The retraction last week of a commentary in the Journal of the National Cancer Institute (JNCI) looks like a nice — and very quick — example of self-correction in science.
The commentary, “Designing a randomized clinical trial to evaluate personalized medicine: a new approach based on risk prediction,” was co-authored by Stuart Baker and Daniel Sargent, and published in the December 1, 2010 issue. The journal published the retraction, and a letter that contained the re-analysis that led to it, on January 12.
The retraction notice:
We thank Drs Simon and Freidlin for their very important letter. Their simulation identified an oversight in our approach (1). We performed a simulation with a binary outcome and also found substantial bias in the presence of many predictors. Therefore, we can no longer recommend our proposed approach.
To the Editor: We respectfully request a retraction of the Commentary.
Here’s what happened: One of Baker’s research interests is making clinical studies more efficient, JNCI editor Barry Kramer tells Retraction Watch:
He was trying to develop an efficient method of measuring effects and effect sizes of treatments within prognostic subgroups in clinical trials. He proposed a statistical method that required increasing the sample size slightly, but getting a reliable answer about subsets.
Richard Simon, who wrote the letter in response to the commentary, likely read Baker’s commentary as soon as it came out, did the calculations himself, and “worked it out that the method was prone to bias,” Kramer said.
When the JNCI sent Baker Simon’s letter, as per protocol, “as soon as he saw that, he obviously saw that he had made a conceptual error, and he’s the one who spontaneously said, ‘you’re absolutely right,'” Kramer recalled.
“It was very, very, very fast,” said Kramer. Baker is
the type of person who says, ‘if I’m wrong I want to know I’m wrong.’ That’s the way scientists should be.
Baker and Simon, we should note, are both statistical editors at the JNCI. In the tightly knit communities of scientific experts, it’s perhaps to be expected that such editorial board members may have to retract periodically. That happened when Marc Hauser retracted a 2002 paper in Cognition, where he was an associate editor when the study was published, and to Naoki Mori, who is on the editorial board of Retrovirology. Both cases involved misconduct.
What’s different here, of course, is that there was no misconduct involved — just good old-fashioned self-correcting science. So we applaud the Baker, Simon, their co-authors, and the JNCI.