In May 2003, the Committee for Proprietary Medicinal Products (CPMP) of the European Agency for the Evaluation of Medicinal Products (EMEA) issued a guidance regarding the adjustment for covariates entitled “Points to Consider on Adjustment for Baseline Covariates”. This article provides a summary of the guidance and specifically discusses the use of dynamic allocation from the viewpoint that advocates the necessity of pre-stratification in clinical trials.
In randomized clinical trials, continuous outcome measures are often used as endpoints. For qualitative covariates associated with the primary endpoint, stratified randomization is frequently used in order to ensure balance between treatment groups. Quantitative covariates associated with the primary endpoint are also frequently prespecified. “Points to Consider on Adjustment for Baseline Covariates” (2003) strongly recommends analysis of covariance (ANCOVA) as a primary analysis for such trials to improve the precision and compensate for imbalance. This paper also recommends ANCOVA, but also accepts the two sample t-test under a small loss of precision and unlikely imbalances of covariates. Interactions between treatment group and covariates will not be generally included in the primary analysis model. Even if they exist, differences between least square means estimated by the ANCOVA model without interactions and differences between arithmetic means of treatment groups are equal in expectation under several assumptions. Therefore, it can be said the ANCOVA without interactions is better than the usual t-test in precision. An example of the preferred way to show ANCOVA results is given. Also the interpretation of least square mean is briefly discussed.
In clinical trials, missing data often happens for a variety of reasons, such as dropouts, making it difficult to analyze the primary variable measured longitudinally and to interpret the results of the primary analysis. While handling missing data sometimes causes bias in the results, there have been no established statistical approaches applied to missing data in appropriate situations. In November 2001, Committee for Proprietary Medical Product of the European Medicines Agency, issued “Points to Consider (PtC) on Missing Data”, which focuses on several points that should be taken into account when handling missing data in clinical trials. In this paper, we review the contents of this PtC, which assumes that the primary analysis is based on the ITT principle, and discuss some of the approaches for handling missing data and the difficulties in interpreting these results.
In any clinical trial or epidemiological study , missing data will occur. The problem in dealing with missing data are discussed from both clinical and statistical view points. Clinical approaches (e.g. LOCF) are easy to understand for physicians and are often appropriate for considering clinical background settings. But they seem to be less objective compared with statistical approaches , including multiple imputation based on “propensity score”. Further efforts should be made to bridge these two approaches in order to deal with missing data appropriately.
The “Point to Consider (PtC) on Missing Data” issued by EMEA implicitly postulates strict intention-to-treat analyses and discusses various methods for imputation of missing data, including parametric imputation methods. Imputation may not be relevant to situations after the occurrence of particular events such as a terminating event like death or clinical events that can largely alter the value of the response variable. Possibility of non-ignorable missing mechanisms when adopting parametric methods with ignorable assumptions is also discussed. Lastly, we stress the need for greater efforts in designing and monitoring clinical trials to prevent missing observations.
“Points to Consider on Multiplicity Issues in Clinical Trials” published by CPMP of EMEA in September 2003 is reviewed and issues to be discussed are identified. First, the examples identified in the PtC as unnecessary cases of “adjusting the type I error level” are grouped into 3 patterns, and the situation and the characteristics of each pattern are explained. After reviewing each pattern, important points for conducting the confirmatory trial according to PtC are summarized. Issues to be discussed about the content and the concrete measures are also addressed.
We would like to provide ideas about multiplicity from the practical point of view of a statistician in a pharmaceutical company. We think that a trial statistician should satisfy not only the requirements, described in ICH-E9, but should also be responsible for explaining statistical analyses, which may be complicated, to non-statisticians in simple language. In addition, certain multiple comparison methods that have been recently developed are hard to understand even for statisticians dealing with clinical trials. When these methods are applied to clinical trials, it is almost impossible to understand the essence of the multiple comparisons and interpret the results correctly. Therefore it is essential to plan a simple confirmatory clinical trial that can be interpreted clearly.
In the development and regulatory review of pharmaceuticals, multiplicity is one of the important issues, and an adequate treatment of multiplicity is required by regulatory guidelines. Though the importance of statistically valid adjustment of multiplicity in drug development is nowadays well understood by practitioners, the more is required in the view of government agencies responsible for drug evaluation. To use the drug effectively and safely in clinical practice, not only the validity of the method itself, but also the validity of the formulation of clinical questions and the adequacy of the interpretation of results are required. In this article, two examples of new drug application review results, one by the US FDA and the other by the Japanese regulatory agency, are reviewed and discussed.
This paper introduces the contents of two documents from the European authority, “Points to Consider on Switching between Superiority and Non-inferiority” and “Guideline on the Choice of the Non-inferiority Margin”. Issues relevant to possible applications in Japanese clinical trials are also discussed.
Two guidelines related to non-inferiority trials were released by the European Medicines Agency (EMEA). One pertains to the choice of non-inferiority margin and the other to switching the objective of a trial between superiority and non-inferiority. These guidelines describe principles for appropriately designing and interpreting non-inferior trials, but it is very difficult to follow these principles in an actual clinical trial setting. In this article, recommendations in interpreting these guidelines are discussed from a sponsor statisticians' point of view. Some issues of non-inferior trials not mentioned in these guidelines are also discussed.
This short note discusses statistical issues in the appropriate design of randomized controlled trials with regards to the recent two documents, “Points to Consider on Switching between Superiority and Non-inferiority” and “Guideline on the Choice of the Non-inferiority Margin” from the European Agency for the Evaluation of Medicinal Products. This paper also points out the inappropriateness of the terminology of “superiority”defined in ICH E9 (Statistical Principles for Clinical Trials) and discusses its relationship with these matters.