01Find the rule that turns several outcomes into one

FDA’s multiple-endpoints guidance describes composite event outcomes that count the occurrence of any specified component, often using time to the first event. It distinguishes these from other multi-component measures that combine scores or require several criteria.FDA — Multiple Endpoints in Clinical Trials, October 2022 (opens in a new tab)

The word combined is therefore insufficient. Find out whether the endpoint means any event, all required criteria, a weighted score or another defined rule. Different rules answer different questions even when they use the same components.

Also identify the time window. A proportion experiencing any event by a fixed date and the time until a first event use related information but are not the same analysis.

In a peptide paper, do not assume that a broad outcome name such as recovery reveals this construction. The methods or protocol should explain how the outcome was calculated.

02Keep overlapping components from becoming extra samples

There are five samples with at least one event, not seven. Adding the A and B counts double-counts the two samples that experienced both. The overlap must be removed when calculating this any-event total.

The example is simple counting, not a clinical trial result or a time-to-event analysis. It shows why a component table cannot always be summed to reproduce the number of people or samples meeting the composite endpoint.

If a study instead tabulates only the first event for each participant, later events may be absent from that particular decomposition. Read the table definition before treating it as a complete record of every component event.

03Ask what drives the overall result

FDA cautions that a composite can be difficult to interpret when components differ in importance and the overall result is chiefly driven by a less important event. Component effects may also differ in direction, so the combined result needs examination alongside them.FDA — Multiple Endpoints in Clinical Trials, October 2022 (opens in a new tab)

For an original hypothetical comparison, suppose the combined endpoint decreases while Event A barely changes and Event B decreases markedly. The supported headline concerns the combined endpoint; it cannot simply be rewritten as a demonstrated reduction in Event A.

Look at the component estimates and their uncertainty, not only their separate significance labels. A component analysis may contain too little information for a precise conclusion even when the overall analysis is informative.

Do not infer equal importance from equal contribution to a count. The scientific or clinical meaning of the components must be considered separately from how frequently they occurred.

04Summarise the composite without overclaiming its parts

Name the composite and briefly define it when first reporting the result. Then describe which components contributed most and whether their patterns support or complicate the overall interpretation.

Keep exploratory component analyses distinct from prespecified claims about those components. Testing several component hypotheses introduces questions about multiplicity that are separate from simply describing the composition of the overall endpoint.

If component information is unavailable, state that limitation. A combined result alone cannot reveal whether all components moved together, whether one dominated or whether opposing patterns were concealed.

For a reader comparing papers, endpoint definitions must match before their results can be treated as the same outcome. Two studies using the same headline label may include different events or apply different combination rules.