thebench.bio #1 Ecosystem for the Bio-Tech Industry
All Clinical Trial Design questions

Question 18 of 20 Advanced Interpretation

Why can a surrogate endpoint improve while patients do worse?

Short answer

A surrogate is a marker on one causal path, and a drug can move that marker while acting on other paths that harm the patient. In the Cardiac Arrhythmia Suppression Trial, encainide and flecainide suppressed ventricular ectopy as hoped, yet patients on them died more often than those on placebo.

Why

A surrogate endpoint is a laboratory measure, image or sign used in place of a direct measure of how patients feel, function or survive. US law, quoted on FDA's surrogate endpoint table, separates surrogates known to predict benefit, which can support traditional approval, from those only reasonably likely to, which can support accelerated approval. A surrogate can fail in several ways.

The disease may cause harm through paths the marker does not capture. The drug may help the marker but have off-target harms. Or the marker may be associated with outcome without causing it.

CAST is the standard example. Suppressing ectopic beats after a heart attack was expected to prevent sudden death. In the published results, 43 patients on encainide or flecainide died of arrhythmia against 16 on placebo.

Fleming and DeMets reviewed this and other cases in 1996. A marker validated for one drug class may not transfer to a new mechanism. That is why accelerated approvals usually require confirmatory trials on clinical outcomes.

A strong answer covers

  • Definition of a surrogate endpoint
  • Known to predict versus reasonably likely to predict benefit
  • Failure paths: off-target harm, other disease pathways, association without causation
  • CAST as a concrete example with numbers
  • Validation does not transfer across mechanisms

Common mistakes

  • Assuming a strong correlation between marker and outcome proves the marker is a valid surrogate.
  • Extending a surrogate validated for one drug class to a new mechanism.
  • Forgetting that safety outcomes are not captured by an efficacy surrogate.