Skip to content
All resources

Reading the evidence

Surrogate Endpoints Explained: When a Marker Stands In for an Outcome

AR

By

· 10 min read

A summary of published research, with citations. It states no amount, schedule or route, and makes no claim about what this compound does in a person. Supplied for laboratory research use only.

On this page · 7 sections
  1. What is the difference between a surrogate and a clinical outcome?
  2. What makes a surrogate valid?
  3. Which surrogate endpoints have failed?
  4. Does that make HbA1c a bad surrogate?
  5. How do surrogates appear in the compounds discussed on this site?
  6. What should you ask of any marker?
  7. Frequently asked questions

A surrogate endpoint is a measurement used in place of the outcome that actually matters. It might be a blood level, an image or a physiological reading, standing in for how a person feels, functions or survives. Surrogates are used because they change sooner and cost less to measure. A surrogate is only trustworthy if changing it, with this kind of intervention, reliably changes the real outcome. Several well-known drugs moved their marker in the right direction and increased deaths.

A monitor showing a biomarker chart beside a printout of a different outcome chart
Illustration: a marker that moves is not an outcome that changed.

What is the difference between a surrogate and a clinical outcome?

A clinical outcome is something a patient would notice or care about without a laboratory: death, a heart attack, a stroke, a fracture, being able to walk further, pain. A surrogate is a measurement that sits on the way to one of those, or is believed to: cholesterol for heart attacks, HbA1c for diabetic complications, tumour shrinkage for survival.

Between the two sits what regulators call an intermediate clinical endpoint. It is a real clinical measure, such as muscle strength or walking distance, but it is used because it is expected to predict something longer-term. The distinction matters because claims slide between the levels. A paper that measured a marker is often reported as if it measured an outcome, and the companion article on how to read a peptide study introduces the problem. This page is about how to tell whether a given marker deserves the trust it is given.

What makes a surrogate valid?

That it is associated with the outcome is not enough. Fleming and DeMets set out the reason in 1996, in a paper still cited as the standard account. A marker can be a perfectly good predictor of risk in untreated people and still be a poor guide to treatment, for three reasons:

  • The marker is not on the causal path. It travels with the disease without driving it, so pushing it around changes nothing.
  • The disease has several paths to the outcome, and the drug acts on only the one the marker tracks.
  • The drug acts on the outcome through other mechanisms, good or bad, that the marker does not capture. Off-target harm is the usual culprit.

A surrogate earns validation when trials of interventions, ideally several in the same class, show that the size of the effect on the marker predicts the size of the effect on the outcome. That evidence is rarer than the use of surrogates suggests. Kim and Prasad examined 55 US cancer drug indications approved on a surrogate between 2009 and 2014. For 14 of the 25 accelerated approvals, they could find no formal analysis of how well the surrogate predicted survival. The four accelerated approvals where the strongest kind of analysis had been done all showed low correlation.

Which surrogate endpoints have failed?

Three cases are taught because each marker was a well-established risk factor and each drug moved it in the intended direction.

Suppressing arrhythmia: the Cardiac Arrhythmia Suppression Trial. After a heart attack, frequent ventricular ectopic beats predict sudden death. CAST tested whether drugs that suppressed those beats reduced it, and enrolled only patients in whom the drug did suppress them. Among 1,498 patients randomised to encainide or flecainide or their placebos, 43 on active drug died of arrhythmia against 16 on placebo, over a mean of 10 months. Those two drugs were withdrawn from the trial. The marker had moved and the outcome had moved the other way.

Raising HDL: torcetrapib. Low HDL cholesterol predicts heart disease. In ILLUMINATE, 15,067 high-risk patients received torcetrapib with atorvastatin or atorvastatin alone. At 12 months HDL was up 72.1% and LDL down 24.9% from baseline on torcetrapib. Cardiovascular events were higher, with a hazard ratio of 1.25, and so was death from any cause, at 1.58. Systolic blood pressure rose by 5.4 mm Hg and aldosterone rose, pointing to an off-target effect, though the authors could not rule out harm from the mechanism itself.

Lowering HbA1c hard: ACCORD. Higher HbA1c predicts cardiovascular events in type 2 diabetes. ACCORD randomised 10,251 patients to an intensive strategy targeting HbA1c below 6.0% or a standard target of 7.0% to 7.9%. The intensive group reached a median of 6.4% against 7.5%. The primary cardiovascular composite did not differ significantly, with a hazard ratio of 0.90 and a confidence interval crossing 1. Deaths were 257 in the intensive group and 203 in the standard group, a hazard ratio of 1.22, and intensive therapy was stopped early.

