Do CT scans cause contrast nephropathy?

Introduction

In April 2013 a series of articles in Radiology debated whether contrast nephropathy still exists using modern contrast dye. Two years later, the controversy remains. This is a daily conundrum when managing critically ill patients: one radiologist will urge us to use contrast, while the next radiologist will caution us against using contrast.
0

0
The existence of clinically significant contrast nephropathy is based upon three suppositions. First, contrast dye should cause an elevation of serum creatinine. Second, this elevation of serum creatinine should indicate genuine kidney injury (rather than random fluctuations in creatinine). Third, kidney injury should result in clinically meaningful outcomes (e.g. dialysis or death). This post examines new evidence about these suppositions.

Background: Understanding types of contrast dye and procedures

Contrast for cardiac catheterization vs. CT scanning

The risk of kidney injury following cardiac catheterization is higher than the risk of a contrasted CT scan for many reasons (e.g., catheterization may dislodge athero-emboli leading to renal failure, and cardiac patients often have tenuous renal perfusion). This post is about the use of intravenous contrast dye for CT scanning.

All contrast dyes are not created equal

Contrast dyes are divided into three groups based on their osmolarity. High-osmolar contrast medium (HOCM), the oldest and most nephrotoxic group, is no longer used. Most studies are currently performed with either low-osmolar contrast medium (LOCM) or iso-osmolar contrast medium (IOCM):
0

Meta-analysis of prospective RCTs comparing LOCM to IOCM shows that the degree of nephrotoxicity varies between different types of LOCM. Iohexol and Ioxaglate are more nephrotoxic than IOCM, while remaining LOCMs are not (figure below from Reed 2009). This was confirmed in the most recent meta-analysis of 42 RCTs that included 10,048 patients (Biondi-Zoccai 2014)(1).

These meta-analyses combined data from cardiac catheterization procedures with data from CT scans. A large retrospective study of CT scans similarly found that Iohexol was associated with higher rates of nephrotoxicity than Iodixanol (Bruce 2009):

(1) Does contrast dye cause an increase in creatinine?

0
The vast majority of papers about contrast nephropathy have focused on whether creatinine increases after administration of contrast dye. Most studies were performed without a control group, based on the assumption that any increase in creatinine must be due to contrast nephropathy. However, some investigators realized that creatinine elevations are common even in the absence of contrast dye.

McDonald 2013 performed a meta-analysis of thirteen observational studies comparing patients who had received contrasted CT scans versus patients who received noncontrasted CT scans. This study found no difference in creatinine elevations between the two groups. However, this data may have been confounded by avoidance of contrast administration in patients at higher risk of kidney injury.

In attempts to make sense out of retrospective observational data, two propensity-matching studies evaluated the relationship between contrast and creatinine elevations (2). Propensity-matching is a complex statistical approach intended to remove multiple confounding variables from observational data. These two studies arrived at opposite conclusions. Davenport 2013 found that contrast dye was nephrotoxic among patients with baseline GFR<30 ml/min, whereas McDonald 2014found no relationship between contrast and changes in creatinine among any subgroup:

Comparing these two studies may explain why they reached different conclusions regarding patients with baseline GFR<30 ml/min. First, Davenport's study is woefully underpowered in this subgroup (with one group as low as n=44, compared to McDonald's groups which are all >700 patients). Second, Davenport's study seems to have broken the 1:1 propensity matching in order to compare two sub-groups of different sizes (unlike McDonald's study, which preserves 1:1 propensity matching throughout). Failing to analyze propensity-matched data in a pair-wise fashion is a common error (Austin 2008). Overall, McDonald's study appears better powered and better designed, with tighter 95% confidence intervals and more credible results. Another possible explanation for the difference in results may be more systematic avoidance of Iohexol among patients at risk for renal failure in the McDonald study (3).

McDonaldalso performed a counterfactual analysis of patients who had received both a contrasted CT scan and a non-contrasted CT scan at different times. Using the patients as their own control, there was no difference in rate of kidney injury following the contrasted versus the uncontrasted scan.

(2) Do increases in creatinine correlate with actual kidney injury?

