1
Introduction

Ye maun understand I found my remarks on figures, whilk . is the only true demonstrable root of human knowledge.

Sir Walter Scott, Rob Roy

Statisticians know that words are important to statistics, yet surely their importance is not fully recognized.

William Kruskal

Opinions are made to be changed - or how is truth to be got at? We don't arrive at it by standing on one leg.

Lord Byron, letter to Murray

1.1 DRUG DEVELOPMENT

Drug development is the process not only of finding and producing therapeutically useful pharmaceuticals and turning them into high-quality formulations of usable, effective and safe medicines, but also of delivering valuable, reliable, and trustworthy information about appropriate doses and dosing intervals and about likely effects and side-effects of these treatments. Drug development is a process carried out by sponsors (mainly pharmaceutical companies) and its acceptability is ultimately judged by regulators. It is an extremely complex business and the risks are high, but the potential rewards are also considerable.

It takes many years for a project to reach development. First, basic research must be undertaken to validate concepts and mechanisms. Assessments of commercial potential for diseases and therapies are also needed and these will continue throughout the life of a project. Next, a lead compound must be identified for a particular indication. This will then be subjected to a battery of screening tests to assess its potential in terms of therapeutic activity. Back-up compounds will also be investigated. If a compound looks promising, it will also be evaluated from both safety and practical points of view. Will it be easy to formulate? How many steps are involved in the synthesis? How difficult will it be to manufacture in large-scale quantities? Before a treatment can go into development, not only must satisfactory answers have been obtained to all these questions but a viable pharmaceutical formulation permitting further study must be available. This can be an extremely delicate matter, involving work to develop suitable solutions, pills, patches or aerosols, as the case may be.

If and when a molecule is accepted into development, animal studies will be undertaken in order to check safety and to establish a dose at which studies in humans may be undertaken. Once basic toxicological work has been undertaken, 'phase I' may begin and the first such studies may start. These will be single-dose studies in which lower doses are tried first and cautiously increased until a maximum tolerated dose may be established. In many indications such studies are carried out on healthy volunteers, but where the treatment is highly aggressive (and hence intended for serious diseases) patients will be used instead. In the meantime, longer-scale toxicological studies with animals will have been completed. Pharmacokinetic studies in humans will be undertaken in which the concentration-time profile of the drug in blood will be measured at frequent intervals in order to establish the rate at which the drug is absorbed and eliminated. These studies together, if successful, will permit multiple-dose studies to be undertaken.

Once maximum tolerated doses have been established, phase II begins and dose finding studies in patients are started. This is usually an extremely difficult phase of development but, if the drug proves acceptable, the object is that preliminary indications of efficacy should be available and that a firm recommendation for doses and dose schedules should emerge. Once these studies have been completed, the pivotal phase III studies can begin. These have the object of proving efficacy to a sceptical regulator and also of obtaining information on the safety and tolerability of the treatment.

A successfully completed development programme results in a dossier - an enormous collection of clinical trial and other reports, as well as expert summaries covering not only the clinical studies as regards efficacy and safety but also pre-clinical studies and other technical reports as well as details of the manufacturing process. If successful, the package leads to registration, but even during the review process, phase IV studies may have been initiated in order to discover more about the effect of the treatment in specialist subpopulations, or perhaps with the object of providing data to cover price negotiations with reimbursers.

Regulatory dossier: A mountain of documents which takes a forest of trees to obscure the wood.

Once a drug has been launched on the market, the process of monitoring and 'pharmacovigilance' begins in earnest, since the drug will now be used by far more persons than was ever the case in the clinical trials in phases I to III, and rare side-effects, which could not be detected earlier, may now appear. Some further phase IV postmarketing studies may be initiated and further work extending indications or preparing new formulations may be undertaken.

1.2 THE ROLE OF STATISTICS IN DRUG DEVELOPMENT

There is no aspect of drug development in which statistics cannot intrude: from screening chemicals for activity to forecasting sales. Because the efficacy and safety of treatments has to be judged against a background of considerable biological variability, all of the judgements of efficacy boil down in the end to a numerical summary of evidence whose message can only be understood with the help of the science of statistics. It is the norm that clinical trials are planned jointly by a statistician and a physician. Statisticians have also slowly been becoming more active in the shaping of projects as a whole. Furthermore, whereas in the past in Europe, the expert reports for a dossier would largely be a subjective qualitative assessment of evidence from individual trials, regulators increasingly expect to see quantitative summaries, so called meta-analyses. Hence statistics has increased its empire throughout the drug development process during the time since the first edition of this book.

Other quantitatively minded disciplines have also made important contributions. There is no question in my mind that statistics had benefitted and would benefit more if more attention was paid to impressive work in pharmacometric modelling, in particular the so-called 'population approach', due to Lewis Sheiner (1940-2004) and various colleagues. This work, using as it did, nonlinear mixed effects models as long ago as the 1970s, was ahead of its time. See (Senn, S.J. 2010) for an appreciation. Recently, there has been much interest in 'big data' and 'analytics' and many pharmaceutical companies have set up departments with or without statisticians involved to cover this new field. This represents both a threat and an opportunity to statisticians. The threat is that statistical lessons will be ignored. The opportunity is to learn from and contribute to the work of others with different perspectives. As an example of the value that statisticians can bring to this field, I would cite particularly the work led by various statisticians at Novartis on using historical data (Schmidli, H. et al. 2014). Since all data that do not use concurrent control are historical, this has profound implications for many analyses.

Decades before the first edition of this book (1997), the Food and Drugs Administration (FDA) in the United States was employing statisticians to assist in its review process, and by the second edition (2007) various national European regulatory agencies had followed suit, although the European Medicines Agency (EMA) had not done so. Well before the third edition, however, the EMA had caught up and, indeed, Olivier Collignon, a member of my group in Luxembourg, the Competence Centre for Methodology and Statistics of the Luxembourg Institute of Health, which I joined in 2011, was seconded to the EMA for four years (2014-2018) to join what was by then an active group of statisticians. The involvement of statisticians in drug regulation in Europe is a very welcome change.

Statistical issues have long been covered by FDA general guidelines and starting in the 1990s separate specific statistical guidelines produced (with the help of European national statisticians) by the Committee for Proprietary Medicinal Products (CPMP) in Europe started to appear, and also by the Japanese Ministry of Health and Welfare. The International Conference on Harmonisation (ICH) was set up to produce general guidelines that would be valid in all three of these regions and these were in turn influential on other countries, outside the regions concerned, that were developing guidelines of their own. There are also many national and international societies specifically for statisticians in the pharmaceutical industry. The exact duties of the statistician working in drug development are covered in some detail in Chapter 5.

1.3 THE OBJECT OF THIS BOOK

As implied by our chapter quotations, this book is about figures, words, and opinions; more precisely, it is an attempt to give a largely verbal account of the various opinions that are held by those who figure. The purpose is not to present an authoritative prescription as to how to deal with each particular application of statistics in...