It's Really Hard to Lie Convincingly

You know how ads for dietary supplements show up on your Facebook feed, and they're so convincing that you feel utterly compelled to buy them? You know how politicians plug their favorite policies, and you are utterly powerless to disbelieve what they are saying? Or have you noticed that whenever celebrities appear in commercials, you automatically believe everything they say about the product?

Insert Homer Simpson saying, "Marge, it takes two to lie. One to lie and one to listen."

Of course you don't actually fall for any of these tricks. It's not that you automatically assume that everything you're told from these sources is false. It's just that you discount the claims. I'm using "discount" in the economist's sense of "decrease by some factor", not "utterly disregard." You treat the truth of these claims as something between 0% and 100% accurate, but probably not close to 100% until you do a more thorough vetting. If the claim seems especially intriguing or credible, you might do some research to get more confident in the claim. But people who don't have the time or sophistication to look into marketing propaganda don't just automatically believe these claims. Mostly, they ignore them. Bryan Caplan puts it well in The Myth of the Rational Voter. On the notion that the public is easily misled by propaganda, he writes:

Ignorant does not mean impressionable. When you walk onto a used car lot, you may be highly ignorant, but you can still discount or ignore the words of the salesmen who shout, “You won’t get a better deal anywhere else!”

Scott Alexander made a similar point here about Ted Cruz, who was very good on the college debate team. Does this make him especially believable? Or should we, in fact, discount everything he says to an even greater degree given that he’s superficially convincing? 

You were on your college debate team, and you were good at it. Really good. You won the national championships and you were pretty widely believed to be the best debater in the country. Quite an achievement. But my worry is – which is more likely? That the best debater in the country would also be the best choice for President? Or that he would be really really really good at making us think that he would be?
…
Senator Cruz, you may not quite be at the superintelligence level, but given that you’ve been recognized as the most convincing person out of all three hundred million Americans, shouldn’t we institute similar precautions with you? Shouldn’t your supporters, even if they agree with everything you are saying, precommit to ignore you as a matter of principle?

Do read the part I skipped over with an ellipse; it’s a very amusing hypothetical about an ultra-persuasive artificial intelligence. (In fact, read the entire piece, which is great.) There is no good way to be superficially convincing or to specialize in being convincing. People simply discount your claims to offset whatever points you gain from being convincing. 

My attempt to say the same thing here:

Even if they do make a good-faith effort to follow along, if you lead them to a conclusion they don’t like, they’ll assume you “tricked” them somehow. After all, you had time to prepare and plan, and they’re seeing your argument for the first time. Keep that in mind when you think you've just presented a killer argument and people don't instantly bow down to your awesomeness. The psychology here is, "That's too easy, so it must be some kind of trap or trick." This puts an upper limit on how convincing any argument can be: the more inherently convincing the argument, the greater the instinctive recoil, and the greater the effort to explain it away.

To take another example, Bill Clinton was an accomplished liar, but he had a tell. Supposedly Paul Ekman (the micro-expressions guy, made famous by Malcolm Gladwell's Blink and the TV series Lie to Me) noticed that he had a facial tick, his tell, that gave away when he was lying. He offered to coach him on it, but the Clinton staffer Ekman dealt with decided that it was imprudent for Clinton to meet with an expert on lie detection. This staffer realized that it would be damaging to Clinton's reputation if the public found out he'd met with a "lying coach." This staffer, whoever s/he was, showed tremendously good judgment. Had Clinton invested in his lying skills, that information would impugn the truth of everything he said. Bad optics, to say the least. This story is very much contrary to the narrative of a deep and secretive state with a sophisticated propaganda machine. The staffer assumed, probably correctly, that no information is truly secret, even a private appointment of a sitting president. If you have employed tactics that make you more superficially convincing, the information that you've done so makes you less convincing. It makes people doubt your claims even more.

