Interscanner variation in brain MRI lesion load measurements in MS: implications for clinical trials

We evaluated the effect of interscanner variation on brain MRI-measured lesion volumes and measurement reproducibility in MS. Twenty clinically definite MS patients were each scanned on two or three scanners (a total of 14 scanners were used). In addition, a formalin-fixed MS brain was studied on eight scanners from different manufacturers and with different field strengths. For the formalin-fixed MS brain, on each machine we obtained two scans with slice thicknesses of 5 and 3 mm. Only 5-mm-thick slices were obtained from patients. The lesion volume present on each scan was evaluated three times by a single observer in random order, using a local thresholding technique. In two groups of eight patients scanned on machines with different field strengths, the mean lesion volumes present on scans obtained at 1.5 T were significantly higher than those measured on scans obtained with machines operating at 0.5 and 1.0 T (p < 0.01). When a single observer repeatedly evaluated the same scan, a median introbserver agreement of 98.7% (95% CI, 97.9 to 99.1) was achieved. However, when the observer evaluated the scans from different MRI scanners, the agreement (an interscanner agreement) fell to 91.1% (CI, 90.2 to 94.1). When only scanners operating at 1.5 T were considered, the median interscanner agreement was 96.7% (CI, 95 to 97.5). Also, for the formalin-fixed MS brain, the intraobserver agreements obtained with both slice thicknesses were significantly higher than the corresponding interscanner agreements. The interscanner agreement, but not the intraobserver agreement, obtained with a slice thickness of 3 mm was higher than that obtained with a slice thickness of 5 mm. Our results indicate that lesion volume measurements in MS are influenced significantly by the use of different MR scanners and that a patient included in a serial study should be always scanned with the same MR machine using 3-mm thick slices.