Abstract

While it is given that some or all of the biases involved in the design and the implementation of a trial are carried to the reporting, our concern in this talk is a list of biases more specific for this stage. I propose that the common denominator in this list is the “urge to prove” (1). One important outcome of this urge is the relative reluctance of the authors to acknowledge the limitations of their work. It is interesting to note that this reluctance is more pronounced in basic science (2) while it has even been said that the main purpose of the discussion sections should be authors discussing limitations of their work, naturally best known to them (3). Another manifestation of the same urge has to do with the pervasive mistaken or ill-use of the p values. As yet again brought up recently (4), the medical literature, for decades, has been replete with articles about this misconception. Physicians and investigators, with no apparent differences between the clinicians and the basic scientists, have singularly been reluctant to correctly understand and use the p value. The main elements of the said misconception seem to be a. Inadequacy in formal probabilistic thinking, the most common being that a p value represents the probability of our trial results being false positive while it actually represents the probability of obtaining the actual trial results had it been they were only due to chance; b. The power of a test is not only important in avoiding a Type II but also a Type I error; c. The information we get from any one p value much depends on the pre-test probabilities of what we try to measure in the study design in which we do this, i.e., an observational or a cross sectional study versus a placebo controlled randomized clinical trial (5). Publication bias traditionally referred to as the athors' and journal editors' propensity to publish positive or otherwise striking results. Recently this has been re-considered (6) and the proposal has been that a myriad of other biases, perhaps even including almost all items related to the “urge to prove” might also be included under this broader definition of a publication bias, and again perhaps now better called a 'dissemination bias'. Although much discussed this bias remains rather difficult to accurately identify and quantitate. Finally it has to be remembered, like all biases, the biases involved when reporting a trial can either be unintentional or intentional and this is equally difficult to differentiate and quantify.