There is a certain type of pundit who thinks that the public is generally quite powerless and easily mislead by clever advertisements and propaganda. Sophisticated operators with lots of resources can simply bend public opinion toward their product or their political machine. Thus we hear so many complaints about misleading advertising, or propaganda machines like the Koch brothers or George Soros. I think this is all bunk. A dietary supplement making pseudo-scientific claims about the effectiveness of its products will mostly be ignored. Those individuals who are actually motivated enough to read their Facebook post or watch a short ad video will be motivated enough to do a quick Google search and turn up some skeptical reviews. Lazy shoppers will attempt to buy on Amazon, where their credit card is already on file and one-click-shopping is available, and where merely scrolling to the bottom of the page will reveal negative 1-star and 2-star reviews. We likewise discount the claims of politicians and pundits. We lazily decline to question our priors, so we reflexively assume that the Soros' and Koch's of the world are somehow out to get us. Mix this with the intellectual laziness of the average person, who barely reads anything at all and who can barely be bothered to challenge their Bayesian priors, and the Soros/Koch message fails to even get a hearing.

It's not that people are easily fooled. It's more that some people want to be fooled. Ideology makes people believe the Koch- or Soros-funded political ad. Our ideologies are comforting to us, so we indulge them. Something similar is happening with people who believe that vaccines cause autism, or people who believe that 9/11 was an inside job or other crazy conspiracy theories. Interestingly enough, these weird beliefs tend to cluster together in a weird "listens-to-Alex-Jones" counter-culture. I've known a few people who fit that profile. Some of them are extremely intelligent, by the way, and not-necessarily-wrong-about-everything. I'm not sneering. But it's hard to escape the fact that weird beliefs are part of these people's identities. It's not that they've been snookered by sophisticated liars. (like...Alex Jones? Like Andrew Wakefield? Are these people really smooth operators?) It's more like they're engaging in a mutually agreed upon hallucination. Likewise, I think that the Theranos investors weren't "fooled" by Elizabeth Holmes, at least not in the sense that Holmes told them a story and they were compelled to utterly believe it. They all had enough information to see that something was amiss, that financial projections were wildly overoptimistic, that Theranos wasn't being totally transparent about these issues. Supposedly Henry Kissinger saw the writing on the wall and wanted out early; the book Bad Blood describes and episode in which other investors are discussing his departure. So they all had clues something was wrong. They were fooled because they decided to be fooled. They chose to believe something that would pay off big-time if true, like a Pascal's Wager. They probably all had portfolios of many Pascal's Wagers, a few of which might actually pay off big time and make all the failed ventures worthwhile. I think a similar thing is happening with Baby Einstein and those video games that claim to enhance your brainpower and stave off Alzheimer's. The FTC hit Luminocity with a $2 million fine. I don't think it's the case that these companies made sophisticated marketing claims and overpowered the judgment of their customers. I think parents wanted to believe there was an easy way to make their babies smarter, and adults wanted to believe that something so simple as playing video games could improve their brain power. These are examples of mutually agreed-upon hallucination, not outright lying or successful deception of unwilling dupes.

Another example that comes to mind is the so-called opioid epidemic. Under the standard narrative, aggressive marketing by pharmaceuticals kicked off the over-prescription of painkillers, which led to rising rates of addiction (or maybe not) and overdose deaths. Supposedly the "sophisticated" actors in this exchange were the pharmaceutical reps, and the poor dupes were the doctors who believed their marketing literature. Something is very wrong with this story. The sophisticated deceivers are marketing people, who don't necessarily have any kind of medical or scientific training, whose job is to basically memorize and recite a marketing script, and all of this is of course known by the doctor. (Yes, remind me that doctors routinely misinterpret the statistical implications of test results, or recite to me the stories of male doctors being beguiled by attractive female reps. These are still very educated people with the wherewithal to challenge an deceitful marketing attempts.) The dupes in this story are, how to put this, the most educated profession in society. Sure, the pharmaceutical reps have the resources of a big corporation to back them. But, once again, this is known to the doctors who are the targets of this marketing. The notion that they just uncritically believe all marketing claims is ludicrous. They know that someone makes a profit off them if they are successfully duped. They know the incentives faced by the reps and the pharmaceutical companies (in this case Johnson and Johnson and Purdue). They discount their claims accordingly. If they were "duped," it's because the claims made by Purdue and Johnson and Johnson were actually credible and supported by the medical literature. Chronic pain was in fact very badly under-treated before the mid-1990s. If these companies convinced doctors to treat pain more seriously and to be less afraid of opioid patients becoming addicts (which by all accounts is quite rare), then they performed an important public service. This narrative would be slightly more believable if the pharmaceutical firms were duping individual patients. But they were supposedly fooling doctors, with their FDA-reviewed marketing materials (which some reps admittedly deviated from) on FDA-approved medications.

