Contents
Preface xi
Authors xvii
List of Abbreviations xix
1 Background and Motivation 1
1.1 Pragmatic approach to problem solving . . . . . . . . . . . . 1
1.2 Problem solving skills . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Mathematics versus statistics . . . . . . . . . . . . . . . . . . 5
1.4 A look at modern drug development . . . . . . . . . . . . . . 8
1.4.1 Stages of drug development . . . . . . . . . . . . . . . 11
1.4.2 Factors that have had an impact on drug development 14
1.5 Statistics and evidence-based science . . . . . . . . . . . . . 15
1.6 In summary: what this book is all about . . . . . . . . . . . 20
Introduction to Chapters 2, 3, and 4 25
2 Statistical Programming 27
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Asking the right questions . . . . . . . . . . . . . . . . . . . 28
2.3 Choice of statistical and presentation software . . . . . . . . 30
2.4 95/5″ rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.1 The sources . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.2 SAS Certi_cation|Is it worth the time and e_orts? . 32
2.5 Data access, data creation and data storage . . . . . . . . . . 33
2.6 Getting data from external _les . . . . . . . . . . . . . . . . 35
2.7 Data handling . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.7.1 The DATA step . . . . . . . . . . . . . . . . . . . . . . 37
2.7.2 Loops and arrays . . . . . . . . . . . . . . . . . . . . . 38
2.7.3 Going from vertical to horizontal datasets
and vice versa . . . . . . . . . . . . . . . . . . . . . . . 39
2.8 Why do we need basic knowledge of the
Macro language? . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.8.1 Open code vs. DATA step . . . . . . . . . . . . . . . . 40
2.8.2 Loops in the open code (inside macros)
and nested macros . . . . . . . . . . . . . . . . . . . . 40
2.8.3 Use of pre-written (by others) macro code . . . . . . . 41
2.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3 Data Management 43
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Design of data collection . . . . . . . . . . . . . . . . . . . . 45
3.3 Organization of data collection . . . . . . . . . . . . . . . . . 46
3.4 Data cleaning or veri_cation . . . . . . . . . . . . . . . . . . 48
3.5 Re-structuring of the data . . . . . . . . . . . . . . . . . . . 50
3.6 First case study . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.7 Second case study . . . . . . . . . . . . . . . . . . . . . . . . 52
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4 Biostatistics 67
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.2 The biostatistician’s role . . . . . . . . . . . . . . . . . . . . 71
4.3 Background assessment: what do we start with? . . . . . . . 78
4.4 A minimal su_cient set of tools for
the biostatistician . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.1 Knowledge of the disease area . . . . . . . . . . . . . . 85
4.4.2 Knowledge of the regulatory landscape . . . . . . . . . 86
4.4.3 Understanding of the clinical trial protocol . . . . . . 87
4.4.4 Knowledge of statistical methodologies for
protocol development . . . . . . . . . . . . . . . . . . 90
4.4.5 Statistical software . . . . . . . . . . . . . . . . . . . . 93
4.4.6 Communication skills . . . . . . . . . . . . . . . . . . 95
4.4.7 Knowledge of processes . . . . . . . . . . . . . . . . . 96
4.5 Advanced biostatistics toolkit . . . . . . . . . . . . . . . . . 97
4.5.1 Adaptive designs . . . . . . . . . . . . . . . . . . . . . 98
4.5.2 Basket, umbrella, platform trials and master protocols 100
4.5.3 Dose-_nding methods . . . . . . . . . . . . . . . . . . 102
4.5.4 Multiplicity issues . . . . . . . . . . . . . . . . . . . . 103
4.5.5 Estimands . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.5.6 Quantitative decision-making support . . . . . . . . . 106
4.5.7 Digital development . . . . . . . . . . . . . . . . . . . 107
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Introduction to Chapters 5, 6, and 7 111
5 Development of New Validated Scoring Systems 115
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2 Recognition of problem existence . . . . . . . . . . . . . . . . 116
5.3 Study of available methods and tools with
consequent realization that they are insu_cient . . . . . . . 117
5.4 Clear formulation and formalization of the
main task to be solved . . . . . . . . . . . . . . . . . . . . . 121
5.5 A solution itself . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.6 Are we _nished? Not in the regulatory setting! . . . . . . . . 129
5.7 Assessment of created by-products as potentially new tools,
skills and methods . . . . . . . . . . . . . . . . . . . . . . . . 131
5.8 Generalization of all achievements and evaluation of potential
applications in the real world . . . . . . . . . . . . . . . . . . 131
6 Resurrecting a Failed Clinical Program 133
6.1 Preamble: what we are dealing with . . . . . . . . . . . . . . 133
6.2 Problems solved . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.2.1 Studying drugs with dosage that depends on needs . . 138
