Errors in a 2017 paper about a new cancer test may have occurred because of a simple typo while performing calculations of the tool’s effectiveness.
According to the last author, the “1” key was likely not pressed hard enough.
The error, however small, affected key values “so greatly that the conclusions of the paper can no longer be supported,” the editor said, which prompted the journal to retract the paper.
The tool was developed by Toronto-based company Proteocyte AI to predict a patient’s risk of developing oral cancer. The study in Oral Oncology by researchers at Proteocyte reported promising results: A test called Straticyte could accurately identify lesions in the mouth most likely to progress to cancer.
Kenneth Pritzker, medical director of Proteocyte AI, told Retraction Watch that the company discovered the error after a reader contacted the journal last year with concerns about the data. Pritzker, the paper’s last author, explained:
The error was a single keystroke error … rather than entering “16”, “6” was entered. In retrospect, probably this happened because the “1” key was not pressed hard enough.
Here’s the retraction notice for “Straticyte demonstrates prognostic value over oral epithelial dysplasia grade for oral potentially malignant lesion assessment:”
This article has been retracted at the request of the Editor-in-Chief. We have recently been made aware that the sensitivity values and the false-negative values reported in the manuscript were erroneous. The sensitivity values were reported as 71% and 60% in the article, but it was later determined that they were 48% and 36%, and the false-negative value was reported as 6 in the article but was 16. The Editors have determined that the values have changed so greatly that the conclusions of the paper can no longer be supported.
Because false-negative is used to calculate sensitivity, the error (6 instead of 16) affected the sensitivity values as well.
Pritzker, who noted that the company currently has two clinical studies underway, told us:
While the corrected data showed sensitivity reduced from 71% to 48%, this was still very much higher than 3% for oral epithelial dysplasia grade using the same test methodology in the same cohort. Therefore, we still consider that the data supports our conclusions that Straticyte is a potentially useful tool or assessment of progression risk for oral dysplasia …
According to Pritzker:
The plan is to offer Straticyte as a Laboratory Developed Test. [U.S. Food and Drug Administration] and Canadian equivalent do not regulate this class of tests at present.
The retraction prompted the editor of the authors’ 2016 paper—which also reported promising results about the tool—to request the data from the study; Pritzker said the authors are complying, but have checked and found no errors.
When asked whether the errors in the retracted paper undermine the utility of Straticyte, Pritzker said “not objectively:”
Different cohorts would be expected to show some variation in sensitivity. … statistical sensitivity is only one criteria for utility and has its limitations. The data is still superior to dysplasia grading …
The company, which changed its name from Proteocyte Diagnostics to Proteocyte AI on March 9, claims:
Straticyte is the only diagnostic predictor that can distinguish, with high accuracy, oral lesions at high risk of becoming cancerous from those with low or no risk. Straticyte sensitivity is >95% and specificity is >85%.
Pritzker noted:
The company has been transparent and timely on this matter with its stakeholders throughout the process.
Like Retraction Watch? You can make a tax-deductible contribution to support our growth, follow us on Twitter, like us on Facebook, add us to your RSS reader, sign up for an email every time there’s a new post (look for the “follow” button at the lower right part of your screen), or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].