Inclusion of Retracted Studies in a Systematic Review and Meta‐Analysis of Prophylactic Dexmedetomidine and Postoperative Junctional Ectopic Tachycardia in Pediatric Cardiac Surgery Patients

Medical decision making is dependent on reliable evidence. Systematic reviews and meta-analyses are cornerstones of this evidence. Meta-analyses however are dependent on the reliability of data from individual studies, and as the number of retractions in the medical literature is increasing [1], the credibility of meta-analyses are affected. An example of this is a meta-analysis published by Ghimire et al. in Pediatric Anesthesia in 2018 [2], evaluating the use of prophylactic dexmedetomidine in preventing junctional ectopic tachycardia (JET) as well as a range of other perioperative adverse outcomes. JET is the most frequent arrhythmia encountered after congenital heart surgery and is associated with a mortality rate as high as 13.5% [3]. It can cause significant hemodynamic instability, sometimes necessitating extracorporeal membrane oxygenation support. Commonly-employed treatment modalities include administration of amiodarone, overdrive pacing, and induction of hypothermia [4]. The meta-analysis authored by Ghimire and colleagues evaluated the prophylactic use of the alpha-2 adrenergic agonist, dexmedetomidine, reporting that patients receiving dexmedetomidine had a significant reduction in the incidence of JET as well as a reduction in the intensive care unit stay, inotropic scores, and total mechanical ventilation time [2]. Seven studies were used to evaluate the primary outcome and three to four studies were used to evaluate the secondary outcomes. However, upon further in-depth review of the studies included in this meta-analysis, we learned that one manuscript [5] had been formally retracted [6], while a second [7] had been the subject of an “expression of concern” [8]. Both the retraction and letter of concern related to data fabrication/duplication because the studies in question had nearly identical data presented in portions of the results. Because data from these two problematic studies were included in the meta-analysis, they impacted the results assessing all of the outcomes. When a problem with a study is recognized soon after publication, a letter to the editor is a mechanism that can be used to identify the issue and give the authors the ability to address it in a response. However, letters to the editor, including those to this journal, are primarily only considered a few months after publication of an article. Because retractions of individual papers may not come to light for years after the publication of a meta-analysis, there is no formal mechanism for correcting these errors. In the present study, we extracted data from the original Ghimere et al. meta-analysis, sequentially removed data from the questionable studies, and recalculated the remaining pooled results. This was performed with the primary outcome and repeated with all secondary outcomes. Pooling of data was performed using RevMan [9], using random effects models similar to what was done in the original meta-analysis. After recalculation of the results, we calculated the effect estimate evolution, which is the percentage change in the effect estimate to quantify the impact of excluding the problematic studies from the original results [10]. The forest plots after excluding the studies in question can be seen in Figures S1–S6. The original and recalculated point estimates, confidence intervals, and effect estimate evolutions are summarized in Table 1. After exclusion of the studies in question, the primary analysis evaluating the association between dexmedetomidine and JET was minimally affected with effect estimate evolutions of 0% with exclusion of the retracted study only [5], and −5.7% when both the retracted and flagged [7] studies were excluded. However, for the secondary outcomes, larger changes were seen with effect estimates changing from −61.9% to 205%. Also, while the original results reported that dexmedetomidine was associated with reductions in total ICU stay, inotropic score, and total mechanical ventilation time, these differences were no longer statistically significant following exclusion of one or both studies. Given that the retraction and expression of concern did not occur until 2020, the authors of the meta-analysis could not have anticipated the problems with these included articles. The incidence of retracted studies changing the results in meta-analyses on other topics is also not uncommon [10], and while problematic scientific articles are flagged by each journal and retracted articles are no longer accessible, a similar system does not exist for meta-analyses. In this letter, we show that after excluding two unreliable studies, there remains an association between dexmedetomidine and reduced incidence of JET. However, dexmedetomidine is no longer associated with a reduction of the secondary outcomes including ICU stay, inotrope score, and mechanical ventilation. Acknowledging the change in the results of this meta-analysis is important. However, it is even more important to recognize that this is a problem throughout the medical literature, and as the number of retractions is rising, more effective solutions for correcting the literature are also needed. Max M. Feinstein: this author helped design the study, analyze the data, write the manuscript, and approve the final version. Richard J. Levy: this author helped design the study, analyze the data, write the manuscript, and approve the final version. Caleb Ing: this author helped design the study, perform statistical analysis, analyze the data, write the manuscript, and approve the final version. Max Feinstein has a consulting relationship with Medtronic to provide guidance on laryngoscope design. R.J.L. was supported by National Institutes of Health grant R01GM148716 The data that support the findings of this study are available from the corresponding author upon reasonable request. Figure S1: pan70125-sup-0001-FiguresS1-S6.docx. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.