Model Predicts Post-PCI Kidney Injury From Contrast

The complex relationship between contrast volume and acute kidney injury (AKI) can be modeled to enable personalized contrast dosing in percutaneous coronary intervention (PCI), according to researchers.

A model was developed for the non-linear association between contrast volume and AKI accounting for variation in baseline preprocedural risk, reported Harlan Krumholz, MD, SM, of Yale School of Medicine, and colleagues in a study published online in JAMA Network Open.

Krumholz and colleagues calculated the following areas under the receiver-operating characteristic curve (AUC) for the model's prediction of increases in creatinine level (within stage 1 AKI), in separate test and validation sets:

• Creatinine level increase of at least 0.3 mg/dL: AUC 0.777 (95% CI 0.775-0.779) in test cohort; AUC 0.794 (95% CI 0.792-0.795) in validation cohort
• Creatinine level increase of at least 0.5 mg/dL: AUC 0.839 (95% CI 0.837-0.841); AUC 0.845 (95% CI 0.843-0.848)
• Creatinine level increase of at least 1.0 mg/dL: AUC 0.870 (95% CI 0.867-0.873); AUC 0.872 (95% CI 0.869-0.875)

"While limiting contrast volume is an acknowledged method of reducing AKI risk, clinicians need to balance reducing contrast volume with providing adequate contrast to permit coronary evaluation and treatment," the authors noted.

"A large body of work on clinical prediction modeling is based on models that exclusively use baseline information. Such a model is useful for both benchmarking quality assessment and planning treatment. With the additional integration of procedural variables, there is opportunity to identify new and helpful information for personalized strategies from the data," according to Krumholz's group.

Having previously built a machine learning model to estimate a person's preprocedural risk of AKI, the researchers fine-tuned it to account for the non-linear relationship between contrast volume and AKI risk and the heterogeneity of this association among different risk groups.

The model incorporated some two dozen aspects of patients' clinical condition and medical history, along with contrast volume, to predict risk of kidney injury from a given volume of contrast. Data were drawn from the National Cardiovascular Data Registry CathPCI Registry.

To develop the model, Krumholz and colleagues identified a "derivation cohort" of more than 2 million people undergoing PCI in 2011-2015 (mean age 65.1, 31.9% women); this cohort was divided into a test set and another set used to evaluate the model's performance. The model was validated with data on more than 960,000 additional individuals getting PCI during 2015-2017 (mean age 65.7, 31.8% women).

The creatinine thresholds used in the study were gradations within stage 1 AKI because there were too few events at higher stages for analysis.

Reliance on observational data was a major limitation of the study, along with the possibility that the model may need to be continually updated to reflect contemporary PCI practice, according to Krumholz and colleagues. "Third, the model cannot provide predictions for patients with very high preprocedural risks and very high contrast volumes owing to lack of data in the cohort for these combinations."

Finally, their prediction model cannot distinguish between contrast agent types.

"The heterogeneity found in the association between AKI and contrast volume among patients could enable personalized decisions on contrast dosing and could even help to inform the decision to have PCI," the investigators suggested.

Early invasive management of stable coronary artery disease notably failed just recently in the landmark ISCHEMIA trial.

PCI and bypass surgery alike did not reduce primary composite outcome -- cardiovascular death, myocardial infarction (MI), hospitalization for unstable angina, hospitalization for heart failure, or resuscitated out-of-hospital cardiac arrest -- compared to optimal medical therapy alone for patients with stable, moderate-to-severe ischemic heart disease.

ISCHEMIA-CKD, a parallel trial of people with estimated glomerular filtration rates <30 mL/min/1.73 m² and moderate or severe ischemia, similarly showed no benefit to routine invasive treatment. Even quality of life was no different between arms.

Comment