Individual-Reader Diagnostic Performance and Between-Reader Agreement in Assessment of Subjects with Parkinsonian Syndrome or Dementia Using 123I-Ioflupane Injection (DaTscan) Imaging
J. Seibyl,A. Kupsch,J. Booij,D. Grosset,D. Costa,R. Hauser,J. Darcourt,N. Bajaj,Z. Walker,K. Marek,I. McKeith,J. O'Brien,K. Tatsch,E. Tolosa,R. Dierckx
DOI: https://doi.org/10.2967/jnumed.114.140228
2014-08-01
Journal of Nuclear Medicine
Abstract:Establishing an early, accurate diagnosis is fundamental for appropriate clinical management of patients with movement disorders or dementia. Ioflupane 123I Injection (DaTscan, 123I-ioflupane) is an important adjunct to support the clinical diagnosis. Understanding individual-reader diagnostic performance of 123I-ioflupane in a variety of clinical scenarios is essential. Methods: Sensitivity, specificity, interreader, and intrareader data from 5 multicenter clinical studies were reviewed. The different study designs offered an assortment of variables to assess the effects on the diagnostic performance of 123I-ioflupane: on-site versus 3–5 blinded image readers, number of image evaluations, early/uncertain versus late/confirmed clinical diagnosis as reference standard, and subjects with movement disorders versus dementia. Results: Eight hundred eighteen subjects had individual-reader efficacy data available for analysis. In general, sensitivity and specificity were high and comparable between on-site versus blinded independent readers. In subjects with dementia, when the clinical diagnosis was made at month 12 versus baseline, specificity improved from 77.4%–91.2% to 81.6%–95.0%. In subjects with movement disorders, this effect was observed to an even greater extent, when diagnostic performance using month-18 diagnosis as a reference standard (sensitivity, 67.0%–73.7%; specificity, 75.0%–83.3%) was compared versus month-36 diagnosis (77.5%–80.3% and 90.3%–96.8%, respectively). Diagnostic performance was similar in subjects with dementia (74.4%–89.9% and 77.4%–95.0%, respectively) and subjects with movement disorders (67.0%–97.9% and 71.4%–98.4%, respectively). In most of the comparisons, between-reader agreement was very good (almost perfect), with κ ranging from 0.81 to 1.00. Within-reader agreement, measured in 1 study, was 100% for 3 blinded readers. Conclusion: Individual-reader diagnostic performance, as assessed by measuring sensitivity and specificity of 123I-ioflupane to detect the presence or absence of striatal dopaminergic deficit, using the clinical diagnosis as a reference standard, was high in subjects with either movement disorders or dementia and was similar in on-site readers versus blinded analyses. Between- and within-reader agreements were very good (almost perfect). Longer follow-up between imaging and clinical diagnosis improved the diagnostic accuracy, most likely due to improvement in the clinical diagnosis reference standard, rather than changes in reader accuracy.