Health by Numbers
MARVIN ZELEN

These days, every trial of a new drug or surgical intervention is closely scrutinized by a statistical "consultant." But before Marvin Zelen came along, biostatistics had something of a Rodney Dangerfield problem: it rarely got the respect it deserved.

"Years ago, clinical trials were often the sole province of surgeons and clinicians," says Professor of Biostatistics Stephen Lagakos. "Many of them didn’t view a clinical trial as what it is–a scientific investigation–and few appreciated the importance of statistics."

Nobody in this country has done more than Zelen, the 69-year-old former chair of the School’s biostatistics department to "elevate the role of the statistician in clinical trials," says Lagakos. Zelen’s theoretical work over his 40-plus year career would make any curriculum vitae shine. But during the 1960s and 70s, he also transformed clinical trial research into a large, well-managed, and statistically sophisticated branch of medical research. First in Buffalo and then at the School, Zelen created what were in effect large biostatistical laboratories–equipped with powerful computers and capable of administering and keeping tabs on hundreds of clinical trials at a time. Randomization techniques, centralized data management, quality control, and other solutions to the practical problems of running mullet-centered clinical trials were worked out by Zelen and his colleagues. "He set the standard for statistical and data management centers in cooperative group studies," says Professor of Biostatistics Richard Gelber. In doing so, Zelen played a key role in the development of cancer treatments that have, just to cite one example, vastly increased the chances of surviving leukemia. Many of the lessons learned in cancer clinical trials were subsequently applied to AIDS clinical trials, including Gelber and his colleague’s landmark study showing that AZT can cut the chances of transmission of hiv from an infected mother to her baby by two-thirds.

Throughout his career, Zelen has been a fighter and defender as well as an intellectual force. "I think I have been very determined and very persistent," says Zelen. "And I never shirk from battle."

Zelen grew up in the East Bronx during the Depression, and when he wasn’t working long hours in his father’s candy store, he enjoyed doing many of the things that The Music Man’s Professor Harold Hill warned were trouble for River City: playing pool, poker, pinochle, and 21. When Zelen stumbled into a probability class as an undergraduate at the City College of New York, he says one reason the subject was so appealing to him was his familiarity with games of chance.

The post-war years were halcyon days for all kinds of science research in the United States, including math and statistics. Wielding seemingly arcane formulas, the nation’s math and statistical whizzes solved all kinds of practical problems during the war, calculating everything from the optimum formation for groups of bomber planes to the best way to search for submarines. (Another former chairman of the School’s biostatistics department, Frederick Mosteller, was a member of an elite group of Princeton mathematicians whose understanding of sampling procedures helped the government answer questions like whether an Army base would overtax a local public transportation system.) The federal government established two large applied mathematics laboratories after the war, one at the National Bureau of Standards in Washington and the other at ucla. After earning a masters degree in statistics from the University of North Carolina (Chapel Hill) in 1952, Zelen was hired by the National Bureau of Standard’s Statistical Engineering Laboratory, which was part of the National Applied Mathematics Laboratory.

For Zelen, those 10 years at the Bureau are a delightful memory. At age 25, he says he was the "baby of the group" and one of the few people without the imprimatur of a Ph.D. (which he remedied in 1957 by taking evening classes at American University). Zelen thrived in the lively, intellectually rambunctious atmosphere of the Bureau. Zelen says he pursued some theoretical problems, but also gained the invaluable experience of designing research projects up front so that when the results came in they would pass statistical muster. His first real taste of biostatistics came in the early 1960s when Zelen was a visiting professor at the University of Wisconsin’s Mathematics Research Center. Asked to work out a statistical problem related to a trial of a childhood leukemia drug, Zelen and his collaborator George Weiss came up with something called the semi-Markov process (Markov was a Russian mathematician), which now stands as one of Zelen’s prime contributions to the biostatistical canon. Essentially, the semi-Markov process is a statistical model used for both calculating "sojourn time," or how long a patient might stay in a certain "phase" of cancer treatment (say remission or relapse) and using that sojourn time to predict the next phase the patient will go through.

Now dedicated to biostatistics, Zelen was invited to head the National Cancer Institute’s Applied Mathematics and Statistical section in 1963, where he immersed himself in cancer and clinical research for the next four years. After a year in London as a Fulbright Scholar, he was lured back to academe and into the arms of the State University of New York at Buffalo. Zelen spent the next decade on the snowy eastern shore of Lake Erie effectively inventing the biostatistics that make possible today’s sophisticated clinical testing of cancer, AIDS, and all kinds of other treatment drugs. Zelen says he wasn’t happy dispensing advice to his nci colleagues. "I decided it was important to have some kind of demonstration project to show how things could be done," he says. So he formed the Statistical Laboratory at the University of Buffalo.

