In 2006, The New England Journal of Medicine ran a review on the renal dangers of radiographic contrast, titled, “Preventing Nephropathy Induced by Contrast Medium.” When the journal revisited the topic in 2019, the kidney damage was no longer induced by the contrast but merely associated with the contrast. This change in language indicates a loss of confidence in the long-standing belief that radiographic contrast causes the kidney injury. However, two recent additions to the contrast-associated acute kidney injury (CA-AKI) literature refuse to accept this change and bring us back to adjudicating meaningless wanderings in serum creatinine.
Intravenous Contrast and Kidney Injury
Some of the first observations to shake confidence in the causal relationship between contrast and AKI were retrospective comparisons of patients who had contrasted CT scans with those who didn’t receive contrast. These repeatedly found higher rates of AKI in the patients who did not receive contrast. While this sounds damning, it’s probably due to physicians choosing to avoid contrast in patients at very high risk for AKI, such as those with low glomerular filtration rate (GFR), advanced age, diabetes, etc.
Researchers tried to control for this using propensity scoring to balance the risk of AKI between patients who did and did not get contrast. In multiple propensity-scored analyses, there was no signal of increased AKI in patients who received contrasted CT scans compared with people who did not receive contrast.
This finding was buttressed by a novel statistical technique called regression discontinuity analysis in which a threshold is set to assign a treatment (in this case, contrast use) while typical confounding variables (eg, age, GFR, diabetes status) don’t meaningfully change.
Patients admitted to Canadian emergency departments with suspected pulmonary embolism are regularly tested for D-dimer. The threshold for a positive test is 500 ng/mL, and fewer than 10% of patients with a D-Dimer below that threshold get a CT pulmonary angiogram (with contrast), while over 30% of patients with a D-dimer over 500 ng/mL get a contrasted angiogram. If contrast causes long-term changes in GFR, AKI, or dialysis, then the frequency of those outcomes should spike as the D-dimer rises above 500 ng/mL. When Goulden and colleagues performed such a regression discontinuity analysis on 156,000 Albertans, there was no signal that a D-dimer value over 500 ng/mL predicted AKI, dialysis, or decreased GFR at 6 months.
A randomized controlled trial on IV contrast and AKI has never been completed and probably never will be (one was started but abandoned after they were unable to recruit patients). These observational data supplemented with clever statistics may be the best data we will ever get, and they’re enough to persuade me that IV contrast is not a meaningful cause of AKI.
Arterial Contrast and AKI
The more interesting question concerns arterial contrast and AKI. Arterial contrast involves not only diagnostic studies but therapeutic interventions where fear of CA-AKI may prevent patients from receiving therapies with proven benefit.
One compelling trail of logic loosening the association of contrast with AKI comes from the MATRIX trial which randomly assigned patients undergoing left heart catheterization to a radial vs femoral approach. Both groups received an equal amount of radiocontrast agent (183 mL), but the rate of CA-AKI was reduced by 13% with the radial approach.
CA-AKI was defined as an absolute (> 0.5 mg/dL) or a relative (> 25%) increase from baseline in serum creatinine (sCr) — a definition that I do not find clinically relevant. More important is that cases of severe AKI, defined as at least a tripling of baseline sCr or oliguria for 24 hours — ie, the cases that really matter — were reduced by 63% with a radial approach. This dramatic drop despite similar doses of contrast indicates what should have always been obvious: Not all AKI that occurs after exposure to contrast is due to the contrast.
In the arterial contrast arena, we have further evidence from randomized controlled trials that contrast is not the culprit. Advances in imaging techniques for percutaneous coronary intervention (PCI) allow for trials comparing low vs usual volumes of contrast. In one study, patients randomly assigned to receive intravascular ultrasound–guided PCI received a third as much contrast as the control group who underwent angiography-guided PCI. Despite this, there was no difference in the rate of CA-AKI. Another randomized controlled trial of an injection system that successfully reduced exposure to contrast during cardiac catheterization found that this did not result in lower CA-AKI rates compared with the control group.
