Stefano Garzon

Research · Oct 23, 2026

Surrogate endpoints and biomarkers

Concept

A surrogate endpoint is a variable measured in place of the clinical outcome of interest. Examples: angiographic atherosclerosis instead of infarction or death, ectopy on Holter monitoring instead of sudden death, tumor size instead of survival, CD4 count instead of AIDS or death, bone mineral density instead of fracture.

The rationale is practical. Continuous variables need smaller samples. The biomarker changes before the clinical event, which shortens the trial.

Prentice’s criteria

Prentice proposed a restrictive criterion. The surrogate must yield a valid test of the null hypothesis of no treatment effect on the true outcome. In practice this requires two conditions:

  1. The surrogate correlates with the clinical outcome.
  2. The surrogate captures the full effect of the intervention on the clinical outcome. Given the surrogate’s history, the risk of the true outcome must be independent of treatment.

Most proposed biomarkers meet the first condition. The second often fails. An intervention can act on the outcome through pathways that bypass the biomarker, or have adverse effects that cancel the benefit.

Examples where the surrogate failed

CAST. Ventricular ectopy is associated with sudden death after infarction. Antiarrhythmics that suppressed it were prescribed to reduce mortality. In CAST, patients with documented suppression were randomized to drug or placebo. With encainide or flecainide there were 43 arrhythmic deaths versus 16 with placebo, and 17 versus 5 nonarrhythmic cardiac deaths.

ILLUMINATE. Torcetrapib plus atorvastatin raised HDL by 72.1% and lowered LDL by 24.9% at 12 months in 15,067 high-risk patients. The trial was stopped for increased mortality and cardiovascular events (HR 1.25). The drug also raised systolic pressure by 5.4 mmHg and altered electrolytes and aldosterone.

Sodium fluoride. In 202 women with postmenopausal osteoporosis, fluoride increased lumbar spine bone density by 35%. Vertebral fractures did not decrease (163 versus 136) and nonvertebral fractures increased (72 versus 24).

CD4 in HIV infection. Fleming summarized eight trials with a positive clinical result: CD4 was favorable in seven. In eight trials with a negative clinical result, CD4 was also favorable in six.

Inotropes in heart failure. They improved exercise tolerance and symptoms. Subsequent trials showed increased mortality.

Other practical requirements

Beyond biological validity, Friedman lists measurement requirements:

When a surrogate is useful

Friedman considers surrogates appropriate in three settings: very serious diseases, where speed matters more; serious and rare diseases, where there are not enough participants for clinical events; and early phase studies, to choose a dose and confirm the biological effect. When approval rests on a surrogate, studies with clinical outcomes should follow. For diabetes drugs, the FDA now requires evidence that cardiovascular events are not increased.

Common errors when reading a paper

What to check in a trial

  1. Whether the primary outcome is clinical or a surrogate.
  2. Whether the surrogate was validated for the same drug class and mechanism.
  3. Whether the intervention acts through pathways that bypass the biomarker.
  4. How the surrogate was measured, with what precision and what losses.
  5. The number of clinical events and serious adverse events, even when secondary.
  6. Whether a clinical outcome trial is ongoing or completed.

Related posts: composite outcomes, clinical trial phases.

References