Definition
The primary endpoint is the variable carrying the answer to the research question. It designates three things at once: what is measured, in whom, and at what point. "The complication" is not an endpoint; "the occurrence of a Clavien-Dindo grade ≥ II complication within thirty days of the procedure" is.
It is singular. A study declaring three has none: it reserves the right to conclude on whichever turns out favourable, and a reviewer reads it that way. Other variables of interest do exist — those are secondary endpoints, declared as such, with the statistical consequences that follow.
Why before collection, and not after
An endpoint chosen after seeing the data is no longer an endpoint: it is a result. The difference is not moral but arithmetic. On any dataset, testing enough variables eventually produces one that crosses the significance threshold by chance alone; picking that variable afterwards and presenting it as the starting hypothesis turns noise into a conclusion.
This is why ICH E6(R3) makes the protocol a document that precedes collection, and why reporting guidelines — STROBE for observational work, the EQUATOR network more broadly — ask that what was planned be told apart explicitly from what was found along the way.
How to phrase one
A usable phrasing answers four questions, leaving no room for interpretation at analysis time:
| Element | Question | Example |
|---|---|---|
| Population | In whom? | Patients operated on for BPH in the department |
| Measure | What, exactly? | Highest Clavien-Dindo grade |
| Threshold | From what point is it an event? | Grade ≥ II |
| Window | Measured when? | Within thirty postoperative days |
The measure must exist in the variable dictionary with its type, scale and possible values. An endpoint pointing at an untyped variable gets settled at analysis time — that is, too late: this is where "grade ≥ II" quietly becomes "grade > II".
Changing it mid-study
It happens, and it is not always a fault: an endpoint may prove unmeasurable, or the attainable sample may collapse. What separates a defensible amendment from an arranged result comes down to three points.
- The date. Is the change earlier than any reading of the data? That is the reviewer's first question.
- The reason. It is written down, and it stands without invoking the results.
- The trace. The old endpoint, the new one, the date and the author stay legible. A protocol rewritten without keeping its previous version proves nothing at all.
In practice this is what locking is for: once the protocol is locked, the endpoint cannot change without the change leaving a timestamped entry in the audit trail.
In AnzarSeha
The primary endpoint is declared on the case report form by designating an existing dictionary variable — not by typing free text. The chosen variable then carries a visible mark, and the protocol cannot be locked while no endpoint is declared.
Applicable frameworks
ICH E6(R3) for the protocol's place in good clinical practice; STROBE and the EQUATOR network for endpoint reporting in publications; CDISC CDASH for the structure of collected variables.