The PRESERVE Trial
The highest-quality data we have on arterial contrast and AKI come from the PRESERVE trial. Almost 5000 people at high risk for adverse renal outcomes undergoing coronary angiography were randomly assigned to isotonic bicarbonate or acetyl cysteine to see if either could protect against CA-AKI.
The authors were rightly skeptical of the standard definition of CA-AKI that the MATRIX trial and others used because it captures a lot of meaningless creatinine meandering rather than significant outcomes. If the creatinine goes from 1.0 mg/dL to 1.3 mg/dL, then back to 1.0 mg/dL in the 10 days following a cardiac catheterization, should we care?
The PRESERVE trialists selected a more meaningful clinical endpoint: a composite of death, dialysis, or a persistent (90 days) 50% bump in creatinine. While the primary outcome showed that neither isotonic bicarbonate nor acetyl cysteine was protective, the trial also provided unique insight into the traditional CA-AKI definition. In one analysis, investigators looked to see how often CA-AKI predicted the more important composite endpoint. Only 53 of the 429 people meeting the CA-AKI definition ultimately developed the composite outcome, while over 70% of the people who experienced the composite outcome never even had CA-AKI.
CA-AKI was neither sensitive nor specific for the more important, long-term, and patient-oriented outcome.
More damning was a substudy of PRESERVE that looked at kidney injury biomarkers. Because CA-AKI requires such small changes in creatinine, clinically meaningless changes in perfusion can trip a diagnosis. Focusing on kidney injury biomarkers can delineate real tissue damage from these hemodynamic changes. The PRESERVE researchers found that biomarkers of patients with CA-AKI were not significantly different from those of patients without CA-AKI. This makes the definition of CA-AKI analogous to having a definition of myocardial infarction where the troponin is negative.
New Risk Score and the Peril of CA-AKI
This brings us to two new manuscripts in the realm of CA-AKI. The first is a new risk score. In 2004, Mehran and colleagues published a widely used risk score to predict the development of CA-AKI. It was criticized for excluding patients with shock or acute myocardial infarction and for requiring knowing the contrast volume in advance, something that cannot be known until after the procedure. Their second bite of the apple brings us the Mehran-2 score. This looks at PCI procedures and only PCI, so this is not validated for other sources of contrast. These PCIs were performed at Mount Sinai Hospital from 2012 through 2020 and statistical regression was used to build two models. Model 1 uses variables available prior to PCI, while Model 2 includes procedural information, such as contrast volume, bleeding (drop in Hgb of > 3 g/dL), slow or low flow post-PCI, and complex anatomy.
The new risk score was derived and validated using independent cohorts and performed well. However, I question its usefulness. The authors use the conventional definition of CA-AKI, the same definition that did such a poor job of predicting long-term kidney outcomes and that did not trigger positive renal biomarkers of AKI in PRESERVE. What’s the use of measuring the risk for CA-AKI if CA-AKI isn’t a clinically meaningful outcome?
Additionally, for a score that is supposed to measure the toxicity of contrast, the contribution of contrast volume to the score is nearly insignificant. Zero points are assigned for a dose of 100 mL, and 1 point is assigned for 100-199 mL (80% of the derivation cohort received < 200 mL). By comparison, an ST-segment myocardial infarction counts for 8 points, unstable angina gives 2, hemoglobin < 11 g/dL yields 1, and chronic kidney disease (CKD) stage 3 adds a single point. According to the score, reducing contrast from 190 mL to 30 mL would have a negligible effect on the risk of developing CA-AKI. It seems that the score is more a measure of the risk of PCI-associated AKI rather than contrast-associated AKI .
Last, the authors provide no guidance on what to do with the results. Does low risk mean we no longer need to pretreat patients with IV fluids, as was shown in the AMACING trial? Should we avoid PCI in patients with high risk scores? Given the uncertainty regarding the effects of CA-AKI on long-term morbidity, I would hope that cardiologists would proceed with an indicated PCI regardless of the Mehran-2 score.
