Counting your chickens before they hatch: improvements in an untreated chronic pain population, beyond regression to the mean and the placebo effect

Abstract

AbstractBackground and aimsIsolating the effect of an intervention from the natural course and fluctuations of a condition is a challenge in any clinical trial, particularly in the field of pain. Regression to the mean (RTM), wherein extreme scores are more likely to be followed by more average scores, may explain some of those observed fluctuations. However, while this phenomenon is relatively well-known, its effect on outcome measures is rarely quantified, and often only evoked as a potential confound. In this paper, we describe and quantify such symptom fluctuations in a chronic pain population in the absence of treatment, and compare the relative stability of various self-reported outcome measures in untreated chronic low back pain (CLBP) patients and healthy controls (HC).MethodsTwenty-three untreated CLBP patients and 25 HC took part in this observational study, wherein they were asked to complete an array of commonly used questionnaires in pain studies (including the Pain Catastrophizing Scale [PCS], State and Trait Anxiety Inventory [STAI], Central Sensitization Inventory [CSI], Pain Disability Index [PDI], Brief Pain Inventory [BPI] etc.) during each of three visits (V1, V2, V3) at 2-month intervals. Scores at V1 were classified into three subgroups (extremely high, normal and extremely low), based on z-scores. The average delta (Δ=V2-V1) was calculated for each subgroup, for each questionnaire, to describe the evolution of scores over time. This analysis was repeated with the data for V2 and V3.ResultsHigh initial scores were likely to be followed by more average scores; for instance, the “extremely high” subgroup for the PCS (a reputably ‘stable’ questionnaire) had an average decrease of 12/52 from V1 to V2. Participants with “average” initial scores tended to show a small decrease over time, and participants with “extremely low” initial scores tended to remain stable. However, while the pattern of fluctuation in the three subgroups was similar across questionnaires, the magnitude of these fluctuations varied greatly. The STAI and CSI were the most stable questionnaires of all, with even the “extremely high” subgroup showing little or no improvement over time. The least stable questionnaires were the PCS, PDI and BPI.Discussion and conclusionThese pain trajectories in untreated patients cannot be attributable to RTM alone because of their asymmetry, nor to the placebo effect as they occurred in the absence of any intervention. We propose that the observed improvements could be the result of an Effect of Care, wherein participants had meaningful improvements simply from taking part in a study and talking about their pain to benevolent research staff, despite the absence of active or sham treatment. These findings have important clinical ramifications. Beyond simply raising a flag as to the existence (and significance) of Effect of Care, we provide a questionnaire-specific baseline of expected fluctuations based on initial score, against which researchers can compare results from clinical trials when trying to isolate the effect of their intervention.