<< Back to Agenda

Slide 1: Recent Approaches to Surrogate Endpoint Validation

Stuart G. Baker, Sc.D.
National Cancer Institute
sb16i@nih.gov

Slide 2: What do we mean by validation of a marker or endpoint?

• marker or endpoint is validated if there is confidence in proceeding to the next stage of evaluation
• Methodology depends on the application
• Biomarkers for the early detection of cancer
• Biomarkers for predicting cancer recurrence
• Biomarkers for targeting intervention in treatment trials
• Surrogate endpoints

Slide 3: Examples of validation

• Biomarkers for early detection of cancer
• Next stage: further study as a trigger for early intervention with cancer-mortality endpoint
• Validation: high sensitivity and specificity in an independent test sample
• Biomarkers for predicting cancer recurrence
• Next stage: randomized trial of standard versus new treatment for those with poor prognosis
• Validation: high predictive value positive and negative in an independent test sample

Slide 4: Examples of validation (continued)

• Biomarkers for targeting intervention in a treatment trial
• Next stage: recommendation for population with the biomarker
• Validation: randomized trial of intervention in subjects with the biomarker
• Surrogate endpoints
• Next stage: recommendation for population (although sometimes further study)
• Validation: correct conclusions about effect of intervention on true endpoint (in trial with both surrogate and true endpoints)

Slide 5: Surrogate endpoint: definition

Measure or indicator of a biological process that is

1. (a) used to make conclusions about the effect of intervention on a true endpoint that is a health outcome
2. obtained sooner, at less cost, or less invasively than a true endpoint

Slide 6: Validating a surrogate endpoint

• NEXT STAGE: Application trial: surrogate but not true endpoint is observed
• What is the effect of intervention on true endpoint?
• VALIDATION: Validation trial in which both surrogate and true endpoints are observed
• Are the conclusions about the effect of intervention on true endpoint the same when based on
• V(i) only surrogate endpoint
• (ii) only the true endpoint ?

Slide 7: Issues in validating a surrogate endpoint

• Hypothesis testing framework
• Estimation framework
• Caveats
• Recommendations

Slide 8: Hypothesis Testing Framework (Chart)

Slide9 : Prentice Criterion "holds" (Identical lines for A and B) (Charts)

Slide 10: Prentice Criterion does NOT "holds" (Identical lines for A and B) (Charts)

Slide 11: Drawbacks of hypothesis testing framework:

• If Prentice Criterion is not rejected, it is not clear how "close" is good enough.
• With estimation, it is easier to include data from previous trials of surrogate and true endpoint
• Generally estimation is preferred to hypothesis testing (e.g. for weighing harms and benefits)

Slide 12: Estimation of Framework (Flow Chart)

Slide 13: Data Scheme (Table)

Slide 14: "Meta-analytic" methods for combining data from previous trials

• Method 1 Regression on trial-level statistics
• Method 1 e.g. fraction with surrogate and true endpoint in each previous trial
• Method 1 Gail et al 2000; Buyse et al 2000
• Method 2. Combine predicted intervention effects
• Method 1 effect of intervention on true endpoint based on data from each previous trial
• Method 1 Baker 2005
• Method 1 Simpler computations

Slide 15: Method 1: Regression" on Trial-level statistics (chart)

Slide 16: Method 2: Predicated effect of intervention on true endpoint based on surrogates A and B in new trial and data from previous trial 1 (Chart)

Slide 17: Method 2: Predicated effect of intervention on true endpoint based on surrogates A and B in new trial and data from previous trial 2 (Chart)

Slide 18: Method 2: Predicated effect of intervention on true endpoint based on surrogates A and B in new trial and data from previous trial 3 (Chart)

Slide 19: Hypothetical Data Method 1 (Chart)

Slide 20: Be aware of caveats even if surrogate is validated

• Method 1 Extrapolation to a new intervention
• Method 1 Surrogate endpoint for benefit does not predict harms that might arise after surrogate is observed

Slide 21: Caveats are less critical if

• Preliminary drug development when next stage is further testing
• Evaluating a different dose or timing of an intervention previously shown effective (using a true endpoint) at another dose or timing

Slide 22: Recommendations

• Use meta-analytic estimation approach for validation of surrogate endpoint
• Check if same conclusion about effect of intervention on true endpoint using (i) surrogate endpoint and (ii) true endpoint
• But hard to get data from previous trials!
• Even if validated, remember caveats
• extrapolation to a new intervention
• unknown effect of intervention on harms

<< Back to Agenda