Scientific Fraud

Between 1999 and 2001, Kranke et al.1–3 published 3 articles highlighting concerns with data from the numerous studies published by Dr. Yoshitaka Fujii on postoperative nausea and vomiting (PONV) management. Approximately 10 years later, Toho University Faculty of Medicine started to investigate the credibility of 9 publications by Dr. Fujii and announced in March 2012 that 8 of those studies lacked appropriate ethics committee approval. The Japanese Society of Anesthesiologists (JSA) subsequently set up a special investigation committee to assess the allegations of scientific fraud. The committee investigated 212 articles and concluded that 172 studies were fabricated and 3 were not. For 37 articles, the JSA stated that the evidence was inconclusive as to whether or not the studies were fabricated. Based on fabrication of at least 172 studies, this likely represents the largest fraud in the history of anesthesia literature. The purpose of this editorial is to provide an overview of the scope of Dr. Fujii’s publications relating to PONV, and the extent to which our knowledge and management of PONV have been affected by removing these data from the literature. Of the 249 articles published by Fujii, 122 dealt with PONV. Of these, 110 articles were randomized controlled trials (RCTs), and the remaining 12 articles were reviews. Of the 110 PONV RCTs, all but 1 were determined to be fabricated by the JSA investigation. The sole study not found to be fabricated was among those with no evidence to prove whether it was fabricated. There are several common themes in Dr. Fujii’s publications. Many of his studies were repeated in different patient populations using the same drugs or combination of drugs, and reporting very similar results. Using a meta-analytic approach, Carlisle4 demonstrated that there was almost no heterogeneity among the majority of outcomes in the studies published by Fujii (I2 = 0). The I2 test assesses the degree of heterogeneity among included studies. It can be interpreted as the percentage of the total variability in a set of effect sizes due to true heterogeneity, that is, between-studies variability. When testing similar drugs in different patient populations, some degree of heterogeneity is expected. For instance, among RCTs comparing granisetron versus metoclopramide, the I2 was 0% in Fujii’s studies for all the end points tested (nausea, vomiting, nausea or vomiting, and need for rescue antiemetics) compared with I2 ranging from 5% to 51% for studies by all other authors.4 The main focus of Dr. Fujii’s fabricated publications has been granisetron, which was allegedly investigated in 70 of his published RCTs. This included single-dose studies, dose-finding studies, and comparisons with other antiemetics mainly ramosetron, droperidol, and metoclopramide, as well as comparisons of the combination of granisetron with dexamethasone or droperidol versus single agents. In Dr. Fujii’s articles, granisetron was reported to be more effective than all those other antiemetics with the exception of ramosetron, which was more effective than granisetron from 24 to 48 hours after surgery, but had similar effectiveness from 0 to 24 hours.5 In his dose-ranging studies, the minimum effective dose for PONV prophylaxis was 40 μg/kg in adults6 and children7 when given IV, 2 mg in adults8 and 40 μg/kg in children9 when given orally, and 20 μg/kg for the treatment of PONV in adults.10 In 2001, Kranke et al.3 performed a meta-analysis on granisetron for PONV prophylaxis and found that Fujii’s results differed from those reported by other centers. Specifically, pooled results from Fujii’s studies suggested that low-dose granisetron (≤20 μg/kg) was ineffective for PONV prophylaxis and only high doses (>20 μg/kg) were effective. This was in contrast to pooled results from other centers where there was no difference in effectiveness between low-dose and high-dose granisetron. Furthermore, pooled results from the Fujii studies suggested much higher effectiveness of high-dose granisetron compared with other centers. More recent studies by other investigators have confirmed that granisetron doses as low as 0.1 mg/kg are effective for PONV prophylaxis in adults.11 In his meta-analysis comparing the results of the studies by Fujii with other authors, Carlisle4 reported that granisetron was 1.3 to 1.5 times more effective in RCTs by Fujii than by other authors. Similar to other literature, Fujii reported that the combination of granisetron with other antiemetics was more effective than each antiemetic alone. However, his publications suggested that granisetron was unique in terms of acting synergistically with other antiemetics, a finding that was not supported by the work of other authors.4 Another focus of Dr. Fujii’s publications was ramosetron, the subject of 15 published RCTs. Fujii reported that ramosetron was effective for PONV prophylaxis for 48 hours after surgery, with a minimum effective dose of 0.3 mg in adults12 and 6 μg/kg in children.13 In 7 studies, he compared granisetron with ramosetron and reported that both had similar effectiveness from 0 to 24 hours, but ramosetron was more effective from 24 to 48 hours.5 The doses used in these studies are similar to those used by other investigators. Unfortunately, the only dose versus response studies published for ramosetron seem to come from Fujii. As a result, ramosetron dosing needs to be