(My take on the opioid narrative is written down here.)

A clearer example of outright dishonesty would be the makers of Vioxx basically deleting three heart attacks from their clinical trial data to make the drug appear safer than it really was. Lying with a secret cover-up or hard-to-discover sleight of hand is very different from "lying" in the light of day by exaggerating or misstating the conclusions of publicly available research. A victim of the former can legitimately claim to have been duped; a victim of the latter is deliberately swallowing the marketing hype, which everyone knows damn well is exaggerated. (An aside: If Vioxx had remained a viable alternative for pain treatment, it's possible that not quite as many people would have been prescribed opioids. Even if the marketing was deceptive, there were probably some patients for whom it was an appropriate treatment, even given the risks. Some trade-offs have multiple bad options. Taking one of them off the table hardly helps.)

I think there needs to be a space for this concept in our legal system and our moral judgments. There is such a thing as a believable but not-strictly-supported exaggeration, which in retrospect turns out not to be true. If anyone's worldview has unsophisticated consumers and voters being helplessly duped by well-financed lie machines, this requires at least a gentle correction. Maybe even a serious overhaul.

_______________________________

I wasn't sure how to insert this into the flow of the post, so I'll put it here at the end. See this post by David Henderson on Obama's whopper "If you like your current health insurance plan, you can keep it." Read the comments, where some of Obama's defenders claim (paraphrasing here) "C'mon! Nobody really believed that!" It was such an obvious whopper, even at the time, that the statement must be interpreted in the narrow context in which he made it. On the one hand, I feel like this kind of reflexive lying by politicians is reprehensible. On the other hand, it's hard to believe that anyone was legitimately fooled. It's still important to hold the line and call this stuff out when it happens, but this notion that sophisticated liars rule the world with their masterful deceptions is way overblown.

Chronic Illness and Opioid Poisoning Deaths

An unappreciated piece of the opioid narrative is the fact that a large proportion of the people who die from taking prescription opioids are sick with some sort of chronic illness. I wrote about this a few years ago, but wanted to reiterate the point and tell you how I'm arriving at this conclusion. I'll lead with my conclusion: About 1/4 to 1/3 of prescription opioid deaths also involve some kind of chronic illness. (More precisely, in 2017, filtering for "accidental" deaths and filtering out death involving cocaine, heroin, and synthetic narcotics, 31.8% of deaths involving prescription opioids also had a chronic illness listed on the death certificate. More details about the filtering at the bottom of the post.)

If you look at the CDC's mortality data, you will see that most records contains several causes of death. The death certificate is filled out free-form, in the sense that the people filling them out aren't forced to pick from a drop-down menu of lists of causes. The medical examiner can write whatever they want on it, so long as it's reasonably comprehensible and in plain English and generally complies with the structure of a death certificate (I won't get into details here).  When the CDC gets these death certificates, it encodes the free-form death certificates with pre-specified codings. So "OD'd on oxycontin", "empty bottle of hydrocodone", and "toxicology found extremely high levels of oxymorphone" would all be coded as T40.2, the ICD-10 code for "other opioids." (I haven't seen a sample of actual death certificates, so I don't know what they typically look like when filled out long-form. But I've spent a lot of time examining the coded file.)

Something that was apparent right away was that many of these deaths have contributing factors beyond the drugs themselves, possibly alternative causes of death. The CDC file has 20 columns for specifying contributing causes of death. (Most only use, say, three to five columns, and I've seen up to 16 used. I'm not sure the full 20 are ever used.) I created a list of all the overdose deaths from the CDC's mortality files (details here; I do a write-up every year when these come out). Then I stacked all 20 of the contributing cause of death columns into one column, and I counted the frequency of each cause. I end up with a list of causes of death that looks like this:

