An effective dose has meaning only after the investigator defines an effect. A dose estimated to improve breathing during one sedation procedure does not automatically establish the dose for full awakening, preserved memory, or every other patient population.
The response the investigators actually measured
Kim and colleagues studied flumazenil during gastric endoscopic submucosal dissection under monitored anesthesia care with remimazolam and fentanyl. Of 132 patients who consented before the procedure, 60 developed the study-defined respiratory distress and entered the dose-finding analysis. This was a prospective study at one center, not a comparison against a placebo group.
The entry definition included sustained low oxygen saturation or disruptive irregular breathing and chest or abdominal movement. Improvement required recovery of oxygenation and respiratory pattern within 30 seconds to the point that the procedure could resume. Physician judgment contributed to the movement assessment. The result therefore depends on a specific clinical endpoint, rather than a direct measurement of receptor occupancy or a complete physiological description of ventilation.
ED90 is a modeled probability, not a guarantee
The investigators used a biased-coin up-and-down design: each patient’s response influenced the dose assigned to the next patient. Centered isotonic regression produced an estimated ED90 of 76.72 micrograms, with a 95% confidence interval of 68.07 to 102.62 micrograms. These are reported research estimates, not dosing instructions.
Here, ED90 identifies the estimated dose associated with a 90% probability of the defined response under the study conditions. It does not mean that every patient received that dose, that 90% of every benzodiazepine effect disappeared, or that success is certain for a particular patient. The confidence interval expresses uncertainty in the estimated dose; it is not an individual patient’s dose range.
Breathing, consciousness and recall remain separate outcomes
The researchers sought respiratory improvement without full awakening. Four of the 60 patients reported procedural recall. Two of 13 patients at the dose levels adjacent to the estimated ED90 reported recall. Those small denominators matter: the findings do not establish a precise general risk of awareness or prove that amnesia was preserved in every responder.
The authors proposed that limited reversal could move sedation to a level compatible with improved breathing while consciousness remained suppressed. That is a mechanistic hypothesis. Their clinical observations do not directly demonstrate selective receptor antagonism in separate respiratory and consciousness pathways.
The limits travel with the result
All patients received a particular remimazolam-fentanyl regimen during a particular procedure. Fentanyl could also contribute to respiratory effects. The study’s lack of a placebo comparator limits attribution of every change solely to flumazenil. Its single-center population, partly judgment-based endpoint, and imprecise dose estimate also limit generalization.
This is a pharmacodynamic lesson about the relationship between dose and a defined response. A defensible interpretation preserves the endpoint, the observation window, the co-administered drugs and the uncertainty. Changing any of those changes the question the dose estimate can answer.
Source
Kim HI and colleagues. Determining ED90 of Flumazenil for Selective Respiratory Distress Improvement Using Remimazolam During Endoscopic Submucosal Dissection of Gastric Neoplasms: A Prospective Study. Cancers. 2025;17:321. Read the primary study. Original analysis of the supplied paper; no tables or figures reproduced.
Leave a Reply