Franklin and Mesmer had different ideas about how to answer this question. Mesmer suggested an experiment much like Lind’s: take two patients with the same disease, mesmerize one of them and not the other, make all other conditions equal, and see who fared better. But Franklin, the wily rationalist, understood that there was a bigger obstacle to the truth: self-interest, especially that of the doctor and his patient. He designed a test to eliminate human subjectivity from the experiment.
Franklin’s proposal also eliminated Mesmer, who, upon hearing of it, withdrew from the proceedings. He sent another mesmerist to Franklin’s house on the appointed night, willing patients in tow. In front of the commission, he focused the magnetism on parts of their bodies. Asked to locate where he was directing the energy, the patients—all women—responded accurately. Then they were blindfolded. When the mesmerist repeated the procedure, the women located the sensations, according to the commission’s report to the king, “at hazard, and in parts very distant from those which were the object of magnetism.” Other variations of the blindfold test yielded the same results. “It was natural to conclude,” Franklin told the king, “that these sensations, real or pretended, were determined by the imagination.”
Lind claimed to have proved that citrus cured scurvy, but Franklin seems to have understood that this was more than an experiment could say, and that there is an inexhaustible supply of variables, known unknowns and unknown unknowns alike, that might have been at work in mesmerism. He controlled for the one he deemed most likely—imagination—and when he did so, there was a difference in outcome. Or, to put it another way, he started with the idea that there would be no difference between a blindfolded and a non-blindfolded treatment—and disproved it. He didn’t prove anything; his conclusion from the proceedings might have been natural, but it was also inferential.
This may seem like a distinction without a difference, especially when you consider the different purposes of these experiments: Lind’s to ratify and Franklin’s to debunk. But what Franklin seems to have understood was that enthusiasm for a treatment wasn’t just another variable in the pursuit of scientific knowledge. It was the enemy. Self-interest, hope, the ineffable qualities of the doctor-patient relationship—in short, subjectivity—would always haunt our attempts to understand the world, and the role of the experiment was to rein in its effects, whether by tying on a blindfold or acknowledging the limits of a controlled experiment to establish the truth.
The modern RCT is much more sophisticated than Franklin’s experiment. But it begins with the same recognition of our limited ability to circumscribe credulity and follows the same logic; it starts with a null hypothesis—that the treatment won’t work—attempts to confirm it, draws inferences from the results, and then tries to strengthen those inferences by replicating the experiment. In citing Lind, Lasagna seems to have forgotten that an RCT is much more suited to disproving than to proving, that it can only give us probabilities, that its primary purpose is negative: to rain on an experimenter’s parade, to put the kibosh on therapeutic enthusiasm rather than to inflame it.
In 1928, twenty-five years before Lasagna touted the virtues of the blinded RCT, and just a year before the stock market crashed, nervous investors wondered if something was wrong in the house of Morgan. Anne Morgan, sister of J.P. and usually no less retiring than the rest of the family, had suddenly turned up as a paid spokesman for Old Gold cigarettes. She reported to newspaper readers that she had “taken the blindfold test, smoked four brands of cigarets [sic], and found that ‘the smoothness’ of one cigaret [sic] was ‘so obvious.’” Miss Morgan urged other smokers to repeat the experiment in the privacy of their own parlors. Advertisers had evidently figured out that by reassuring the consumer that his tastes were not a figment of his imagination, that his own ever-unreliable subjectivity had been neutralized, they could build brand loyalty. They had hit upon a way to use the blindfold test to stoke enthusiasm rather than to curb it. (Miss Morgan’s finances, as it turned out, were sound; she apparently donated her thousand-dollar honorarium to charity.)
A decade later, Cornell scientist Harry Gold (no relation to Old) was trying to figure out whether or not xanthines, a group of stimulants that included caffeine, really deserved their reputation as a remedy for angina. He realized that he couldn’t trust the data provided by doctors studying the question. They asked leading questions, assigned patients nonrandomly to get the drug or placebo, and interpreted ambiguous results in a way that favored the drugs. It wasn’t enough, Gold concluded, to control for patients’ credulity; doctors also had to be placed in the dark. Experiments had to be double-blind, Gold said. He cited the Old Gold campaign as the inspiration for the method’s name.
Gold was using the blindfold test as Franklin had intended—to impose restraint. He would probably be discomfited to see the ease with which his method has been used to create certainty rather than to limit uncertainty. But perhaps no one would be more upset at the way that RCTs have become one of the drug industry’s greatest marketing tools than the British geneticist who invented the mathematical language in which RCTs are reported—a language intended, like the experiments themselves, to rule out rather than to rule in.
Ronald Aylmer Fisher, who was so blind that he had to do his calculations in his head, developed modern statistics while working for an agricultural research institute just after the First World War. Fisher was trying to sort out fact from opinion when it came to crop yields. A farmer could plant two different varieties of grain and at harvesttime reap a much bigger crop from one of them, but most scientists understood that this didn’t necessarily mean one strain was more vigorous than the other. Maybe the soil varied from plot to plot, or the exposure to sunlight, or the population of varmints. Without isolating those variables, there was no way to know if the change was due to what the farmer did on purpose or to what just happened to occur.
