Surrogate Endpoints vs Clinical Outcomes in Ageing Trials

Key Takeaways

A clinical outcome directly describes how participants feel, function, or survive; a surrogate endpoint substitutes an indirect measure for that outcome. [1] [8]
Surrogates can make trials shorter and smaller, but a change in a biomarker does not by itself establish a change in healthspan, disability, disease, or survival. [2] [5]
Validation asks whether intervention-induced effects on the surrogate reliably predict intervention effects on a specified clinical outcome, usually across multiple randomized trials. [3] [4]
No biomarker is a universal surrogate for ageing merely because it predicts age-related risk or changes after an intervention. [5] [7]

Who This Is Useful For

This page is useful for readers assessing trials that report changes in epigenetic clocks, inflammatory proteins, metabolic measures, imaging results, physical-performance tests, or other ageing-related endpoints. It focuses on what those results can establish and when they do not yet demonstrate a clinical benefit.

Every trial needs outcomes that match the question it is intended to answer. In an ageing trial, the outcome might be survival, onset of major disease, disability, independence, cognition, or recovery after a stressor. These outcomes may take years to occur. Researchers therefore often also measure earlier biological or physiological changes, but an earlier measure becomes a surrogate only when it is used to stand in for a specified later clinical outcome. [1] [5] [6]

The Core Distinction

Endpoint Type	What It Measures	Ageing-Trial Example	What a Favorable Result Supports
Clinical outcome	A direct aspect of how a participant feels, functions, or survives	Mortality, incident dementia, persistent disability, or ability to live independently	An effect on the measured health outcome
Biomarker or response measure	A biological, imaging, or physiological characteristic	DNA-methylation measure, inflammatory protein, or molecular marker of target engagement	An effect on that measure, without automatically establishing clinical benefit
Surrogate endpoint	An indirect endpoint used to substitute for a specified clinical outcome	A validated intermediate measure used in place of later disease, disability, or survival	An inferred clinical effect only within the context for which the surrogate is valid

These labels describe roles, not fixed properties of a measurement. The same measurement might be a mechanistic outcome in one trial, a risk predictor in another, and a proposed surrogate in a third. Its interpretation depends on the target outcome, population, intervention, and intended use. [1] [7] [8]

Why Ageing Trials Use Indirect Endpoints

Clinically important ageing outcomes can be slow, uncommon, and heterogeneous. A mortality trial may need a large sample and long follow-up, while trials of disability or multimorbidity must define and adjudicate several kinds of event. Earlier biological or functional endpoints can reduce duration and sample-size requirements and can show whether an intervention engaged its intended target. [5] [6] [8]

That efficiency answers a narrower question. A short trial can provide evidence that an intervention changed a molecular pathway or a candidate ageing marker, but it may be too small or brief to determine effects on major clinical outcomes or uncommon harms. Trials using surrogates can therefore produce earlier evidence while leaving greater uncertainty about net clinical benefit. [2] [8] [9]

Prediction Is Not Surrogacy

A marker may correlate with chronological age, predict mortality, or distinguish healthier from less healthy populations and still fail as a surrogate. Those observations concern associations between the marker and outcomes. Surrogacy requires evidence about treatment effects: when an intervention changes the marker, does the magnitude and direction of that change reliably predict what the same intervention does to the clinical outcome? [3] [4] [5]

The distinction matters because an intervention can influence a marker through a pathway that does not cause the clinical outcome, or it can have additional beneficial and harmful effects that the marker does not capture. A strong marker-outcome association among individuals is therefore not enough to show that treatment effects on the marker reproduce treatment effects on health. [2] [4]

What Validation Requires

Early statistical criteria asked whether a surrogate captured the intervention's effect on the true endpoint. Later approaches separated individual-level association from trial-level association and evaluated whether differences in treatment effects on the surrogate predict differences in treatment effects on the clinical endpoint across randomized trials. This generally requires substantially more evidence than showing that one biomarker predicts outcomes in one cohort. [3] [4] [10]

Validation Question	Why It Matters	Insufficient Evidence
What clinical outcome is being replaced?	Surrogacy must refer to a defined target outcome	The paper refers only to “slower ageing” or “better healthspan”
Does treatment change the proposed surrogate?	A response is necessary before the marker can transmit useful information about treatment effect	The marker predicts risk only in observational data
Do treatment effects track across trials?	Trial-level evidence tests whether effects on the surrogate predict effects on the target outcome	One small trial reports only the surrogate
Is the context of use similar?	Validity may not transfer across interventions, populations, doses, or outcomes	Evidence from one drug class or disease is generalized to ageing overall
Are harms and off-pathway effects measured?	A surrogate may capture only one route by which an intervention affects health	Short follow-up and a small sample provide little clinical safety information

Surrogate validity is consequently conditional rather than universal. Evidence may support a measure for one target outcome and treatment class without supporting it for another. Transparent trial reports should name the surrogate, the target outcome, the supporting validation evidence, and the limits of that evidence. [4] [8]

Biological Age Is Not Automatically a Clinical Outcome

Biological-age measures, including epigenetic clocks, are attractive because they compress complex information into a single score and some predict age-related outcomes. However, predictive performance does not establish that an intervention-induced change in a clock predicts an intervention-induced change in disease, disability, function, or survival. Different clocks can also represent different biological signals and may respond differently to an intervention. [5] [7]

