Introduction

Article Outline

• Reference
• Copyright

This special issue focuses on strengths and weaknesses of conducting secondary analyses in clinical trials and other studies. A discussion of the analytic approaches and limitations of such analyses are presented. Secondary analyses are defined as any analyses that are not concerned with the primary outcome(s) and the primary hypothesis on which the study sample size is based.

The initial article in this special issue (Furberg and Friedman) presents general thoughts on secondary analyses and on the need to use the intent-to-treat principle (“as randomized so analyzed”¹) to guide primary trial analyses when comparing intervention outcomes. Furberg and Friedman also stress the importance of collaborating with those who conducted the primary study. This might include both discussing ideas for the analyses as well as asking for a review of the manuscript or report before it is finalized. The second article by Marler addresses in depth the limitations of secondary analyses that need to be considered in their conduct and reporting. Marler also stresses caution in making treatment decisions based on secondary analyses alone.

In addition, the supplement focuses on more advanced statistical concepts including

A case study (Dickstein) is also included to provide a guide and lessons learned for those embarking on secondary analyses. This special issue ends with a novel perspective. Most secondary analyses focus on using data from clinical trials to address questions that involve randomized comparisons in subsets of the trial cohort. Howard and Howard consider the usefulness of clinical trials for more traditional epidemiologic studies.

Many of the articles address the issue of sample size directly or indirectly. Studies are designed for the primary outcome. Thus, sample size and issues of power can be critical in secondary analyses where the lack of findings may be attributable to either no effect or small sample sizes. Of course, if the secondary analyses demonstrate statistically significant and clinically meaningful, interpretable findings, they have to be considered as a starting point for further replication, either in a new study or from other data.

The supplement is not meant to supplant a basic course in biostatistics or to be a primer on how to develop sample size or power estimates. The reader is referred to standard texts on the subject such as the book Fundamentals of Biostatistics by Rosner, 7th edition, Brooks/Cole CENGAGE Learning, Boston, 2011.

It is our hope that new clinical investigators with an interest in cardiovascular disease will find this special issue particularly informative and that more experienced investigators may find some new ideas to enhance their future analyses and interpretation of secondary data.

Back to Article Outline

Reference 

1. Olsen J, Kaare C, Murray J, et al. An introduction to epidemiology for health professionals. New York: Springer; 2010;
   • View In Article