The Significance of ‘No Significant Difference’

Abstract

It is generally accepted that patients experience less discomfort with low osmolality contrast media (LOM) than with high osmolar media (HOM). Hard statistical facts from so called ‘high quality’ controlled trials, proving that more significant reactions such as vomiting, hives, urticaria or anaphylactic complications also are less common with LOM are, however, not readily available (3). One reason for this may be that most of the well designed controlled studies performed may have been tailored by the drug manufacturer for a specific purpose: to fulfil the format requirements for registration by the licensing governmental authorities. For this the sponsor, to save time, usually engages several medical centres, each only performing 15 to 60 studies (4). Materials of such a size are of course much too small to reveal any change in the frequency of a complication occurring with an incidence of only a few per cent or less. The absence of a statistically significant difference in such low incidences of complications does not justify any conclusion. The question then arises: how big a material would be needed to obtain a fair chance to statistically verify a clinically highly important decrease in the incidence of a complication from, for instance, 10 to 5 per cent? This paper deals with such questions.