1.3.1: The Pharmaceutical Industry

The pharmaceutical industry has a long history of the use of statistics and data science, and the importance of making data-based decisions in drug development is becoming more important each year. From assessing the effectiveness of proposed treatments for diseases to the development of marketing plans for new treatments, nearly every step of the process that a drug or treatment takes from discovery to final deployment is based on decisions made by using data.

The development of new and innovative medical treatments is a time-consuming and costly process that requires very carefully designed medical experiments. Treatment development often begins with preliminary trials based on either human or nonhuman subjects, depending on the conditional being treated and the type of treatment being administered. The main challenge is how to identify what treatments have potential based on these preliminary trials. The stakes are high as moving a possible treatment from the preliminary stage to later stages is expensive and time-consuming. There is a great deal of risk involved with spending too much time and resources on a treatment that is ultimately not useful or has side effects that outweigh the potential benefits of the treatment.

As a potential treatment moves through the process, the stakes become higher, and the complexity and costs of the experiments increase. Moving a treatment to human trials is a serious proposition. At this point there must be some evidence that the treatment has some potential benefit and is not harmful to the participants. This is particularly true when treating a serious disease for which a standard treatment exists. The decision to give someone a new treatment that may not work as well as a standard treatment for a condition that is life-threatening has serious ethical implications. In the final phase, when a treatment is ready to be released to the public, studies of possible side effects from the treatment provide practitioners and patients with important information that they need to know when considering using the treatment for their condition.

Statisticians and scientists have been working on methodology for performing these experiments for years. Specialized methods are used to determine what treatments might be useful early on, and what conditions are required for a potential treatment to proceed to the next phase of testing. All this methodology is based on data observed from experiments, where the important issue is how much evidence must the data provide before a decision can be made on whether to continue studying a treatment. One of the most important parts of this methodology is the assessment of risk. Of course, there is potential risk in moving an ineffective treatment to the next phase in terms of costs both to human lives and company resources. There is also risk involved with not identifying an effective treatment in terms of lost revenue to the company, and the obvious loss to society of an effective medical treatment.

As the methodology of statistics is developed in this book, the concept of risk is an important issue. Statistical methodologies are not useful solely because they can be used to identify trends and structure within data—they are useful because they include an assessment of risk. In a statistical decision-making process, the practitioner can specify the amount of risk they are willing to tolerate when making certain decisions.