The statistical laboratory was a master stroke. Zelen saw that the statistical aspects of the large, complex trials of treatment drugs needed to be more centrally administered, particularly when it came to treatment assignments and randomization. Leaving treatment choice to clinicians and treatment centers created too many opportunities for cutting corners, steering certain patients to certain treatments, and robbing randomization of its randomness. "There were a lot of complaints in the beginning," Zelen says. "People would say, ‘We used to hold the sealed envelope up to the window to see what the treatment would be, and now we can’t do that anymore.’ " He also saw that poor record keeping–what would be today called data management–was undermining high quality research. Untrained secretaries were often left in charge of reviewing patient records and filling out the forms for a study. Proper training and some "pretty smart people" were needed to do the job right, Zelen believed, so he created a new job and called it "data manager" because managers are paid more. "We made data management into a profession," he says. "There was really no point in applying sophisticated statistical techniques to data that was garbage."

The emergence of the large, multi-center randomized clinical trial also posed some new statistical challenges. Zelen designed "dynamic randomization" techniques so that a roughly equal number of patients would be allocated to each treatment option under study. He also wrote one of the first papers on adaptive design of clinical trials, a novel randomization strategy that would have an investigator stick with a treatment until it fails and then switch over to the alternative. Zelen also did some groundbreaking work in the field of survival analysis that preceded Sir David Cox’s proportional hazard model, a mainstay of contemporary biostatistics.

Meanwhile, Zelen was building suny-Buffalo into a clinical trials juggernaut by getting grants from NCI and attracting young, smart statisticians and mathematicians. For 19 years, Zelen was the top statistician in the Eastern Cooperative Oncology Program (ECOG), which became–partly because of Zelen–the largest program in the world for testing various cancer treatments.

By the mid-70s, Zelen was prominent enough to capture the attention of Mosteller, who, at Dean Howard Hiatt’s request, had come to the School with the express purpose of building up the biostatistics department. In his negotiations with the university and the School, he asked for just a little more than usual. "I said I would like 10 faculty appointments in addition to my own. They were used to hiring faculty. They weren’t used to hiring a basketball team," says Zelen, adding that he believed bringing the "team" with him to Boston was essential to the viability of ongoing clinical trials. When Zelen came to the School in 1977, he brought with him an entourage that included 10 faculty members, 17 non-faculty members, a huge dec 20 computer (they had to knock down a wall to fit it into the basement of the Dana-Farber’s Jimmy Fund building) and 150 cancer trials involving several thousand patients.

Zelen succeeded Mosteller as chairman of the biostatistics department in 1980 and held the post for 10 years. He kept the ecog trial work going (Professor Dave Harrington has replaced Zelen as the lead statistician) and laid the groundwork for the department’s pre-eminence in AIDS clinical trials, now manifest in the Center for Biostatistics in AIDS Research (CBAR). He beefed up the biostatistics curriculum. When one leading professor at the School expressed puzzlement, tinged with disapproval, that there could be that much to teach in biostatistics, Zelen says he answered, "We have only just scratched the surface!" Lagakos says Zelen "really made this department into a wonderful place," fostering creativity and confidence in his colleagues. "He made you feel like you could do anything."

In the early 1980s, the wider world caught a glimpse of Zelen’s tenacity. Flying in the face of conventional wisdom that such a study could ever produce significant results, Zelen, with Lagakos’s help, launched an investigation of the connection between what seemed to be a cluster of childhood leukemia cases in the Boston suburb of Woburn, and the town’s contaminated water supply. Three years later, the Harvard Health Study, as it came to be known, showed for the first time a connection between Woburn’s contaminated water supply and a variety of adverse health effects, including leukemia. As described in A Civil Action, the best-selling account of the lawsuit that grew out the Woburn cancer cluster, when Zelen announced the study’s results in the basement of a Woburn church in February 1984, someone in the audience said "Thank God for Marvin Zelen." The crowd burst into applause. A hero to those Woburn parents, Zelen and his study were attacked elsewhere. In fact, the chair of the School’s own epidemiology department, Brian MacMahon, was sharply critical of Zelen and Lagakos for what he said was an over-interpretation of their data. But Zelen didn’t waiver. "He champions the underdog," says Lagakos. "And if he feels something is wrong, he doesn’t worry about getting in trouble."

In 1990, Zelen stepped down as chairman of the department, but he continues to be a presence at the School, teaching, attending seminars, and supervising post-doctoral students. His fourth-floor office at Dana-Farber overlooking busy Brookline Avenue has the delightful aura of the active, accomplished mind: the haphazard photos of family and colleagues, an eclectic collection of academic journals and textbooks, the desk and tabletop choked by a shifting layer of papers and correspondence. No longer involved in the day-to-day running of clinical trials of cancer treatments, Zelen has started thinking about how cancer screening studies could be improved, diplomatically describing the current approaches as "suboptimal." And though his back-of-the-candy-shop, card-playing days in the Bronx are long gone, Zelen has given a friend some statistically-based advice on how to play one of the biggest games of chance–mutual fund investment. Zelen’s tip is to "play the winner": stick with the biggest gainer until it is surpassed, and then switch to the new leader.

- Peter Wehrwein