Contrast-induced acute kidney injury (CI-AKI) is defined as either a 25% increase in or an absolute elevation in serum creatinine (SCr) of 0.5 mg/dL, 48 to 72 hours after parenteral contrast exposure. The objective of this study was to compare the incidence and complications of AKI between patients exposed and those unexposed to intravenous (IV) contrast.
This was a retrospective cohort study using the electronic medical record of adult patients (>18 years) with and without contrast-enhanced abdominal or chest computed tomography (CT) between May 2008 and April 2009. Inclusion criteria were emergency department (ED) patients with normal renal function who received either a contrast-enhanced abdominal or a contrast-enhanced chest CT, compared to those unexposed to IV contrast, with a repeat SCr within 48 to 72 hours. Exclusion criteria were contrast exposure within 7 days before the index visit. CI-AKI in the contrast-exposed group and AKI in the contrast-unexposed group were defined by the same changes in SCr 48 to 72 hours after contrast or ED admission. Data were described by proportions or medians with 95% confidence intervals (CIs) or interquartile ranges (IQR; 25% to 75%). Group comparisons were by Mann-Whitney U or Fisher's exact test (α = 0.05, two tails).
The contrast-exposed (n = 773) and contrast-unexposed (n = 2,956) patients were evenly matched for initial demographic, renal, and metabolic parameters. The incidence of CI-AKI/AKI was significantly higher for the patients unexposed versus exposed to contrast (8.96% vs. 5.69%, p = 0.003). There was no significant difference in mortality rates between contrast-exposed and unexposed patients (9.09% vs. 6.79%, p = 0.533).
The definition of CI-AKI for ED patients with normal renal function may not represent a true clinical entity and the definition warrants revision.
Refleja la Definición Actual de Lesión Renal Aguda Inducida por Contraste una Entidad Clínica Real?
La lesión renal aguda inducida por contraste (LRA-IC) se define indistintamente como un incremento del 25% o una elevación absoluta de 0,5 mg/dl de la creatinina plasmática (Crpm) entre las 48 a 72 horas tras la exposición a contraste parenteral. El objetivo de este estudio fue comparar la incidencia y las complicaciones de la LRA entre los pacientes expuestos y aquéllos no expuestos a contraste intravenoso (IV).
Estudio de cohorte retrospectivo mediante la revisión de las historias clínicas electrónicas de los pacientes adultos (>18 años) con y sin exposición a contraste durante una tomografía computerizada (TC) torácica o abdominal entre mayo de 2008 y abril de 2009. Los criterios de inclusión fueron pacientes del servicio de urgencias (SU) con función renal normal que recibieron contraste en la TC de tórax o abdominal comparados con aquéllos no expuestos a contraste IV, con Crpm repetida entre las 48 a 72 horas. El criterio de exclusión fueron la administración de contraste en los últimos siete días anteriores a la consulta actual. La LRA-IC en el grupo expuesto a contraste y la LRA en el grupo no expuesto a contraste se definieron por los mismos cambios en la Crpm a las 48 a 72 horas tras el contraste o la llegada al SU. Los datos se describieron como proporciones o medianas con intervalos de confianza (IC) al 95% o rangos intercuartílicos (RIC 25% a 75%). Las comparaciones de los grupos se realizaron mediante el test de la U de Mann-Whitney o el test exacto de Fisher (error alfa = 0,05).
Los pacientes expuestos a contraste (n = 773) y los no expuestos a contraste (n = 2.956) estaban distribuidos de forma homogénea según los parámetros demográficos, renales y metabólicos. La incidencia de LRA-CI/LRA fue significativamente mayor en los pacientes no expuestos que en los expuestos a contraste (8,96% vs. 5,69%, p = 0,003). No hubo diferencias estadísticamente significativas en los porcentajes de mortalidad entre los pacientes expuestos y no expuestos a contraste (9,09% vs. 6,79%, p = 0,533).
La definición de LRA-IC para los pacientes del SU con función renal normal puede no representar una entidad clínica real y la definición requiere ser revisada.
