Search

Recently, the degree to which scientific publications provides a reliable source of information has come under intense scrutiny. Reports suggest that a substantial amount of published literature is likely to be biased, distorted, and non-reproducible. This cuts across basic, pre-clinical, and clinical research. It has been estimated that non-reproducible preclinical research consumes $28 billion/year (USD) and that 85% of biomedical research resources are wasted on biased research. The response from the scientific and policy community has been to identify common practices that contribute to the problem, and develop methods to counteract them. This matter is very relevant to the pain field. Understanding the causes of non-reproducible and non-replicable research and its ultimate impact on how we prevent and treat pain should assist pain researchers to improve the reproducibility and replicability of their work. The distinction between reproducibility and replicability is presented in Figure 1 and defined elsewhere. This paper aims to: (1) define drivers of non-reproducible and non-replicable research with examples from pain sciences and broader research fields; and (2) provide an overview of potential solutions and practices that could improve reproducibility and replicability of pain research.

Evidence from randomized controlled studies shows that reconceptualizing pain improves patients' knowledge of pain biology, reduces catastrophizing thoughts, and improves pain and function. However, causal relationships between these variables remain untested. It is hypothesized that reductions in catastrophizing could mediate the relationship between improvements in pain knowledge and improvements in pain and function. To test this causal mechanism, we conducted longitudinal mediation analyses on a cohort of 799 patients who were exposed to a pain education intervention. Patients provided responses to the neurophysiology of pain questionnaire, catastrophic thoughts about pain scale, visual analogue pain scale, and the patient specific functional scale, at baseline, 1-month, 6-month, and 12-month follow-up. With adjustment for potential confounding variables, an improvement in pain biology knowledge was significantly associated with a reduction in pain intensity (total effect =-2.20, 95% confidence interval [CI] =-2.96 to-1.44). However, this effect was not mediated by a reduction in catastrophizing (indirect effect =-0.16, 95% CI =-0.36 to 0.02). This might be due to a weak, nonsignificant relationship between changes in catastrophizing and pain intensity (path b = 0.19, 95% CI =-0.03 to 0.41). Similar trends were found in models with function as the outcome. Our findings indicate that change in catastrophizing did not mediate the effect of pain knowledge acquisition on change in pain or function. The strength of this conclusion is moderated, however, if patient-clinician relational factors are conceptualized as a consequence of catastrophizing, rather than a cause. Refereed/Peer-reviewed