By Heidi Ledford of Nature magazine

A prominent paper that claimed to reveal the genetic factors that help people live to 100 or older has been retracted, a year after it was first released.

The study, published in Science, reported 150 genetic variations that could be used to predict whether a person was genetically inclined to see their 100th birthday. The results were based on a search through the genomes of more than 1,000 centenarians.

But shortly after the paper was published, a host of criticisms arose. In particular, geneticists noted that the control samples and the samples from centenarians were analyzed in slightly different ways. Last November, Science editor Bruce Alberts published an editorial expression of concern and noted that the authors were working to address the issue.

Today, the authors are officially retracting the paper, acknowledging that the original analysis was flawed. The team, led by Paola Sebastiani, a biostatistician at Boston University School of Public Health in Massachusetts, and Thomas Perls, a gerontologist at Boston University School of Medicine, collaborated with an outside laboratory to identify and remove affected data, and used a different genotyping platform to confirm their results. Ultimately, the reanalysis yielded a new set of genetic variations that might contribute to extreme longevity.

"However," the authors wrote in their retraction, "the specific details of the new analysis change substantially from those originally published online to the point of becoming a new report. Therefore, we retract the original manuscript and will pursue alternative publication of the new findings."

The incident could haunt the work of others who are searching for the genetic underpinnings of longevity, says Thomas Kirkwood, director of the Institute for Ageing and Health at Newcastle University, UK. "It means that people are going to be more cautious about future studies," he says.

"On the other hand," he added, "I think a high degree of caution is justified when we're dealing with a trait that's going to be as complex in its genetic underpinnings as longevity."

Not just genetics

Previous studies have suggested that tiny contributions from hundreds of genetic factors combine to account for about 25% of variation in human longevity. The remaining 75% is attributable to environmental influences.

Knowing this, Anatoli Yashin, a biostatistician at Duke University in Durham, North Carolina, was skeptical of Sebastiani and Perl's paper from the start. "I didn't believe it," he says. "Any model that plans to predict lifespan based only on genetic components is not quite correct. The environmental contribution to longevity is huge."

Indeed, some were immediately concerned that the paper's conclusions had been oversold. "It was picked up by almost every newspaper and broadcast channel as saying this was a genetic test that would allow people to predict whether they would live to be a centenarian," says Kirkwood.

But it was the use of multiple types of genotyping platform, with control DNA analyzed on one and centenarian DNA on another, that prompted the most ire. Although they work on the same principles, different platforms have distinct biases that could influence the results.

Cross-platform comparisons are becoming common as researchers try to cut costs and save time, says Peter Kraft, a biostatistician at the Harvard School of Public Health in Boston. "But I think folks are largely cognizant of the potential problems," he adds, noting that most research teams take steps to filter out errors caused by different platforms.

Replication and confirmation

Science issued a statement about the retraction, explicitly noting that the journal had found no evidence of misconduct by the study's authors. But, it said, after careful peer review of the resubmitted data, the journal concluded that the new analysis would not support a paper of the caliber necessary for publication in Science.

In particular, reviewers were concerned that the results of the reanalysis had not been replicated in a separate sample of patients. Many journals impose this requirement on genome-wide studies of this sort, says Anne Newman, director of the Center for Ageing and Population Health at the University of Pittsburgh in Pennsylvania.

But Yashin says that this might not be the right approach when it comes to longevity. Because environment is so influential on variations in lifespan, he worries that expecting researchers to find the same genetic variations in different populations could be setting the bar too high. Instead, he says, the goal should be to find genetic variations that could affect similar cellular pathways or functions.

To highlight the complexity of working with longevity, Kirkwood points to experiments in nematodes, or roundworms. He says that longevity varies widely, even among nematodes that are genetically identical and grown under the same environmental conditions. This suggests that chance plays a significant part in determining lifespan.

All this makes unraveling the genetic contributors to longevity challenging. "But it's really important that we do it," says Kirkwood. "As we begin to understand the factors that influence life expectancy, they may give us clues to what may be done to improve health and healthy life expectancy in the future."

This article is reproduced with permission from the magazine Nature. The article was first published on July 21, 2011.