Since the first reported case of acute kidney injury (AKI) following intravenous (IV) contrast by Bartels et al. in 1954, a substantial medical literature has developed to describe the incidence, risks, prognosis, and prevention of contrast-induced acute kidney injury (CI-AKI). A PubMed search combining the terms “contrast media” with “acute kidney injury” found over 1,100 publications from 1966 to December 2011 in the Medline database. This intense focus on CI-AKI by the medical community seems justified by the fact that this completely iatrogenic cause of renal failure represents one of the most common causes of hospital acquired AKI, accounting for 11% to 12% of cases. Acute renal replacement will be required for 3% to 15% of CI-AKI patients. Hospitalized patients complicated by CI-AKI, compared to patients with similar premorbid characteristics, experience longer hospital stays[6-8]; and suffer significantly higher in-hospital[9-11] and long-term mortality rates. Those patients who survive CI-AKI are then predisposed to continued loss of kidney function.[13, 14]
The standard case definition of CI-AKI is not based on patient-centered outcomes such as requirements for renal replacement, long-term renal failure, or mortality. Instead, CI-AKI is defined by a surrogate measure of declining renal function by a rapid and mostly reversible change in serum creatinine (SCr) after contrast exposure. The most commonly quoted definition of CI-AKI is based on an absolute increase of 0.5 mg/dL or a 25% increase in SCr from baseline 48 to 72 hours after contrast exposure. The validation for this definition of CI-AKI, as described by Chertow et al., has been the association of relatively small changes in SCr (>0.5 mg/dL) in hospitalized patients with significant mortality risks not predicted by their initial SCr.
Most CI-AKI studies, however, lack adequate control groups of hospitalized patients unexposed to contrast, where changes in SCr may be ascribed to the progression of their comorbid conditions. Without a control group of unexposed patients, the association of AKI with contrast administration is possibly confounded by the independent risks for renal failure seen in all hospitalized patients. When the incidence of CI-AKI, as defined by changes in SCr, was tested between hospitalized patients exposed and unexposed to contrast, studies by Cramer et al., Heller et al., and Bruce et al. all failed to find a significant difference in the incidence of renal failure. These studies,[17-19] comparing the SCr definition of CI-AKI in contrast-exposed and contrast-unexposed patients, included a broad range of contrast studies, in all subjects regardless of their initial renal function.
We tested the association of IV contrast exposure with AKI using changes in SCr over time. A significant change was defined as an increase in SCr greater than 25%, or an absolute increase of at least 0.5 mg/dL for emergency department (ED) patients' exposed and unexposed to contrast. We specifically limited our study population to patients with SCr < 1.5 mg/dL or an estimated glomerular filtration rate (GFR) > 60 mL/min/1.73 m2. For the contrast-exposed group, we limited the exams to computed tomography (CT) scans of the abdomen and chest, the most common ED contrast exams.
We conducted a retrospective cohort study using the electronic medical records of ED patients to evaluate the incidence of CI-AKI after abdominal or chest CT. The study was approved with exemption from informed consent by the joint institutional review boards of SUNY-Downstate Medical Center and Kings County Hospital Center, in Brooklyn, New York.
Study Setting and Population
The electronic medical records of ED patients seen between May 1, 2008, and April 30, 2009, were reviewed. Patients in the contrast-exposed group were included based on the following criteria: 1) given an IV contrast-enhanced CT exam during their index ED visits, 2) 18 years of age or older, and 3) initial SCr (must have been drawn prior to receiving contrast-enhanced CT for the exposed group) with a repeat multiple SCrs within 48 to 72 hours. If multiple SCr's were measured during the 48- to 72-hour period post contrast, the first SCr in that 48-hour time frame was then recorded for our analysis. Patients were excluded based on the following criteria: 1) those who were receiving hemodialysis at the time of the study, 2) those with an initial SCr > 1.5 mg/dL or estimated GFR < 60 mL/min/1.73 m2, 3) those exposed to contrast up to 7 days before their index ED visits, or 4) patients who were re-exposed to contrast in the 48- to 72-hour period between repeat SCr measurements.
Patients selected for the contrast-unexposed group had the same inclusion/exclusion criteria as the contrast-exposed group with the exception of a contrast-enhanced CT. For both groups, the repeat SCr in 48 to 72 hours was timed from the initial SCr in the ED.