Nothing too surprising. You see all the drugs and drug categories for which ICD-10 has a code. The drugs most commonly involved in drug poisonings are at the top of the list (synthetic narcotics has 33,985 mentions in 2017, Heroin has 18,276 mentions, etc.). And various codings for persistent drug problems, whose descriptions all start with "Mental and behavioral disorders due to...", are highly represented. But then you start seeing things like "Atherosclerotic cardiovascular disease, so described." Why is this on here? Why is it so common? Is it showing up because medical examiners are writing "overdose was exacerbated by ..." and then listing such a medical condition? Or are the death certificates clearly specifying that this was a drug overdose, then gratuitously adding details about the health status of the decedent? There is supposed to be a structure and a logical flow to a death certificate. They aren't generally supposed to add gratuitous information, but rather they are supposed to specify a causal chain. (If there is a causal chain, like cocaine intoxication -> car accident or opioid overdose -> anoxic brain damage, then these details are supposed to be filled out on the various lines of a death certificate in a certain standard order.)

I took the full list and attempted to whittle it down to just the chronic illnesses. I deleted the things that are obviously drug categories (Heroin, Cocaine, etc.) and the "Mental and behavioral disorders..." listings. I also tried to eliminate anything that looked like the physical effects of a drug overdose. For example, "anoxic brain damage, not elsewhere classified" shows up quite a lot. The mechanism by which opioids kill people is suppression of respiration. The body breaths to slowly to get enough oxygen. "Anoxic brain damage" on, say, a heroin overdose is just specifying in gory detail how a drug overdose kills someone. So I deleted terms like this from my list (also, "asphyxiation", "respiratory failure, unspecified" and "asphyxia"). My intention is to get a list of chronic illnesses or other alternative causes of death that were listed on the certificates.

There are a couple of things going on here. Some of these conditions are obviously risk factors in an opioid overdose. People with obesity and sleep apnea have trouble breathing properly, so they should be especially susceptible to a drug that suppresses respiration. I'm not as sure about "cardiomegaly", or "atherosclerotic heart disease." For whatever reason, the medical examiners thought these conditions were relevant enough to list on the death certificate. I don't know if they're saying this condition exacerbated the effects of an opioid overdose, or if maybe these are alternative causes explaining the death, or if they're somehow involved in the chain of events leading up to the person's death, or if the examiner is just gratuitously listing information not relevant to establishing cause of death. What is clear is that many of these people were already sick, and moreover sick enough that their illness warranted mention. And that makes me question how many of these were actually "opioid poisoning" deaths versus deaths from chronic illness in which the patient happened to be using opioids.

It does not seem to be the case that medical examiners gratuitously mention details about drugs on the death certificate unless they think it's a drug poisoning. Take a look at my earlier post, which is mainly about drug poisonings by intent (suicide vs. accidental). The "other" category includes non-poisoning deaths. These make up about 6% to 12% of drug-involved deaths, depending on the year. Inspecting some of these, you see a lot of deaths where drugs were obviously some kind of factor. There are a lot of automobile accidents and bath-tub drownings, for instance. But there aren't a whole lot of death certificates that mention drugs when they are just incidental. A mention of drugs almost always implies that they are causally connected to the death. Given that, I'm scratching my head at all these "incidental" mentions of chronic diseases on drug poisoning deaths. Is there a kind of reporting bias in which any finding of drugs means the death ends up getting labeled a drug overdose? Or do medical examiners scrupulously avoid mentioning incidental findings of drugs when they think there is some other cause present? (If "yes" to that last one, why do they list incidental information about chronic illnesses so frequently?)

It would be an interesting project for someone with access to the CDC's internal systems to take a closer look at how these chronic illnesses are ending up on the death certificates. It's possible that a representative sample of the actual death certificates would paint a clearer picture. I have to settle for the data after it's been converted from raw text and coded under ICD-10. I think there is a very rich vein of research here that would illuminate a not-often-discussed driver of the opioid crisis, or possibly uncover a reporting bias that's inflating the official numbers.

