Integrated longitudinal analysis does not compromise precision and reduces bias in the study of imaging outcomes: A comparative 5-year analysis in the DESIR cohort

Abstract

ObjectiveTo assess if an integrated longitudinal analysis using all available imaging data affects the precision of estimates of change in patients with axial spondyloarthritis (axSpA), with completers analysis as reference standard. MethodsPatients from the DESIR cohort fulfilling the ASAS axSpA criteria were included. Radiographs and MRIs of the sacroiliac joints and spine were obtained at baseline, 1, 2 and 5 years. Each image was scored by 2 or 3 readers in 3 ‘reading-waves’ (or campaigns). Each outcome was analyzed: i. According to a ‘combination algorithm’ (e.g. ‘2 out of 3′ for binary scores); and ii. Per reader. Change over time was analyzed with generalized estimating equations by 3 approaches: (a)‘integrated-analysis’ (all patients with ≥1 score from ≥1 reader from all waves); (b1)Completers-only analysis (patients with 5-year follow-up, using scores from individual readers); (b2)Completers analysis using a ‘combination algorithm’ (as (b1) but with combined scores). Approaches (b1) and (b2) were considered the ‘reference’. ResultsIn total, 413 patients were included. The ‘integrated analysis’ was more inclusive with similar levels of precision of the change estimates as compared to both completers analyses. In fact, for low-incident outcomes (e.g.% mNY-positive over 5-years), an increased incidence was ‘captured’, with more precision, by the ‘integrated analysis’ compared to the completers analysis with combined scores (% change/year (95%CI): 1.1 (0.7; 1.5) vs 1.2 (0.5; 1.8), respectively). ConclusionAn efficient and entirely assumption-free ‘integrated analysis’ does not jeopardize precision of the estimates of change in imaging parameters and may yield increased statistical power for detecting changes with low incidence.