Several decades ago, statisticians who viewed the field of medical trials saw that it was a field that used imprecise and often inappropriate methods for designing trials, analyzing data, and interpreting results. They proposed new and modified methodologies that were implemented by the clinicians to raise standards of clinical study designs. Their contributions to this field are extremely important and form a major part of the foundation of modern clinical trials.

It is a basic principle that statisticians can statistically interpret clinical data. That is a clear and unarguable point. However, some statisticians suggest that their role in clinical trials includes the clinical interpretation of clinical data as well as its analysis and statistical interpretation. One possible reason that some statisticians believe they are qualified to provide clinical interpretations of data appears to be their acknowledged expertise with numbers and their ability to analyze numbers in many ways, combined with the fact that clinical interpretations are based primarily on the clinical data analyses that the statistician has made. However, the author does not believe that most statisticians are prepared by training or by experience to clinically interpret clinical data, even though they may interpret their statistical analyses of clinical data from a statistical perspective.

The clinical interpretation of data by statisticians has been claimed to be part of their role by some statisticians and is found in various books and chapters on clinical trial methods written by statisticians. For example, Moses (1985) defined the field of statistics as referring “to the aspects of interpreting quantitative data that tend to be independent of the specific data at hand and thus to carry over from one problem to another.” A.A. Nelson (1980) stated, “Statistics are the tools with which one analyzes and interprets research data to develop scientific theories.” Although it is possible that these authors were referring to only statistical interpretations, the context of their comments in clinical journals suggests that they included the possibility of statisticians making clinical interpretations as well. A scientifically and medically accurate and appropriate approach to the process of data analysis and interpretation is described in this chapter.

The two processes of statistical analysis/interpretation and clinical interpretation of data from a clinical trial are separate functions, although both processes are related, and both are required to fully understand the results (Fig. 78.1). The primary role of statisticians is to both help design a clinical trial and to analyze and statistically interpret the data after a trial is completed and the data have been entered into computers and presented in tables, figures, and listings that the statistician has previously determined would be created [in the statistical analysis plan (SAP)]. Most statisticians prefer to have the clinician involved with this SAP or, at the minimum, to review it after it is in draft form. It is the role of trained and experienced clinicians to develop a medical interpretation of the data. This usually requires using a statistical report of the analyses, assuming that one has been written. The clinical interpretation is largely based on the statistical analyses and to a variable degree on other analyses or data (e.g., pathology reports, toxicology reports, analytical chemistry reports). There are exceptions to this principle if, for example, a proof-of-concept trial (e.g., a pilot trial) is conducted using a much smaller number of patients than the number required to have 80% or greater power of finding an effect if it is present or, alternatively, if no power analysis has been done and clinical judgment was used to decide on the number of patients to include in the trial. In those situations, the clinical interpretation of the data will not be based on a statistical analysis of the results and none may have been conducted. (Nonetheless, the statistician may still make a contribution to understanding what the data mean.) The number of patients might have been chosen based on the number believed by the clinician(s) involved with the
trial’s design to provide some assurances of finding a clinically significant effect if one was present or to answer another question. This approach is based on the belief that good data in a well-designed and well-controlled clinical trial are better than a much larger amount of mediocre data from a larger number of patients in a poorly designed trial (e.g., open label or single blind).

Statisticians are generally asked to review and critique clinical interpretations. Statisticians commonly suggest ideas or modifications for the clinician(s) to evaluate. When data and results of a trial are extrapolated to new patient groups or new clinical hypotheses are created, the processes of extrapolation and hypothesis generation are also primarily roles of clinicians.