DIRECT AND INDIRECT EFFECT OF TNF INHIBITORS ON SPINAL MOBILITY IN PEOPLE WITH AXIAL SPONDYLOARTHRITIS AND THE MEDIATOR ROLE OF DISEASE ACTIVITY

BackgroundAlthough it may be difficult to detect changes in spinal mobility on the short term, spinal mobility is considered an important measure to assess the efficacy of drugs used to treat axial spondyloarthritis (axSpA). However, few studies evaluated the long-term impact of biologic treatment on spinal mobility.ObjectivesTo describe the long-term effect of TNF inhibitors (TNFi) on spinal mobility in patients with axSpA, and to determine whether the use of TNFi treatment influences spinal mobility, and if this due to a direct or indirect effect (mediated by disease activity).MethodsWe performed a retrospective observational study, using data collected from patients with a clinical diagnosis of axSpA treated with TNFi at a tertiary care centre where disease activity and metrology assessments are routinely done. Adult patients with at least two Bath Ankylosing Spondylitis Metrology Index (BASMI) measurements were included. Disease activity was measured using the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) and Ankylosing Spondylitis Disease Activity Score C-reactive protein (ASDAS). The longitudinal association between TNFi and improvement in BASDAI/ASDAS was tested using a linear mixed effects model with BASMI as dependent variable. To test whether TNFi had a direct effect on BASMI, not mediated by disease activity, we tested that TNFi treatment was not conditionally independent of BASMI given BASDAI/ASDAS (Figure 1). We tested whether the nodes TNFi and BASMI were disconnected if we removed BASDAI and ASDAS. To test this conditional independence, we first built a linear mixed effects model for BASMI given BASDAI or ASDAS when the patient was under TNFi and used this model to predict a 95% confidence interval (CI) for BASMI given the data for BASDAI/ASDAS when the patient was without TNFi. We checked whether the true value of BASMI lay within this 95% CI and performed a hypothesis test for binomial distribution where H0: p=0.95. To test for the indirect effect of TNFi on BASMI reduction, mediated through the disease activity, we regressed BASMI on BASDAI/ASDAS, TNFi (if there was a direct effect), demographics, presence of radiographic (r-) axSpA and HLA-B27 positivity, using a linear mixed effects model adjusted for within-patient correlation.Figure 1.Indirect effect of TNFi on BASMI (represented by the full line), through the influence of TNFi on disease activity, adjusted by other confounders and direct effect of TNFi on BASMI (dashed line), independently of disease activity.ResultsData from 188 patients and 1326 visits were analysed. Mean age was 45.6 (SD 11.6) years, mean disease duration was 15.8 (SD 9.64) years, 152 (80.9%) were male, 120 (73.6%) had r-axSpA, and 83 (74.8%) were HLA-B27 positive. Mean follow-up time was 8.0 (SD 4.4) years, ranging from 0.8 to 18.2 years. Treatment with TNFi was significantly associated with long-term improvement in BASMI (B=-0.423, 95% CI=[-0.553,-0.292], p<0.001). An indirect effect of TNFi on BASMI improvement was observed, mediated by reduction in disease activity, measured by BASDAI (B=0.146, 95% CI=[0.092, 0.200], p<0.001) or ASDAS (B=0.405, 95% CI=[0.260, 0.549], p<0.001). Using conditional independence tests, a direct effect of TNFi on BASMI improvement was also observed, independently of disease activity, when BASDAI was used (p<0.001) as a covariate, but not when ASDAS was used (p=0.3104). The direct effect of TNFi (B=-0.300, 95% CI=[-0.576,-0.025], p<0.001) on BASMI was estimated in the BASDAI-adjusted mixed effects model.ConclusionTNFi are effective at improving BASMI in patients with axSpA, in a real-life setting. This effect is mainly explained by the reduction in disease activity. However, a direct effect of TNFi on BASMI could also be demonstrated, when disease activity was measured by BASDAI, suggesting that ASDAS captures additional factors that can influence spinal mobility. These potential factors deserve further investigation, but they could for example include biomechanical properties of tendons and myofascial tissue.Disclosure of InterestsAna Sofia Pinto: None declared, Bohao Yao: None declared, Claire Harris: None declared, Rhys Hayward: None declared, Andrew Keat: None declared, Pedro Machado Speakers bureau: Received consulting/speaker’s fees from Abbvie, BMS, Celgene, Eli Lilly, Galapagos, Janssen, MSD, Novartis, Orphazyme, Pfizer, Roche and UCB, all unrelated to thismanuscript, Consultant of: Received consulting/speaker’s fees from Abbvie, BMS, Celgene, Eli Lilly, Galapagos, Janssen, MSD, Novartis, Orphazyme, Pfizer, Roche and UCB, all unrelated to thismanuscript