0
Defining contrast nephropathy on the basis of elevated creatinine 2-3 days after receiving contrast is convenient for investigators, but it is unclear what these creatinine elevations really mean (4). Many studies have reported that although some patients develop contrast nephropathy, the average creatinine among all patients is either stable or improved (e.g., Azzouz 2014, Lencioni 2010, Sandstede 2007). This raises the possibility that we may be observing random variations in creatinine following a normal distribution, and labeling the outliers as having “contrast nephropathy:” (5)
0

Schmalfuss 2014 found that among 508 patients receiving IOCM, 14 had an increase in creatinine satisfying their definition of contrast nephropathy. However, eight patients had a decrease in creatinine of the same magnitude. There was no significant difference between the number of patients who had increased creatinine versus decreased creatinine. Kooiman 2014 reported similar results following contrast exposure: on average there were small decreases in creatinine, with similar numbers of patients experiencing increases or decreases in creatinine (figure below). Unfortunately, the vast majority of studies focus only on patients with elevated creatinine, creating the illusion that contrast dye causes an increase in average creatinine.
0

0
Another way to investigate the significance of changes in creatinine is to compare them to biomarkers of renal injury, for example neutrophil gelatinase-associated lipocalin (NGAL). NGAL has been shown to rise rapidly and predict subsequent renal failure due to a broad variety of renal injuries, including cardiac catheterization. Kooiman just published the largest study of biomarkers, describing 511 patients with chronic kidney disease who received contrast-enhanced CT scans. These authors detected no change in two renal biomarkers (NGAL and KIM-1) following contrast administration. This held true even in subgroups at higher risk for renal injury, including 36 patients with GFR<30 ml/min.

4% of patients in this study (20/501) met their definition of contrast nephropathy based on creatinine elevations. However, there was no difference in biomarker levels between these patients and patients with stable creatinine levels. This indicates that elevated creatinine among patients diagnosed with “contrast nephropathy” may simply reflects fluctuations in renal perfusion or vascular tone, rather than genuine kidney injury.

(3) Does contrast dye effect patient-centered outcomes?

0
Most descriptions of contrast nephropathy indicate that the creatinine peaks around three days after receiving contrast and then will often normalize within the next two weeks. Supposing this is true, what is its clinical significance? Ultimately patient-centered outcomes are what matters (e.g. dialysis and death).

There are many uncontrolledstudies in the literature showing that acute kidney injury following cardiac catheterization or contrasted CT scanning of hospitalized patients correlates with worse outcomes (e.g., death and dialysis; Weisbord 2011). However, these studies lacked a control group that didn't receive contrast, and therefore fail to establish causality between contrast administration and kidney injury. Indeed, within any group of sick patients, the development of renal failure prognosticates worse outcomes.

McDonald’s meta-analysis in 2013 of controlled studies did not find any relationship between contrast administration and death or dialysis. However, as discussed earlier, these observational studies failed to account for confounding factors. To address this, in 2014 McDonald performed a propensity-matching analysis involving 21,346 patients including 1,725 deaths and 52 dialysis initiations within 30 days of the CT scan. Following propensity-matching, there was no independent relationship between these outcomes and contrast administration. This study confirmed the correlation between AKI and mortality, but demonstrated that neither outcome is independently associated with contrast administration (figure below). This debunks uncontrolled studies which attempted to imply causation between contrast administration, AKI, and mortality.
0

Conclusions

For decades, the incidence and consequence of contrast nephropathy have been systematically inflated by poor research methodology. For example:

Contrast nephropathy was defined in terms of small transient creatinine elevations, a definition which is sensitive but nonspecific.
Although contrast nephropathy is intended to be a diagnosis of exclusion, studies attribute all elevations in creatinine to contrast nephropathy, causing contrast to be blamed for every renal insult.
Most studies have been uncontrolled, based on the assumption that creatinine never changes in the absence of contrast dye.
Among sick patients, renal failure correlates with mortality. Studies of cohorts receiving contrast dye have replicated this correlation, with the implication that contrast was to blame for both renal failure and mortality.

Among contrast dyes commonly used today, there is only evidence to support that Iohexol and Ioxaglate cause elevated creatinine. This seems to be limited to patients with pre-existing renal dysfunction. There is no convincing data that IOCM or safer LOCMs cause elevations in creatinine when used for CT scanning.
0

0
Contrast nephropathy has been defined in terms of short-term elevations in serum creatinine. However, many patients meeting the definition of “contrast nephropathy” may merely have random fluctuations in serum creatinine. A recent study revealed that outpatients with increased creatinine following contrast administration lacked elevations in the renal biomarkers, suggesting that these creatinine elevations may not reflect genuine kidney injury.

