetes, the relationship between pre-procedural glucose levels and the risk for CI-AKI differed according to baseline GFR (Figure 2). There was a graded, significant relationship between higher pre-procedural glucose levels and greater risk for CI-AKI in patients with GFR ≥60 ml/min (CI-AKI rates across the 5 glucose groups from lowest to highest: 6%, 7%, 9%, 10%, and 18%; p < 0.001) and in those with GFRs of 30 to 59 ml/min (11%, 16%, 19%, 18%, and 27%, respectively; p = 0.003). However, the relationship between glucose and CI-AKI was not significant among patients with severely impaired renal function (GFR <30 ml/min), as risk remained high regardless of glucose level (CI-AKI rates across the 5 glucose groups from lowest to highest: 39%, 30%, 33%, 44%, and 44%; p = 0.66).

Among patients with established diabetes, there was no significant association between pre-procedural glucose levels and CI-AKI rates in any of the GFR subgroups (data not shown).

In this large sample of AMI patients undergoing coronary angiography, we found that the nature of the relationship between pre-procedural glucose levels and the risk for CI-AKI was markedly different among patients with and without established diabetes. Elevated glucose was common and associated with a steep incremental increase in the risk for CI-AKI after coronary angiography among patients who did not have known diabetes. In contrast, no such association was observed among patients with established diabetes, who experienced high CI-AKI rates regardless of pre-procedural glucose levels. Importantly, glucose-associated CI-AKI risk was as high or even higher in patients without diabetes who had significant pre-procedural hyperglycemia, as in those with established diabetes. Moreover, the relationship between higher pre-procedural glucose and greater CI-AKI risk was particularly pronounced in those patients without diabetes who did not have significantly impaired renal function, a group that is not currently considered to be at high risk for CI-AKI. Our findings imply that, at a minimum, hyperglycemic patients without known diabetes should be recognized as a high-risk group for CI-AKI and should be considered for prophylactic measures similar to those used in other high-risk patients.

This study is the first, to our knowledge, to demonstrate that elevated pre-procedural glucose levels are associated with increased risk for CI-AKI in AMI patients without diabetes undergoing coronary angiography. However, the relationship between hyperglycemia and renal injury has been previously described in other patient populations. Specifically, prior studies have demonstrated a relationship between higher glucose levels and greater risk for acute renal failure among critically ill patients hospitalized in intensive care units (27). Elevated pre- and perioperative glucose levels are also independently associated with higher risk for post-operative renal dysfunction among patients undergoing cardiac surgery (28- 29); in fact, pre-operative glucose >140 mg/dl has been incorporated into a prediction model of post-operative kidney injury in this patient population (28). A few relatively small, single-center studies also suggested that pre-diabetes may increase the risk for contrast-induced nephropathy in patients with chronic kidney disease who undergo elective coronary angiography (30) and that metabolic syndrome may increase the risk for acute renal failure in patients with AMIs (31). Our study adds substantially to these data by demonstrating in a large, multicenter cohort that pre-procedural hyperglycemia markedly elevates the risk for CI-AKI in patients without established diabetes, even after accounting for baseline renal function and a multitude of other factors.

Our findings have several potentially important clinical implications. First, although diabetes is a well-known predictor of contrast nephropathy (10,12,14), elevated pre-procedural glucose levels in patients without established diabetes are not currently recognized by clinicians as a risk factor for contrast-mediated renal injury. Since hyperglycemia occurs in over 40% of AMI patients, nearly half of whom have no known diabetes (19,22,32), this group of patients should at least be considered to be at as high a risk for CI-AKI as those with established diabetes. CI-AKI prophylaxis may need to be considered in these patients, as well as careful monitoring of post-procedural renal function. Second, elevated pre-procedural glucose may represent a modifiable intervention target for CI-AKI prevention. Whether hyperglycemia is a marker of increased comorbidity burden and disease severity or a direct mediator of CI-AKI is unknown, but prior studies have demonstrated increased oxidative stress, blunting of free radical scavengers, decreased levels of nitric oxide, and endothelial dysfunction in the setting of elevated glucose (33- 42). Other investigators have shown enhanced susceptibility to ischemic and reperfusion injury in animal models of experimentally induced hyperglycemia (43). Since numerous studies have shown that similar pathophysiologic mechanisms are also involved in contrast nephropathy (3,14- 15), the effects of hyperglycemia in our analysis may reflect a “double insult” to renal function that previously has not been described in patients without established diabetes. Although no conclusions regarding a possible cause-and-effect relationship between hyperglycemia and CI-AKI can be drawn from our study, several randomized trials in critically ill patients have demonstrated that glycemic control with insulin may reduce the rates of acute renal failure (44- 46). Because the findings of these clinical trials cannot be automatically extrapolated to AMI patients undergoing coronary angiography, prospective randomized studies are needed to determine whether glucose control can reduce the rates of CI-AKI in the setting of AMI.

