Background The use of patient-reported outcome measures has become routine in clinical practice and in clinical trials. In order to use a specific outcome measure in the daily clinic, the natural variation of the outcome measure must be known. Natural variation may also be characterized as measurement error and is assessed in individuals who are considered to be in “steady state”.
Objectives To examine natural variation of BASFI, patient global assessment (PaGl), pain, fatigue in patients with stable spondyloarthropathy (SpA).
Methods 107 SpA patients treated in the daily clinic with a TNF-inhibitor and characterized by stable disease were identified in the Danish rheumatology registry (DANBIO). According to ASAS response criteria for biological treatment in SpA  and to previous reports on BASDAI measurement errors , stable disease was defined as a change in BASDAI ≤20 between two consecutive visits. Paired data from a single set of such two visits were extracted for each patient. Data comprised BASDAI, BASFI, PaGl, pain and fatigue scored on 0–100 VAS scales. Natural variation was examined using the Bland-Altman method with calculations of lower and upper 95% limits of agreement (LLoA;ULoA) between two consecutive assessments and the corresponding bias (mean of individual differences). Associations between intra-individual inter-visit differences (Δ) were described by linear correlation (r) and stepwise multiple regression analyses (partial regression coefficients (rp) and standard errors of estimation (SEE)).
Results Mean age was 44±14 years, mean inter-visit time duration 16±13 weeks (range 3 – 91) and mean ΔBASDAI 0.0±10.5 (range -20 – 20, ns). Biases were close to 0 for all the variables indicating stable conditions on the group level between the two consecutive visits. On the individual level, differences were more pronounced (Table). LoA for BASFI corresponded to the predefined accepted limits for BASDAI (±20). However, LoA for fatigue, pain and PaGl approximated ±35. No significant correlations were found between the absolute Δvalues of BASDAI, BASFI, fatigue, pain or PaGl and the inter-visit time duration (r range -0.1 – 0.2, ns). ΔBASFI, Δfatigue, Δpain and ΔPaGl were weakly correlated with ΔBASDAI (r range 0.30 – 0.60, p<0.01). In multiple regression analyses, ΔBASFI was best predicted by Δpain, and Δfatigue, Δpain and ΔPaGl by ΔBASDAI (rp range 0.40 – 0.49, p<0.0001, SEE). Although the associations were highly significant, SEEs were high (range 10.8 – 15.6) illustrating poor agreement in individuals.
Conclusions Independently of time duration, fatigue, pain and PaGl fluctuated substantially and unpredictable in SpA patients with stable BASDAI. In the daily clinic, a change in these outcome measures of at least 35 may be considered as natural variation or “measurement error” with no clinically important relation to the disease activity.
Braun J et al. Ann Rheum Dis 2003; 62: 817–24.
Madsen OR et al. Clin Rheumatol 2010; 29: 849–54.
Disclosure of Interest None declared