Fisher’s solution was to divide the field into strips, randomly assign different strains (or different fertilizers or tilling practices or some other independent variable) to different strips, and then measure the differences in yield (or color or time to maturity or some other dependent variable). This approach may seem obvious, and indeed farmers had been doing something like it for centuries, but Fisher was the first to figure out how to systematize the procedure and to express the results in numbers.
To do this, however, Fisher first had to change his colleagues’ minds about what they were doing. While most scientists thought that ignorance was their greatest enemy, that with their experiments they were establishing the bedrock facts about how nature works, Fisher, like Franklin before him, saw it differently. To him, the adversary of knowledge was chance. Nature’s randomness was a wily foe, one that could never be fully accounted for. Indeed, as Fisher’s biographer put it, the best a scientist could do was
to equalize the chance that any treatment shall fall on any plot by determining it by chance himself. Then if all the plots with a particular treatment have the higher yields, it may still be due to the devil’s arrangement, but then and only then will the experimenter know how often his chance arrangement will coincide with the devil’s.
The reason to plant one kind of wheat in one row and another in the next is not so much to prove that one variety is superior. Rather, it is to minimize the possibility that the differences between them were the result of random events, which in turn increases the possibility that what the humans did, rather than what just happened to occur, is responsible for the outcome.
Randomization may have put nature’s perversity into the hands of the scientist, but only at a cost. You could crunch the numbers all day long—and Fisher developed a series of formulas for just this purpose—but all they were going to tell you was the probability that you had merely caught the devil at play. That’s why statistics-guided studies are more like Franklin’s experiment than Lind’s. They are not designed to confirm a positive hypothesis—that, say, basmati rice grows faster than long grain brown�
��but rather to disconfirm the hypothesis that any differences between the strains are due to random chance, which is to say that there is no real difference. Repeatedly disprove the null hypothesis and you will gain some certainty that the result most likely wasn’t random. In a fallen world, inference was the best that a scientist could do, and the job of the numbers was to specify how confident he could be about those conclusions.
But when Lasagna and other champions of statistics-guided RCTs adopted Fisher’s methods, they did not also adopt his humility. Researchers using statistics-guided double-blind methods placed credulity in the devil’s camp—as part of the background noise from which the drug’s signal could be discerned. They claimed that the method would “free a researcher from the accusation that his beliefs had affected a study’s execution” and that therefore the study itself had uncovered the plain facts. It also meant that Miss Morgan’s confidence could now be expressed in numbers—a language that, like the method itself, seems beyond the manipulations of self-interest and the vagaries of imagination. Drug scientists didn’t have to settle for an alliance with the devil, statistics that could only measure the uncertainty turned up by an experiment designed to dampen enthusiasm. Instead, they could defeat the devil, claim certainty about their results, and use them to assert that their drugs really work.
Some doctors saw immediately how easily the combination of statistics and the RCT could be used to make experiments say more than they really could. Some of them were even bothered by this. At a drug evaluation conference in 1958, one participant said:
I have an intense prejudice against the mass of statistics that accompanies the introduction of new drugs. It seems to me that statistics…are too often misleading. Many people believe that they eliminate chance when in fact they merely give an idea as to the probability of the results being due to chance.
The difference between “eliminat[ing] chance” and “the probability of the results being due to chance”—or between proving efficacy and disproving a null hypothesis—may seem minor. And sometimes it is. Virtually any diabetic given the right amount of insulin is going to survive, just as almost anyone with bacterial pneumonia will get better with an antibiotic. You don’t need an RCT to prove this. If you ran one, you wouldn’t need statistics to interpret the results. And if you did use statistics, there would be no meaningful difference between saying you had disproved the null hypothesis and that you had confirmed that the drugs worked. The restored health of nearly all the patients would speak for itself.
But if you had a drug-approval scheme that allowed regulators to ignore the preponderance of evidence and a drug that was no more effective than an inert substance at fixing the problem it claimed to fix, that achieved these meager results only after the best minds had concocted the conditions under which the drug was most likely to succeed, and that impaired a patient’s sex life and caused him to gain weight and sometimes even made him suicidal—if, in short, you had antidepressants, then you would really want to pay attention to that distinction. You would want to bear in mind the warning of one early statistician:
Let the experimenter who is driven to use statistical methods not forget this, that the very fact that he is compelled to use statistical methods is a reflection of his experimental work. It shows that he has failed to attain the very object of his experiment and exclude disturbing causes.
You would want to remember that when it comes to disturbing causes there’s nothing more disturbing, or harder to exclude, than consciousness, especially when it comes to drugs that change it, and that if those statistics mean anything, if they have any significance at all, it is not that your drugs work. It is, at best, that sometimes they don’t not work, and that even when that happens it may not be the result of the drug inside the pill.
