Background and Aims
Baseline Pain Variability (PV) is often cited as a key predictor of placebo response (1-3). This variability can be caused both by actual fluctuating pain levels and by an inability to accurately assess one’s pain. Some suggest that this latter source may indicate patients more prone to contextual influence and thus more likely to respond strongly to placebo (3). However, a recent analysis (4) suggests that when accounting for regression-to-the-mean, PV is no longer prognostic of the response. This analysis aims therefore to explore if PV remains a relevant prognostic factor to consider in the analysis of randomized clinical trials (RCTs). Additionally, we had previously established in a previous analysis (5) that a self-traindaily pain assessment over an extended period could significantly reduce PV. It was therefore interesting to study whether extended baseline assessment increased or reduced the prognostic association, potentially explaining certain disparities between studies.
Methods
Patient data from three osteoarthritis studies were analyzed, totaling 469 patients. All patients recorded daily pain levels. Firstly, the change in weekly PV during baseline, relative to continuous self-training duration, was assessed to confirm self-training’s effect and determine the optimal duration needed for a reduction of PV. Secondly, the impact of self-training on the prognostic performance of PV was evaluated by measuring the partial correlation of PV with the measured response after excluding the effect of baseline measures, study differences, and regression-to-the-mean. Three endpoints were used: change from baseline of average pain score (APS), as well as the WOMAC Pain and Physical subscales. Correlations were assessed for the entire population and separately for patients with long and short self-training. For comparison, these partial correlations were recalculated considering the first week of pain measurement as a baseline to eliminate the effect of self-training.
Results
Patients who continuously assessed their pain during the baseline period saw their PV significantly decrease during the first 3 weeks of assessment (a reduction of 17.3% in the average standard deviation, p <0.01). The PV of patients who had a self-training period longer than 3 weeks did not decrease further thereafter. The partial correlation of PV with the response was significant (p < 0.01) but weak, with values between 0.13 and 0.14, depending on the endpoints considered. This correlation was higher for patients who had undergone longer self-training (0.20-0.21 for the 3 endpoints, p < 0.001), allowing a 4% reduction in response variability. Conversely, the correlation was almost zero for patients who had little or no opportunity to train in pain assessment. For comparison, if the first week of assessment had been used to measure the baseline APS and BPV, the correlation of PV would have been null and not significant for any of the endpoints and any of the analyzed population.
Conclusions
Counterintuitively, reducing PV, through longer self-training, increases its correlation with the response. This underscores the need for high-quality, noise-free measurements to best evaluate correlations. Indeed, without accustoming the patient to the assessment scale, the prognostic value of PV might be undetectable. Moreover, this analysis seems to confirm that when accounting for baseline and regression-to-the-mean, PV keeps indeed playing a significant prognostic role in patient response. The consistency of partial correlations across different endpoints appears to validate this finding. However, this prognostic role is limited since only 4% of the endpoints’ variance could be attributed to PV. It therefore seems relevant to take it into account in the analysis of RCTs but as a simple element of a multivariate strategy, accounting for other prognostic effects. However, this relies on high-quality measurements, with patients trained to accurately assess their pain.
Ethical Permissions
The data used in this analysis was obtain through three studies which were approved by the respective regional ethics committees and written informed consent was obtained from all participants.
Relevance to Patient Care
Understanding prognostic factors for patient response in RCTs is crucial. Accurately determining patient prognosis improves the precision of clinical trials, increasing the likelihood of approval for beneficial treatments while minimizing the number of subjects exposed to potentially ineffective substances. If PV as an adjustment covariate can enhance this precision, it is therefore important to grasp its actual benefit and the prerequisites for its proper utilization. Our analysis thus confirms the relevance of considering PV in adjusted analyses and sets conditions for it to be truly informative.
References
- Harris RE, Williams DA, McLean SA, Sen A, Hufford M, Gendreau RM, et al. Characterization and consequences of pain variability in individuals with fibromyalgia. Arthritis Rheum. (2005) 52:3670–4. doi: 10.1002/art.21407
- Farrar JT, Troxel AB, Haynes K, Gilron I, Kerns RD, Katz NP, et al. Effect of variability in the 7-day baseline pain diary on the assay sensitivity of neuropathic pain randomized clinical trials: an ACTTION study. Pain. (2014) 155:1622–31. doi: 10.1016/j.pain.2014.05.009
- Treister R, Honigman L, Lawal OD, Lanier RK, Katz NP. A deeper look at pain variability and its relationship with the placebo response: results from a randomized, double-blind, placebo-controlled clinical trial of naproxen in osteoarthritis of the knee. Pain. (2019) 160:1522–8. doi: 10.1097/j.pain.0000000000001538
- Tiwari, Siddharth R et al. On the Relationship Between Pain Variability and Relief in Randomized Clinical Trials. Frontiers in pain research (Lausanne, Switzerland) vol. 3 844309. 8 Apr. 2022, doi:10.3389/fpain.2022.844309
- Ooghe A, Branders S, Pereira A. Can daily self-assessment induce a learning effect mitigating pain evaluation error in clinical trials?. Osteoarthritis and Cartilage. (2021). 29. S262-S263. 10.1016/j.joca.2021.02.346.