Emergency department patients were divided into contrast-exposed and -unexposed groups. Contrast-exposed was defined as an ED patient who received a contrast-enhanced abdominal or chest CT. For the abdominal/pelvis CT scans, patients received IV 100 mL of low-osmolal Iohexol (300 mg iodine/mL; Omnipaque, GE Healthcare, Princeton, NJ), whereas CT angiograms of the chest used 110 mL of isoosmolal Iodixanol (320 mg iodine/mL; Visipaque, GE Healthcare). The authors (RAS, SW) were trained in data abstraction using a defined data collection sheet. Random records were re-reviewed for accuracy (RAS). The patients' demographics, as well as laboratory values and times ordered, were obtained by reviewing the patients' electronic medical records for initial SCr, repeat SCr, blood urea nitrogen (BUN), anion gap, lactate level, base excess, and bicarbonate. Each patient's medical records were then reviewed to determine if the patient met inclusion criteria as stated. The electronic records were reviewed to determine if the patient received emergent hemodialysis or if death occurred during the incident hospital stay.
Serum electrolytes, BUN, and SCr were all measured on a modular autoanalyzer (Roche Corp., Indianapolis, IN). A venous blood gas sample was obtained shortly after each patient's arrival for analysis (ABL 725, Radiometer, Copenhagen, Denmark), which included the pH, bicarbonate, base excess, and lactate levels.
Data were described by medians with interquartile ranges (IQRs; 25% to 75%) or by proportions with 95% confidence intervals (CIs) as appropriate. Group comparisons were made by Mann-Whitney U or Fisher's exact test (α = 0.05, two tails). Statistics were calculated by SPSS version 18.0 (SPSS, Inc., Chicago, IL). Sample size was based on a previous study, conducted at our institution limited to only trauma patients exposed to contrast, where we found a 5% incidence of CI-AKI. Setting the precision at ± 2% around an expected incidence of 5%, an estimated sample size of 457 contrast-exposed patients was needed to screen retrospectively for CI-AKI. To assure we would have a sufficient sample size for complete data, we requested 12 months of electronic medical records of ED patients.
In our 1-year retrospective analysis of ED patients, 61,431 patients accounted for 90,708 adult visits. Of these, the contrast-exposed group consisted of 3,788 adult patients of whom 1,024 had a repeat creatinine in 48 to 72 hours. Of these, 773 hospitalized patients met the inclusion–exclusion criteria of the study and comprised the data set for the contrast-exposed group (Figure 1). There were 4,292 adult patients during the same period who had an initial SCr and a repeat SCr between 48 and 72 hours after the initial one. Of those with a timely follow up SCr, 2,956 hospitalized patients met all the other inclusion–exclusion criteria for the study and comprised the data set for the contrast-unexposed group (Figure 2; see Table 1 for initial group characteristics). The groups were evenly matched for demographic, renal, and metabolic parameters.
Table 1. Baseline Demographic and Clinical Characteristics of the Patients
Non–contrast-exposed Patients (n = 2,956)
Contrast-exposed Patients (n = 773)
BUN =blood urea nitrogen; IQR = interquartile range.
Table 2 compares the initial renal and metabolic parameters between patients unexposed to contrast with and without AKI. We only found statistically (but not clinically) significantly higher estimated GFRs, with lower SCr and BUNs in the patients who subsequently developed AKI.
The contrast-exposed group included 773 patients with a median age of 52 years (IQR = 37 to 65 years) and chief complaints including abdominal pain (429 of 773, 55%), trauma (186 of 773, 24%), chest pain/shortness of breath (48 of 773, 6%), and other (107 of 773, 14%; see Table 3). The contrast-exposed group included 126 (16%) with diabetes mellitus and 186 (21%) patients greater than 70 years of age.
In patients given contrast (Table 3), initial renal function was similar in both the AKI and the non-AKI groups. Only a slightly higher (statistically but not clinically significant) anion gap was common to both AKI groups in the contrast-exposed and contrast-unexposed patients. We found no differences in patients' age, prevalence of diabetes mellitus, initial bicarbonate, or lactate between AKI compared to non-AKI patients in the contrast-exposed group.
In both groups, the mean SCr declined, with a significant difference (p < 0.05) noted between the two groups. SCr declined slightly more in the contrast-exposed group (−0.104 mg/dL, 95% CI = −0.086 to −0.121 mg/dL) compared with the contrast-unexposed group (−0.036 mg/dL, 95% CI = −0.027 to −0.045 mg/dL).
