Expert prior elicitation and Bayesian analysis of the association between chiropractic care and acute lumbar disc herniation (LDH) with early surgical intervention

Chiropractic care, involving spinal manipulation therapy, is popular for low back pain, but little is known about its potential link to acute LDH with radiculopathy conceptualised as an adverse event. Several systematic reviews suggest that spinal manipulation is a viable therapeutic option for low back pain, but the summarized studies are of varying quality and too small to examine rare serious adverse events. Moreover, randomized clinical trials have shown benefit of spinal manipulation for the management of LDH with radiculopathy, yet little is known about serious adverse events possibly related to this treatment. This void of knowledge represents a problem because an effective treatment is one that cost-effectively improves health outcomes while first doing no harm.

A Bayesian approach to health services research provides an opportunity to advance knowledge despite a scarcity of information and evidence. Bayesian methods start with existing “prior” beliefs, formally quantified as probability distributions, and update these using new data (the likelihood) to arrive at “posterior” updated beliefs or knowledge. Quantifying currently held beliefs can determine the magnitude of a potential risk expected by experts and describe the presence of uncertainty about an exposure-outcome association. Experts in a field can have knowledge of the risk of using a treatment through years of clinical experience. When quantified, the knowledge gained from their clinical experience can be included in models estimating risk, which may help to bridge the gap between beliefs and evidence. Although infrequently used in epidemiologic research, a Bayesian approach may have important utility in the study of treatment-related rare serious adverse events, particularly in the absence of definitive scientific evidence.

Our objective is to undertake a Bayesian analysis of the association between chiropractic care and acute LDH with early surgical intervention. The findings of two published studies will be used to estimate the posterior probability of the association of interest using Bayes’ theorem. Expert prior elicitation from a clinician belief elicitation study will serve as informative Bayesian priors, and likelihood estimates will be sourced from a population-based epidemiologic study from Ontario, Canada.