___________________________
Here was my list of chronic illnesses and other alternative causes of death: Hypertensive heart disease without (congestive) heart failure ; Atherosclerotic cardiovascular disease, so described ; Cardiomegaly ; Atherosclerotic heart disease ; Essential (primary) hypertension ; Obesity, unspecified ; Chronic obstructive pulmonary disease, unspecified ; Cardiac arrest, unspecified ; Unspecified diabetes mellitus, without complications ; Drowning and nonfatal submersion ; Cardiac arrhythmia, unspecified ; Other and unspecified cirrhosis of liver ; Other obesity ; Pneumonia, unspecified ; Emphysema, unspecified ; Congestive heart failure ; Acute myocardial infarction, unspecified ; Asthma, unspecified ; Chronic viral hepatitis C ; Bronchopneumonia, unspecified ; Dilated cardiomyopathy ; Septicaemia, unspecified ; Fatty (change of) liver, not elsewhere classified ; Other specified cerebrovascular diseases ; Pulmonary oedema ; Other chronic pain ; Injury, unspecified ; Toxic effect of carbon monoxide ; Anxiety disorder, unspecified ; Insulin and oral hypoglycaemic [antidiabetic] drugs ; Bipolar affective disorder, unspecified ; Pulmonary embolism without mention of acute cor pulmonale ; Unspecified multiple injuries ; Intracerebral haemorrhage, unspecified ; Cardiomyopathy, unspecified ; Heart disease, unspecified ; Atrial fibrillation and flutter ; Hypothermia ; Exposure to excessive natural cold (hypothermia) ; Sleep apnoea ; Chronic ischaemic heart disease, unspecified ; Non-insulin-dependent diabetes mellitus, without complications ; Exposure to excessive natural heat (hyperthermia) ; Gastrointestinal haemorrhage, unspecified ; Hyperlipidaemia, unspecified ; Acute renal failure, unspecified ; Other specified symptoms and signs involving the circulatory and respiratory systems ; Chronic renal failure, unspecified ; Hepatic failure, unspecified ; Effect of heat and light, unspecified ; Stroke, not specified as haemorrhage or infarction ; Cardiovascular disease, unspecified ; Unspecified renal failure ; Heart failure, unspecified ; Chronic kidney disease, stage 5 ; Open wound of head, part unspecified ; Unspecified diabetes mellitus, with ketoacidosis ; Endocarditis, valve unspecified ; Other hypertrophic cardiomyopathy ; Effect of reduced temperature, unspecified ; Obstetric death of unspecified cause ; Dorsalgia, unspecified ; Dissection of aorta [any part].

I sorted the list from most common to least common and went down the list, deleting obvious drug-related causes and such. I stopped when I got 100 or so, because the full list is in the thousands. I would have to go through by hand and manually curate the list to get a more accurate number. So the 31.8% given at the top of this post is a bit low; it's not counting the full list of chronic illnesses found on these death certificates. But it's counting the most common ones, so it's not an egregious under-count.

Medical Nihilism

I recently read an excellent book called Medical Nihilism by Jacob Stegenga. I strongly recommend this book. Here is the recent Econtalk where Russ Roberts interviews Stegenga, which gives a good overview of the general theme. The book is relatively short and readable with about 200 pages of text.

Stegenga's story is that most medicine is not very helpful. There are some obvious treatments that have clear benefits: antibiotics for an active infection, emergency medicine (such as setting broken bones or stabilizing a trauma victim), vaccinations, pre-natal care, and so on. He certainly is not denying the benefits of modern medicine wholesale, nor is he some sort of anti-pharmaceutical malcontent or natural medicine enthusiast. But he does make a strong case that a lot of medical interventions are unnecessary and even harmful. Not in a "It was the right call at the time but we had an unpredictable medical misadventure" sense. Rather in a "This whole class of interventions is not helpful and in fact causes harm" sense. This is even true of interventions that have supposedly been vetted by modern "science" and randomized clinical trials.

Stegenga starts with a series of quotes from ancient and modern philosophers, mostly before the era of modern medicine, remarking on the uselessness or harmfulness of medicine. (He quotes Voltaire, for example: "...the art of medicine consists of amusing the patient while nature cures the disease.") He then proceeds to argue that the message of these quotes still applies to modern medicine. It's not just that discredited medical practices, like blood-letting and laying-of-hands, used to cause harm before we smart, sophisticated moderns figured out they were bunk. Our scientifically derived medicines, whose biochemistry is thoroughly understood and whose efficacy has been tested in clinical trials, so often turn out to be useless. They either have no effect, or they have harmful side effects that would counter-indicate using them if we did any kind of rational cost-benefit analysis.

