PNAS Retracts Paper That Contributed to Lung Cancer Trial

Retraction Watch Staff

November 15, 2021

A paper that was the subject of a four-page correction in 2018, and which helped inform a now-halted clinical trial of a drug for lung cancer, has been retracted following an institutional investigation concluded that one of the researchers had falsified the data in that article and at least four others.

And we have learned that Springer Nature should be acting on a different article by the researcher shortly, and has just begun an investigation of yet another.

The PNAS article, which appeared in 2015, was written by Takashi Nojiri, formerly of Osaka University and the National Cerebral and Cardiovascular Center Research Institute, in Japan. As we reported in June, an August 2020 report from National University Corporation Osaka University and National Cerebral and Cardiovascular Center Hospital concluded that Nojiri committed:

specific misconduct (fabrication, falsification) because the researcher could not obtain a valid and scientifically grounded explanation.

The inquiry flagged five papers by Nojiri, one of which has been retracted and one — a 2014 article in Pulmonary Pharmacology & Therapeutics titled "Atrial natriuretic peptide inhibits lipopolysaccharide-induced acute lung injury," has received a correction. The PNAS paper was not among that batch.

According to the Asahi newspaper, in January 2021, Osaka University halted a clinical trial by Nojiri's group of a drug to treat lung cancer over concerns that his misconduct might be leading to patient harm:

Osaka University decided to discontinue the clinical study, judging that it was a clinical study for which the scientific basis was not clear. Ten health hazards were reported for which a causal relationship with clinical studies cannot be completely ruled out. The hospital apologizes to the patients who participated in the study and continues to observe their health.

In 2018, PNAS issued a remarkable "mega" correction for a 2015 paper by Nojiri and colleagues. Running four pages, the addendum was two-thirds as long as the article itself. At the time, a spokesperson for the journal told us that, despite the extensive nature of the errors, the paper did not require retraction:

The editor and the Board member considered this Correction. The editor felt that the errors did not substantially affect the main findings of the article. The Board member concurred and concluded that a correction was sufficient in this case and the article did not need to be retracted.

What changed? The institutional investigation found that those errors weren't so inconsequential after all.

The PNAS retraction notice reads:

The editors note that, based on the recommendation of the National Cerebral and Cardiovascular Center (NCVC) in Osaka, Japan, we are retracting this paper. An NCVC investigative committee has identified several areas of concern related to Fig. 4E, Fig. S8, and Table S3. The NCVC committee notes, "We believe this paper should be withdrawn, and we have communicated our stance to the lead and corresponding authors."
Specific concerns of the committee are noted below:
Fig. 4E in the originally published version of the manuscript: "The source data used to calculate the secondary measures (i.e., second-order data) for the LPS(0)/Control and LPS(0)/ANP groups are missing. Further, certain values were excluded from the calculation of mean E-selectin mRNA expression in the remaining four groups. Similarly, associated standard error (SE) values were calculated based on data obtained from different experiments."
Fig. S8 in the originally published version of the manuscript: "Statistical analyses were conducted in an improper manner. The experiments are presented as having been conducted on five groups. However, in reality, the dataset consisted of measurements from four groups in an experiment performed on October 23, 2014 [vehicle iv./ANP iv. (0.025 γ)/vehicle sc./ANP sc. (0.5 γ)] that were combined with partial data of a vehicle iv. condition and complete data of an ANP iv. (0.1 γ) condition from a prior experiment dated March 14, 2014…. Therefore, it was inappropriate to run statistical comparisons between the vehicle iv. and ANP iv. (0.1 γ) groups."
Table S3 in the originally published version of the manuscript: "The paper purports to calculate several hemodynamic parameters at the individual and group levels…the five raw (zero-order) measurements from the same mouse were averaged to obtain the individual (first-order) mean, which were then averaged across the four mice in the respective groups to obtain the group (second-order) mean. Second-order data for heart rate (HR) have also been reported for each group; however, these were calculated as the mean ± SE of six values apparently chosen at random from the raw dataset (which consisted of five measurements per mouse × four individuals = 20 data points in total), and not of the four individual means."
The undersigned authors have agreed to this retraction:
Hiroshi Hosoda, Takeshi Tokudome, Koichi Miura, Shin Ishikane, Kentaro Otani, Ichiro Kishimoto, Yasushi Shintani, Masayoshi Inoue, Toru Kimura, Noriyoshi Sawabata, Masato Minami, Tomoyuki Nakagiri, Soichiro Funaki, Yukiyasu Takeuchi, Hajime Maeda, Hiroyasu Kidoya, Hiroshi Kiyonari, Go Shioi, Yuji Arai, Takeshi Hasegawa, Nobuyuki Takakura, Megumi Hori, Yuko Ohno, Mikiya Miyazato, Naoki Mochizuki, Meinoshin Okumura, and Kenji Kangawa.

Missing from that list: Takashi Nojiri.

Here are the articles by Nojiri that have yet to be retracted or corrected:

A spokesperson for Springer Nature told us:

Surgery Today are completing the COPE-compliant investigation and expect to take appropriate editorial action shortly.
The editors of Cancer Chemotherapy and Pharmacology were not aware of the institute's investigation and have now started an investigation.