The Number Crunchers Drugmakers Fear And LoveRandi Hutter Epstein
Richard Peto has never treated a patient or shaken a test tube. Yet the 54-year-old Oxford University professor of statistics loves to talk about the millions of people he has spared from death. The impressive thing is, others tend to agree with him. "If you look at his work in heart, breast, and lung disease, it adds up to a lot of lives saved," says Dr. Thomas Chalmers, a sometime critic of Peto who is an adjunct professor of medicine at Tufts University.
Peto and his partner, Dr. Rory Collins, are the kings of a new statistical science called meta-analysis--literally, beyond analysis. At Oxford's Clinical Trial Service Unit, which they co-direct, they amass heaps of data from studies medical researchers have done on big killers and after some tweaking add them together. Voila! Headline-grabbing, hard-to-refute conclusions emerge that can earn--or lose--hundreds of millions for drugmakers. Hoechst, Merck, Bristol-Myers Squibb, Wellcome Foundation, and Roche have sponsored Peto's independent reviews of their drugs--despite the risk of unfavorable findings. "Few investigators have the impact" of Peto and Collins, says Steven Weisman, research director at Sterling Winthrop, maker of Bayer aspirin in the U.S.
THE NEXT ACT. That's for sure. Since 1985, when Peto declared that tamoxifen, made by Zeneca Group PLC, boosted the five-year survival rate for breast cancer patients to 55% from 50%, sales of the drug have doubled. It was Peto and Collins who, in studies done in 1988 and 1994, found that aspirin helps treat--and prevent--heart attacks and strokes. Some 72% of U.S. heart attack victims now get aspirin vs. 38% in 1987.
Another splash, in 1991, involvedt-PA, the blood-clot-busting drug that Genentech Inc. sells for $2,200 a dose. After the Oxford duo found that Hoechst's $300-a-dose streptokinase was as effective and a bit safer, t-PA's market share fell to 53% from 65%. It cost Genentech $50 million in new studies to tip the balance back and recover t-PA's market share. Although Peto disagrees with the decision, in June a U.S. Food & Drug Administration advisory panel approved Genentech's proposed labeling changes, in effect endorsing its claim that t-PA is slightly more effective than cheaper brands.
For their next act, Peto and Collins plan a probe of cholesterol-lowering drugs. And come September, Peto may make waves with a new book in which he predicts that worldwide cancer deaths from smoking will reach 10 million a year by the 2020s. The book contradicts Devra Lee Davis of the U.S. Public Health Service, who argues that a sharp rise in U.S. cancers of the brain, breast, and testicles over 20 years may be linked to pesticides, air pollution, and other chemicals in the environment. Peto, by contrast, attributes the rise in these--and all other cancers--to smoking.
A former pack-and-a-half-a-day smoker himself, Peto earned a degree in mathematics and natural sciences from Cambridge University, plus a master's in statistics from Imperial College in London in 1967. By then he longed for real work. But his mentor, Sir Richard Doll--who in the 1950s confirmed the link between smoking and lung cancer--won a position at Oxford and persuaded Peto to join him. Collins arrived 12 years later as a new M.D. Intending to stay a year, he never left.
TOGETHERNESS. In an office crammed with stacks of papers and with memos taped to walls, the lanky Peto focuses on the big picture--conceiving and publicizing studies. Across the hall sits Collins, 39, who deals with researchers and designs drug reviews. They live with their girlfriends, 100 yards apart and a five-minute walk from the office, "in case we get desperate to talk," says Collins.
Both are motivated by something other than a desire for wealth--they each earn about $60,000 a year. "Anyone who names his sons Karl after Marx, Russell after Bertrand, Tom after Paine, and Leon after Trotsky is not after money," says Dr. Charles Hennekens, professor of preventative medicine at the Harvard School of Medicine, referring to Peto's four boys. "We must be saving a few tens of thousands of lives a year in breast cancer; another few tens of thousands from aspirin," says Peto of what drives him. His smoking estimates "could save hundreds of thousands of more lives a year," he adds. All three cases illustrate his strategy: Concentrate on common ailments, where minute drops in mortality benefit masses of patients.