MarkerOutcome it stands in forWhat the trial showed
Ventricular ectopySudden cardiac deathSuppressing it with encainide or flecainide increased arrhythmic death
HDL cholesterolHeart attack, strokeRaising it with torcetrapib increased events and deaths
HbA1c (pushed low)Cardiovascular eventsIntensive lowering did not reduce events, and deaths rose
Body weightCardiovascular and kidney eventsUntested for retatrutide; TRIUMPH-Outcomes is the test
Lean mass (DXA)Muscle function, frailtyMeasured in GLP-1 trials; function largely unmeasured
Plasma BDNFBrain function and recoveryMeasured in Semax studies; link to outcome not established
Telomere lengthAgeing and disease riskGenetically longer telomeres associated with higher risk of several cancers

Does that make HbA1c a bad surrogate?

Not in general, and the nuance is the useful part. HbA1c remains the standard primary endpoint of diabetes drug trials, including retatrutide's TRANSCEND-T2D programme. What ACCORD showed is narrower: how the marker is lowered, how far, and in whom, can matter more than the number reached. The same marker can be a reasonable guide for one question and misleading for another.

That is why diabetes and obesity drugs are now tested in large cardiovascular outcome trials that count events directly. For tirzepatide, SURPASS-CVOT randomised 13,299 patients with type 2 diabetes and cardiovascular disease against dulaglutide. A primary event occurred in 12.2% against 13.1%, and tirzepatide met the trial's noninferiority test. How to read that result is covered in tirzepatide cardiovascular outcomes, and what the percentages mean in absolute terms in relative vs absolute risk. Retatrutide's phase 3 weight and HbA1c results are surrogate results until its outcome trial reports, as the phase guide explains.

How do surrogates appear in the compounds discussed on this site?

BDNF and Semax. Dolotov and colleagues reported that Semax raised BDNF protein in the rat basal forebrain within three hours, and found specific binding sites there. That is a mechanism in rats. A Russian clinical study by Gusev and colleagues followed 110 patients after ischaemic stroke, split into early and late rehabilitation groups, each subdivided by whether they received Semax, and measured plasma BDNF alongside motor scores and the Barthel index. A circulating BDNF level is several steps from brain recovery, and the abstract describes no randomisation, placebo or blinding. The wider literature is in Semax and cerebral ischaemia. Semax is a registered medicine in Russia only and is not approved by Health Canada, the FDA or the EMA.

Telomere length and Epitalon. Epitalon's best-known findings are increases in telomerase activity and telomere length in cultured cells. Those are often read as "anti-ageing", on the assumption that longer is better. A Mendelian randomisation study by Haycock and colleagues used genetic variants as a natural experiment. Genetically longer telomeres were associated with higher risk of several cancers, including lung adenocarcinoma (odds ratio 3.19 per standard deviation), and lower risk of coronary heart disease (0.78). A marker whose direction is good for one outcome and bad for another cannot be read as simply good. Epitalon and cancer risk covers the evidence; Epitalon is approved by no major regulator.

Lean mass and GLP-1 drugs. Body-composition scans report lean mass, which includes water, organs and connective tissue as well as muscle. In a SURMOUNT-1 substudy of 160 participants scanned by DXA, lean mass fell 10.9% on tirzepatide against 2.6% on placebo. Roughly a quarter of weight lost was lean mass in both groups. Whether that loss matters depends on strength and function, which DXA does not measure. GLP-1 muscle loss sets out what was and was not measured.

Muscle strength and elamipretide. Elamipretide's US accelerated approval for Barth syndrome rests on muscle strength. Its confirmatory trial, 4TAZPower (NCT07531251), describes knee extensor strength as an intermediate clinical endpoint. Its own primary endpoint is a composite of three functional tests over 72 weeks. That is the accelerated approval logic in miniature: approve on a measure reasonably likely to predict benefit, then check.

What should you ask of any marker?

  • Is it a clinical outcome, an intermediate clinical endpoint, or a laboratory or imaging marker?
  • Has changing this marker, with an intervention of this kind, been shown to change the outcome in trials? Or is the evidence that the two are associated in untreated people?
  • Could the compound affect the outcome through routes the marker does not capture?
  • Was it measured in the tissue that matters, or in blood as a proxy?
  • Is a "better" direction for the marker actually established, or assumed?

Every compound named here is discussed as a research subject. Nothing here is medical, dosing, or therapeutic guidance.

