Article Text
Abstract
Objectives Before proved by evidence, plausible hypotheses are more likely to be false than true. Therefore, one should expect more true negative studies than true positive studies. However, mechanisms such as imprecision, bias in study design, outcome reporting bias, spin and publication bias fabricates a higher prevalence of positive studies. Scientific integrity constitutes a series of mechanisms to prevent such bias towards positive results. Based on this rational, a higher number of negative studies in comparison with positive studies indicates scientific integrity of an author, laboratory, scientific field or journal.
To propose the ‘index of integrity’ and use it to describe scientific integrity among first authors of papers published in major journals.
Method We selected the first authors of any original article published during the first month of 2019 in three major medical journals: Journal of American Medical Association, The Lancet and British Medical Journal. Lifetime articles published by these individuals as first authors in Pubmed indexed journals were selected and defined as positive or negative according to abstract conclusion.
A positive article was defined as those with (1) a positive conclusion based on the primary analysis or (2) a negative primary conclusion spun into positive by emphasis on secondary positive results. Otherwise, the article was defined as negative. The index of integrity was calculated by the ratio between negative articles (numerator) and positive articles (denominator). An index > 1 is desired as a marker of scientific integrity.
Results A total of 27 authors were selected, with a median of 5 articles (interquartile range = 1 – 7, minimum of 1 and maximal of 55) published during lifetime in Pubmed indexed journals. The prevalence of positive articles was 87% (95% CI = 82% – 91%) and the utilization of spin to positivize the conclusion was present in 7.9% of positive articles.
Sixteen authors (59%) had no negative studies and leading to an index of integrity equal to zero. Only 4 authors (15%) had an index higher than 1. The average estimation for the index of integrity was 0.40 (95% CI = 0.10 - 0.70). There was no association between the index and total number of publication or the practice of spin.
Conclusions In this preliminary report, an especially higher number of positive articles compared with negative articles suggests undesired level of scientific integrity among first authors of major publications in medical science. The index of integrity, first utilized in this report, should be validated in larger samples of authors.