×

Dose-finding designs for early-phase cancer clinical trials. A brief guidebook to theory and practice. (English) Zbl 1427.92001

SpringerBriefs in Statistics. JSS Research Series in Statistics. Tokyo: Springer (ISBN 978-4-431-55584-1/pbk; 978-4-431-55585-8/ebook). xv, 133 p. (2019).
The book under review is intended to serve as a brief handbook for graduate students and biostatistics, as well as for oncologists who are involved in the analysis of methods for developing optimal doses for treating cancer in the early stages.
In Chapter 1, the authors describe the basic concepts for early-phase trials, outlining chemotherapeutic, immunotherapeutic agents and dose-finding designs.
Chapter 2 dedicates to the limitations of the 3+3 design, because it poorly identifies maximum tolerated dose (MTD) despite its simplicity and transparency. Then the here overviewed alternative rule-based designs that can improve the performance of the 3+3 design and 10 related topics are discussed. However, the authors consider that, although the 3+3 approach is a conventional use in practice, they do not recommend its use, as the statistical community agrees on its limitations.
Chapter 3 is the continuation of the second chapter: if rules-based algorithms are discussed there, then algorithms that implement dose-search based on statistical models are presented here. The chapter provide full description of the concepts, theories, properties, advantages and disadvantages of the continual reassessment method (CRM). Phase 1 trials in oncology are designed to determine the maximum tolerated doses of agents of interest. The designs traditionally are based on systems of rules or on statistical models. But in recently developed model-assisted designs, the rules of escalation, de-escalation, and maintenance of the current dose are assisted by the model for the toxicity outcome, and are simple and transparent before the trial initiation. This approach considers toxicity alone by the following ways: the modified toxicity probability interval (mTPI) design and its improved version (the TPI-2), Bayesian optimal interval design, and keyboard design. In Chapter 4, these problems are discussed, including getting an answer to a question – what is all the same better? At the same time, comparative studies conducted in 2017–2019 are mentioned.
Chapter 5 dedicates designs considering toxicity and efficacy. As far as MTD is expected to produce maximal efficacy under admissible toxicity, this dose is adopted as optimal dose in the subsequent phase I and III trials. But this concept is not suitable for a cytostatic, molecularly targeted, or biologically agent. So, for combined phases I/II, it is helpful to consider both efficacy and toxicity. This issue is discussed in Chapter 5.
In the final Chapter 6, the problem of cancer immunotherapy, in which the determination of biologically optimal doses is critical to the success of treatment of the patient is described. Such doses are found by considering the outcome related to the immune response in addition to the toxicity and efficacy outcomes. The book provides links to the open source software for the described algorithms.

MSC:

92-02 Research exposition (monographs, survey articles) pertaining to biology
92C50 Medical applications (general)
62P10 Applications of statistics to biology and medical sciences; meta analysis
92-08 Computational methods for problems pertaining to biology
92-04 Software, source code, etc. for problems pertaining to biology

Software:

R; SAS
PDFBibTeX XMLCite
Full Text: DOI Link