Improving interobserver variation in reporting gadolinium-enhanced MRI lesions in multiple sclerosis.

Abstract

Gadolinium-enhanced MRI is a sensitive and objective means to monitor disease activity in multiple sclerosis (MS). We evaluated the interobserver agreement and the value of observer training in reporting enhancing lesions from serial MRI. Scans of 16 MS patients were evaluated by five inexperienced and five experienced observers before and after consensus formation and training. The number of lesions at baseline, and the number of new and persistent lesions at follow-up were scored. For each condition, weighted kappa values (kappa) and the mean average difference to the median (MADM) scores were calculated. Without training, the experienced readers showed good agreement on number of lesions at baseline and new lesions at follow-up, and moderate agreement for persistent lesions. The inexperienced readers showed poor agreement for baseline and persistent lesions, and moderate agreement for new lesions. After training, both groups reported lower absolute numbers of lesions, especially the inexperienced readers. The experienced readers showed good agreement for all lesion types, the inexperienced readers showed agreement for baseline and new lesions, and agreement was moderate for persistent lesions. In both groups MADM scores were < 0.72 for baseline and new lesions, but > 1.2 for persistent lesions. Interobserver agreement is improved by training, especially in inexperienced readers. Interobserver agreement in reporting gadolinium-enhanced lesions is high, which validates the use of serial, enhanced MRI as an outcome parameter in treatment trials in MS.