Harvard Public Health Review Winter 2007
Review home
next story
Previous story- The America Next Door
HSPH home page

Leaders worth folowing: HSPH celebrates distinguished faculty and alumni
Does that mean A causes B? Or might chemical A be just an innocent bystander, linked in some way to as-yet-unknown cause C ... or even, perhaps, causes C, D, and E? Robins’s novel statistical methods can help researchers sort spurious associations from true causes and effects by accounting for confounding factors and intermediate variables clouding the pathway from cause to effect.

Moreover, Robins’s statistical wizardry can explain why observational and randomized studies of the same health problem sometimes appear at first glance to yield conflicting results. For example, while results from a number of large observational studies in post-menopausal women have indicated that hormone replacement therapy (HRT) involving estrogen and progesterone can prevent heart attacks, results from the Women’s Health Initiative’s randomized trial found that HRT appears to cause heart attacks.

Similarly, a number of observational analyses of HIV-infected study subjects showed that highly active retroviral therapy (HAART) slowed the rate of their progression to AIDS or death only marginally. A randomized trial, however, revealed a much larger drop in the disease progression rate.

Such disparities have led some scientists to question the reliability and utility of observational trials. Explains Robins: “Usually it’s assumed that observational studies are biased, because they may not account for important, and possibly also unknown, common causal factors.” In the case of both HRT and HAART, however, reanalyzing the observational trials using Robin’s methods has yielded results consistent with the randomized trials.

Given that so much of research is observational, colleagues say Robins’s statistical methods could steer biomedical research down a very different road, influencing not only the precision of their findings but ultimately also health practices and policies. HSPH Dean Barry R. Bloom foresees nothing short of a revolution in the way observational studies are analyzed, predicting that causal inferences derived from observational studies using Robins’s methods “will come to be regarded as a new standard, second only to the gold standard of randomized trials.”

“THE POTENTIAL FOR JAMIE’S WORK is huge,” says Miguel Hernan, a former student of Robins’s who is now an associate professor of epidemiology at HSPH. “Every time the doctor tells you to take a certain drug, every time you hear a nutritional recommendation, every time a new policy is implemented, it is because someone conducted a study and analyzed the data using a statistical method. Jamie is changing the way data are collected and analyzed, and in the past five years or so we’ve begun to see the practical impact of these new methods. The only problem with his work is, it’s 20 years ahead of its time.”

Though Robins’s insights are “respected at the very highest level of statistical science,” says Jim Ware, they have not yet caught on universally.
“There’s no controversy over whether Jamie is correct,” says Mark van der Laan, a biostatistician at the University of California at Berkeley who has worked with Robins. “His theories are revolutionary. But his methods are truly challenging to learn. There’s a certain element of, ‘What I’m doing is working fine for me. I don’t need the more complicated stuff.’”
“I’m not a particularly energetic salesman,” Robins admits. The work “has to spread to enough people who are good at selling new ideas, if it ever will.” As his theories of causality slowly wend their way through the scientific community, Robins—whose professorship was endowed in 2001 by longtime admirers Mitchell L. Dong and his wife, Robin LaFoley Dong—has moved on to a project he expects will take up his remaining years: a unified theory of statistics.

“The field of statistics is currently divided into parametric, nonparametric, and semiparametric schools of thought,” he says, launching into an explanation that quickly moves beyond the average listener’s grasp. “I just have a feeling that there should be one unified story for everything. I think it will allow for more accurate estimates of uncertainty. But I don’t know if it will be useful yet; that’s part of the research.”

Asked to comment on Robins’s chances of succeeding at this latest self-imposed challenge, Jim Ware smiles. “In physics, Einstein spent the latter part of his life trying to develop a unified theory,” he says. “Einstein didn’t actually succeed. But I’m not prepared to say Jamie won’t.”

Elizabeth Gehrman writes about science, medicine, and public health

1 | 2

Review home
next story
Previous story- The America Next Door
HSPH home page
top of page

Related Links

Department of Biostatistics

James Robins

 

 

 

 

This page is maintained by Development Communications in the Office of Resource Development.
To contact us with suggestions, comments, and questions, please e-mail: review@hsph.harvard.edu

Copyright 2007, President and Fellows of Harvard College