Endpoints and Surrogate Outcomes

#What does a study measure?

An endpoint is a defined outcome used to assess what happened in a study. It might be a symptom score, a hospital admission, a laboratory measurement or death. Its definition should include how and when it is measured, because different definitions can lead to different interpretations.

Studies usually identify a primary endpoint that addresses the main research question. Secondary endpoints explore additional effects. Specifying these before results are known helps limit selective reporting. When researchers test many outcomes, some apparently positive findings can arise by chance and need particularly cautious interpretation.

#Direct outcomes and substitutes

Patient-important outcomes concern how people feel, function or survive. Examples include pain, ability to carry out daily activities and quality of life. These can be measured through participant reports or other methods. Being personally meaningful does not remove the need for reliable measurement and fair comparisons.

A surrogate endpoint is used as a substitute for an outcome of direct interest. A laboratory marker or imaging change may provide an earlier signal, making research more practical. However, an intervention can improve that marker without producing the expected health benefit, or can cause harms through another pathway.

#Check the link to benefit

Confidence in a surrogate depends on evidence that intervention-related changes in it reliably predict changes in the clinical outcome for the relevant setting. A marker associated with prognosis is not automatically a valid surrogate. Support may differ across diseases, populations and types of treatment.

#Look inside combined outcomes

Some studies use a composite endpoint that combines several events. This can make analysis more efficient, but a frequent, less serious event may drive the overall result. Readers should examine the individual components rather than assuming that every component improved to a similar extent.

Also check the timing and size of any change, not just whether a result was statistically significant. A small score difference may have little practical importance. Benefits, harms and the burden of treatment together provide a fuller picture than any single measurement.

#Common misunderstandings

A result can be statistically significant without being large enough to matter in everyday life. Statistical significance does not establish that people feel better, function better, or live longer. Equally, a study that finds no clear difference has not necessarily proved that two treatments are equivalent: it may have been too small or too short to detect an important difference.

An endpoint described as “objective” is not automatically more useful than one reported by patients. Laboratory tests can be measured precisely but still leave uncertainty about benefit, while well-designed symptom questionnaires can capture experiences that matter directly. Nor does a favourable result for one endpoint establish that a treatment is beneficial overall. Side effects, treatment burden, and other outcomes also count. Findings are generally more convincing when the main endpoint and analysis were specified before researchers examined the results, rather than selected afterwards because they appeared promising.

#Questions worth asking a clinician

• Which outcomes was this study designed to measure, and what can those endpoints establish about how people feel, function, or survive?
• If the study showed improvement in a surrogate measurement, did it also show benefits that patients would notice or value?
• What evidence shows that treatment-induced changes in this surrogate predict clinical benefit, rather than simply being associated with prognosis?
• For this composite endpoint, how much did each component contribute to the overall result, and were the most important components improved?
• Could this treatment improve the surrogate measurement while leaving symptoms, quality of life, or survival unchanged or worse?