6.2.2 Separation of toxicity and e_cacy e_ects
in safety outcome misbalance . . . . . . . . . . . . . . 139
6.2.3 Creation of a PK model for the transfusion _eld . . . 142
6.2.4 Mystery of the transfusion trigger . . . . . . . . . . . 149
6.2.5 The rise and fall of the HBOC _eld . . . . . . . . . . . 153
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7 Can One Predict Unpredictable? 159
7.1 Personal disclaimer/preamble . . . . . . . . . . . . . . . . . . 159
7.2 First, what can we do? . . . . . . . . . . . . . . . . . . . . . 160
7.3 Problems in planning of the open-ended projects . . . . . . . 161
7.3.1 Extraneous vs. overlooked parts in
preliminary planning . . . . . . . . . . . . . . . . . . . 162
7.3.2 Level of uncertainty of elementary tasks . . . . . . . . 166
7.3.3 Terminology and de_nitions . . . . . . . . . . . . . . . 172
7.4 Estimating distribution of time to completion
of an open-ended project . . . . . . . . . . . . . . . . . . . . 176
7.4.1 Surprising results of _rst test runs of the algorithm . . 177
7.4.2 The nature of estimates for elementary tasks . . . . . 180
7.4.3 Estimation for a single branch . . . . . . . . . . . . . 183
7.4.4 How to analyze the results? . . . . . . . . . . . . . . . 186
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Appendix A: Relativistic and Probabilistic Functions 189
Appendix B: Manual for Successful Crusade in Defense of
Patients’ Rights 193
Afterword 197
Final Remark 201
Bibliography 203
Index 213
Preface
The current book is a product of three authors: the _rst author (AP) is a mathematician, the second author (OS) is a statistician, and the third author (LBP) is a pharmacologist. As a result, the book is slightly eclectic, since di_erent parts represent views and experiences of di_erent people. The main subject of the book is application of mathematical and statistical skills in modern clinical drug development (mostly by biostatisticians). Most commonly, the development of new drugs is carried out by pharmaceutical or biotechnology companies. The whole process is a long, expensive, and complex enterprise, which should be performed keeping in mind such aspects as medical ethics, reproducibility and reliability of the results, e_ciency in decision making (to maximize return-on-investment), and, of course, compliance with the regulatory standards.
Biostatisticians play a very important role in the drug development process|this is even documented in the regulatory guidelines. However, what does it really mean to be a biostatistician? Job descriptions may vary depending on the place of work, seniority level, etc. For instance, a biostatistican in a small one-drug company can be charged with numerous tasks, including data management, statistical programming (including data cleaning, analysis, and reporting), and biostatistics (trial design, submissions, publications, etc.) In a big pharma company, biostatistician may be focused more on biostatistics itself, while supporting a much larger number of studies and projects within the company portfolio. In contract research organizations (CROs), the work may be tailored to support the client’s (big pharma) speci_c requests, which can be very broad. One can also mention academia and regulatory agencies, where biostatisticians play major roles as well, and their job descriptions are di_erent from the ones in the biopharmaceutical sector.
Regardless of the place of work and assigned duties, a biostatistician typically has background in mathematics/statistics/biostatistics/computer science/ data science/etc., and he or she has to solve various problems that arise in the context of drug development. Importantly, biostatisticians do not operate in a vacuum”|they have to collaborate with many stakeholders including medical doctors, pharmacologists, clinical scientists, regulatory scientists, etc. As such, an important attribute for a biostatistician (in addition to technical background) is statistical consulting skills. In addition, the biopharmaceutical industry has a distinct feature|regulatory health requirements to ensure that research and development are carried out in full compliance with medixi cal ethics, technical standards, and quality, to protect patients (present and future).
In this book, we describe a philosophy of problem solving based on a system of principles for pragmatic problem solving, speci_cally in the context of clinical drug development. The examples of applications vary from selection of necessary toolkits (Chapters 2 and 4) to providing help to a struggling neighbor (Chapter 3); from a general evaluation of the safety outcome of a randomized controlled trial (Chapter 5) to planning of open-ended projects (Chapter 7); from an attempt to salvage a failed clinical development program to some generalizations of encountered problems to the entire _eld and even entire drug development (Chapter 6). These ideas stem from the authors’ many years of work in the biopharmaceutical industry.