Risk of contrast induced nephropathy in the critically ill: a prospective, case matched study

Computerized tomography (CT) is invaluable for the management of critically ill patients. While the intravenous administration of iodinated radiographic contrast media (RCM) may be helpful or essential for adequate imaging, the potential complication of contrast induced nephropathy (CIN) is a significant concern. While it has no universally accepted definition, CIN typically refers to an at least modest decline in the glomerular filtration rate (GFR) occurring in the first days following RCM exposure. Its pathophysiology remains uncertain, with proposed mechanisms including medullary hypoxia, free radical generation, and direct tubular toxicity [1].

Depending upon the population studied, CIN incidence has ranged from 1% to more than 30% [2‐12]. Risk factors include impaired renal function, diabetes, anemia, and hypotension [13‐17], all common among ICU patients. Critically ill patients have therefore been presumed to be at relatively high risk for CIN [18‐22]. However, analyses of CIN risk factors have focused predominantly on patients undergoing cardiac catheterization and have generally not included unexposed comparator populations. The applicability of these findings to critically ill patients undergoing CT scanning is therefore unknown.

The presumption that critically ill patients are at high risk for CIN has clinical consequences. Perhaps most importantly, it may influence how often physicians avoid exposing their critically ill patients to RCM during CT scanning even when it would provide better imaging. Furthermore, when contrast enhancement is utilized a variety of measures may be employed in an attempt to avert CIN, occasionally with unwanted consequences of their own. It is axiomatic that development of rational strategies to prevent CIN, either by prophylaxis or avoidance of RCM exposure altogether, requires accurate estimates of CIN incidence in populations of interest.

CIN is usually recognized by a rise in serum creatinine, with many investigators using as a threshold a 44 μmol/L (0.5 mg/dL) absolute increase or a 25% relative increase from before exposure. Identification of CIN among ICU patients using serum creatinine is problematic for several reasons: volatility of filtration with dynamic clinical states; rapidly changing volumes of distribution, from both disease and therapy; and varying creatinine production. These may in part explain disparate reports of CIN incidence among ICU patients [6, 23, 24]. The measured creatinine clearance (mCr_Cl), a routinely available test used in trials of critically ill patients [25], intrinsically compensates for these phenomena and may provide better estimates of GFR in the ICU setting [26‐28].

As the initial phase of a planned program to investigate the possible utility of very early markers of CIN in the critically ill, we undertook a prospective study of ICU patients requiring CT scanning to better quantify their risk of developing the syndrome and, in particular, to determine whether it could be identified reliably in individual patients. We opted to perform serial measurements of mCr_Cl, before and after scanning, to identify changes in renal function with high accuracy. Because lability of renal function may lead to over-diagnosis of CIN even among less acutely ill patients [29‐31], we used matched ICU patients scanned without contrast enhancement as comparators.

Subjects were recruited from the medical and surgical ICUs of the Miami Department of Veterans Affairs Medical Center from 12 October 2004 through 2 December 2006. The Human Studies Subcommittee of the Research and Development Committee (the Institutional Review Board of the medical center) approved the study, permitting urine collections before informed consent, but with consent required for sample analysis and inclusion in the study.

CTs were considered evaluable if done while the patient was in the ICU, a urine collection on that day was completed before the scan and the patient was expected to remain in the ICU with an indwelling urinary catheter for the succeeding three days.

We studied critically ill patients receiving RCM for CT scanning to determine their risk for CIN. Anticipating lability in renal function from critical illness itself, we compared these patients to a matched population scanned without RCM. Significant declines in renal function were frequent, but no more so among those scanned with RCM than those scanned without. It is sobering to note that absent a control population, the declines in renal function observed in the RCM exposed patients would have suggested a high CIN incidence.

To the contrary, CIN appears to have been, at most, rare in the population we studied. Our data suggest that among similar patients fewer than two per 100 should suffer a 50% loss of filtration persisting even three days as a consequence of RCM use. It is clear from our study that CIN occurs infrequently compared to other renal insults in the critically ill and loss of GFR after RCM administration, when it occurs, cannot reliably be ascribed to RCM exposure. The observed magnitude of effect is not a consequence of study size. Even a much larger study of similar patients could only be reasonably anticipated to narrow the confidence bands for significant adverse effects from RCM, but still within the upper or lower bounds of the confidence intervals we found.

Our population had many characteristics considered risk factors for CIN. Most had mCr_Cl less than 60 mL/minute/1.73 m² before scanning. Anemia, diabetes, congestive heart failure and advanced age were common. A widely cited tool [14] for predicting CIN incidence, developed from a population undergoing cardiac catheterization, classified most of our patients as being at moderate or high risk. The predicted incidence was 17%, even without accounting for the relative hypotension and recent renal compromise prevalent among our patients. Indeed, the true incidence of nephropathy within the contrast exposed group was reasonably approximated by the model, but our control group, with similar risks but no contrast exposure, had comparable nephropathy rates.