There is much discussion in the book of the vetting process for new medicines. Clinical trials attempt to prove or discredit the effectiveness of new pharmaceuticals. Stegenga argues that there are a lot of biases in this process that lead to approving drugs that aren't effective. Trials conducted by the pharmaceutical companies themselves have a massive bias in favor of finding an effect, while trials for the same medicine conducted by disinterested parties often find no effect. At any rate, even when there is an effect, it is often quite small. Rarely do we find a "magic bullet" that cures the illness. More often, we find medicines that have a mild statistical effect on the illness, but which require treating a large number of individuals for each person helped. For some medicines, you may need to treat 100 patients in order to (statistically speaking) prevent one heart attack. Cancer medicines may add only a few weeks or months to the average patient's life. And that's assuming we aren't mistaking statistical noise for a real (and admittedly small) effect. This problem is compounded by the fact that many medicines also have negative side effects. A small benefit could easily be offset by small harms so as to make the drug not worth taking. Clinical trials often are not looking for specific side effects, so they fail to uncover them. These side effects only become known later, when the drug is released and used on hundreds of thousands of patients. The book also points out that the screening process for drug trials tends to select for healthier individuals who will tend to be more responsive to therapy and less susceptible to nasty side effects. When the drug is given to a more general population "in the wild", without this filter for the relatively healthy, it will not be as effective. (Now that I'm thinking of it, doesn't this argument imply that there is a sub-population that does benefit, and furthermore that doctors are able to identify patients that will or won't benefit from a new drug? If there's a selection effect that makes medicines seem more effective than they actually are, I'm very curious about who is doing that selecting and how. If doctors have some kind of deep medical knowledge or freaky sixth sense to determine who will benefit from which drugs, let's try to leverage that for the good.)

Some of the examples Stegenga uses are interesting, even amusing. One study found lower rates of mortality for patients with severe heart problems during cardiology conferences, when senior cardiologists were out of the office (and presumably not practicing iatrogenic medicine on their patients). Another study discontinued medications in a population of elderly patients. The patients in the study were on an average of 7.7 medications each. They discontinued an average of 4.2 medicines per patient. Only a few of the discontinued medications were re-administered after deciding that was the wrong call, and the vast majority of patients reported an improvement in health.

Stegenga ends with his "Master Argument", which is a straightforward application of Bayesian reasoning. He argues that we should barely update our confidence in medical intervention given new evidence that the intervention is effective. There are so many biases built into medical research that a positive result is barely informative.

I was primed to welcome the message of Medical Nihilism because it jives well with much of the literature I have read. In his classic Cato Unbound piece titled Cut Medicine In Half, Robin Hanson argues that marginal medicine is largely ineffective at the population level. When we subsidize people to buy a lot of extra medicine, they buy more (30-40% more). But they don't appear to get any healthier due to this extra (or marginal) medicine. It must be the case that there are a lot of treatments that seem like they'd be helpful at the time, and that doctors and patients agree will help, but which don't have any kind of net health benefits in the aggregate. By the way, that's not just the weird contrarian opinion of Robin Hanson. It's also the view of one of the architects of the RAND Health Insurance Experiment, and frankly it's a straightforward reading of the evidence. I'm also familiar with John Ionnidis' work on the failure of many medical studies to replicate. It's hardly radical to suggest that something is amiss here, and it shouldn't be radical to suggest that we should think about cutting back on medical interventions, given that so many of them are either useless or even harmful.

I'm not intending here to write a definitive review or adequate summary of Medical Nihilism. For the full details, read the book, or at least listen to the Econtalk which will give you the overall outline of Stegenga's arguments.

