Consensus approach for 3D joint space width of metacarpophalangeal joints of rheumatoid arthritis patients using high-resolution peripheral quantitative computed tomography

Kathryn S. Stok, Andrew J. Burghardt, Stephanie Boutroy, Michiel P. H. Peters, Sarah L. Manske, Vincent Stadelmann, Nicolas Vilayphiou, Joop P. van den Bergh, Piet Geusens, Xiaojuan Li, Hubert Marotte, Bert van Rietbergen, Steven K. Boyd, Cheryl Barnabe, for the SPECTRA Collaboration

Abstract

Background: Joint space assessment for rheumatoid arthritis (RA) by ordinal conventional radiographic scales is susceptible to floor and ceiling effects. High-resolution peripheral quantitative computed tomography (HR-pQCT) provides superior resolution, and may detect earlier changes. The goal of this work was to compare existing 3D methods to calculate joint space width (JSW) metrics in human metacarpophalangeal (MCP) joints with HR-pQCT and reach consensus for future studies. Using the consensus method, we established reproducibility with repositioning as well as feasibility for use in second- generation HR-pQCT scanners.
Methods: Three published JSW methods were compared using datasets from individuals with RA from three research centers. A SPECTRA consensus method was developed to take advantage of strengths of the individual methods. Using the SPECTRA method, reproducibility after repositioning was tested and agreement between scanner generations was also established.
Results: When comparing existing JSW methods, excellent agreement was shown for JSW minimum and mean (ICC 0.987–0.996) but not maximum and volume (ICC 0.000–0.897). Differences were identified as variations in volume definitions and algorithmic differences that generated high sensitivity to boundary conditions. The SPECTRA consensus method reduced this sensitivity, demonstrating good scan-rescan reliability (ICC >0.911) except for minimum JSW (ICC 0.656). There was strong agreement between results from first- and second-generation HR-pQCT (ICC >0.833).
Conclusions: The SPECTRA consensus method combines unique strengths of three independently- developed algorithms and leverages underlying software updates to provide a mature analysis to measure 3D JSW. This method is robust with respect to repositioning and scanner generations, suggesting its suitability for detecting change.