Background Structural progression in rheumatoid arthritis (RA) is a major outcome. Radiographic scoring should be performed by one or two readers, blinded to clinical data. This method may be limited in terms of feasibility in large observational longitudinal research including large number of patients and multiple timepoints.
Objectives To compare a new method for radiographic scoring involving multi-reader evaluation to a mono-reader evaluation taken as reference.
Methods 11 rheumatologists were trained for radiographic reading according to Sharp-van der Heijde Score (SHS) (objective: ICC≥0.8 for intra and inter-reliabilities). ESPOIR cohort is a longitudinal observational study of early arthritis (ie; at least 2 swollen joints, symptom duration:6 weeks-6 months, no DMARDs). Xrays of hands and feet were performed each year. All patients with complete radiographic data at M0 and M60 were included in this study.
Xrays were scored (SHS) according to 2 different methods (monoreader or multireader evaluations).The “mono-reader evaluation” was taken as reference and consisted in scoring all the patients by one reader (M0 and M60). A “multi-reader evaluation” was compared to the monoreader evaluation. For this multireader evaluation, patients were randomly divided in equal subgroups which were randomly allocated to the readers. Each reader had to read all available Xrays (M0 to M60) of the patients allocated to him. Xrays were scored according to SHS, blindly from clinical data, in chronological order.
Statistical analysis was performed using R.2 software. Intraclass correlation coefficients (ICC) for SHS scores and kappa coefficients for rapid radiographic progression (RRP: change (M60-M0) of SHS≥5) between multireader and monoreader evaluations were calculated.
Results Patients characteristics in the ESPOIR cohort have already been published (1). 524 patients had complete radiographic data at M0 and M60 and were included in this study. 144 patients (34.6%) had RRP. While correlations between the multireader and monoreader readings were moderate at M0, good correlations of the change of SHS scores between M0 and M60 were observed. Moderate agreements for the evaluation of structural progression were observed between the 2 methods, with good agreement for erosion progression.
Conclusions Multireader evaluation for radiographic structural progression could be an alternative to Monoreader evaluation in large observational longitudinal cohort.
Tobon et al. Arthritis Care & Research Vol. 65, No. 12, December 2013, pp 1907–1915.
Disclosure of Interest None declared