Few drug-trial studies have the time or money to study the actual health outcomes that people care about, such as whether the middle-aged man avoids a heart attack after a few decades, the hyperactive first-grader holds down a good job someday, or the menopausal woman remains free from a hip fracture when she's elderly. Waiting for these events would stifle any meaningful innovation, so doctors pick surrogate endpoints, which they hope serve as short-term checkpoints. Thus drugs trials for the preceding examples may just decide to measure the middle-aged man's cholesterol level, the youngster's symptom checklist for hyperactivity, and the woman's bone density with a DEXA scan.
Unfortunately, surrogate endpoints can be problematic for two key reasons. First, their relation to the actual outcome of interest may be weak or nonexistent (just as impressing Donald Trump may fail to predict later business success); second, the intervention can improve the surrogate outcome but have bad side effects (just as the nonstop fancy dates that titillate viewers later lead to the couple's inevitable romantic disappointment).
Consider high serum cholesterol levels, which are a surrogate marker for later heart attacks. In the 1970s, a drug called clofibrate was discovered to markedly reduce cholesterol levels; unfortunately, a World Health Organization trial found it increased heart attacks. In 2002, the Women's Health Initiative trial showed hormone replacement caused blood clots and increased heart attacks by 29 percent, though it also made cholesterol numbers better. In 2006, Pfizer abruptly pulled the plug on its drug torcetrapib, which had cost roughly $1 billion to develop and had great effects on cholesterol levels—but also caused more heart attacks. Keep in mind that cholesterol levels, like blood pressure or levels of tumor markers in the blood, are at least considered "validated" markers.
In part, the increased focus on surrogate markers instead of hard outcomes came about in the 1990s, when the U.S. Food and Drug Administration was pressured to get HIV drugs to market rapidly for desperate patients, at the expense of using even "unvalidated" surrogate endpoints (like CD4 cell counts) that at the time were "reasonably likely" to predict benefit. For every turkey like clofibrate or torcetrapib, there are occasionally life-saving breakthroughs like protease inhibitors for HIV that may depend on a dizzying array of complicated surrogate endpoints.
How, then, can we encourage approval of drugs like protease inhibitors and cut the number of failures like clofibrate? The truth is that you can't, just like you can't easily figure out how to guarantee a happy marriage or find a surefire pop icon. You do the best with the information you have, then wait and see. In his terrific 2004 book Powerful Medicines, Jerry Avorn proposed a two-phased drug evaluation process in which "the initial FDA approval of a drug should be seen as the beginning of an intensive period of assessment, not the end." No doctor, he writes, would ever start a patient on a new medicine without scheduling any kind of follow-up, but that's exactly what the U.S. health care system now does.
More ...
