Chapter 9 Estimands

The importance of Estimands is introduced in the ICH E9 addendum.

9.1 Estimands, estimator and estimates

  • Estimands: the clearly stated, non-ambiguous variable to be estimated

  • Estimator: the statistics tool used to arrive at the estimands

  • Estimates: the statistics results for the estimation of the intended estimand

9.2 Strategies for Estimands

  • Treatment policy strategy

  • Composite strategy

  • Principal stratum strategy

  • Hypothetical strategy

9.3 Estimand thinking for biomarkers

  • Causal thinking is key in coming up the right strategies for the estimands. It is relatively simple in terms of computational/statistics methods involved for treatment policy and composite strategies. A few examples:

    • Treatment discontinuation due to biomarker monitoring

    • Composite endpoints involving biomarkers surrogates

  • For principal stratum and hypothetical strategies, relatively complicated causal inference methods need to be applied. We will go into details and give some application examples in biomarkers:

    • anti-drug antibody appearance in the treatment arm

    • Use of rescue medication in the trial

9.4 R vignettes with examples to follow

Many of the relevant examples can only be demonstrated with non-public clinical trial data. I will probably include R vignettes later using simulation data.