Clinically important difference thresholds of the visual analog scale: a conceptual model for identifying meaningful intraindividual changes for pain intensity.

The aim of this study was to estimate a range of clinically important difference (CID) values of the visual analog scale for pain intensity (VAS-PI), and to assess the effect of patient baseline characteristics on VAS change scores. Data from a prospective cohort study with 678 patients with subacute and chronic temporomandibular disorder pain were analyzed. Patients were divided into 9 cohorts on the basis of the baseline VAS score and the duration of pain. The CID was estimated over a 12-week period, and 2 different methods were used: (1) mean change scores, and (2) optimal cutoff point in receiver operator characteristic curves. The patient's global impression of change was used as an external criterion. The general linear model univariate analysis was applied to assess the effect of baseline pain level and duration of pain on the raw VAS change scores, while adjusting for age and sex. The CID mean change ranged from 20.9 to 57.5 mm (64.1-76.3%), and the CID optimal cutoff point from 11.5 to 28.5 mm (29.9-47.7%). For the VAS change scores, the main effect of the variable baseline pain level was significant (F=107.09, P<.001). However, there was no significant baseline pain level by duration of pain interaction effect (F=1.13, P=.340). On the basis of the results, we advocate the choice of a single CID value according to the context of the patient's baseline level of pain.