5. Trial Endpoints: Primary, Composite, Surrogate, and Safety
After this lesson you can
- Turn an outcome into an endpoint measured in everyone at a fixed time by a fixed rule
- Classify endpoints and state the risk ratio, mean difference or hazard ratio for each
- Explain censoring in a time-to-event endpoint
- Judge when a composite or surrogate endpoint is justified
- Classify adverse and serious adverse events and build an endpoint hierarchy
Video coming soon
Lessons publish daily from November 17, 2026.
This lesson defines an endpoint and classifies endpoints as binary, continuous or time-to-event, each with its summary measure. It explains the composite and surrogate traps, patient-reported outcomes and safety endpoints, and builds the fictional Oakhaven endpoint hierarchy and visit schedule.
Questions and answers
Questions a learner would ask about this lesson. Think through your own answer before you open each one.
- What makes something an endpoint rather than just an outcome?
Show answer
An endpoint is a measured quantity, taken in every participant, at a fixed time, by a fixed rule, that answers the trial's question. "Feeling better" is an outcome, the general idea. The symptom score, measured with the same instrument at a fixed visit, is an endpoint.
- What are the three statistical types of endpoint and how is each summarized?
Show answer
Binary endpoints are yes or no at a fixed time and are summarized by proportions, compared with a risk ratio or risk difference. Continuous endpoints are numbers on a scale, summarized by means and compared with a mean difference. Time-to-event endpoints record whether and when an event happens, summarized by an event curve and a hazard ratio.
- Why are composite endpoints tempting, and what is the trap?
Show answer
Combining several events gives more events in the same follow-up, and because time-to-event sample size depends on events, the trial can be smaller. The trap is that the components may not move together: a serious component may be unchanged, or components may move in opposite directions. A composite result is a result about the mixture, so every component must be reported separately.
- What does a surrogate endpoint assume?
Show answer
It assumes a chain: the program changes the marker, and the change in the marker changes the outcome that matters, so effects on the marker predict effects on the outcome. The chain can break at either link. In Oakhaven, poor control at twelve months is a surrogate for long-term harm years beyond the trial, and the protocol must say so.
- What is the Oakhaven endpoint hierarchy?
Show answer
The primary endpoint is poor control at twelve months, binary, a control value of 60 or above measured by the blinded outcome assessor. The three secondaries are the symptom score at twelve months, time to first flare over twelve months, and adverse events. Visits are at baseline and months three, six, nine and twelve, each with a window of plus or minus two weeks.
Practice quiz
- Question 1"Poor control at twelve months" is which type of endpoint?
- Question 2The usual summary measure for a time-to-event endpoint is:
- Question 3If 20 percent of the program arm and 40 percent of the usual-care arm are in poor control, what are the risk ratio and risk difference?
- Question 4The symptom score has a standard deviation of 20 and the hoped difference is 8 points. What is the standardized difference?
- Question 5A participant leaves the trial at month five without having a flare. In a time-to-event analysis, that participant is:
- Question 6A composite endpoint is attractive mainly because it:
- Question 7What does cutting a continuous measure into a binary one cost?
- Question 8A surrogate endpoint assumes:
- Question 9Which of these is a serious adverse event as the lesson defines it?
- Question 10Why did the Oakhaven unit decline a composite of flare or hospital visit?
Saved in your browser only: no account, no server.
Design of Clinical Studies is education only. Every lesson follows the fictional Oakhaven Adherence Trial with invented numbers; the course uses no real data and cites no real studies.