The incidence of AKI was significantly (p = 0.003) higher for the patients unexposed to contrast (265 of 2,956, 8.96%, 95% CI = 7.98% to 10.24%) compared to those exposed to contrast (44 of 773, 5.69%, 95% CI = 4.27% to 7.55%). No additional AKI patients were identified using the 0.5 mg/dL absolute change in SCr criteria that were not identified with the incremental 25% change in SCr. No patient, whether exposed or unexposed to contrast, developed AKI requiring dialysis during the index hospitalization. For patients meeting the criteria for AKI, we found no significant (p = 0.533) difference in mortality rates between contrast-exposed (4 of 44, 9.09%, 95% CI = 3.59% to 21.16%) and contrast-unexposed (18 of 265, 6.79%, 95% CI = 4.34% to 10.48%) patients.
Our finding of a 5.69% incidence of CI-AKI in ED patients receiving contrast-enhanced CT is consistent with our previous study of CI-AKI (5.1%) limited to only trauma patients. Unlike our previous study of CI-AKI in trauma patients, we did not find that older patients had a significantly higher risk for renal failure in either the contrast-exposed or contrast-unexposed groups.
In the current study, differences in initial renal function or metabolic markers of illness (anion gap, bicarbonate, or lactate) failed to predict AKI in either the contrast-exposed or the contrast-unexposed group. Since this was a retrospective study, it is unlikely that emergency physicians ordering contrast were selecting patients at low risk for CI-AKI. It also does not appear that our sample of contrast-exposed subjects was biased against the risks of CI-AKI by avoiding patients with diabetes mellitus or advanced age > 70 years.[22, 23] Of particular interest in our study was a statistically greater incidence of AKI in the contrast-unexposed compared to contrast-exposed patients. The patients in the contrast-exposed and -unexposed groups who developed AKI by the CI-AKI criteria had similar outcomes with respect for dialysis requirements and mortality rates.
In 2010, Li and Solomon reviewed the literature on the incidence of CI-AKI after IV contrast, including 16 observational and five prospective randomized trials with a range of CI-AKI between 0.3 and 25%. They concluded that for a broad range of in- and outpatients, CI-AKI was relatively uncommon, with an estimated pooled incidence of 5%, consistent with our study. A meta-analysis in 2012, by Kooiman et al., reviewed 40 studies (n = 19,563) of the incidence of CI-AKI after contrast-enhanced CT and found an incidence of 6.4%, similar to our results of 5.7%. Similar to our study in which we failed to find any cases of CI-AKI requiring dialysis, Kooiman et al. also found an extremely low incidence (0.06%) of CI-AKI requiring any form of renal replacement therapy.
Recently, studies by Mitchell et al. and Kim et al. focused specifically on CI-AKI in ED patients. While Mitchell et al. found a significantly higher incidence of CI-AKI of 11% (95% CI = 8.8% to 13.8%), Kim et al. reported only 4% (95% CI = 3.2% to 6.3%), similar to our study. Both studies[26, 27] used the same definition for CI-AKI (25% or 0.5 mg/dL absolute increase of SCr) as we used. The study by Mitchell et al. allowed inclusion of patients with a repeat SCr up to 7 days, compared to the usual standard of three days postcontrast exposure. This longer inclusion time frame by Mitchell et al. may possibly explain their higher CI-AKI incidence by allowing a greater probability for patients to develop nephropathy secondary to other hospital-acquired causes (antibiotics, sepsis, surgery, etc.) of kidney injury in addition to contrast exposure.
Newhouse et al. studied this issue of fluctuations in SCr over time in hospitalized patients not exposed to contrast. This study examined the electronic medical records for the change in SCr over a 5-day period of 32,161 admitted patients and found that greater than 50% of noncontrast patients developed an increase in SCr of at least 25%. This increase in SCr over the 5-day period was not dependent on initial SCr. Patients with a normal range baseline SCr (0.6 to 1.2 mg/dL) compared to those with renal insufficiency (SCr > 2.0 mg/dL) actually had a higher incidence (27% vs. 16%) of a 25% increase in SCr after 5 inpatient days. In a study prior to this study, Rao and Newhouse concluded that the attribution of a change in SCr as a de facto standard for CI-AKI should be considered a logical fallacy: post hoc, ergo propter hoc (after this, therefore because of this).
Multiple other risks of kidney injury encountered by hospitalized patients have confounded the changes in SCr in similar observational studies without an unexposed control group. Lipsitch et al. stated that noncausal associations between outcomes and exposures are the result of mismeasurement (recall bias), confounding, or selection bias. To prevent confounding, Lipsitch et al. suggests designing a negative control experiment where the observation is repeated under conditions not expected to produce the outcome of interest. If the outcome is encountered without the exposure, then a confounding bias may exist.
