Erratum: Expression of concern regarding âIndividual human scent as a forensic identifier using mantrailingâ by Woidtke L, DreÃler J and Babian C. Forensic Sci Int. 2018 Jan;282:111â121 (Forensic Science International (2018) 282 (111â121), (S0379073817304796), (10.1016/j.forsciint.2017.11.021))
Jackowski, C; Cattaneo, C; Broccard, A; Duembgen, L
With respect to the publication âIndividual human scent as a forensic identifier using mantrailingâ authored by Woidtke L, DreÃler J and Babian C. and published in Forensic Sci Int. 2018 Jan;282:111â121 [1] the Editors-in-Chief of Forensic Science International are issuing the following expression of concern to our readership, alongside the statisticians who examined this manuscript on behalf of FSI. The above-mentioned publication was submitted to FSI in February 2017 and underwent two rounds of double-blind review by two experts in the field prior to its acceptance and online publication in November 2017. After publication, the scientific community raised a series of concerns related to the publication to the Editors, which led to two published comments and a published response from the authors in 2019 [2â4]. In 2020, the Editors received another letter regarding the initial publication, which pointed to an issue within the statistics that had been undetected so far [5]. FSI then initiated an independent re-evaluation of the statistics in the article. Within the publication, the choice of negative samples is described as follows âAt each run, the dog handler randomly chose one out of three identically looking ziplock bags, containing either the scent article of runner A or runner B, or the scent article of an absent person (negative).â Based on this approach the number of trails with negative samples should roughly meet one third of all trials. However, out of all 675 trails only 158 (23.4%) negative samples were chosen. Statistically this is an extremely unlikely event since the probability of getting not more than 158 negatives out of 675 random decision processes as described within the publication is 1 in 88.4 million. These numbers may raise the suspicion either that the decision process did not occur in the random manner as described or that a relevant number of negative trails has been excluded from the statistical evaluation of the results. The authors were asked for an explanation for the low number of negative trails as well as their approval to use the existing study documentation (the video documentation of each trail) for an independent re-evaluation. Regarding the low number of negatives the authors referred to the methods section of the publication again as cited above and in a second response they suggested a color coding issue possibly affecting the choice of the negative samples, which they finally assessed as ââ¦unlikely expression of a stochastic process that follows a uniform distribution. A possible cause could be an uneven, non-random distribution of the colour marking of the respective sample (negative samples were marked green or blue less frequently and red more frequently)â [6]. Our own analyses of the numbers via multiple testing revealed indeed, that the distribution of all 3 à 3 = 9 possible combinations of color (red, green, blue) and target (negative, runner A, runner B) deviates significantly from a uniform distribution. The authors also suggest that samples with red color might have been avoided by the dog handlers. Following up on that potential explanation, we analyzed the frequency of negative samples among samples with a green or blue dot only. Again, significant deviations from a uniform distribution crystallized, with negative samples under- and green samples overrepresented. Given these uncertainties in the experimental design, it is unclear which null hypotheses should be tested. It should also be noted that a highly significant deviation from the null hypothesis of a ârandom decisionâ by the dogs (decision meaning to follow runner A, to follow runner B or to indicate a negative), does not imply that in a particular case, a dog's decision is highly reliable. However, in references to this paper, this misinterpretation of the small p-values seems to prevail too. The authors did not consent to an independent re-evaluation of the entire study documentation arguing that local and international data protection regulations prevented them from releasing the video footage. Thereby the Editors-in-Chief are not able to dispel the doubts related to the validity of the methods and results of the study. Finally, the Editors-in-Chief concluded that the described concern raised by Goss [5] and confirmed by the independent statistical reevaluation is comprehensible. The study numbers as given in Table 2 [1] arouse suspicion that the selection process of the negatives has been manipulated (intentionally or unintentionally) or that a relevant number of negative trails was removed from the statistical evaluation of the data. The authors did not agree to an independent reevaluation of their study data, which would have been needed to definitively address this and the other concerns raised by the community. As such, without this original data, it is not possible to make a definitive judgement as to the soundness of the methods and the results obtained in the study and the investigation of the so far published (and thereby available) content leaves inconclusive evidence of publication or research misconduct [7]. Therefore, in line with the Committee on Publication Ethics (COPE) guidelines [7], the Editors-in-Chief are issuing this expression of concern and suggest that the study results are taken with care especially for the application to forensic casework. The Editors have linked this expression of concern with the aforementioned original publication in the online scholarly record and reserve the right to take further action on this publication in the light of any new pertinent evidence that might emerge in the future. We apologize to our readership that this and the other issues were not spotted during the peer review process and solved prior to publication. © 2020