Predicting Post-Contrast Nephropathy
Search Search Strategy: You conduct a clinical prediction guide Pubmed Clinical Query “contrast induced nephropathy” broad/sensitive search revealing 611 articles (see http://tinyurl.com/3rlh2to) including all four of the manuscripts below.
While working an overnight in the ED you evaluate a patient inthe trauma bay who complains of severe low back pain. She is a 49year old female and her pain started abruptly 2 hours ago. Upon initial evaluation your gestalt tells you ‘this is a sick one’. She has a history of diabetes and a kidney transplant with a baselinecreatinine of 1.9 mg/dL. Your inner epidemiology meter screams“vasculopath” based upon her co-morbidities so you immediately begin to contemplate aortic dissection. After reviewing the August 2009 Journal Club, you assess bilateralupper extremity blood pressures and note a 20 mm Hg difference.
The definitive test of choice for aortic dissection in 2011 is a dissection protocol CT, but Radiology is concerned about her underlying renal disease. Initially, they refuse to do her CT with contrast because of the risk of contrast induced nephropathy (CIN). They suggest a non-contrast CT which will identifyperiaortic hematoma or rupture. Another alternative is MRI. Sinceperiaortic hematoma does not rule out dissection and a frank rupture would be immediately fatal and since she is too unstable for an MRI, you again request a contrast CT.
As you wait for the CT tech to scan the patient, you wonder if you will kill her transplanted kidney in theprocess of confirming your diagnosis. You analyze the research evidence forCIN in patients receiving contrast for any reason. You focus upon the common reasons for getting contrast and possible alternative diagnostic studies, what you can do to estimate CIN risk before contrast exposure, and the best way to reduce the risk after contrast is given.
Population: ED patients with indications for a CT with contrast
Intervention: Risk assessment for contrast nephropathy
Comparison: Not applicable
Outcome: Creatinine elevation, dialysis, mortality
Fourth years: Probability of reduced renal function after contrast-enhanced CT: A model based on serum creatinine level, patient age, and estimated glomerular filtration rate, AJR 2009; 193: 494-500 (http://pmid.us/19620448)
Article 1: A simple risk score for prediction of contrast-induced nephropathy after percutaneous coronary intervention, J Am Coll Card 2004; 44: 1393-1399.
Article 2: Impact of nephropathy after percutaneous coronary intervention and a method for risk stratification, Am J Cardiol 2004; 93: 1515-1519.
Article 3: Risk stratification nomogram for nephropathy after abdominal contrast-enhanced computed tomography, Am J Emerg Med 2011; 29: 412-417.
Article 4: Probability of reduced renal function after contrast-enhanced CT: A model based on serum creatinine level, patient age, and estimated glomerular filtration rate, AJR 2009; 193: 494-500.
Bartels reported the first case of contrast-induced nephropathy (CIN) in 1954. CIN is believed to account for ~12% of hospital acquired acute kidney injuries today (Nash 2002, Hou 1983) and dialysis will be required for 3%-15% of these cases (Nikolsky 2004, Marenzi 2004). Even if CIN does not necessitate dialysis, CIN is independently associated with prolonged hospitalizations (Wickenbrock 2009, Shema 2009) and increased in-hospital (Levy 1996, Senoo 2010, Medalion 2010) and long-term mortality (Rihal 2002). However, before we assess the literature for the questions of whether pre-contrast CT CIN can be predicted in the ED, clinicians need to think critically about CIN by first asking two questions:
The most commonly quoted definition of CIN (and the one used in the studies we evaluated) is an absolute increase of serum creatinine (SCr) of 0.5 mg/dL or a 25% increase in SCr from baseline at 48-72 hours post-contrast. Chertow et al validated this definition by noting an association of relatively small changes in SCr in hospitalized patients with mortality above that predicted by their initial SCr. SCr has been shown to fluctuate (Bruce 2009) for a variety of reasons in hospitalized patients (i.e. those most likely to have a 48- to 72-hour post-CT SCr and thus be included in these studies) and in the Nephrology literature Newhouse/Rao have suggested that the attribution of a change in SCr as a de facto standard for CIN should be considered flawed logic. Other studies that have compared post-contrast patients to adequate inpatient control groups (admitted patients not exposed to contrast) have failed to find a significant difference in the incidence of renal failure (Cramer 1985, Heller 1991, Bruce 2009). Furthermore, most CIN studies do not include patient-centric outcomes in the definition of CIN such as need for short- or long-term dialysis or CIN-related mortality. Therefore, the very existence of CIN as currently defined in most studies is not a foregone conclusion. However, for the remainder of the discussion we will proceed under the assumption that CIN is real for a subset of ED patients, produces outcomes that are important to patients/clinicians/decision-makers/society, and that if a subset of high-risk ED patients can be identified then viable CIN-prevention strategies (Sinert 2007) might someday be employed.
Based upon 16 observational and 5 prospective randomized trials, the incidence of CIN is 0.3-25% with a pooled incidence point-estimate of 5%. (Li and Solomon 2010) ED-based studies have also demonstrated a 5% incidence of CIN. Our review of the literature found very few clinical prediction instruments for CT-associated CIN so we also assessed the closely related phenomenon of PCI-associated CIN in Cardiology patients. Even when the PCI-based trials are included, no risk-stratification tools exist that are ready for routine application to a wide variety of ED patients since each was only internally validated against the same populations in which they were derived. In addition, none of the derivation/validation trials followed Stiell’s methods for the development of valid, reproducible, ED-appropriate clinical decision rules. Specifically, investigators did not consider the full spectrum of CIN risk factors in the derivation of their decision aids, did not explicitly define all predictor variables, did not use recursive partitioning to construct their decision-models, and failed to report diagnostic accuracy parameters (sensitivity, specificity, likelihood ratios) that clinicians can use at the bedside to risk-stratify individual patients. Therefore, the reliability and accuracy of these instruments when actually used prospectively in the busy ED are undefined as are questions of clinician acceptability and the instruments’ impact on clinician behavior. For undifferentiated patients pending a contrast CT, the best rule is an exponential equation that is not easy to apply at the bedside:
The attached Excel sheet can be used to help compute the probability of a post-contrast CT GFR < 60 based upon age, gender, and the initial creatinine using this equation. However, this decision aid has yet to be validated on other populations and tends to over-estimate the risk of CIN. Further research is needed to assess the internal validity (additional risk factors), external validity (same results when applied to different populations), and reliability of this equation.
Ultimately, CT-contrast diagnostic questions will need to weigh the risks & benefits of contrast exposure on an individual patient basis. Pauker and Kassirer provided one method to do so that is being employed in evidence based diagnostics (see below).
However, until a mutually accepted (by Nephrology, Radiology, Surgery, Emergency Medicine, etc.) definition of CIN is proposed and without a validated clinical decision aid for CIN with estimates of sensitivity and specificity, the test- and treatment-thresholds for individual diagnostic questions like aortic dissection cannot be estimated.