1  The question

1.1 What the client asked

The PI, an anesthesiologist, and their fellow ran a randomized trial of a licorice gargle against a sugar-water gargle, given shortly before induction of anesthesia in thoracic surgery with a double-lumen tube. At the intake meeting on the tenth of October (notes/2026-10-16-intake-meeting.md; see notes/2026-10-16-intake-date-correction.md for the date) the PI asked:

“did it reduce sore throat afterwards?”

and also:

“does it last past the recovery room?”

The intake notes record that they care a lot about the second question and expect reviewers to ask it. The results are intended for a manuscript.

On a follow-up call (notes/2026-10-16-pi-call.md) the PI confirmed the primary outcome and chose “any” sore throat over a higher threshold:

“any is what patients notice”

1.2 How we restated it

• Primary question. In patients having thoracic surgery with a double-lumen tube, does a licorice gargle five minutes before induction, compared with a sugar-water gargle, lower the odds of any sore throat at rest thirty minutes after arrival in the recovery room (PACU)? “Any” means a score above zero on the zero-to-ten sore-throat-at-rest scale (pacu30min_throatPain). The PI confirmed this as the primary outcome, with no higher threshold, on the follow-up call (D-001).
• Secondary question: does it last? The population-averaged odds ratio and risk difference for any sore throat at rest at each later assessment: ninety minutes after PACU arrival (pacu90min_throatPain), four hours after surgery (postOp4hour_throatPain) and the first postoperative morning (pod1am_throatPain). Durability is judged from these per-time estimates and their confidence intervals. They are descriptive, with no formal test, and a non-significant later assessment is not evidence that the effect wore off (D-002, D-003, D-013). A common odds ratio across all four assessments is reported as a summary, but it assumes one common effect and does not answer this question (D-014).
• Not analysed. Cough, at extubation and at each assessment. The PI said on the follow-up call that they are not interested in it (D-004).

1.3 Open questions

Questions we could not answer from the data alone. Each is resolved here or logged in the decision record.

• Swallowing pain. Should sore throat on swallowing (pacu30min_swallowPain, recorded at thirty minutes only) be a secondary outcome? The PI’s answer on the follow-up call: “maybe, let me think”.
• Stratified randomization. Was randomization stratified? The PI does not remember and will check the protocol. If it was, whether the strata belong in the adjusted model is a decision to make before any outcome is examined.
• Preoperative sore throat in the protocol. Did the trial protocol exclude patients with a sore throat before surgery? This plan keeps them (D-005); the PI is checking.
• Enrolled versus analysed. The enrollment count was uncertain at intake; the codebook gives one. The delivered file (data/raw/licorice_gargle.csv) has fewer rows than the codebook’s enrollment count; the counts are in the research log entry for the delivery. Ask the client which patients are missing and why: withdrawal, exclusion after randomization, or a difference between the public and the trial dataset. If a randomized patient is absent, the analysis is not fully intention to treat (D-005).
• Patients without postoperative data. Checked at data receipt. A small number of patients in the file have no recorded value for any postoperative variable; the count is in the research log. This is under the revisit trigger (D-015). Ask the client why these patients have no postoperative data.