The evaluation of surrogate endpoints. (Q1773294)

This recently appeared volume deals with the potentials and limitations of surrogate endpoints in clinical studies. In spite of some erroneous and momentous results in the past based on invalid surrogate endpoints there persists the need of their implementation in clinical trials, e.g., when the true endpoint might be difficult to use or accelerated approval of a new drug is an imperative. This multi\,-\,author monograph contributes to the evaluation of surrogate endpoints a comprehensive overview of statistical modeling and applications accounting also positions taken by the pharmaceutical industry and regulatory authorities (Chapter 2: ``Setting the Scene''; Chapter 3: ``Regulatory Aspects in Using Surrogate Markers in Clinical Trials''). Furthermore, the motivation for the use of surrogates, validation and assessment of the quality of a surrogate as well as its usage for prediction of the true outcome are dealt with in this book. To specify a good surrogate, Prentice's criteria serve as definition and provide operational criteria for the validation of a surrogate endpoint. Starting from the case of a single trial and a single surrogate, different measures are introduced to asses the quality of the surrogate and its association to the true endpoint at the individual as well as at the trial level. See Chapter 5 on ``The History of Surrogate Endpoint Validation'', where the results are described under the assumption of the same data type for both binary and normal endpoints; and see also Chapter 6 on ``Validation Using Single-trial Data: Mixed Binary and Continuous Outcomes'', which tackles mixed binary and continuous outcomes. In Chapter 7 (``A Meta-Analytic Validation Framework for Continuous Outcomes''), the model for normally distributed endpoints is amplified to a meta-analytic or hierarchical two-stage model based on fixed-effects as well as a random-effects representations. Corresponding measures for the assessment of the quality of the surrogate at the individual and the trial level are defined. These models are generalized to a three-stage model where the different units ``patient'', ``center'' and ``trial'' are considered, followed by a discussion of several model strategies (Chapter 8: ``The Choice of Units''). Considering surrogates in clinical trials, one is not only interested in testing for treatment-effects on the true endpoint but also in the estimation of true treatment-effects. In the meta-analytic framework several approaches are proposed for normal, binary and failure-time endpoints, and precision estimates for some predictions are suggested, see the chapters 9: ``Extensions of the Meta-Analytic Approach to Surrogate Endpoints''; 10: ``Meta-analytic Validation with Binary Outcomes''; and 11: ``Validation in the Case of two Failure-time Endpoints''. Next, measures of surrogacy are presented for combinations of ordinal-survival endpoints (Chapter 12: ``An ordinal Surrogate for a Survival True Endpoint'') as well as longitudinal survival endpoints (Chapter 13: ``A Combination of Longitudinal and Survival Endpoints''). Encountering repeated measurements on either one or on both endpoints requires the crossover to a fully multivariate model where even the points in time of measurement at the individual patient are variable (Chapter 14: ``Repeated Measures and Surrogate Endpoint Validation''). Further measures of surrogacy and approaches for evaluating surrogates are presented using, e.g., two-stage and fully hierarchical Bayesian models (Chapter 18: ``An Alternative Measure for Meta-analytic Surrogate Endpoint Validation''; Chapter 19: ``Surrogate Endpoint Definition and Evaluation''; and Chapter 15: ``Bayesian Evaluation of Surrogate Endpoints''). A variety of case and simulation studies helps to illustrate the results, to demonstrate the performance of specific models and to discuss computational issues. This practical illustration is augmented by an assortment of studies (Chapter 4: ``Notation and Motivating Studies'') which are utilised as running examples throughout the book and make the different approaches comparable. Especially in the context of cancer a multiplicity of case studies and discussions are presented (see Chapter 20: ``The Promise and Peril of Surrogate Endpoints in Cancer Research''), likewise for HIV infection (see Chapter 17: ``The Evaluation of Surrogate Endpoints in Practice: Experience in HIV''), and a whole chapter was dedicated to the adaptation of the developed framework for the use in mental health to assess the so-called ``criterion-validity'' of psychiatric symptom scales (Chapter 16: ``Surrogate Marker Validation in Mental Health'').