This form of negative control experiment, in which the incidence of AKI is compared across patients exposed and unexposed to contrast, is exactly the design of our current study (exposed n = 773/unexposed n = 2,956) and similar studies by Cramer et al. (exposed n = 193/unexposed n = 233), Heller et al. (exposed n = 292/unexposed n = 405), and Bruce et al. (exposed n = 5,790/unexposed n = 7,484). Our finding of an incidence of AKI of 5.69% in contrast exposed and of 8.96% in contrast-unexposed ED patients is consistent with the findings studied by Cramer et al. (2.1% vs. 1.3%), Heller et al. (4% vs. 4%), and Bruce et al. (9% vs. 9%). The findings of these studies together with our own suggest a significant risk of a confounding bias in the current definition of CI-AKI, based on timed changes of SCr. Also, similar to these previous studies, we could find no clinically significant predictor for AKI when we compared initial patient characteristics: renal function, comorbidities, or metabolic parameters, in either the contrast-exposed or the contrast-unexposed groups.
Future studies of CI-AKI should include suitable negative controls of unexposed patients (with the addition that they be adequately matched or stratified based on medical acuity) and be appropriately powered to detect clinically significant outcomes such as rates of renal replacement therapy and mortality. These studies should also attempt to isolate markers of kidney injury that may be specific to CI-AKI and predict morbidity and mortality.
The present study is limited by several factors. It is possible that contrast-exposed patients could have bounced back to another hospital and thus would not have been included in this study, biasing to the null hypothesis. We did not identify in our 1-year retrospective database any contrast-exposed patients who bounced back to our ED.
This study was not conducted in a randomized prospective fashion. We looked into several statistical methodologies to better compare the risks of AKI between those contrast-exposed and -unexposed. Multivariable logistic regression would have risked overfitting by the small number (n = 44) of CI-AKI patients. Propensity scoring may have addressed the issue of model overfitting, but required comprehensive matching for AKI risks in both groups. Unfortunately, a retrospective study such as ours does not provide the granularity of detail necessary to identify all the risks of AKI pre and post admission in the contrast-exposed and -unexposed groups.
In addition, our study did not compare patients who needed a CT and received contrast and were admitted versus patients who needed a CT and did not receive contrast and were admitted. We compared patients who needed a CT and received contrast and were admitted versus patients who were admitted. Although these groups were clearly selected from two different populations and represent a selection bias, the two groups' acuity level was comparable by demographic and initial metabolic parameters such as BUN/SCr, base excess, bicarbonate, anion gap, and lactate levels.
We were also limited in our ability to search for other possible causes of renal failure. In our study, the reasons for the decreasing creatinine are unclear. Patients who have a repeat SCr 48 to 72 hours are likely to have remained in the hospital after their initial ED evaluation. Similarly, they are likely to be receiving IV hydration, a putative treatment, or preventative measure for CI-AKI.[31, 32] Furthermore, our study was limited to patients with normal renal function, e.g., creatinine clearance rate > 60 mL/min/1.73 m2.
The 18th century Scottish empiricist philosopher David Hume defined causal inference as “we may define cause to be an object followed by another … where, if the first object had not been, the second never had existed.” This tenet of causation has formed the basis for the counterfactual model[34, 35] of causation, where the association between an exposure and outcome is tested against the frequency of the outcome without the exposure. Our data, consistent with those of Cramer et al., Heller et al., and Bruce et al., further bring into question the current definition of contrast-induced acute kidney injury to differentiate the outcomes of contrast-exposed and contrast-unexposed patients.
The implications of questioning the definition of contrast-induced acute kidney injury to define a specific disease are multiple. First, it would seem from our data and from previous studies that it is safe to administer IV contrast for CT to patients with normal renal function. There is little evidence to suggest that significant acute kidney injury will result from contrast administration in patients with previously normal kidneys. It is therefore not unreasonable to administer IV contrast to patients with glomerular filtration rates greater than 60 mL/min/1.73m2 without fear of contrast-induced acute kidney. Further additional studies are clearly indicated in this group and in those with a creatinine clearance less than 60 mL/min/1.73 m2.
The authors thank Jonathan Mortimer, Nachama Abdelhak, and Shana Laborde for their assistance in data collection