Who remained in the comparison?
In the feasibility trial of a tongue spray containing cetylpyridinium chloride, 14 of the 17 randomised patients entered the per-protocol analysis. Two excluded participants developed reconstructive-flap necrosis. The third, in the control arm, was excluded for poor postoperative oral hygiene. That last exclusion touches the outcome being measured: removing a participant with poor oral care changes the meaning of a comparison investigating oral hygiene.[1]
Improvement from baseline in the spray arm does not establish that the treatment outperformed control care. The difference between the groups’ changes was not statistically significant. Structured oral care given to both arms, the perioperative care setting and variability in a small sample may also have contributed to the observation. The authors describe this limitation. Random allocation is a strong starting point for taking the result seriously; who remains in the comparison when outcomes are measured matters as much.[1]
Each measurement has its own denominator
The denominator changed across outcomes. One control participant had no teeth, making the oral-hygiene index unavailable; that exclusion was not prespecified. Some tracheal samples could not be collected or fell below the detection limit. Without a prospective rule for imputing missing values, the authors did not generate numerical efficacy estimates covering everyone randomised. Explicitly identifying who contributed to each measurement is a real strength of this small trial. Explaining missing outcomes does not supply those outcomes.[1]
Another boundary concerns the primary endpoint. The original protocol foregrounded changes in oropharyngeal bacterial composition, while the manuscript presents feasibility measures as primary. The authors disclose that this is a later reporting reclassification. A new label does not create a prospectively selected efficacy test. The study’s usable contribution is experience in running a larger trial. A subsequent design assessing clinical benefit needs participant follow-up, missing-sample rules and the primary clinical endpoint specified together in advance.[1]