Frequently asked questions

Is a surrogate endpoint the same as a biomarker? Every surrogate is a biomarker, but most biomarkers are not surrogates. A biomarker is anything measurable that indicates a biological state. It becomes a surrogate endpoint only when it is used in place of a clinical outcome to judge whether an intervention works, and it becomes a validated one only when trial evidence supports that substitution.

Why do regulators accept surrogates at all? Because waiting for hard outcomes can take years and thousands of patients, which is unworkable for rare or rapidly fatal diseases. Accelerated approval accepts a surrogate reasonably likely to predict benefit, on condition that a confirmatory trial follows. The cost is that some approvals are never confirmed on the outcome itself.

Can a validated surrogate stop being valid? Yes, for a new drug class. A marker that tracked the outcome with drugs acting one way may not hold for a drug acting another way, because the new drug may have effects the marker does not register. In the torcetrapib trial, HDL moved as intended while blood pressure and aldosterone, which HDL does not reflect, moved too.

Is weight loss a surrogate endpoint? For cardiovascular and kidney disease, yes. Body weight is a clinical measure in its own right for obesity trials, but heart attacks, strokes and deaths are the outcomes it is expected to predict. That is why obesity drugs are being tested in separate outcome trials, such as TRIUMPH-Outcomes for retatrutide.

Selected Research References

  1. Fleming TR, DeMets DL. Surrogate end points in clinical trials: are we being misled? Annals of Internal Medicine. 1996;125(7):605-613. pubmed.ncbi.nlm.nih.gov/8815760
  2. Echt DS, Liebson PR, Mitchell LB, et al. Mortality and morbidity in patients receiving encainide, flecainide, or placebo. The Cardiac Arrhythmia Suppression Trial. New England Journal of Medicine. 1991;324(12):781-788. pubmed.ncbi.nlm.nih.gov/1900101
  3. Barter PJ, Caulfield M, Eriksson M, et al.; ILLUMINATE Investigators. Effects of torcetrapib in patients at high risk for coronary events. New England Journal of Medicine. 2007;357(21):2109-2122. pubmed.ncbi.nlm.nih.gov/17984165
  4. Gerstein HC, Miller ME, Byington RP, et al.; Action to Control Cardiovascular Risk in Diabetes Study Group. Effects of intensive glucose lowering in type 2 diabetes. New England Journal of Medicine. 2008;358(24):2545-2559. pubmed.ncbi.nlm.nih.gov/18539917
  5. Kim C, Prasad V. Strength of Validation for Surrogate End Points Used in the US Food and Drug Administration's Approval of Oncology Drugs. Mayo Clinic Proceedings. 2016. pubmed.ncbi.nlm.nih.gov/27236424
  6. Nicholls SJ, Pavo I, Bhatt DL, et al.; SURPASS-CVOT Investigators. Cardiovascular Outcomes with Tirzepatide versus Dulaglutide in Type 2 Diabetes. New England Journal of Medicine. 2025;393(24):2409-2420. pubmed.ncbi.nlm.nih.gov/41406444
  7. Dolotov OV, Karpenko EA, Inozemtseva LS, et al. Semax, an analogue of adrenocorticotropin (4-10), binds specifically and increases levels of brain-derived neurotrophic factor protein in rat basal forebrain. Journal of Neurochemistry. 2006;97(Suppl 1):82-86. pubmed.ncbi.nlm.nih.gov/16635254
  8. Gusev EI, Martynov MY, Kostenko EV, et al. The efficacy of semax in the treatment of patients at different stages of ischemic stroke. Zhurnal Nevrologii i Psikhiatrii imeni S.S. Korsakova. 2018;118(3 Vyp 2):61-68. pubmed.ncbi.nlm.nih.gov/29798983
  9. Haycock PC, Burgess S, Nounu A, et al.; Telomeres Mendelian Randomization Collaboration. Association Between Telomere Length and Risk of Cancer and Non-Neoplastic Diseases: A Mendelian Randomization Study. JAMA Oncology. 2017;3(5):636-651. pubmed.ncbi.nlm.nih.gov/28241208
  10. Look M, Dunn JP, Kushner RF, et al. Body composition changes during weight reduction with tirzepatide in the SURMOUNT-1 study of adults with obesity or overweight. Diabetes, Obesity and Metabolism. 2025;27(5):2720-2729. pubmed.ncbi.nlm.nih.gov/39996356

All products sold by NovaEvum are for research and laboratory use only. They are not intended for human or animal consumption, diagnosis, treatment, cure, or prevention of any disease.