臨床研究で必要となる統計学

抄録

Understanding how to verify a hypothesis is a prerequisite for engaging in high–quality clinical research. Unless this is understood, no amount of knowledge regarding experimental design and methods of data analysis by itself will enable the selection of the appropriate types of design and analysis to verify the hypothesis to be tested.

Establishing the appropriate primary endpoint, control group, and sample size is important for proper hypothesis verification in clinical research. A deficiency in any of these makes the interpretation of experimental results problematic and means that the hypothesis cannot be verified. I will outline the basic concepts pertaining to these three matters.

Once these have been determined, the next step is to decide methods for subject enrollment and data collection. For a randomized comparative trial, the method for randomization must be decided. The appropriate method for the study concerned must be chosen in light of the number of subjects and adjustment factors for allocation. In particular, guaranteeing comparability between the experimental groups in a randomized comparative trial is extremely important for guaranteeing the validity of the study results.

In terms of data collection, attention must also be paid to ensuring that data are collected in a manner that is appropriate for the data analysis methods used. In retrospective studies, on the other hand, the analysis population consists of non–randomized patients, and the areas requiring particular care or adjustment are different from those that concern randomized comparative studies. For example, it may be necessary to adjust for confounders and other patient background factors in data analysis. I will explain the different methods used for verifying hypotheses in randomized comparative trials and retrospective studies and describe the data analysis methods appropriate to each.

Given the irregularities seen in some clinical studies in recent years, there is a need to ensure that the quality of clinical research at least meets certain standards, as evidenced by the guidelines on quality assurance for clinical studies (including monitoring and auditing) set out by the Ministry of Health, Labour and Welfare, even before it reaches the stage of a clinical trial. Initiatives to improve the quality of clinical studies are now underway in a variety of fields, including consideration by the Biometric Society of Japan, the professional association for Japanese biostatisticians, of the introduction of a certification system for trial statisticians who take part in general clinical research. Properly guaranteeing the quality of clinical studies will require the cooperation of professionals such as biostatisticians and data managers in addition to that of doctors, pharmacists, nurses, clinical research coordinators, and other medical professionals engaged in research.