surrogate endpoint

A trial would usually rather measure the thing that matters: whether people live longer, feel better or function better. Those outcomes take years and large numbers of participants to observe. A surrogate endpoint is what gets measured instead. The FDA and the National Institutes of Health, in the joint BEST resource that fixes this vocabulary for both agencies, define it as an endpoint used in clinical trials as a substitute for "a direct measure of how a patient feels, functions, or survives." Blood pressure standing in for stroke, or fat in the liver standing in for liver disease, are surrogate endpoints in that sense.

The same resource grades them, and the grade is the useful part. A validated surrogate endpoint has both a clear mechanistic rationale and clinical data giving strong evidence that an effect on it predicts a specific clinical benefit. A reasonably likely surrogate endpoint has the mechanistic or epidemiological rationale, and is expected to correlate with clinical benefit, but lacks the clinical data that would make it validated. A candidate surrogate endpoint is still under evaluation for whether it predicts clinical benefit at all. The grades say how much of the link between the marker and the outcome has been demonstrated rather than assumed.

That is why the word is worth keeping in view when reading a result. A surrogate stands in for an outcome because the two usually travel together, and usually is not always. A treatment can move a marker convincingly and leave the outcome where it was, and until the clinical endpoint itself is measured nobody can tell which has happened. A study reporting only surrogate endpoints is therefore making a smaller claim than its headline tends to suggest, however large the effect on the marker.