Scientists who asserted last year that they could predict with 77 percent accuracy who would live past 100 have retracted their report in the journal Science, yet say they are right anyway.
The researchers, Paola Sebastiani and Dr. Thomas T. Perls of Boston University, wrote in Science last July that they had found 150 genetic variants that correlated with extreme longevity.
Their claim was immediately assailed by geneticists who said the report was highly implausible and should not have been published by a leading journal like Science. Perls, the senior author, is a geriatrics physician, not a geneticist, and Sebastiani is a biostatistician.
Within a week of publication, Dr. Kari Stefansson, of the gene-hunting company Decode Genetics in Iceland, had pinpointed the probable source of error: The Boston scientists had used a particular model of gene-detecting chip, called an Illumina 610, that was known to have a specific flaw. Stefansson calculated that 10 percent of the centenarians the Boston team studied had been tested with the flawed chip, a figure not reported in the paper, and Sebastiani confirmed that that had been the case.
Nevertheless, Perls and Sebastiani did not retract their article, and the editors of Science, evidently concerned about the unresolved situation, published an "expression of concern" in November that mentioned the flawed chip and said the authors were rechecking their results.
The retraction published by Science on Friday is unusual because the authors concede that they used a flawed chip but say that they were right anyway. When data from the flawed chip are excluded, "the main scientific findings remain supported by the available data," they write.
Asked why the paper was being withdrawn if its conclusions were correct, Sebastiani laughed and said, "That's a good question."
Scientists often find that there are errors in published reports, and unless the errors are fatal, they simply ask the journal to publish an erratum. Perls and Sebastiani have instead retracted their entire report and say they will seek to publish the corrected version elsewhere.
Asked why he did not just publish an erratum, Perls said in an e-mail that he could not comment because "I just don't want to do anything that could jeopardize the current review of the corrected paper."
Natasha Pinol, a Science press officer, said that even a revised version, without the flawed data, would not be adequate for the journal. "A paper built on the corrected data would not meet the journal's standards for genomewide association studies," she said.