Time to vasopressor discontinuation

Method

Elapsed time from randomisation until every vasopressor is stopped and stays stopped for the trial's stated washout — 48 consecutive hours in both trials that motivated this node. Reported in hours or days. Three things are recorded on the edge and none is optional: the washout, the horizon at which the analysis is truncated, and the censoring rule. The censoring rule decides what the number means; the reason is in the notes and a curator must read it.

The method is part of the endpoint identity. Studies measuring the same quantity by a different method are held as separate endpoints.

Notes

WHY THE CENSORING RULE IS NOT OPTIONAL. A patient who dies early stops needing vasopressors, so an analysis that treats death as discontinuation rewards killing the patient faster; one that censors at death answers a different question from one that assigns the worst rank. PMID 36915146 truncates at 28 days. This argument was written in `measurement` and moved here on 2026-08-27, because `measurement` is scored: the node whose notes say NOT A MORTALITY ENDPOINT held "patients", "die" and "death" in its own positive vocabulary. The instruction the extractor needs — record all three, none is optional — stays above. SIGN INVERSION against VASOPRESSOR-FREE DAYS, exactly as CAT:outcome/mechanical-ventilation-duration inverts against ventilator-free days, and for the same reason: free-days improve by RISING and this duration improves by FALLING. A review reporting vasopressor-free days must NOT be mapped here. That trap has already been documented once on the ventilation node and is repeated here rather than cross-referenced, because the reader who needs it is reading this entry. AND NOTHING MECHANICAL ENFORCES IT — measured 2026-08-26, spanNamesOutcome accepts "vasopressor-free days" for this node and "ventilator-free days" for the ventilation node, with no rival beating either, because "days" is the only token the span shares with any vocabulary and nothing matches "vasopressor-free" at all. A bare "days" passes every duration node for the same reason. That is a property of the whole duration family and predates this entry; it means the sign-inversion defence here is the refuter reading this paragraph, not the validator. Do not read the prohibition above as something the pipeline checks. Distinct from CAT:outcome/mean-arterial-pressure-in-vasodilatory-shock, which is the pressure itself rather than the time spent supporting it — an agent can raise MAP at 6 h without shortening the infusion, and the two are reported in the same table of the same papers. Distinct from CAT:outcome/icu-length-of-stay, a longer window that contains this one: vasopressors stop well before ICU discharge and discharge is governed by weaning, delirium and bed availability. Distinct from CAT:outcome/mechanical-ventilation-duration, a different device window inside the same admission. NOT a mortality endpoint, and see the censoring warning in the measurement for why the two are entangled. Minted 2026-08-26 under ADR-0046 rule 3. ADR-0014's two-paper floor is cleared and was checked rather than assumed: PMID 36915146 ("time to vasopressor discontinuation at 28 days, defined as discontinuation of all vasopressors for at least 48 consecutive hours") and PMID 39780053 ("time to vasopressor discontinuation in septic shock, hours from randomisation until all vasopressors are stopped for at least 48 consecutive hours"), both declined for a vocabulary gap in the methylene blue batch. HONEST LIMIT ON ITS REACH: both are methylene blue papers, so unlike the MAP-in-shock node this one is not yet demonstrated across subjects. PMID 23432501, "Methylene blue as a vasopressor: a meta-analysis of randomised trials", is a likely third and is also methylene blue. If it never attracts a second subject that is a fact about the vasopressor literature this corpus holds, not a reason to have withheld the node — but it should be revisited rather than assumed to have broadened. Found only because the mint queue learned to cluster. The two papers word the endpoint differently, and the queue grouped on the exact string, so it reported two singletons and ADR-0014 rule 2 refuses to mint a singleton.