Peto's method, inspired by a 1976 lunch with Tufts's Chalmers, who is often called the father of meta-analysis, is built around gathering data from existing studies and trying to make it all comparable. Some of the studies have thousands of patients, some a handful. Peto uses only randomized trials, in which half the patients get the drug being investigated and the rest another drug or placebo.
The controversial part of meta-analysis is merging the disparate studies. Say there have been a dozen trials examining one drug's effectiveness against cancer, all inconclusive because too few patients were involved in each one. If all 12 are lumped together, and some shrewd fudge factors are used to make the data more similar, the drug may be shown to work. The downside is that whether the result is legitimate or not depends heavily on the fudge factors.
Critics of Peto and Collins say their method is chancier than some, assuming greater homogeneity in trials and producing more definitive results. "I've disagreed with them on techniques," says Nan M. Laird, chair of the department of biostatistics at the Harvard School of Public Health. Not many complain publicly: Peto "intimidates statisticians because he's learned the clinical stuff, and clinicians because he knows statistics," says Laird. His regard for nonbelievers is low. "I'm sick of them saying I'm mixing apples and oranges. Can't they even find a different metaphor?"
Much of the company-sponsored work that Peto and Collins do isn't meta-analysis--it's traditional trials involving hundreds of doctors worldwide. Usually, "it's quite a long process to persuade companies" to sponsor such work, says Collins, since they don't get early peeks at the data. "But it's in their interest for us to be seen as independent. If we can show unequivocally that a drug works, it's a hell of a lot easier to sell."
The two see increasing threats, particularly in the U.S., to both large randomized trials and meta-analysis. The first is a new statistical tool called outcomes research, which includes analyzing databases, such as hospital records, to see what treatments work best. The U.S. Agency for Health Care Policy & Research, created in 1989, is funding such studies, which cost much less than randomized trials. Wrong move, argues Peto: "If you try to review nonrandomized evidence, you can't avoid the possibility of bias." Dr. Richard J. Greene, director of the agency's Center for Medical Effectiveness Research, disagrees. He calls outcomes research "useful for understanding current patterns in clinical treatment. It is not a substitute for randomized clinical trials."
More threatening, Peto worries, is the brouhaha over alleged fraud in a 2,100-patient National Cancer Institute trial comparing two surgical procedures for breast cancer. In March, a doctor at a Montreal hospital was found to have altered the ages of 6 of 354 women. It didn't affect the study's results, but the deception prompted Congress to double the budget for auditing such trials. Frets Peto: "They're going to make all sorts of extra rules. It'll reduce the number of patients in random trials and the number of doctors who do them. Unless these people are stupid, they've got to realize this."
Last March, Peto pleaded with medical writers to do nice stories on Dr. Bernard P. Fisher of the University of Pittsburgh, who led the breast cancer trial. He even urged Dr. Jerome Kassirer, editor of the New England Journal of Medicine, to dedicate an issue to Fisher and to "explain why large-scale randomized evidence...is so important." Kassirer declined.
Most of Peto's peers agree that his flair for figures more than compensates for such antics. "Just because he goes off the deep end sometimes shouldn't diminish the credit he gets," says Chalmers. Not much chance of that. Peto and Collins say they'll keep stirring up trouble--and trying, indirectly, to save lives.
PETO'S GREATEST HITS
Oxford University Professor Richard Peto and his partner, Dr. Rory Collins, use statistical studies to determine what causes killer diseases--and how to treat or prevent them. Some of their most influential findings:
They find that tamoxifen from ICI, now Zeneca Group, boosts the 5-year survival rate for breast cancer patients from 50% to 55%. Sales of the drug quickly double.
Their analysis of treatments involving 17,000 heart attack patients shows that deaths plunge 40% when victims are treated with a combination of Hoechst's streptokinase and aspirin. Use of both drugs for heart patients zooms.
They find that a $300 dose of strepto-kinase is as effective as t-PA, which Genentech sells for $2,200 a dose, for dissolving blood clots. t-PA sales suffer until subsequent research by Genentech shows it to be marginally more effective.
An analysis based on 120,000 patients shows that aspirin helps prevent heart attacks and strokes.A new Peto book, due in September, argues that the increase in cancers caused by smoking--which he predicts will reach 10 million annually by the 2020s--accounts for all of the increase in world cancer rates.