Endpoints & Outcomes Terminology

A trial’s primary outcome is the single pre-specified result that drives the sample-size and power calculation — it is what the trial is built to answer. Secondary outcomes are additional and largely hypothesis-generating. A surrogate endpoint is a substitute marker (e.g. a lab value) for a patient-important event; a composite bundles several events into one count. Safety outcomes track harm. Good outcomes are pre-specified, patient-important and measured objectively and blind.

A hierarchy of weight. The primary outcome sits alone at the top because the whole trial — its size, its power, its claim — is anchored to it. Secondary, surrogate and composite outcomes sit beneath as supporting evidence, and safety outcomes run alongside to catch harm.

Find the protocol/registry primary outcome and read that result first — it is the verdict. Treat significant secondaries as leads for future trials, not conclusions. For a surrogate, ask whether it reliably tracks the outcome patients care about; for a composite, check that its components are of similar importance and frequency.

Outcome choice is where spin enters. Testing many secondaries inflates the chance of a false positive (multiplicity); leaning on a surrogate can flatter a drug that never improves real outcomes; and a composite can be driven entirely by its softest component. Knowing the vocabulary lets you read the result the trial actually earned.

• Primary = pre-specified & powers the trial (read it first)
• Secondary = hypothesis-generating, not confirmatory
• Outcomes should be pre-specified, patient-important, objective & blind

COMPare project (CEBM, Oxford, 2016) — Goldacre’s team checked all 67 trials published over six weeks in the top five medical journals against their registered protocols. Only 9 reported their outcomes perfectly; across the rest, 301 pre-specified outcomes went unreported while 357 new ones were silently added — on average each trial reported only ~58–62% of its planned outcomes and slipped in ~5 extras. Outcome switching is not rare or hypothetical — it was the norm in elite journals. The defence is to read the trial against its registry/protocol, not against its abstract.

• Outcome switching — accepting a promoted secondary as if it were the pre-specified primary.
• Multiplicity — cherry-picking one “significant” result from many secondaries without correction.
• Surrogate reliance — treating a marker change as proof of patient benefit.