The Laws of Medicine (6 page)

Around the first week of my fellowship, I learned that another such drug, a molecular cousin of Gleevec's, was being
tested in our hospital for a different form of cancer. The drug had shown promising effects in animal models and in early human experiments—and an early trial was forging ahead with human patients.

I had inherited a group of patients on the trial from a former fellow who had graduated from the program. Even a cursory examination of the trial patients on my roster indicated a spectacular response rate. One woman, with a massive tumor in her belly, found the masses melting away in a few weeks. Another patient had a dramatic reduction in pain from his metastasis. The other fellows, too, were witnessing similarly dramatic responses in their patients. We spoke reverentially about the drug, its striking response rate, and how it might change the landscape for the treatment of cancer.

Yet six months later, the overall results of the study revealed a surprising disappointment. Far from the 70 or 80 percent response rates that we had been expecting from our data, the overall rate was an abysmal 15 percent. The mysterious discrepancy made no sense, but the reason behind it became evident over the next few weeks when we looked deeply at the data. The oncology fellowship runs for three years, and every graduating batch of fellows passes on some patients from his or her roster to the new batch and assigns the rest to the more experienced attending physicians in the hospital. Whether a patient gets passed on to a fellow or an attending doctor is a personal decision. The only injunction is that a patient who get reassigned to a new fellow must be a case of “educational value.”

In fact, every patient moved to the new fellows was a drug
responder, while all patients shunted to the attending physicians were nonresponders. Concerned that the new fellows would be unable to handle the more complex medical needs of men and women with no drug response—patients with the most treatment-resistant, recalcitrant variants of the disease—the graduating fellows had moved all the nonresponding patients to more experienced attending physicians. The assignment had no premeditated bias, yet the simple desire to help patients had sharply distorted the experiment.

....

Every science suffers from human biases. Even as we train massive machines to collect, store, and manipulate data for us, humans are the final observers, interpreters, and arbiters of that data. In medicine, the biases are particularly acute for two reasons. The first is hope: we
want
our medicines to work. Hope is a beautiful thing in medicine—its most tender center—but it is also the most dangerous. Few stories involving the mix of hope and illusion in medicine are more tragic, or more long-drawn, than that of the radical mastectomy.

By the early 1900s, during the brisk efflorescence of modern surgery, surgeons had devised meticulous operations to remove malignant tumors from the breast. Many women with cancer were cured by these surgical “extirpations”—yet, despite surgery, some women still relapsed with metastasis all over their bodies. This postsurgical relapse preoccupied great surgical minds. In Baltimore, the furiously productive surgeon William Halsted argued that malignant tissue left behind during the original surgery caused this relapse. He described breast-cancer surgery as an “unclean operation.” Scattered scraps of tumor left behind by the surgeon, he argued, were the reason for the metastatic spread.

Halsted's hypothesis was logically coherent—but incorrect. For most women with breast cancer, the real reason for postsurgical relapse was not the local outgrowth of remnant scraps of malignant tissue. Rather, the cancer had migrated out of the breast long
before
surgery. Cancer cells, contrary to Halsted's expectations, did not circle in orderly metastatic parabolas around the original tumor; their spread through the body was
more capricious and unpredictable. But Halsted was haunted by the “unclean operation.” To test his theory of the local spread of cancer, he amputated not just the breast, but a vast mass of underlying tissue, including the muscles that move the arm and the shoulders and the deep lymph nodes in the chest, all in an effort to “cleanse” the site of the operation.

Halsted called the procedure a
radical
mastectomy, using the word
radical
in its original meaning from the Latin word for “root”; his aggressive mastectomy was meant to pull cancer out by its roots from the body. In time, though, the word itself would metastasize in meaning and transform into one of the most inscrutable sources of bias. Halsted's students—and women with breast cancer—came to think of the word
radical
in its second meaning: “brazen, innovative, bold.” What surgeon or woman, faced with a lethal, relapsing disease, would choose the
non
radical mastectomy? Untested and uncontested, a theory became a law: no surgeon was willing to run a trial for a surgical operation that he
knew
would work. Halsted's proposition ossified into surgical doctrine. Cutting more had to translate into curing more.

