Abstract 110: A Novel Method for Estimating the Optimal Contrast Amount Needed to Minimize Acute Kidney Injury After Percutaneous Coronary Intervention
Introduction: Risk models are the foundation of personalized medicine. Converting risk into clinically actionable strategies to improve care, however, can be difficult. Acute kidney injury (AKI) after percutaneous coronary intervention (PCI) occurs in 7% of cases and contrast use has a linear relationship with risk. As the amount of contrast use is readily modifiable, we developed a novel approach to convert the National Cardiovascular Data Registry (NCDR) AKI risk model into an actionable guide for defining ‘safe contrast limits’. Methods: Because some patients are at low risk for AKI and the amount of contrast does not markedly increase their absolute risk, we reasoned that only patients with an above average pre-PCI risk (>7%) for AKI would be targets for contrast minimization. Providers could then define the magnitude of risk reduction they wanted to achieve (e.g. 5%, 10% or 15%). Given this target, we were able to back-calculate the safe contrast limits to achieve the respective magnitude of risk reductions. Results: 25% of patients were estimated to have above average risk (>7%) for developing AKI after PCI. The safe contrast limits for alternative magnitudes of relative risk reduction (RRR), across the range of patients’ pre-procedural risk, and the percentage of patients are shown (Figure-A). The number needed to treat (NNT) to prevent 1 AKI episode was <50 for patients with baseline risk of ≥14% for a 15% RRR and for patients with baseline risk of ≥21% for a 10% RRR, however it was never <50 for a 5% RRR (Figure-B). Conclusion: We were able to convert a complex prediction model into a clinically-actionable guide for prospectively improving treatment. Testing whether prospective targets of safe contrast limits (achieved through avoiding left ventriculograms, staging multi-vessel PCI or using bare wire lesions/injection minimization) could improve safety and costs should be prospectively tested.