Interobserver variability for interpretation of DMSA scans in the RIVUR trial

Introduction Technetium-99m dimercaptosuccinic acid (DMSA) renal scans are used in the diagnosis of renal scarring. In the Randomized Intervention for Children with Vesicoureteral Reflux (RIVUR) trial that randomized 607 children, DMSA renal scans were used for evaluating the presence and the severity of renal scarring. Objective The aim was to determine interobserver variability in reporting of DMSA renal scans in the RIVUR trial. Study design We compared DMSA renal scan reports for renal scarring and acute pyelonephritis from all non-reference local radiologists (ALRs) at study sites with adjudicated as well as non-adjudicated reports from two reference radiologists (RRs) of the RIVUR trial. Two-way comparisons of concordant and discrepant responses were analyzed using an unweighted kappa statistic between the ALR and the adjudicated RR interpretations. All analyses were performed using SAS v 9.4 (SAS institute 2015) and significance was determined at the 0.05 level. Results Of the 2872 kidneys evaluated, adjudicated RR reports had 119 (4%) kidneys with renal scarring compared with 212 (7%) by the ALRs. For 79% kidneys the grading for scarring reported by ALRs was either upgraded (24%) or downgraded (55%) by RRs. For acute pyelonephritis (n = 2924), adjudicated RR reports had 85 (3%) kidneys with pyelonephritis compared with 151 (5%) by the ALRs. For 85% kidneys, the grading for pyelonephritis reported by the ALRs was either upgraded (28%) or downgraded (57%) by the RRs. A three-way comparison revealed that all three (RR1, RR2, and ALR) agreed over presence of renal scarring in 19% cases and two of the three agreed in 80% cases. The respective numbers for pyelonephritis were 13% and 84%. The agreement rate for all DMSA scan reports between the RRs and the ALRs was 93%. Discussion The study revealed significant interobserver variability in the reporting of abnormal DMSA renal scans compared with the previously published studies. A noteworthy limitation was a lack of uniformity in local reporting of the scans. Conclusions Our study highlights the need for optimizing the clinical yield of DMSA renal scans by more specific guidelines, particularly for standardized and uniform interpretation.