A trial showing that a biological-age score moved in a favorable direction therefore demonstrates a result on that score. Interpreting it as evidence of longer life or better health requires separate validation against the particular clinical outcome being claimed. [5] [7]

Functional Measures Need Context Too

Not every non-mortality endpoint is a biochemical surrogate. Measures such as gait speed, muscle strength, cognition, or ability to perform daily activities can directly reflect function that matters to older adults. Whether a measure is treated as a clinical outcome, an intermediate outcome, or a surrogate depends on the trial question and on what broader benefit investigators claim it represents. [1] [5] [6]

How to Read an Ageing Trial

Name the primary endpoint: determine whether it is a clinical, functional, mechanistic, or surrogate endpoint. [1] [8]
Identify the target outcome: ask exactly which later health outcome the surrogate is expected to predict. [3] [8]
Check the validation claim: look for evidence based on treatment effects across randomized trials, not only correlation with age or disease risk. [4] [10]
Compare the context: check whether validation involved similar participants, interventions, follow-up, and clinical outcomes. [4] [7]
Inspect clinical secondary outcomes and harms: they may be imprecise in a short trial, but they show what was and was not observed directly. [8]
Keep the conclusion aligned with the endpoint: a biomarker effect supports a biomarker conclusion unless stronger validation justifies inference about health. [2] [5]

What This Does Not Mean

It does not mean surrogate endpoints are inherently unreliable; well-validated surrogates can make informative trials more feasible. [4] [8]
It does not mean biomarkers are useful only after surrogacy validation; they can measure exposure, target engagement, biological response, prognosis, or risk. [1] [7]
It does not mean mortality is the only meaningful clinical outcome; disease, function, disability, independence, symptoms, and quality of life can also be directly relevant. [5] [6]
It does not mean every ageing trial must be long enough to prove clinical benefit; early trials can answer narrower mechanistic and feasibility questions. [6]

Practical Interpretation Examples

If an epigenetic clock decreases after treatment: the trial shows an effect on that clock; it does not by itself show that disability or mortality will decrease. [5] [7]
If an intervention lowers an inflammatory marker: this may support target engagement, but clinical benefit depends on whether treatment effects on that marker validly predict the claimed health outcome. [1] [2]
If gait speed improves: that may be a directly meaningful functional result, but it should not automatically be translated into a claim about longer survival or delayed multimorbidity. [5] [6]
If a trial calls a marker “validated”: check what it was validated to do; reliable measurement, risk prediction, and surrogate validation are different standards. [1] [7]

Summary

Surrogate endpoints can make ageing trials faster and more practical, but their interpretation depends on a demanding link between intervention effects on the surrogate and intervention effects on a defined clinical outcome. Association with age, prediction of future risk, and response to treatment are each useful properties, but none alone establishes surrogacy. Until that link is demonstrated for a specific context, the most accurate conclusion is the one stated at the level of the endpoint actually measured. [2] [4] [5]

References

Biomarkers Definitions Working Group. (2001). Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clinical Pharmacology & Therapeutics. https://pubmed.ncbi.nlm.nih.gov/11240971/
Fleming, T. R., & DeMets, D. L. (1996). Surrogate end points in clinical trials: are we being misled? Annals of Internal Medicine. https://pubmed.ncbi.nlm.nih.gov/8815760/
Prentice, R. L. (1989). Surrogate endpoints in clinical trials: definition and operational criteria. Statistics in Medicine. https://pubmed.ncbi.nlm.nih.gov/2727467/
Buyse, M., et al. (2000). The validation of surrogate endpoints in meta-analyses of randomized experiments. Biostatistics. https://pubmed.ncbi.nlm.nih.gov/12933525/
Cummings, S. R., & Kritchevsky, S. B. (2022). Endpoints for geroscience clinical trials: health outcomes, biomarkers, and biologic age. GeroScience. https://doi.org/10.1007/s11357-022-00671-8
Justice, J. N., et al. (2016). Frameworks for proof-of-concept clinical trials of interventions that target fundamental aging processes. The Journals of Gerontology: Series A. https://pmc.ncbi.nlm.nih.gov/articles/PMC5055651/
Moqri, M., et al. (2024). Validation of biomarkers of aging. Nature Medicine. https://pubmed.ncbi.nlm.nih.gov/38355974/
Manyara, A. M., et al. (2024). Reporting of surrogate endpoints in randomised controlled trial reports (CONSORT-Surrogate): extension checklist with explanation and elaboration. BMJ. https://pmc.ncbi.nlm.nih.gov/articles/PMC11231881/
Ciani, O., et al. (2013). Comparison of treatment effect sizes associated with surrogate and final patient relevant outcomes in randomised controlled trials: meta-epidemiological study. BMJ. https://pubmed.ncbi.nlm.nih.gov/23360719/
Buyse, M., & Molenberghs, G. (1998). Criteria for the validation of surrogate endpoints in randomized experiments. Biometrics. https://pubmed.ncbi.nlm.nih.gov/9840970/

Educational Disclaimer

This content is provided for educational purposes only and does not constitute medical advice.