Utility of Urine Eosinophils in the Diagnosis of Acute Interstitial Nephritis

Abstract

Urine eosinophils (UEs) have been shown to correlate with acute interstitial nephritis (AIN) but the four largest series that investigated the test characteristics did not use kidney biopsy as the gold standard. This is a retrospective study of adult patients with biopsy-proven diagnoses and UE tests performed from 1994 to 2011. UEs were tested using Hansel's stain. Both 1% and 5% UE cutoffs were compared. This study identified 566 patients with both a UE test and a native kidney biopsy performed within a week of each other. Of these patients, 322 were men and the mean age was 59 years. There were 467 patients with pyuria, defined as at least one white cell per high-power field. There were 91 patients with AIN (80% was drug induced). A variety of kidney diseases had UEs. Using a 1% UE cutoff, the comparison of all patients with AIN to those with all other diagnoses showed 30.8% sensitivity and 68.2% specificity, giving positive and negative likelihood ratios of 0.97 and 1.01, respectively. Given this study's 16% prevalence of AIN, the positive and negative predictive values were 15.6% and 83.7%, respectively. At the 5% UE cutoff, sensitivity declined, but specificity improved. The presence of pyuria improved the sensitivity somewhat, with a decrease in specificity. UEs were no better at distinguishing AIN from acute tubular necrosis compared with other kidney diseases. UEs were found in a variety of kidney diseases besides AIN. At the commonly used 1% UE cutoff, the test does not shift pretest probability of AIN in any direction. Even at a 5% cutoff, UEs performed poorly in distinguishing AIN from acute tubular necrosis or other kidney diseases.