Yet women relapsed—not occasionally, either, but in large numbers. In the 1940s, a small band of insurgent surgeons—most prominently Geoffrey Keynes in London—tried to challenge the core logic of the radical mastectomy, but to little avail. In 1980, nearly eight decades after Halsted's first operation, a randomized trial comparing radical mastectomy with a more conservative surgery was formally launched. (Bernie Fisher, the surgeon leading the trial, wrote, “In God we trust. All others
must bring data.”) Even that trial barely limped to its conclusion. Captivated by the logic and bravura of radical surgery, American surgeons were so reluctant to put the procedure to test that enrollment in the control arm faltered. Surgeons from Canada and other nations had to be persuaded to help complete the study.

The results were strikingly negative. Women with the radical procedure suffered a host of debilitating complications, but received no benefits: their chance of relapsing with metastatic disease was identical to that of women treated with more conservative surgery, coupled with local radiation. Breast cancer patients had been ground in the crucible of radical surgery for no real reason. The result was so destabilizing to the field that the trial was revisited in the 1990s, and again in 2000; more than two decades later, there was still no difference in outcome. It is hard to measure the full breadth of its effects, but roughly one hundred thousand to five hundred thousand women were treated with radical mastectomies between 1900 and 1985. The procedure is rarely, if ever, performed today.

....

In retrospect, the sources of bias in radical surgery are easy to spot: a powerful surgeon obsessed with innovation, a word that mutated in meaning, a generation of women forced to trust a physician's commands, and a culture of perfection that was often resistant to criticism. But other sources of bias in medicine are far more difficult to identify because they are more subtle. Unlike in virtually any of the other sciences, in medicine the subject—i.e., the patient—is not passive, but an active participant in an experiment. In the atomic world, Heisenberg's uncertainty principle holds that the position and momentum of a particle cannot simultaneously be measured with absolute accuracy. If you send a wave of light to measure the position of a particle, Heisenberg reasoned, then the wave's hitting the particle changes its momentum, and thereby its position, and so forth ad infinitum; you cannot measure both with absolute certainty. Medicine has its own version of “Heisenbergian” uncertainty: when you enroll a patient in a study, you inevitably alter the nature of the patient's psyche and, therefore, alter the study. The device used to measure the subject transforms the nature of the subject.

The active psyche of a patient, for instance, makes it particularly treacherous to run studies or trials that depend on memory. In 1993, a Harvard researcher named Edward Giovannucci set out to determine whether high-fat diets altered the risk of breast cancer. He identified a set of women with breast cancer and an age-matched cohort without breast cancer, then asked each group about their dietary habits over the last decade. The survey produced a pronounced signal: women with breast
cancer were much more likely to have consumed diets higher in fat.

But this study had a twist: the women in Giovannucci's study had also completed a survey of their diets nearly a decade
before
this study, and the data had safely been stored away in a computer. When the two surveys were compared, in women without breast cancer, the actual diet and the recalled diet were largely identical. In women with breast cancer, however, the actual diet had no excess of fat. Only the “remembered” diet was high in fat. These women had unconsciously searched their memories for a cause of their cancers and had invented a culprit: their own bad habits. What better blame than self-blame?

But don't prospective, controlled, randomized, double-blind studies eliminate all these biases? The very existence of such a study—in which both control and experimental groups are randomly assigned, patients are treated prospectively and both doctors and patients are ignorant of the treatment—is a testament to how seriously medicine takes its own biases, and what contortions we must perform to guard against them (in few other scientific disciplines are such drastic measures used to eliminate systematic biases). The importance of such studies cannot be overemphasized. Several medical treatments thought to be deeply beneficial to patients based on strong anecdotal evidence, or decades of nonrandomized studies, were ultimately proved to be
harmful
based on randomized studies. These include, among other examples, the use of high-dose oxygen therapy for neonates, antiarrhythmic drugs after heart attacks, and hormone-replacement therapy in women.

