Serious renal dysfunction after percutaneous coronary interventions can be predicted.

Background: A prediction rule for determining the post-percutaneous coronary intervention (PCI) risk of developing contrast-induced nephropathy (> or = 25% or > or = 0.5 mg/dL increase in creatinine) has been reported. However, little work has been done on predicting pre-PCI patient-specific risk for developing more serious renal dysfunction (SRD; new dialysis, > or = 2.0 mg/dL absolute increase in creatinine, or a > or = 50% increase in creatinine). We hypothesized that preprocedural patient characteristics could be used to predict the risk of post-PCI SRD.

Methods: Data were prospectively collected on a consecutive series of 11141 patients undergoing PCI without dialysis in northern New England from 2003 to 2005. Multivariate logistic regression model was used to identify the combination of patient characteristics most predictive of developing post-PCI SRD. The ability of the model to discriminate was quantified using a bootstrap validated C-Index (area under the receiver operating characteristic [ROC] curve). Its calibration was tested with a Hosmer-Lemeshow statistic. The model was validated on PCI procedures in 2006.

Results: Serious renal dysfunction occurred in 0.74% of patients (83/11141) with an associated inhospital mortality of 19.3% versus 0.9% in those without SRD. The model discriminated well between patients who did and did not develop SRD after PCI (ROC 0.87, 95% CI 0.82-0.91). Preprocedural creatinine (37%), congestive heart failure (24%), and diabetes (15%) accounted for 76% of the predictive ability of the model. The other factors contributed 24%: urgent and emergent priority (10%), preprocedural intra-aortic balloon pump use (8%), age > or = 80 years (5%), and female sex (1%). Validation of the model was successful with ROC: 0.84 (95% CI 0.80-0.89).

Conclusions: Although infrequent, the occurrence of SRD after PCI is associated with a very high inhospital mortality. We developed and validated a robust clinical prediction rule to determine which patients are at high risk for SRD. Use of this model may help physicians perform targeted interventions to reduce this risk.