Background We have previously described a model of fatigue in patients with primary Sjogren's syndrome (PSS) based on the levels of serum cytokines, pain and depression scores. Importantly, removal of cytokines from this model substantially reduced the accuracy suggesting that cytokines may have a key role in the biological basis of fatigue . However, interpreting the model is complicated by the complexity of the immune system and the likely multiple interactions between numerous cytokines and other variables . Structural equation modelling (SEM) is a statistical technique that allows for analysis of one or multiple independent variables with one or multiple dependent variables. SEM consists of two components – the structural model, which represents the relationships between the theoretical variables, and the measurement model, which are the relationships between the latent variables and their measures .
Objectives To use SEM to test our hypothesis that the balance between pro-inflammatory and anti-inflammatory cytokines play an important role in determining severity of fatigue in patients with PSS.
Methods We used Canonical Correpondence Analysis (CCA) to investigate the variation in cytokine expression across our spectrum of fatigue patietns to explore interactions and dependencies between cytokines. We then built a conceptual model based on the literature representing the likely relationships between fatigue and various proinflammatory and anti-inflammatory cytokines and other soluble molecules in the serum. This conceptual model was then challenged using serum data and fatigue scores of 161 PSS patients from the UK primary Sjogren's syndrome registry. Model fit was assessed using the Confirmatory Factor Index, the Root Mean Square Error of Association and the Standardised Root Mean Square Residual. We also analysed changes in fatigue scores over a period between 1–4 years.
Results CCA revealed the first axis of ordination (CCA1) broadly correlates with fatigue, consists of many pro-inflammatory cytokines including TNFα, LTα and IFNγ, IL17, which were negatively correlated with fatigue while IL-6 and MCP1, which were positively associated with increased fatigue severity. The second axis (CCA2) reflects a trend in cytokines which appear to relate to patients' age. Fatigue scores were largely stable over time and therefore data were not included in the SEM analysis. The main pro-inflammatory SEM model showed fatigue was negatively associated with pro-inflammatory cytokine activity (p=0.019); IL-10 drove IP-10 (p=0.000); and IL-10 was driven by IL-6 (p=0.006) (Fig. 1)
Conclusions Chronic fatigue in PSS is negatively associated with many pro-inflammatory cytokines. We hypothesize that it reflects adaptive biological processes, which occurs after chronic exposure to inflammation in conditions such as PSS.
Howard-Tripp N, et al. RMD Open. 2016;2:e000282.
Genser B, et al. BMC immunology. 2007;8:1.
Beran TN & Violato C. BMC Research Notes. 2010;3:267.
Acknowledgements All UKPSSR participants.
Disclosure of Interest D. Gardiner: None declared, W. Reid: None declared, J. Tarn: None declared, D. Lendrem: None declared, N. Howard-Tripp: None declared, S. Bowman Consultant for: Cellgene, Glenmark, GSK, Eli Lilly, Novartis, Roche, Takeda, UCB, B. Griffiths: None declared, S. Rushton: None declared, W.-F. Ng Consultant for: Pfizer, UCB, MedImmune, Takeda and Sanofi