Spin and bias: the tip of the iceberg.

Quantitative PCR analysis for Bcl-2/IgH in a phase III study of Yttrium-90 ibritumomab tiuxetan as consolidation of first remission in patients with follicular lymphoma. combination with radiation primes durable antitumor immune responses in mouse models of lymphoma. Spin and bias: the tip of the iceberg Vera-Badillo et al. [1] have to be congratulated for uncovering a problem with reporting of trial results, a problem that distorts the record, impedes inference, and can lead to flawed medical decisions. It is important to note that the associated problems of biased reporting and spin remain problems no matter how often they occur; once is once too often. Hence, the proportions presented should be understood in this context, and we should not lose sight of the forest for the trees. Nevertheless, I would propose that the proportions presented should have been based on the subset of 30 trials with complete information (including the primary end point that was prospectively specified), either instead of or in addition to the proportions that were presented based on the full sample of 164 trials. We are told, for example, that 7 of the 30 trials reported a different primary end point, and we can only imagine that this switch would have occurred in a higher proportion of the other 134 trials, given that the authors of those trials did not need to worry about getting caught. This uncertainty undermines our ability to offer credible statements about how often these 134 trials were biased or spun. But we do see bias in at least 7 of the 30, and have to wonder how often there was bias and/or spin in the other 23 among these 30. This would be a rather telling statistic, arguably more telling than the one based on the full set of 164. Moreover, the problems reported, though certainly of substantial importance, may still represent only the tip of the iceberg, and not only because we can check for changes in the primary end point in only 30 of the 164 trials. Beyond this, we also have the fact that there are many more biases, far too many to list here, that may have gone into producing spurious statistical significance in the primary end point. So what we have here is a best-case scenario, and a comprehensive look at spin, plus bias as defined here, plus bias defined more broadly, might reveal a far worse state …