Development of a simple risk score model to predict renal artery steno-sis

Renal artery stenosis (RAS) is a known cause of secondary hypertension and renal failure. The most patients with renal artery stenosis are asymptomatic. So, the exact prevalence of this disease is unknown. The gold standard of diagnosis of RAS is renal angiography that is an expensive somewhat hazardous procedure and may revealed nothing. The aim of this study was to develop a simple risk model score to predict significant RAS based on known risk factors. This may enable us to select patients with high probability of having RAS to perform angiography.
A total of 4177 patients whom underwent renal angiography from April 2001 to March 2016, were randomly assigned to a development and a validation dataset in ratio of 2:1 respectively. The clinical and laboratory data of patients were analyzed by multivariate regression analysis. The factors of female sex, history of hypertension and glomerular filtration rate were determined as predicting factors and they were assigned a weighted integer, the sum of the integers was a total risk score for each patient. This model was examined at validation set.
We retrospectively evaluated all patients undergoing renal artery angiography since 15 years ago. We extracted all risk factors of RAS including age, sex, height, weight, and history of diabetes, hypertension and hyperlipidemia. We also looked at coronary or peripheral vascular diseases and presence of heart failure. The age of patients was 63.5±11.2 years and 40% of the patients were female. The significant RAS was defined as 70% or more narrowing of renal artery. The prevalence of renal artery stenosis was 14.4% and 13.5% in development and validation dataset respectively. The area under curve and confidence interval for final mode in development dataset was 67.9% (65.0-70.8%). The rates of RAS increased with increasing risk score. In 1402 patients in validation dataset the model showed good discrimination power (cstatistic= 0.76)
This model simply assesses the risk of RAS using available information. This model can be used both in clinical and research purposes. The power of model for diagnosis of RAS is estimated to be 72.6% (68.8%-76.4%).