The Boston team did not report in its article how many patients had been typed with which chip. But responding to a request, Paola Sebastiani, the group’s statistician, said in an e-mail message that 108 of the 1,055 centenarians had been tested with the Illumina 610 array, almost exactly the 10 percent that Dr. Stefansson had predicted.
Dr. Sebastiani said that “the problem should be limited to the top two SNPs,” referring to the genetic variants the team found to be associated with extreme longevity.
“The conclusions of this analysis, even if one or two SNPs have bad genotype data, are not going to change,” she said.
Dr. Stefansson, however, believes the problem affects all the results, including the 33 genetic variants that the Boston University team said had been confirmed in a second study. “It is absolutely clear that there is nothing to this,” he said.
Scientific journals cannot guarantee the truth of the reports they publish, but they do try to weed out obvious error by submitting manuscripts to expert reviewers. Dr. Sebastiani said she had performed extra analyses that “were suggested by the anonymous reviewers of the article, who were then satisfied with the results.”
But whether because the reviewers were not rigorous enough, or because Science’s editors ignored their cautions, the result was a publication that leading experts felt had not been sufficiently scrutinized. “Accurate typing of the associated SNPs should have been required, and hence there is a very specific failing in the review process,” Dr. Goldstein said.
Keywords: longevity
