Five years ago, at the American Heart Assn.'s annual scientific meeting, reporters crowded into a room at the Dallas convention center for details of an exciting finding in women's health. A $20 million government-funded study had found that giving estrogen and another hormone to women after menopause boosted their levels of so-called good cholesterol, cutting their heart disease risks by a spectacular 25%. "If my mother were alive now, I would want her on the hormones," said one of the study's authors.
The finding made perfect sense: Women's estrogen levels fall at menopause, and the incidence of heart disease climbs. Giving estrogen seemed the obvious solution.
The story is not so simple, however. The study showed only that the women had improved cholesterol levels. It did not show that they had fewer heart attacks or deaths. When the follow-up study was done, researchers were surprised to find no reduction in heart attacks or heart-related deaths. The point is that beneficial effects on cholesterol, blood pressure, or other indicators are not sure signs that the drugs will actually reduce disease and death.
DANGEROUS. The case demonstrates why the FDA needs to tighten its rules for drug testing. Drugs to treat heart disease, obesity, and other ailments are routinely approved by the Food & Drug Administration on the basis of their beneficial effects on such physiological measures. For many of those drugs, the follow-up studies--looking at effects on disease and death--have never been done. "In the late 1990s, millions of Americans are taking antihypertensive and antidiabetic therapies that have not been adequately evaluated in large-term clinical trials," Dr. Bruce M. Psaty of the University of Washington and his colleagues wrote in the Aug. 25 issue of the Journal of the American Medical Assn.
Among the examples, Psaty says, are the treatment of hypertension with so-called ACE inhibitors, such as AstraZeneca PLC's Zestril and Merck's Vasotec, and calcium-channel blockers, including top-selling Norvasc from Pfizer. According to IMS Health Inc., which tracks drug usage, doctors wrote 99.7 million prescriptions for ACE inhibitors and 91.5 million prescriptions for calcium-channel blockers in the year that ended in June.
Some drugs turn out to be dangerous upon further testing. A cholesterol-lowering drug called clofibrate was found to lower the risk of heart disease, but it also boosted overall mortality by a staggering 44%, Psaty notes. Even therapies such as the estrogen treatment, which had no effect, can be a public-health problem. They consume scarce health-care dollars, and patients may be given these drugs as a substitute for other, better treatments.
SHORTCUTS. The issue arises from the FDA's practice of approving drugs on the basis of scientific shortcuts. If a drug is shown to reduce cholesterol or to lower high blood pressure, the FDA assumes it will reduce deaths and heart attacks. So the drug is approved. Physiological changes such as reductions of cholesterol or blood pressure are referred to as "surrogate endpoints."
These surrogate studies are of great value in drug testing. It is cheaper and faster to determine a drug's effect on cholesterol or high blood pressure than its effect on disease and deaths. As Psaty notes, a study showing cholesterol reduction might require 100 patients and take less than a year. A study showing effects on heart attacks could require several thousand patients and take five years to complete--because that's what's needed to tally enough heart attacks for the study to be valid. Such studies are consequently far more expensive: Costs can reach tens of millions of dollars.
Use of the surrogate endpoints can get new drugs to patients much sooner. That was critically important in drugs for AIDS, for which there were no effective treatments. Drugs were approved on the basis of their favorable effects on the immune system. Only later was it shown that the treatment effectively fought the disease itself.
RISKY RELIANCE. Critics say, however, that reliance on surrogate endpoints is risky. "It's probably best not to rely on surrogate endpoints unless you have extremely strong evidence to support doing so," says Dr. Gordon H. Guyatt of McMaster University in Hamilton, Ont., who also wrote on the subject in the Aug. 25 issue of the Journal of the American Medical Assn. The only way to be certain that drugs reduce disease and deaths is to study the question with what are called randomized, double-blind clinical trials. In these trials, patients are randomly given either the drug in question or a placebo. Neither the patients nor the doctors treating them know who is getting the drug until the clinical trial is completed.
Asked to comment about the two articles, Pfizer spokesman Andrew B. McCormick said: "We certainly think lowering blood pressure is a good endpoint, and we believe the data has validated that position." Dr. Laurence Hirsch, vice-president of public affairs for Merck Research Laboratories, noted that many follow-up studies would have to compare new drugs with existing drugs, because giving a placebo is unethical when effective treatments exist. "It complicates things because the studies have to be enormous," Hirsch added. "It's very, very difficult."
Dr. Robert Temple, associate director for medical policy at the FDA's Center for Drug Evaluation & Research, thinks the concern about surrogate endpoints is somewhat overblown. He acknowledges that ACE inhibitors have not been proven to provide better outcomes for people with high blood pressure but notes that they've been tested in many other circumstances. "They've been shown to improve survival after heart attack and virtually every degree of heart failure there is. Do I think they are likely to not have the expected benefit in hypertension? Of course not," he says. Temple worries that if large trials were required for all drugs, some might never reach the market. "It would discourage development," he says. "I think people wouldn't bother."
That's the essence of the problem: Requiring proof of a real health effect for every drug before FDA approval may delay the arrival of some effective drugs for years, or perhaps prevent some of them from ever being approved. On the other hand, continuing the current practice could allow harmful drugs to reach the market, causing needless suffering and deaths. Or it could mean that
billions are wasted on treatments that are harmless but ineffective.
There is, however, a way out. The FDA itself has suggested an alternative by putting some drugs into what it calls "accelerated approval." When there is an extraordinary need for a new drug to reach the market to save lives, the FDA requires that larger trials be done after the treatment is approved on the basis of surrogates. Psaty and his colleagues favor extending that approach to other drugs that are not part of the accelerated-approval process. It's a kind of middle road: Drugs are not delayed on the way to the market, but pharmaceutical companies would be required to follow up with the larger, expensive studies. In this case, they would be doing the costly studies when revenues from a drug have already begun to flow in.
Under this approach, drug companies would be allowed to rely on surrogate testing only in cases where there is no doubt that reliance on the surrogates is safe. The cost of testing drugs would rise, and the pharmaceutical industry might see that as a burden. But those costs are likely to be far outweighed by the savings from pulling ineffective or harmful drugs off of the market.