Comment on: “Radiation-Free Assessment of Scoliosis: A Reliability and Validity Study for Ultrasound Angles”

Abstract

We read with interest the study by Zhu et al 1 evaluating the reliability and validity of ultrasound-derived angles for radiation-free scoliosis screening. The intra-and interobserver reliability data are a useful contribution, but before the reported minimal detectable change (MDC) and validity estimates are applied to longitudinal curve monitoring, several methodological points deserve comment. Validity was framed largely through simple linear regression coefficients (R), with agreement categorized as poor, moderate, or excellent by R thresholds, even though Bland-Altman plots were also generated. Correlation coefficients are a misleading index of agreement between two measurement methods, because two techniques can correlate strongly while differing by a clinically meaningful, systematic margin. 2 Framing the validity conclusions mainly around R risks overstating how interchangeable ultrasound and radiographic Cobb angle measurements are. Second, the MDC of 4.93 degrees (automatic) and up to 4.43 degrees (manual) was derived from repeat scans obtained by two observers within the same visit. This same-session design captures inter-rater and immediate re-scan variability but not the between-visit sources of error, such as differences in probe angulation, coupling-gel pressure, and patient posture across weeks to months, that longitudinal monitoring, the technique’s stated principal use, would actually encounter. Standard error of measurement and MDC estimates depend heavily on the conditions under which they are derived, and same-day data may understate the error relevant to serial follow-up. 3 Two further points bear on generalizability. Subjects were stratified into seven overlapping subgroups, by apical location, three BMI categories, and scoliosis status, from a cohort of only 94 participants, without correction for multiple comparisons; the low R of 0.425 in the overweight subgroup may reflect small-cell instability as much as a genuine BMI effect, a pattern that subgroup-analysis guidance cautions