Adherence Rules Complicate the CBD Jaw-Pain Trial’s Analysis Label
The jaw-pain paper described an intention-to-treat analysis while also stating that follow-up visits were excluded when reported adherence fell below 80%. Both statements matter when determining which observations contributed to its results.
The report describes assignment and visit availability
The authors said primary analyses followed original treatment assignment and included all randomized subjects. Elsewhere, the methods state that participants attested to daily intake at monthly follow-up and that the corresponding visit was excluded if reported compliance for that interval was below 80%.
The paper also reports participants without follow-up and removals for screening errors. These details caution against using the analysis label alone as a complete description of the analyzed data.
The rule does not identify the direction of bias
Adherence was self-reported rather than established by a measured CBD concentration. From the rule alone, one cannot infer how many visits were removed in each group or whether excluding them favored CBD, the comparator or neither.
Those questions require the participant flow, visit-level denominators and handling of missing observations. A sensitivity analysis retaining or otherwise accounting for excluded observations would address a different question from simply repeating the “intention-to-treat” description.
Retain the reported clinical result
The trial did not demonstrate a significant between-group pain advantage for its 20 mg daily CBD regimen. The adherence qualification neither creates a hidden benefit nor proves the null result was caused by nonadherence. It clarifies what additional information is needed to assess the robustness of this small human trial.
Related analysis: the CBD jaw-pain trial’s pain and function results.
Research cited
Reeve and colleagues (2026), Oral Cannabidiol in the Treatment of Myofascial Pain Disorder of the Temporomandibular Region: A Placebo-Controlled Randomized Clinical Trial. DOI: 10.1159/000551040.
