Quantitative bowel apparent diffusion coefficient measurements in children with inflammatory bowel disease are not reproducible

• Published in: Clinical radiology 2018-06, Vol.73 (6), p.574-579
• Main Authors: Watson, T., Calder, A., Barber, J.L.
• Format: Article
• Language: English
• Subjects: Child; Colonic Diseases - pathology; Diffusion Magnetic Resonance Imaging; Humans; Ileal Diseases - pathology; Inflammatory Bowel Diseases - pathology; Jejunal Diseases - pathology; Observer Variation; Reference Standards; Retrospective Studies
• ISSN: 0009-9260; 1365-229X
• DOI: 10.1016/j.crad.2018.01.015

To investigate the intra-observer and interobserver variability of quantitative apparent diffusion coefficient (ADC) measurements in children with inflammatory bowel disease. Nine readers were recruited. Six magnetic resonance imaging (MRI) enterography cases with known active disease in the jejunum, terminal ileum, or colon were analysed. Readers measured repeat ADC values from the known diseased site and an unaffected site, at two sittings. Seven readers completed the study. The Lin concordance coefficient for intra-observer agreement was poor (0.844, 95% confidence interval [CI]: 0.77, 0.896). Bland–Altman limits of agreement for intra-observer agreement were 0.66×10−3 mm2/s (95% CI: 0.46, 0.86), and −0.56×10−3 mm2/s (95% CI: −0.36, −0.76). Therefore, a single measured value would be compatible with no disease, superficial ulceration, or deep ulceration according to published thresholds. Interobserver variability was poor to moderate across all observers (intraclass correlation coefficient [ICC]: 0.51, 95% CI: 0.27, 0.77). Between the two best-agreeing observers, agreement was good using the ICC (ICC 0.85, 95% CI: 0.43, 1.0), but poor using the Lin correlation coefficient (Lin 0.83, 95% CI: 0.65, 0.93), and Bland–Altman. The intra-observer and interobserver agreement is inadequate to allow accurate characterisation of disease activity using previously published thresholds. Qualitative ADC assessment may be preferable. •Quantitative ADC measurement is proposed in the literature to assess severity of IBD.•There is wide inter and intra-observer variability for quantitative ADC measurement.•Measurement variability is too great for reliable use of ADC severity thresholds.•Bland-Altman plots are best for assessing clinical impact of measurement variability.