The Relative Merits of Observational and Experimental Research: Four Key Principles for Optimising Observational Research Designs

The main barrier to the publication of observational research is a perceived inferiority to randomised designs with regard to the reliability of their conclusions. This commentary addresses this issue and makes a set of recommendations. It analyses the issue of research reliability in detail and fully describes the three sources of research unreliability (certainty, risk and uncertainty). Two of these (certainty and uncertainty) are not adequately addressed in most research texts. It establishes that randomised designs are vulnerable as observation studies to these two sources of unreliability, and are therefore not automatically superior to observational research in all research situations. Two key principles for reducing research unreliability are taken from R.A. Fisher's early work on agricultural research. These principles and their application are described in detail. The principles are then developed into four key principles that observational researchers should follow when they are designing observational research exercises in nutrition. It notes that there is an optimal sample size for any particular research exercise that should not be exceeded. It concludes that best practice in observational research is to replicate this optimal sized observational exercise multiple times in order to establish reliability and credibility.