Abstract
74
Objectives 123I-ioflupane SPECT is a well-established radiotracer for the assessment of the striatal dopamine transporter. Visual image interpretation has a high diagnostic accuracy for the differentiation of degenerative parkinsonian syndromes (PS) from essential tremor (ET)[1] and probable dementia with Lewy bodies (pDLB) from Alzheimer’s disease (AD)[2]. This study investigated impact on accuracy and reader confidence offered by the addition of image quantification compared to visual interpretation alone.
Methods We collected 304 123I-ioflupane SPECT (DaTscan™) images from 3 multi-center, Phase 3/4 trials, enrolling subjects with a clinical diagnosis of PS, non-PS, pDLB and AD established after a minimum follow-up of 12 months. Images were reconstructed with standardised parameters before striatal binding ratios were quantified against a normal database with the DaTQUANT™ software. Images were assessed by five blinded readers (Nuclear Medicine physicians in the US) who had limited prior clinical experience with 123I-ioflupane interpretation (i.e. between 5 and 50 prior assessments). Two reads were performed, separated by at least 1 month, in which readers were presented with either visual only or combined (i.e. visual plus quantitative) data. Readers were not given any information on the patient or the image source. Readers were asked to rate confidence of image interpretation (from 1 - very challenging to 5 - very easy) and judge scans as easy or difficult to read. Diagnostic accuracy was assessed comparing image results with the standard of truth by measuring the positive percent agreement (PPA, equivalent to sensitivity) and the negative percent agreement (NPA, equivalent to specificity).
Results Combined read compared to visual read analysis showed a small, non-significant increase of mean NPA (87.9% vs 89.9%) and equivalent PPA (80.1% vs 80.2%). Readers who initially read in the combined analysis had significantly greater accuracy (85.8% vs 79.2%, p=0.0178) and were closer to the accuracy of the more expert readers in the original 3 studies (range 83.3% to 87.2%). Mean reader confidence in the interpretation of images was high and showed a significant improvement using combined analysis in the total population (i.e. 4.25 visual vs 4.37 combined, p<0.0001) as well as in the subsets of movement disorders and dementia. More images met at least 3 out of 8 difficult to read criteria in the visual only analysis compared to the combined read (12.2% vs 7.9%). In difficult to read cases, PPA remained high for both visual only and combined assessment while NPA trended to increase with the combined assessment, with differences becoming greater at increased levels of difficulty. Intra- and inter-reader agreement was very high (kappa >0.85). The combined read improved the percentage of agreement among all readers from 82.8% (visual) vs 89.9% (combined) and kappa values from 0.85 to 0.91. Similar trends were observed in the movement disorders and dementia populations.
Conclusions Addition of quantification allowed readers with limited experience in the interpretation of 123I-ioflupane SPECT scans to perform as well as the more experienced readers used for the initial clinical studies. We observed minor but consistent improvements in diagnostic accuracy while offering an increase in reader confidence that may result in fewer scans being considered difficult to read.