Proposal of a linear definition of the Bath Ankylosing Spondylitis Metrology Index (BASMI) and comparison with the 2-step and 10-step definitions

Background: The Bath Ankylosing Spondylitis Metrology Index (BASMI) characterises the spinal mobility of patients with ankylosing spondylitis. Two versions have been published using categorical scores with either scores 0-2 for each of the five assessments, or scores 0-10. For metric purposes, we recently defined a BASMI version with linear score definitions. Aim: to evaluate agreement between three BASMI definitions and to test sensitivity to change. Patients and methods: The performance of the BASMI(2) (based on the 2-step definition), BASMI(10) (based on the 10-step definition), and BASMI(lin) (based on the linear definition), are compared in 598 status assessments and 222 follow-up assessments with a 24-week interval after an intervention with either placebo or a tumour necrosis factor (TNF) blocker from various cohorts of patients with ankylosing spondylitis (AS). Descriptive statistics and Bland-Altman plots were applied to compare the pairwise agreement of the three definitions. To assess sensitivity to change, Guyatt effect size using change data from the placebo and actively treated patients were used. Results: Bland-Altman analysis showed that the differences between BASMI(2) scores and scores obtained by either of the two other definitions were highly dependent on the magnitude of the measurement. Guyatt effect sizes were 0.66 for the BASMI(2), 0.95 for the BASMI(10), and 1.04 for the BASMI(lin), respectively, demonstrating best sensitivity to change for the newly-developed BASMI(lin). Conclusions: The BASMI10 and BASMI(lin) have clear metric advantages as compared to BASMI(2), among which are their superior sensitivity to change and feasibility of BASMI(lin) in computer evaluations. The BASMI(10) and BASMI(2) are not interchangeable.