But that kind of caution is not going to move product. It is better to cite the numbers as proof that the drugs cure depression and leave out all those picky details—not to mention all those other numbers from all those studies that don’t count, the ones that say that the drugs don’t work. Then the RCT becomes the opposite of what it was intended to be—not a device to debunk the claims of parties with a stake in the outcome, but a way to harness the power of science to justify their enthusiasm by putting it into the persuasive language of numbers. As it had for Kraepelin, and as it had for the advocates of the catecholamine hypothesis, the rhetoric of science, if not science itself, was going to lend authority to the pronouncements of the drug experts. It would give them the appearance of objectivity, allow them to claim that they were not expressing mere opinions but rather revealing what was really in nature.
And that’s not all. What the FDA meant by efficacy was proof that a drug was effective with a particular disease. Suddenly, it was more important than ever to find an indication, the specific illness on which the drug could be tested and for which it could be approved (and advertised) as a treatment. The fate of a new drug was now tied to the skill with which doctors identified and isolated a disease for it to treat. For a disease like depression, whose definition had proved changeable and whose very existence had been questioned, those government-sanctioned numbers now had a hidden function: they could give the public confidence not only that a drug worked, but that its target was a genuine disease. It was only a matter of time before an official, statistics-friendly definition of depression emerged, one that indicated that vast numbers of people harbored that disease in their brains and should turn their discontents over to their doctors.
CHAPTER 11
DIAGNOSING FOR DOLLARS
Sometime during my first session with a new patient, usually toward the end, the touchiest subject in therapy comes up. Which is not what you might think it is. People will pour out, sometimes unbidden and in more detail than you may want, their illicit romps, their most ignoble or kinky wishes, their peccadilloes and deceptions, and other carnal secrets long dammed by shame and other family traditions. But ask these same people how much money they earn or have in the bank and they will seize up like a frozen pipe. Sometimes they even tell me that it’s none of my business, and, as a rule, the more affluent they are, the more likely they are to feel that way. It is as if they are putting me on notice that there is a nakedness that even I’m not allowed to see.
So when the m-word arises in that first session, it’s not because I’m asking. I’ve learned that much in twenty-five years. It’s because despite the intimacy of the encounter—which, by the way, is real; nothing draws people closer than simple, honest talk—we are doing business. I’m renting myself out by the hour, so as the clock ticks down, just as they might wonder if a first date will end with a handshake or a kiss or a wave good-bye, patients find themselves thinking about the transaction that will bring our time to a close. And more often than not, the way they let me know this is to say, “Do you take my insurance?”
This often ends up being a much more interesting conversation than patients bargain on. I tell them that I would be glad to submit the paperwork to their insurance company so that they can get reimbursed for what they have paid me. If they can convince me that their cash flow can’t handle that approach, I’ll even wait for the insurance company to pay up, although, I remind them, the responsibility to pay me is ultimately theirs, and they need to bear in mind that insurance companies make their profits by paying for as little health care as the law and, occasionally, common decency will allow. And then I tell them what I must do to have a chance of opening up the insurance coffers. “I’ll have to tell your insurance company that you have a mental illness.”
You wouldn’t think that would be a big surprise; they don’t call it medical insurance for nothing. But very often it is. In general, the people who come to see me for therapy don’t think that they are mentally ill. Mostly, they think they are worried or unhappy or reeling from setbacks. So they are often nonplussed when I point out the obvious—that they will soon be officially sick in the head, that this fact will be part of their permanent medical record, that shou
ld they seek life insurance or a security clearance or high political office, should the nosy manager in the human resources office happen to get a look at their file, they might come to regret having received that diagnosis.
This discussion sometimes ends with a decision to skip the insurance, which spares me paperwork and which I may reward with a reduction in fee. But in many cases, people feel like they must rely on their benefits, so the discussion next turns to the question of which mental illness they have. Sometimes I tell a patient what I think. If he has come in unshaven and rumpled and tells me it’s the first time he’s been out of bed in a couple of weeks and that he’s been feeling sad and guilty and apathetic, that he doesn’t eat and is considering suicide, then that’s an easy call: if we’re using his insurance, he’s got to have major depressive disorder. But other times, I take out the DSM and give a patient some choices.
Thanks to the authors’ foresight—some of them were private practice psychiatrists, after all—we can usually find a disorder that fits the facts without sounding too terrible. Adjustment disorder—which is what the DSM calls “a psychological response to an identifiable stressor or stressors that results in the development of clinically significant emotional or behavioral symptoms,” and which you can have in various emotional flavors, including with depression, with anxiety, or unspecified—is an obvious favorite; there is no shortage of identifiable stressors in everyone’s life, and simply coming to my office indicates clinical significance. So is generalized anxiety disorder, in which a patient has “excessive anxiety and worry, occurring more days than not for a period of at least six months”; Dr. Dording has already provided an excellent example of how slippery “excessive” can be in capable hands, and on the days you are short of things to worry about, there’s always global warming and the coming exhaustion of the oil supply. The only problem with these diagnoses, I explain, is that they are not on the relatively short list of disorders that insurance companies are obliged by the law in my state to pay for in the same fashion that they pay for any other illness. So the benign diagnosis might lead to reduced benefits.
Manufacturing depression Page 24