The second recent study is a follow-up of the ADAPT-DES trial that looked at the impact of CA-AKI on long-term outcomes. This examination, like many before it, showed that developing CA-AKI is associated with devastating long-term outcomes. At 2 years, the multivariable adjusted association between CA-AKI and all-cause mortality was a hazard ratio of 1.77 and 2.08 for cardiac mortality.
The problem is that the people who developed CA-AKI look completely different from those who did not. The 6.5% of patients who developed CA-AKI were older and had more peripheral artery disease, heart failure, diabetes, hypertension, chronic kidney disease, and anemia. Multivariable adjustment attempts to account for these confounding factors, but the authors’ decision to use dichotomous variables diminishes its power to squash residual confounding. For example, they measured GFR but then use CKD or no CKD in their multivariable adjustment. The 30-year-old with a GFR of 31 mL/min is wildly different from the 67-year-old with a GFR of 57 mL/min, yet this analysis treats them the same. Similarly for hypertension: 137/84 mm Hg does not equal 188/104 mm Hg, and so on.
I believe that these latest papers and others that show an association between CA-AKI and adverse outcomes are examples of reverse causality. We know that diabetes, heart failure, and CKD are risk factors for CA-AKI, and that CA-AKI is associated with all kinds of devastating outcomes, but the jump from association to causation is a bridge too far.
Do you think a creatinine bump from 1.1 mg/dL to 1.4 mg/dL that returns to normal within a week of cardiac catheterization is a more important cause of mortality than the fact that the patient had heart failure or diabetes? Instead of thinking of the heart failure being a risk factor for CA-AKI, think of the CA-AKI as a convoluted way to find patients with heart failure, diabetes, and peripheral artery disease. The CA-AKI is not the cause of the devastating outcomes but a indicator of patients at high risk for those outcomes.
We can’t get a definitive answer from the CA-AKI trials because they were not powered for mortality and we have yet to identify effective therapies to prevent CA-AKI. But we do have trials on interventions that cause acute, mild to moderate, and often temporary elevations in sCr. Coca and colleagues looked at 14 such trials to see if the spread in AKI between the intervention and control group predicted future mortality. It did not. This is further evidence that spending time worrying about tiny changes in serum creatinine is a fool’s errand that does not benefit patients.
Could Subclinical Cholesterol Emboli Be a Factor?
If contrast is not the cause of AKI during cardiac catheterization, what could be? I wonder if subclinical cholesterol emboli syndrome may be the true culprit.
Cholesterol emboli syndrome is a rare complication of arterial catheterization where the fibrous caps of plaques are disrupted and cholesterol emboli are showered downstream. These bits of cholesterol cause ischemic damage wherever they finally lodge. They can cause ischemic colitis, purple toes, and AKI. While cholesterol emboli are reported following 1%-2% of cardiac catheterizations, I suspect that subclinical cholesterol emboli syndrome is more common and may be responsible for some, perhaps many, cases of so-called CA-AKI.
Supporting my hunch is a study looking at AKI following transcatheter aortic valve replacement (TAVR). AKI is relatively common following TAVR, occurring in 10%–30% of interventions. Shishikura and colleagues looked at risk factors for AKI, and while the usual suspects of low GFR and anemia were found to be significant predictors, contrast volume was not. The investigators then used CT to estimate aortic atherosclerotic burden and found that increased plaque volume above the renal arteries was a powerful predictor of AKI, while aortic plaque volume below the renal arteries was not. The authors suggest that cholesterol emboli could be an underrecognized but important cause of AKI.
Cholesterol emboli from aortic plaques would also explain the decreased CA-AKI found with radial catheterizations in the MATRIX study.
Stop Chasing Creatinine
Nephrology, cardiology, and radiology have been trying to reduce and prevent contrast from damaging kidneys for decades. I had hoped that the PRESERVE trial would lead to a change in how people evaluated contrast and other nephrotoxins. Or that the cholesterol emboli theory would get more attention. However, looking at the two most recent additions to this field, we are back to adjudicating meaningless wanderings in serum creatinine.
Follow Joel Topf, MD, on Twitter