Yet even that desperate experimental contortion cannot eliminate the subtlest of biases. It's the Heisenbergian principle at work again: when patients are enrolled in a study, they are inevitably affected by that enrollment. A man's decision to
enroll
in a study to measure the effect of exercise on diabetic management, say, is an active decision. It means that he participates in the medical process, follows certain instructions, or lives in particular neighborhoods with accessible health care and so forth. It might mean that he belongs to a certain race or ethnic group or a particular socioeconomic class. A randomized study might make particular conclusions about the effectiveness of a medicine—but in truth it has only judged that effectiveness in the subset of people who were randomized. The power of the experiment is critically dependent on its strong limits—and this is the very thing that makes it limited. The experiment may be perfect, but whether it is
generalizable
is a question.

The reverential status of randomized, controlled trials in medicine is its own source of bias. The BCG vaccine against tuberculosis was shown to have a potent protective effect in a randomized trial, but the effectiveness of the vaccine seems to decrease almost linearly as we move in latitude from the North to the South—where, incidentally, TB is the most prevalent (we still don't understand the basis for this effect, although genetic variation is the most obvious culprit). These distortions—call them heuristic biases—are not peripheral to the practice of medicine. Virtually every day I'm asked to decide whether a particular drug will work for a patient—an African-American man, say—when the trial was run on a population of predominantly
white men in Kansas. Women are notoriously underrepresented in randomized studies. In fact, female
mice
are notoriously underrepresented in laboratory studies. Extracting medical wisdom from a randomized study thus involves much more than blithely reading the last line of the study published in some august medical journal. It involves human perception, arbitration, and interpretation—and hence involves bias.

The advent of new medical technologies will not diminish bias. They will amplify it. More human arbitration and interpretation will be needed to make sense of studies—and thus more biases will be introduced. Big data is not the solution to the bias problem; it is merely a source of more subtle (or even bigger) biases.

Perhaps the simplest way to tackle the bias problem is to confront it head-on and incorporate it into the very definition of medicine. The romantic view of medicine, particularly popular in the nineteenth century, is of the doctor as a “disease hunter” (in 1926, Paul de Kruif's book
Microbe Hunters
ignited the imagination of an entire generation). But most doctors don't really hunt diseases these days. The greatest clinicians who I know seem to have a sixth sense for biases. They understand, almost instinctively, when prior bits of scattered knowledge apply to their patients—but, more important, when they don't apply to their patients. They understand the importance of data and trials and randomized studies, but are thoughtful enough to resist their seductions. What doctors really hunt is bias.

....

Priors. Outliers. Biases. That all three laws of medicine involve limits and constraints on human knowledge is instructive. Lewis Thomas would not have predicted this stickiness of uncertainties and constraints; the future of medicine that Thomas had imagined was quite different. “The mechanization of scientific medicine is here to stay,” he wrote optimistically in
The Youngest Science
. Thomas presaged a time when all-knowing, high-precision instruments would measure and map all the functions of the human body, leaving little uncertainty and even fewer constraints or gaps in knowledge. “The new medicine works,” he wrote. “The physician has the same obligations that he carried, overworked and often despairingly, fifty years ago—but now with any number of technological maneuvers to be undertaken quickly and with precision. . . . The hospitalized patient feels, for a time, like a working part of an immense, automated apparatus. He is admitted and discharged by batteries of computers, sometimes without even learning the doctors' names. Many patients go home speedily, in good health, cured of their diseases. . . . If I were a medical student or an intern, just getting ready to begin, I would be more worried about this aspect of my profession. I would be apprehensive that my real job, taking care of sick people, might soon be taken away, leaving me with the quite different occupation of looking after machines.”