Is it time to reconsider how we document pediatric uroflow studies?: A study from the SPU Voiding Dysfunction task force

Abstract

Uroflowmetry is a non-invasive study used in the diagnosis and monitoring of treatment response for lower urinary tract disorders. For optimal clinical utility, uroflow studies rely on careful clinical interpretation by a trained provider, but currently there is a lack of accepted standardized normal values for the measured uroflow parameters in children. The International Children's Continence Society proposed standardizing the terminology for uroflow curve shapes. However, the patterning of curves is largely at the physician's subjective discretion. The aims of this study were to understand inter-rater reliability in interpreting uroflow curves and to define characteristics of uroflow curves that could be used to provide definitive criteria for uroflowmetry parameters. All members of the SPU Voiding dysfunction Task Force were invited to contribute deidentified uroflows to a centralized HIPAA complaint receiving database. All studies were then distributed to all raters for review. Each observer's findings were recorded according to ICCS criteria (ICCS), additional readings were done using a previously reported system which defined curves as smooth or fractionated (SF) and whether the shape resembled a bell, tower, or plateau (BTP). Flow indexes (Qact/Qest) (FI) for Qmax and Qavg were generated using formulas previously reported for children 4-12 and for patients≥12 years. A total of 119 uroflow studies were read by 7 raters and curves were contributed from 5 sites. Kappa scores for the 5 readers from different institutions were 0.34 and 0.28 for the ICCS and BTP methods, respectively (both considered fair agreement). Kappa for smooth and fractionated curves was 0.70 (for each; considered substantial agreement), which were the two highest agreement scores obtained throughout the study. Discriminant analysis (DA) revealed that the FI Qmax was the dominant vector, and that the ICCS uroflow parameters have a total of 42.8% prediction rate in the training sample. Using DA of a smooth/fractionated system, the total prediction rates were 72% and 65.5% for smooth and fractionated, respectively. Given the poor inter-rater agreement for analyzing uroflow curve pattern based on ICCS criteria in this study and others, one may find it reasonable to consider alternative approaches to describing and characterizing uroflow curves. Our study is limited by lack of EMG and post-void residual data. For a more objective uroflow interpretation and comparison of studies among different centers, we recommend using our proposed system (based on FI, and smooth vs. fractionated curve pattern), which is more reliable.