The reasons behind the different relationships between glucose levels and CI-AKI risk among patients with and without known diabetes are unclear but have been previously described with other outcomes, such as mortality (19- 22). Several possible explanations exist for this phenomenon. First, as previously demonstrated in other studies (19,32), hyperglycemic AMI patients without known diabetes typically receive much less aggressive therapy for glucose control compared with those who have known diabetes. Since intensive glucose control with insulin has been previously shown to reduce the rates of renal injury in critically ill patients (44,46- 47), this significant discrepancy in treatment rates may have an effect on glucose-associated CI-AKI rates. Second, since diabetes is a well-recognized risk factor for CI-AKI, patients with diabetes may receive more aggressive pre-procedural CI-AKI prophylaxis, attenuating the potential effect of hyperglycemia on renal injury. Third, some hyperglycemic AMI patients without known diabetes (particularly those with glucose >200 mg/dl) may actually have diabetes that has been neither recognized nor adequately managed and thus may represent a higher-risk cohort for the development of CI-AKI. Even if not definitively diagnosed with diabetes by hospital discharge, hyperglycemia in AMI patients likely represents glucose dysmetabolism, and the simple act of evaluating pre-catheterization glucose should alert clinicians to the elevated risk for renal injury. Finally, it is also possible that a greater degree of stress (illness severity) is needed for patients without established diabetes to achieve the same degree of hyperglycemia as their counterparts with diabetes. However, the difference in the association between hyperglycemia and CI-AKI among patients with and without known diabetes persisted even after accounting for disease severity indicators, including infarct size.

The results of our study should be interpreted in the context of several potential limitations. First, the volume of contrast administered in each procedure was not available; however, we believe that by using the performance of PCI (including the differentiation between single-vessel and multivessel PCI) in the multivariable models, we introduced a surrogate for presumed higher contrast volume (compared with diagnostic angiography alone). Moreover, there is no a priori reason to expect that hyperglycemic patients would be more likely to receive larger or smaller doses of contrast material than normoglycemic patients. Second, although we rigorously attempted to account for baseline differences between patients in different pre-procedural glucose groups, residual confounding cannot be excluded. Specifically, we were unable to control for certain clinical variables, such as the presence of ST-segment elevation and left ventricular ejection fraction. However, adjustment for ST-segment elevation and ejection fraction did not affect the nature of glucose-associated risk for mortality in our prior studies (19), and we were able to control for other variables, such as infarct size (using peak troponin levels) and numerous other patient factors. Third, our data are limited to AMI patients undergoing angiography, and the relationship between hyperglycemia and CI-AKI in other patient populations or with other diagnostic modalities requiring intravenous contrast (e.g., computed tomography scans, other angiograms) remains unclear. Fourth, we do not know how many of those patients without known diabetes were subsequently diagnosed with diabetes after discharge; however, our specific focus was to evaluate the CI-AKI risk associated with pre-procedural glucose values in patients who were not recognized as having diabetes at the time of AMI hospitalization. Finally, although we adjusted for sodium bicarbonate administration in our analyses, we were unable to determine whether glucose lowering with insulin or the use of other prophylactic CI-AKI therapies (e.g., intravenous hydration or N-acetylcysteine) could reduce the risk for CI-AKI in hyperglycemic AMI patients without established diabetes. This will need to be determined in future prospective investigations.

In summary, elevated pre-procedural glucose levels are associated with greater risk for CI-AKI in AMI patients without established diabetes who undergo coronary angiography. Close monitoring of post-procedural renal function and prophylactic therapies used to prevent CI-AKI should be considered in this patient group.