Variability of "optimal" cut points for mild, moderate, and severe pain: neglected problems when comparing groups.

Defining cut points for mild, moderate, and severe pain intensity on the basis of differences in functional interference has an intuitive appeal. The statistical procedure to derive them proposed in 1995 by Serlin et al. has been widely used. Contrasting cut points between populations have been interpreted as meaningful differences between different chronic pain populations. We explore the variability associated with optimally defined cut points in a large sample of chronic pain patients and in homogeneous subsamples. Ratings of maximal pain intensity (0-10 numeric rating scale, NRS) and pain-related disability were collected in a sample of 2249 children with chronic pain managed in a tertiary pain clinic. First, the "optimal" cut points for the whole sample were determined. Second, the variability of these cut points was quantified by the bootstrap technique. Third, this variability was also assessed in homogeneous subsamples of 650 children with constant pain, 430 children with chronic daily headache, and 295 children with musculoskeletal pain. Our study revealed 3 main findings: (1) The optimal cut points for mild, moderate, and severe pain in the whole sample were 4 and 8 (0-10 NRS). (2) The variability of these cut points within the whole sample was very high, identifying the optimal cut points in only 40% of the time. (3) Similarly large variability was also found in subsamples of patients with a homogeneous pain etiology. Optimal cut points are strongly influenced by random fluctuations within a sample. Differences in optimal cut points between study groups may be explained by chance variation; no other substantial explanation is required. Future studies that aim to interpret differences between groups need to include measures of variability for optimal cut points.