Surrogate markers and what they do not tell you

Tracing the path from biomarker to clinical benefit — and where it breaks

1
Step 1 — Preclinical science
A mechanism is identified
Laboratory or animal research shows that a compound interacts with a biological pathway. A plausible mechanism connecting the compound to a health-relevant process is established.
often assumed to translate to humans
2
Step 2 — Biomarker trial
A marker changes in a short trial
A randomised trial shows the compound changes a measurable variable — a blood marker, a physiological score, an imaging result — in the expected direction. This is the most common type of evidence in supplement research.
often assumed to predict clinical outcome
3
Step 3 — Clinical outcome
The outcome that actually matters
Reduced disease incidence, improved symptoms, better function, avoided hospitalisation, longer survival. These are what the person taking the supplement is hoping for. This step is rarely tested directly in supplement research.
Where the logic breaks
The arrows between each step are assumptions, not established facts. A mechanism that works in cells may not operate the same way at supplemental doses in people. A biomarker that is associated with disease risk in observational data may not reliably predict what happens when an intervention changes it. The history of clinical medicine contains numerous examples of interventions that moved a biomarker in the right direction without producing the expected clinical benefit — and in some cases while causing harm.

The reliability of a surrogate marker exists on a spectrum. The critical question is always the same: has the relationship between this biomarker and this clinical outcome been rigorously tested in large intervention trials? Most biomarkers used in supplement research have not been through this process.

Homocysteine
Antioxidant markers
CRP / inflammation
LDL (statins)
Blood pressure → stroke
Not validated — association only Well-validated surrogate
Systolic blood pressure
Decades of randomised trials across multiple drug classes have established that reducing systolic blood pressure reduces stroke risk, with a reasonably consistent dose-response relationship.
Well-validated for stroke
LDL cholesterol
Multiple statin trials support LDL as a surrogate for cardiovascular events in that drug class. Whether the same relationship holds for other mechanisms of LDL reduction is more contested.
Partial — class-specific
Homocysteine
Strongly associated with cardiovascular risk in observational data. B vitamins reliably lower it. Large randomised trials consistently show no reduction in cardiovascular events.
Not validated — failed in RCTs
Antioxidant markers
Many supplements raise measures of antioxidant activity. The relationship between antioxidant marker changes and clinical disease outcomes has not been established and in some cases has been reversed.
Not validated

These cases illustrate how the assumption that changing a biomarker translates to clinical benefit has been tested — and what happened when large trials measured actual outcomes.

✗
Homocysteine and B vitamins
Cardiovascular prevention — surrogate lowered, outcome unchanged
▾
The claim
Elevated homocysteine is associated with cardiovascular risk. B vitamins lower homocysteine. Therefore B vitamins reduce cardiovascular events.
The biomarker
Plasma homocysteine — reduced by 25–30% with folic acid supplementation.
The trial
HOPE-2 (Lonn et al., NEJM 2006): 5,522 patients with vascular disease, five-year follow-up, independent funding. Homocysteine reduced significantly. No effect on cardiovascular death, MI, or stroke.
The meta-analysis
Clarke et al. (Archives of Internal Medicine, 2010): 8 trials, 37,485 individuals. No reduction in cardiovascular events, cancer, or cause-specific mortality.
Lowering homocysteine through this intervention did not improve clinical outcomes. Whether homocysteine is causal, contributory, or a marker of other processes remains unresolved — what is clear is that changing it via B vitamins did not help in this population and context.
✗
Antioxidant vitamins and cardiovascular disease
Oxidative stress markers improved, clinical outcomes unchanged or worse
▾
The claim
Oxidative stress contributes to atherosclerosis. Antioxidant vitamins reduce markers of oxidative stress. Therefore antioxidant supplementation reduces cardiovascular risk.
The biomarker
Measures of oxidative stress and lipid peroxidation — improved by vitamin E and beta-carotene in laboratory and small clinical studies.
The trials
Multiple large RCTs including HOPE (vitamin E, n=9,541) and the ATBC and CARET trials (beta-carotene) found no cardiovascular benefit and in some cases increased risk of adverse outcomes in specific populations.
Improving a marker of oxidative stress did not translate to clinical cardiovascular protection. The biological plausibility was strong; the clinical translation was not.
~
Omega-3 and triglycerides
A validated biomarker effect — with important scope limits
▾
The biomarker
Plasma triglycerides — reliably reduced by EPA and DHA at doses of 2–4 g daily. This effect is dose-dependent and well-replicated.
The complexity
Triglycerides are a cardiovascular risk marker, but their relationship to clinical outcomes is less well-validated than blood pressure or LDL. The cardiovascular event reduction seen in REDUCE-IT (pure EPA, 4 g daily, high-risk population) has not been replicated by STRENGTH (EPA+DHA, similar dose, similar population), raising questions about whether the benefit is from EPA specifically, from the dose, or from the study design.
The scope limit
Standard supplement doses (1 g EPA+DHA) do not reliably produce clinically meaningful triglyceride reductions in most people at baseline triglyceride levels within the normal range.
The biomarker effect is real at therapeutic doses. Clinical outcome translation is more complex and dose-dependent. Standard supplement doses may not achieve either.
✓
Blood pressure and stroke
A well-validated surrogate relationship
▾
The biomarker
Systolic blood pressure — a consistently validated surrogate for stroke risk across multiple drug classes and populations.
The evidence base
Decades of randomised trials demonstrated a consistent, quantifiable relationship between the magnitude of blood pressure reduction and the reduction in stroke incidence. The FDA accepts blood pressure as a surrogate endpoint for stroke in this context.
The caveat
The validation is strongest for stroke and less robust for heart failure, myocardial infarction, and overall mortality. The relationship was established through specific drug class trials, not supplements.
This is what surrogate validation looks like: a quantified, replicated, mechanism-coherent relationship between biomarker change and clinical outcome across multiple independent trials.

Self-reported outcomes — questionnaires, rating scales, and subjective assessments — are a specific category of outcome that deserves separate consideration. They are not surrogates in the strict sense, but they carry their own interpretation challenges, particularly for supplements marketed around energy, mood, cognition, stress, and sleep.

Lower confidence

Open-label or inadequately blinded trial
Participants aware of product marketing claims
Short duration with attention-focused participants
No blinding integrity check reported
Small sample, single trial, no replication
Outcome scale with poor validated properties

Higher confidence

Double-blind with blinding integrity confirmed
Validated, well-characterised outcome scale
Adequate sample size with pre-specified power
Effect size large enough to be clinically meaningful
Replicated across independent trials
Pre-registered with outcomes matching registration
Self-reported outcomes are not inherently unreliable. For conditions like pain, fatigue, and mood, they are often the most direct available measure of patient experience. The question is always whether the trial design was rigorous enough to isolate the intervention effect from expectation, attention, and placebo. A small, unblinded trial showing improved self-reported energy scores tells us almost nothing. A well-blinded, adequately powered trial showing the same effect, replicated by independent groups, is a different kind of evidence entirely.