assessed by other investigators. The effectiveness of ramosetron was also much higher in Fujii’s publications; for instance, the relative risk (95% confidence interval) for reducing postoperative vomiting in pooled studies by Fujii was 0.39 (0.30–0.51) compared with 0.65 (0.47–0.90) in pooled studies by other investigators.4 Seven of Dr. Fujii’s studies reported on the dose response of perioperative administration of dexamethasone. Most of those studies suggested that dexamethasone 8 mg and 16 mg, but not 4 mg, were effective for PONV prophylaxis.14 Dexamethasone 4 mg was reported to be effective for PONV prophylaxis in only 2 of his 7 studies.15,16 Other large studies have, however, confirmed the antiemetic effectiveness of dexamethasone 4 mg,17 which is the currently the recommended dose by the PONV consensus guidelines of the Society of Ambulatory Anesthesia (SAMBA).18 Sixteen of Dr. Fujii’s studies reported antiemetic effects of propofol given as a small bolus dose of 0.25 mg/kg or 0.5 mg/kg at the end of surgery or as a continuous low-dose infusion at 1 mg/kg/h in women undergoing cesarean delivery under spinal anesthesia. The main findings were that propofol 0.5 mg/kg, but not 0.25 mg/kg, given at skin closure19 was effective for reducing PONV for 24 hours and was more effective than metoclopramide in all reported studies and droperidol in some of the reported studies.20 In Fujii’s studies, infusion of propofol at 1 mg/kg/h was also effective at reducing emetic symptoms during spinal anesthesia for cesarean delivery.21 Data from other investigators suggest that propofol has prophylactic antiemetic effects mainly if used for maintenance of anesthesia.22 Studies investigating the effectiveness of subhypnotic doses of propofol have reported mixed results.23,24 The administration of a small dose of propofol at the end of surgery was not shown to have the significant antiemetic effects reported by Fujii.25 Metoclopramide and droperidol were included, mainly as comparators with granisetron, in many of Fujii’s studies. In his meta-analysis, Carlisle4 reported that droperidol was less effective in studies by Fujii compared with others. Midazolam was investigated in 2 studies with both concluding that preinduction 50 μg/kg and 75 μg/kg doses were effective for PONV prophylaxis for 24 hours after surgery.26 The effectiveness of the 75 μg/kg dose was also reported by other investigators,27 but no other studies compared 50 μg/kg versus 75 μg/kg. One study by Fujii reported that oral clonidine was more effective than diazepam in reducing PONV after strabismus surgery in children.28 It is interesting to note that this was the only PONV study that the JSA could not determine whether it was fabricated. Another study reported that oxygen was effective for reducing PONV in women undergoing gynecologic laparoscopic surgery.29 The current evidence suggests, however, that oxygen supplementation is not an effective modality for PONV prophylaxis.30 Because of the large amount of literature on the topic of PONV, numerous meta-analyses, review articles, and consensus guidelines have been published to summarize this literature and guide the clinician. It is, therefore, important when reading those reviews to assess whether its authors are using articles published by Fujii. While some meta-analyses, particularly earlier ones, have included studies by Fujii,31–33 others have conducted sensitivity analyses excluding his studies.34–36 Similarly, many of the published guidelines such as the SAMBA consensus guidelines for the management of PONV18 and recommendations by French37and German38 experts did not include any of the articles published by Fujii in formulating their recommendations. Fujii’s fabricated data differed from data reported by other authors. His data compromised, at least in part, review articles and meta-analyses that used his fabricated data. Some published systematic reviews performed sensitivity analyses and noted that their conclusions were not significantly different when excluding Fujii’s studies.35,39 Others, however, indicated that including Fujii’s data changed some of the results of their meta-analyses.34,36,40 It is impossible to determine to what extent the reviews that included Fujii’s data, but did not undertake a sensitivity analysis, were compromised. However, it is likely modest, because of the enormous amount of data generated by other investigators. Although Dr. Fujii published extensively, the effect of his fraud on the management of PONV has been minimal. The lack of inclusion of his data in the current commonly used guidelines should minimize the effect of these retractions on the practice of PONV management, particularly in North America, where the SAMBA guidelines have been widely adopted. DISCLOSURES Name: Ashraf S. Habib, MBBCh, MSc, MHSc, FRCA. Contribution: This author helped analyze the data and write the manuscript. Attestation: Ashraf S. Habib approved the final manuscript. Conflicts of Interest: The author has no conflicts of interest to declare. Name: Tong J. Gan, MD, MHS, FRCA. Contribution: This author helped write the manuscript. Attestation: Tong J. Gan approved the final manuscript. Conflicts of Interest: Tong J. Gan received research funding from Cara Therapeutics, Premier Inc., AcelRx., Cheetah Medical, and Acacia, Speaker’s Bureau: Baxter, Hospira, Cadence, Fresenius-Kabi. This manuscript was handled by: Steven L. Shafer, MD.