Automated joint space width quantification of hand and wrist joints: a proof of concept study

Abstract

Objective. To compare as proof of concept the sensitivity to change of automated quantification of radiographic wrist and hand joint space width (JSW) with scoring JSW according to the Sharp/van der Heijde scoring method (SHS) in two strategy groups of a treat-to-target and tight-control early rheumatoid arthritis (RA) study. Methods. Digital radiographs were assessed for JSW changes of 134 patients of the 236 patients participating in the second Computer Assisted Management in Early Rheumatoid Arthritis trial, of whom both baseline and year 2 radiographs were available (year 1 radiographs n = 125). Of those 134 patients, 70 started with methotrexate and prednisone (MTX + Pred) and 64 with MTX and placebo (MTX + Plac). JSW change over 1 and 2 years of the hands and wrists was assessed, applying both the joint space narrowing (JSN) subscore of the SHS by 2 readers and the automated assessment with the JSW quantification software 'JSQ'. For both methods, progression of JSW change of the hand and wrist was analysed using linear mixed modelling (dependent variable 'JSW', factor 'strategy group', covariate 'follow-up time in years', interaction term 'strategy group*follow-up time'; radiographs of baseline, year 1 and year 2 were used). For each method the standardised mean difference (SMD) for the change in JSW from baseline to year 2 between the treatment strategies was obtained using a non-parametric method. Results. Patient characteristics of the current subpopulation were similar to those of the whole study population. JSN of the hand and wrist according to SHS at 2 years was present in 16 vs. 23% in the MTX+Pred group vs. the MTX+Plac group. The mean yearly progression rates of JSW change of the hands and wrists using JSQ were -0.00mm (95% confidence interval (CI) - 0.01; 0.01) for MTX + Pred vs. - 0.02mm (95%CI -0.03; - 0.01) for MTX + Plac, p=0.045, and using SHS JSN they were 0.19 units (95%CI 0.09; 0.30) vs. 0.30 units (95%CI 0.14; 0.45) for MTX + Pred vs. MTX + Plac, p = 0.271. The SMD for the change from baseline to year 2 between the treatment strategies was 0.37 for JSQ and 0.13 for SHS JSN. Conclusion. In this proof of concept study the yearly progression rate of JSW change of hand and wrist joints, according to the automated JSW quantification software package 'JSQ', was higher in the group initiating MTX+Plac than in the group initiating MTX+Pred. A similar trend was seen with the JSN assessment according to the SHS method of the hand and wrist. However, JSN of the hand and wrist according to SHS, the current gold standard to assess radiographic progression, was seen in only about 20%. Therefore, further studies are needed to conclude firmly that JSQ should be incorporated into quantitative scoring of radiographs in RA.