Background:Radiographic spinal progression determinates functional status and mobility in ankylosing spondylitis (AS)1.Objectives:To analyse whether biomarker of inflammation, bone turnover and adipokines at baseline or their change after 3 months or 2 years can predict spinal radiographic progression after 2 years in AS patients treated with TNF-α inhibitors (TNFi).Methods:Consecutive AS patients from the Groningen Leeuwarden Axial Spondyloarthritis (GLAS) cohort2 starting TNFi between 2004 and 2012 were included. The following serum biomarkers were measured at baseline, 3 months and 2 years of follow-up with ELISA: - Markers of inflammation: calprotectin, matrix metalloproteinase-3 (MMP-3), vascular endothelial growth factor (VEGF) - Markers of bone turnover: bone-specific alkaline phosphatase (BALP), serum C-terminal telopeptide (sCTX), osteocalcin (OC), osteoprotegerin (OPG), procollagen typ I and II N-terminal propeptide (PINP; PIINP), sclerostin. - Adipokines: high molecular weight (HMW) adiponectin, leptin, visfatinTwo independent readers assessed spinal radiographs at baseline and 2 years of follow-up according to the modified Stoke Ankylosing Spondylitis Spine Score (mSASSS). Radiographic spinal progression was defined as mSASSS change ≥2 units or the formation of ≥1 new syndesmophyte over 2 years. Logistic regression was performed to examine the association between biomarker values at baseline, their change after 3 months and 2 years and radiographic spinal progression. Multivariable models for each biomarker were adjusted for mSASSS or syndesmophytes at baseline, elevated CRP (≥5mg/l), smoking status, male gender, symptom duration, BMI, and baseline biomarker level (the latter only in models with biomarker change).Results:Of the 137 included AS patients, 72% were male, 79% HLAB27+; mean age at baseline was 42 years (SD 10.8), ASDAScrp 3.8 (0.8) and mSASSS 10.6 (16.1). After 2 years of follow-up, 33% showed mSASSS change ≥2 units and 24% had developed ≥1 new syndesmophyte. Serum levels of biomarkers of inflammation and bone formation showed significant changes under TNFi therapy, whereas adipokine levels were not altered from baseline (Figure 1).Univariable logistic regression revealed a significant association of baseline visfatin (odds ratio OR [95% confidence interval] 1.106 [1.007-1.215]) and sclerostin serum levels (OR 1.006 [1.001-1.011]) with mSASSS progression after 2 years. Baseline sclerostin levels were also associated with syndesmophyte progression (OR 1.007 [1.001-1.013]). In multivariable logistic analysis, only baseline visfatin level remained significantly associated (OR 1.465 [1.137-1.889]) with mSASSS progression. Furthermore, baseline calprotectin showed a positive association with both, mSASSS (OR 1.195 [1.055-1.355]) and syndesmophyte progression (OR 1.107 [1.001-1.225]) when adjusting for known risk factors for radiographic progression.Univariable logistic regression showed that change of sclerostin after 3 months was associated with syndesmophytes progression (OR 1.007 [1.000-1.015), change of PINP level after 2 years was associated with mSASSS progression (OR 1.027 [1.003-1.052]) and change of visfatin after 2 years was associated with both measures of radiographic progression – mSASSS (OR 1.108 [1.004-1.224]) and syndesmophyte formation (OR 1.115; [1.002-1.24]). However, those associations were lost in multivariable analysis.Conclusion:Independent of known risk factors, baseline calprotectin and visfatin levels were associated with radiographic spinal progression after 2 years of TNFi. Although biomarkers of inflammation and bone formation showed significant changes under TNFi therapy, these changes were not significantly related to radiographic spinal progression in our cohort of AS patients.References:[1]Poddubnyy et al 2018[2]Maas et al 2019Acknowledgements:Dr. Judith Rademacher is participant in the BIH-Charité Clinician Scientist Program funded by the Charité –Universitätsmedizin Berlin and the Berlin Institute of Health.Disclosure of Interests:Judith Rademacher: None declared, Mark Siderius: None declared, Laura Gellert: None declared, Freke Wink Consultant of: AbbVie, Maryna Verba: None declared, Fiona Maas: None declared, Lorraine M Tietz: None declared, Denis Poddubnyy: None declared, Anneke Spoorenberg Consultant of: Abbvie, Pfizer, MSD, UCB, Lilly and Novartis, Grant/research support from: Abbvie, Pfizer, UCB, Novartis, Suzanne Arends Grant/research support from: Pfizer.
Prediction of Attrition in Large Longitudinal Studies: Tree-based methods versus Multinomial Logistic Models