The only randomized prospective trial of CIN prophylaxis for ICU patients, which compared theophylline with N-acetylcysteine but employed neither a control arm nor an unexposed arm [8], further underscores the need for caution when ascribing cause. In that study, a 44 μmol/L (0.5 mg/dL) rise in serum creatinine within 48 hours of imaging was defined as CIN and implicitly attributed to contrast exposure irrespective of comorbidities, and in aggregate occurred after 6% of scans. We observed a similar overall incidence of nephropathy satisfying this definition, 4%. However, all of the events in our study happened to have occurred in patients scanned without contrast (who thus had an 8% incidence), a group specifically studied because their risk profiles were comparable to patients who were scanned with contrast.

Unlike other CIN investigations, we used mCr_Cl as the principal assessment of GFR. Both the CG and MDRD formulas presume normal creatinine generation and a static GFR. Assessments of renal function derived solely from serum creatinine suffer the same shortcomings. Among critically ill patients with potentially abnormal creatinine production [36, 37] and labile filtration, mCr_Cl may offer advantages over other commonly used estimates of GFR [38].

However, measured creatinine clearance is not without problems. It depends critically upon urine collections being accurately timed and complete. By studying patients with indwelling urinary catheters errors from incomplete voiding were minimized. Accuracy of timing was assured by recording when collections were started and ended, rather than the common practice of specifying the duration of collections in advance.

In addition to changes in GFR, creatinine based measures of renal function are sensitive to fluctuations in the tubular secretion rate of creatinine. Interval institution or cessation of drugs effecting secretion may falsely imply changes in GFR. Use of these drugs, however, was rare, and balanced between RCM and non-RCM patients. Creatinine based measures may underestimate declines in GFR if tubular secretion remains intact, particularly when GFR is initially low. While other techniques for measuring GFR, such as inulin or iothalamate clearance, avoid this shortcoming, they cannot feasibly be employed on a daily basis as this study required.

Properly obtained, mCr_Cl provides a more accurate picture of GFR in critically ill patients than the serum creatinine or CG or MDRD formulas. However, our key finding remains unchanged irrespective of how GFR is measured, and even what level of dysfunction defines injury: the instability of renal function during critical illness overwhelms any signal from CIN, making it impossible to identify the syndrome accurately. Reinforcing this point, the pre-scanning variability of serum creatinine was of the same magnitude as that required to satisfy CIN definitions of renal injury. Such lability effectively precludes studying interventions to prevent CIN among the critically ill without at least one of the following: a biomarker more specific than GFR for CIN be identified and used as a surrogate endpoint; or very large populations are studied, without attempt to attribute individual events to contrast exposure. This caveat is also applicable to investigations of other potential etiologies of renal injury among the critically ill.

Our study has limitations. While scientifically preferable, a randomized trial of contrast exposure is impracticable, as it would mandate either suboptimal imaging or RCM exposure simply to assess toxicity, violating the principal of beneficence. We therefore employed a matched pairs design. Matching might have inadvertently paired RCM patients with unexposed patients at intrinsically higher risk for deterioration of GFR. Comparisons of other unmatched patient characteristics do not suggest this was the case. Furthermore, brain imaging, more common among non-RCM patients, tests for pathologies not generally associated with renal injury.

Because of the stringency of the matching algorithm employed, matching proved impossible for most patients receiving vasopressors. However, no unmatched RCM enhanced scan of a patient who received vasopressors was associated with even a 33% decline in measured clearance. Somehow including these patients in the analysis could only have diminished further the upper bounds for CIN incidence.

Perhaps more importantly, we could not study patients with RCM exposures shortly after ICU admission, since we required a pre-scan urine collection. These patients may well differ from those scanned later in their ICU stays. For example, there might be salutary effects of the markedly positive pre-scan fluid balances among established ICU patients. We therefore cannot exclude a possibly substantial CIN risk for patients just arriving to the ICU. We also could not study patients whose physicians felt the risk of RCM exposure outweighed the benefits. However, scanning with contrast was judged preferable from an imaging standpoint for fewer than 10% of patients scanned without contrast, including those who could not be matched.

Previously, one has had to extrapolate from other populations when judging the CIN risk faced by ICU patients and how best to manage it. Cardiac catheterization studies have provided most of these data. The critically ill patients we studied often had findings identified as CIN risk factors in cardiology populations and received RCM volumes comparable to those commonly administered during coronary angiography. However, there are potentially important differences between these populations: critically ill patients received considerably more fluid than recommended for cardiac catheterization CIN prophylaxis; exposure to certain drugs effecting kidney function, including diuretics and angiotensin pathway modulators, is likely higher in the cardiac population; and catheterization simultaneously poses a risk for cholesterol embolization. Our study indeed suggests that extending CIN findings from cardiac catheterization to CT scanning of the critically ill is unwarranted.

We performed a prospective case matched study to determine the identifiability and frequency of CIN in the critically ill. Our investigation places a very low upper bound on CIN incidence in established ICU patients. The attributable risk for clinically significant renal dysfunction lasting even three days is unlikely to exceed 2%, which should inform planning for imaging studies. Furthermore, in ICU patients attribution of declining filtration to RCM exposure is unreliable because of the background prevalence of deterioration in the absence of exposure. Pending the validation of etiology specific biomarkers, this limitation will likely apply not only to investigations of CIN but also of other potential causes of acute renal injury in the critically ill as well.