This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Surrogate Markers in Clinical Trials The use of biological markers to replace or complement traditional clinical markers is increasingly common in clinical trials. Reasons include the time necessary to get clinical endpoints (unacceptable in Phase I or II trials) and the measurement variation for some clinical markers (forces a high sample size). Examples: • CD4+ cell counts or viral load in HIV clinical trials. Particularly important after the development of effective antiretroviral therapy that allowed longer survival. • Tumor size to determine response in Phase I or II cancer clinical trials (rather than waiting for death or disease progression). • In development: biomarkers for Alzheimer’s disease trials based on imaging, blood tests, or CSF levels. Needed to complement clinical change such as decline in cognitive function because decline is very gradual and slow, and neuropsych tests have high variation within person. Much of the statistical framework was developed by Ross Prentice. Will review some of the ideas and some examples. Criteria for a surrogate marker Let T be a random variable that is defined to be the true endpoint, for example, time to death, time to disease progression, a vector of cognitive function measurements over time. Let Z denote treatment (we will take this to be dichotomous for now, standard or new treatment). Let S be the proposed surrogate marker. Prentice proposed the following definition of a surrogate marker: a test of the null hypothesis of no relation- ship of surrogate marker to treatment is also a valid test of the null hypothesis of no relationship of true endpoint to treatment. In symbols, S satisfies P ( S | Z ) = P ( S ) ⇔ P ( T | Z ) = P ( T ) . Formally, we can write out two conditions that are jointly sufficient for the definition to hold: i P ( T | S, Z ) = P ( T | S ), that is, S fully captures the effect of Z on T , and ii P ( T | S ) 6 = P ( T ), that is, S is informative about T . The proof is pretty straightforward. We will go over in class. A slightly different way to describe Prentice’s conditions for S to be a surrogate marker for the effect of treatment Z on outcome T : P1. Z has a significant effect on the outcome T and on the surrogate T . P2. S has a significant effect on the outcome T . P3. Z has no significant effect on outcome T given S . This adds the idea (P1) that the treatment actually is effective. P2, that the surrogate has a relationship to the clinical endpoint, is equivalent to (ii) above. P3 is equivalent to (iii) above: once you know the surrogate, the treatment is independent of the outcome. For example, suppose T is survival time in prostate cancer with bone metastases, and S is change in PSA level.is change in PSA level....
View Full Document
This note was uploaded on 12/30/2010 for the course BST 252 taught by Professor Tsodikov during the Winter '06 term at UC Davis.
- Winter '06