Background Rheumatoid arthritis (RA) disease activity is measured using DAS28 – consisting of patient-reported, clinician-observed and laboratory measures. A higher proportion of DAS28 attributed to patient-reported measures (1), fatigue, anxiety and depression might reflect augmented central pain processing.

Objectives We hypothesised that subgroups of people with RA showed different contributions from inflammation and central pain mechanisms, and had different clinical needs.

Methods Data were analysed from (1) The Early RA Network (ERAN) inception cohort (n=828); and British Society for Rheumatology Biologics Register (BSRBR) cohorts (2) commencing anti-TNF (n=9905) or (3) using non-biologic (control) DMARDs (n=2581). Subgroups were defined by latent class analysis (LCA) using erythrocyte sedimentation rate (ESR), swollen joint count (SJC), tender joint count (TJC), visual analogue scale-general health (VAS), and SF36 Bodily Pain, Vitality and Mental Health. Scores were standardised to permit graphical comparison. Latent classes were named with reference to disease activity and discordance/concordance of patient-reported variables with externally-measured variables. Prognosis was examined using MANOVA.

Results Five, 4 and 4 latent classes were found respectively in ERAN, BSRBR anti-TNF and control cohorts. The latent classes displayed either concordance across different measures; discordantly higher patient-reported measures despite less markedly elevated ESR/SJC; or discordantly lower patient-reported measures despite elevated ESR/SJC. Latent classes with discordantly higher patient-reported measures represented 12%, 32% and 22% of the ERAN, BSRBR anti-TNF and BSRBR control cohorts respectively. The 5 latent classes found in the ERAN cohort are shown in figure 1. The proportion of people assigned with >80% probability into latent classes were 76% in ERAN, 58% in the anti-TNF and 72% in the control cohort. The latent classes showing discordantly higher patient-reported variables contained more females, and showed worse function.

In those latent classes with higher scores at baseline, the DAS28, HAQ and SF36 bodily pain scores improved over the first 12 months of follow up (p<0.001 for all comparisons), such that scores differed between latent classes less at follow up than at baseline. Patients in the concordant latent classes with low patient-reported measures, low SJC and ESR at baseline did not further improve after 1 year in ERAN or BSRBR-Control cohorts, and displayed clinically important worsening of Bodily Pain scores in the BSRBR anti-TNF cohort (mean at baseline: 46, at 1 year: 41; p<0.001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Latent classes showing the mean levels of each variable.

Conclusions Discordant latent classes can be identified in people with RA. There might be a sizeable discordant subgroup with active RA (12 to 32%), and also a group of people with mild disease being considered for anti-TNF treatment, for whom pain management strategies might add benefit to DMARD therapy.

1. McWilliams et al Arthritis Care and Research 2012;64(10):1505–13

Acknowledgement Funded by Pfizer iCRP grant

Disclosure of Interest D. Mcwilliams Grant/research support from: Pfizer Ltd, E. Ferguson: None declared, A. Young: None declared, P. Kiely: None declared, D. Walsh Grant/research support from: Pfizer Ltd, Consultant for: Novartis