Ultimately, patient-centered outcomes such as dialysis or death are more important than transient fluctuations in creatinine. By all accounts, these outcomes are far less common than transient elevations in creatinine. Controlled studies have found no causal relationship between contrast dye and patient-centered outcomes.

Of course, the absence of evidence does not constitute evidence of absence. It is nearly impossible to prove that toxicity from a substance does not exist. Thus, it remains possible that newer contrast agents could have some degree of nephrotoxicity. In particular, there is little evidence regarding whether it is safe to use these agents in patients with severe renal failure.
0

Meta-analyses of RCTs comparing different contrast dyes demonstrate that some types of LOCM (especially Iohexol) cause creatinine elevation more often than others (table below).
There is no evidence that safer contrast dyes cause creatinine elevation. The highest quality propensity-matched study of CT scans performed at the Mayo Clinic found no effect of contrast dye on renal function (of note, IOCM was used for patients at higher risk of renal failure).
“Contrast nephropathy” has been defined in terms of transient elevations in creatinine. Its clinical significance is unclear. There is no evidence that contrast dye causes patient-centered outcomes such as death or dialysis.
One recent study found that outpatients with “contrast nephropathy” lack biomarker evidence of kidney injury. This suggests that many studies of contrast nephropathy may have been measuring random fluctuations in creatinine rather than genuine kidney injury.
Overall, there is currently no clear evidence of nephroticity due to IOCM or low-risk LOCMs (i.e., all except Iohexol and Ioxaglate). However, it has not been excluded that they could pose a risk to patients with severe renal failure.

Notes

(1) Unfortunately many investigators incorrectly assumed that all LOCM would have the same degree of nephrotoxicity, leading to multiple studies comparing IOCM to all types of LOCMs pooled together. Most of these studies found no difference between IOCM and pooled LOCMs, leading to the incorrect conclusion that all LOCMs and IOCM are equivalent.

(2) Actually there were a total of five propensity-matched studies released by two research groups (McDonald 2014, McDonald 2013, McDonald 2014, Davenport 2013, Davenport 2013). This post focuses primarily on two papers which represent the most recent and most directly comparable analyses by each research group (indeed, it appears that the paper by McDonald 2014 was written as a direct retort to the publication by Davenport 2013). Both series of papers are successive re-analyses of the same underlying data set.

(3) It is possible that differences between these two studies could relate to the type of contrast dye used. Davenport's study used data from multiple hospitals which utilized different contrast dyes using various protocols (in 42% of cases the type of contrast dye was unknown). McDonald's propensity-matching study was obtained using data from the Mayo Clinic, where Iohexol was utilized for patients at low risk of kidney injury and Iodixanol was utilized for patients at higher risk of kidney injury. It is conceivable that the Mayo Clinic's systematic approach to selecting Iodixanol for higher-risk patients may have allowed them to avoid kidney injury.

(4) To make matters worse, there is no consistent definition of contrast nephropathy. Different studies use a variety of definitions based on short-term elevations in creatinine (e.g. increase in Creatinine of >0.3 mg/dL, >0.5 mg/dL, >25%, >50%, or various combinations of these criteria). Some papers present multiple analyses of the same data using different definitionsof contrast nephropathy, leading to conflicting conclusions (e.g., McCullough 2011).

(5) The higher frequency of “contrast nephropathy” among patients with renal dysfunction could simply be due to the observation that patients with renal dysfunction have greater random variations in creatinine. This likely reflects the shape of the curve relating creatinine to GFR, wherein small fluctuations in GFR cause greater variations in creatinine in the setting of renal dysfunction. This is illustrated in the figure below. Let us imagine, for the sake of argument, that occasionally any patient's GFR may fluctuate as much as +/-10 ml/min due to hydration status. For a patient with normal renal function, this will cause minimal fluctuation in creatinine (blue arrows). However, for a patient with renal dysfunction, this may cause substantial variation in creatinine (green arrows).

COI & Disclaimers: I have no conflicts of interests (e.g., I receive no funds from drug, device, or imaging companies). Opinions expressed herein are mine alone, and do not necessarily reflect the opinions nor policies of my employers or institution. Additional disclaimers are listed here.