In search of questions
If a biostatistician's job is to help investigators reach the right answers, it is, equally, to make sure they ask the right questions. Consider a clinical trial of a hypothetical new breast cancer drug. Researcher and biostatistician must decide how many patients need to be enrolled for results to be valid, what stage of disease they should be in, what previous treatments are allowed, what doses they should receive and how often, and — crucially — what constitutes success: tumor shrinkage? tumor disappearance? a longer survival rate? a better quality of life?
If a biostatistician's job is to help investigators reach the right answers, it is, equally, to make sure they ask the right questions.
A successful study will generate minimal statistical "clutter," or data clouded by ambiguities and uncertainties. Dana-Farber biostatisticians accomplish this in two ways: by advising investigators on what data to pursue and by giving them new statistical tools to capture it.
"As biostatisticians we wear many hats," says Richard Gelber, PhD, one of those who made the transition from SUNY Buffalo to DFCI a quarter century ago. "We work with in-house investigators and do our own research to improve statistical techniques. We're motivated by the problems and needs that come up in our day-to-day collaborations with scientists." Example: One of the conundrums in evaluating new therapies is how to measure the tradeoff between short-term side effects and long-term benefits. Gelber and his colleagues invented a method known as Q-TWiST, for Quality-adjusted Time Without Symptoms and Toxicity, to do just that.
"We broke the treatment process down into three stages: an initial period of toxicity during and after a drug is given, a period we call TWiST when the side effects subside, and a time when the disease recurs," Gelber says. "Q-TWiST provides a way of incorporating quality-of-life considerations into the assessment of new treatments."
Advances in computer technology have made it possible to sort, sift, and interpret statistical data in ways that once would have been unimaginable — to create mathematical formulas, for instance, of the ways genes are switched on and off. Several Dana-Farber biostatisticians are working on projects to harness this power for researchers' benefit.
Among them is Robert Gentleman, PhD, a key architect of computer software called "R" that has become a universal language for statistical work. Available for free over the Internet, it is used by medical research centers, financial institutions, marine investigators, geographers — "anywhere that quantitative work is done," Gentleman remarks.
Robert Gentleman, PhD, is working to harness the power of new computer technology to advance biostatistics.
A newer project promises to transform the way scientists present, communicate, and interpret experimental data. Ever since the dawn of the Scientific Revolution, researchers have published data in predigested form through charts, graphs, and tables in journal articles. This format has persisted despite the advent of electronic publishing, in which scientific studies are posted on the Internet before appearing in print. Gentleman proposes an updated system in which e-published studies would include software that enables readers to see how authors perform their calculations and to recompute the numbers in useful ways.
"As it currently stands, electronic publishing is merely a way of shipping data from one computer to another," he remarks. "This can be improved using the computing power that readers have right on their desktops." He and his colleagues have produced prototype software for this system and demonstrated how it could be used with certain studies.
- Next: Finding patterns
- Page: 1 | 2 | 3 | 4
