Can we believe what we read?


During my mostly enjoyable 14 years as editor of a mainstream surgical publication, one of my less enjoyable but necessary functions was to detect and police author misconduct. Most instances of wrongdoing involved duplicate or redundant publication or the awarding of authorship when it was not deserved. I considered these infractions to be misdemeanors. The more serious sins of plagiarism and fabrication or falsification of data were considerably less common or at least infrequently detected. Based on my experience, one might conclude that most scientific findings are reported accurately and with integrity. In other words, we should be able to believe what we read. Unfortunately this may not be the case.

Compared to prior years, the past decade has seen a 10-fold increase in the number of articles retracted from the scientific literature. Although retractions account for well less than 1% of published articles, it is disturbing that fraudulent research and publishing rather than inadvertent errors underlie up to two-thirds of cases. Of note is that there appears to be a direct correlation between a journal\'s impact factor and the number of articles that are retracted from it. This may be due to a closer scrutiny by the scientific community of what initially were thought to be seminal contributions. Thus the limited number of retracted articles may, and probably does, represent the tip of an iceberg.

Deliberate fraud is an important issue in academic publishing, but an even greater problem is the abundance of poorly designed studies and the misuse of data. A significant fraction of the information available to us for clinical decision-making are underpowered studies with type I or II statistical errors, biased analyses, articles with data that cannot be reproduced by others, inappropriately done meta-analyses based on heterogeneous rather than homogeneous trials, and inaccurate conclusions based on erroneous manipulation of data.

In his analysis of the scientific literature in 2005, John Ioannidis, M.D., Ph.D. (PLoS 2:696-701) postulated that the majority of published articles were inaccurate. Flawed studies were more likely to result when sample sizes were small, the effect between tested variables was small, the number of tested relationships was large (at a P < .05, 5% will be statistically significant by chance), and conflicts of interest were present. Unsound studies were more likely to occur when they took place within popular and highly competitive scientific fields where the timing of publication was of the essence in order to claim primacy.

Compounding this cascade of misleading information is the reluctance of journals to publish negative studies, thereby giving greater weight to positive studies testing the same hypotheses.

These revelations regarding our research enterprise are not widely appreciated. In fact, in surveys of the public who are responsible for funding much of it with their tax dollars, biomedical research occupies an exalted position in comparison to most other endeavors. Why in recent times has it become tainted and what can be done to reverse the present trend? First, and probably most importantly, scientific research has become a highly competitive game. The battle for available research faculty positions in our universities is more intense than ever with room for only one of every six Ph.D. graduates. Once an appointment is attained, promotion and tenure are dependent on obtaining federal funding from an increasingly shrinking pool of money and publishing in high impact journals, many of which have rejection rates in excess of 90%. It is not surprising that minor or even major massaging of data to reach the magic 0.05 P value is probably not uncommon. Surveys of scientists regarding misconduct indicate that up to 20% have either participated in questionable practices themselves or know of colleagues who have.

The onus is not only on the researcher. Journal peer reviewers and editors are exerting much of their effort in looking for the rare seminal paper that would be attractive to their readers. Less attention is paid to the details of scientific rigor. Negative studies and those that are only confirmatory of previously published investigations are generally not given high enough grades to reach the threshold for publication.

Science has and continues to contribute much to the quality of human life. Most research scientists in the academic world operate with integrity and make a sincere effort to uncover truth in their fields. Marginally done and underpowered studies are the culprits leading to misinformation much more often than are issues of scientific honesty.

So what is to be done to right a somewhat listing research enterprise? From the researcher’s perspective, more attention to study design and rigid adherence to it to avoid bias is essential. Pre-study statistical consultation, especially regarding power calculation, is also a key to obtaining reliable results and conclusions. The editorial boards and editors of journals need to be more accepting of negative and confirmatory analyses than they have been in the past. The recent stance of most reputable journals to require registration of all clinical trials and to provide the data from those trials, positive or negative, to one of several web based repositories for review by others is a step in the right direction. It is the responsibility of our academic institutions to detect and police poor research design and implementation in addition to outright academic misconduct and, when necessary, to change the culture within their research establishments. Finally, it behooves us as readers and consumers of new information to realize that science marches forward in only small incremental steps. Important new findings need to be confirmed before they are adopted. The adage of never being the first to accept the new or the last to abandon the old still conveys a great deal of wisdom.

Next Article:

   Comments ()