I want to instead deal with Stegenga's policy conclusions. He suggests that the FDA should be more conservative about approving new drugs. He's obviously thinking something like: "If all of these new medicines have little benefit at the cost of nasty side-effects, let's stop people from taking them." It makes sense if you think that the history of medicine up to present day reflects the future of medicine. But I actually think it's a huge mistake. It makes a big assumption about how medical advances will play out in the future. FDA policy should not simply be asking, "Is this particular intervention worth doing for this class of disease?" It should be asking, "What meta-policy for drug approval will give us the optimum rate of medical advances?" (I say "optimum" rather than "fastest", because we are optimizing along several dimensions. Forcing people to be unwitting subjects in medical trials could give us "faster" medical progress at the expense of other things, things that we justifiably care about more.) It's quite possible that there are a lot of useful medicines, magic bullets even, that simply haven't been discovered yet. How will we ever find them if we don't go through a period of cavalier experimentation with new treatments? The FDA has made it very expensive to discover effective new pharmaceuticals. It has effectively put the brakes on medical discovery, and Steganga is suggesting we hit those brakes even harder.

Imagine a future of Star Trek level medical technology. All-purpose miniature scanning devices. A device that sets and knits a broken bone instantly. Manipulation of tissues at the microscopic level. Vaccinations for all known viruses. Maybe it's childish to think that's even possible, and maybe some of the specific technologies are ruled out by common sense or the laws of physics. But imagine a future that's even half-way between here and there. So much suffering would be alleviated, but in the meantime we have to tough it out and make do without those technologies. Whether that future is one hundred years off or a full two hundred years off matters very much. The longer we have to wait for those medical advances, the more suffering humanity must endure. So Stegenga's policy prescriptions only make sense if we assume that medical nihilism is and forever will be an accurate description of healthcare technology. If there is a path between current state and a fantastic future state, it almost certainly entails a lot of experimentation. Generations in the interim must "take the hit" in terms of trying experimental medicines, some of which are ineffective and some of which have nasty side effects. They will have to pay the cost, in terms of suffering through ineffective treatments and in terms of the dollars and cents that finance drug development, if we're to have any hope of traversing that path. Drug companies will not bother to develop new drugs if they can't turn a profit on the pathway to discovering effective drugs. (Alex Tabarrok has even made the argument that allowing ineffective but safe drugs on to the market incentivizes drug development. I think he meant this as a positive statement rather than a normative policy prescription. Plainly the more we allow drug makers to recoup their costs, the more innovation they will do. As much as we bristle at the thought of letting pharmaceuticals sell us snake-oil, we need to think about the incentives that will produce the most innovation.) Stegenga's policy prescriptions, back-fit to history, might have spared us a lot of unnecessary costs and needless suffering, but it would also shut off progress. You should only swallow his policy prescriptions if you think progress is utterly negligible and always will be, and on top of that is unresponsive to the incentives created by the FDA.

There is also a policy prescription that he fails to mention. We should stop subsidizing medicine, given that so much of the medicine we consume is useless. There is this progressive/populist narrative of health care that goes something like this:

Poor people don't have sufficient resources to purchase medicine, and anyway medicine is so expensive that even people with middle incomes would find it unaffordable without insurance or government subsidies. We need government programs that cover the full cost of medicine for poor people. And for middle-class people, we need to make rich insurance companies cover all their expenses, and on top of that make employers pay for those insurance premiums. If we don't do this, people will do without necessary medicine and needlessly suffer and die from treatable conditions.

There are so many problems with this story. See the paragraph above about Robin Hanson and his discussion of the RAND healthcare experiment. Or see the Oregon Medicaid Study, another instance in which very poor people were given a ton of free healthcare without getting any healthier. Government hands out a lot of subsidies to purchase medicine, but it's very hard to point to any positive causal effect on aggregate indicators of health. This ceases to be a mystery if you buy Stegenga's general thesis about the ineffectiveness of modern medicine. "Stop wasting money on bad medicine" might be a useful policy conclusion, but unless I missed it, Stegenga fails to mention it.

Explicit government subsidies aside, we consume a lot of unnecessary medicine because we are over-insured. Bad tax policy encourages companies to offer health insurance to their employees instead of directly paying out in wages or salary. State and federal laws mandate that companies purchase health insurance for their employees, and other laws mandate that insurance policies must pay for basically everything. Conditions and procedures that are excluded from coverage are constantly being demagogued or becoming the political football in the next iteration of the culture war, with the end result being that they end up getting covered because of some new mandate. If bad policy is encouraging us to buy more health insurance than a rational person would want to buy, and if another bad policy makes bare-bones catastrophic health insurance illegal, that means we're all getting a big implicit subsidy to buy medicine. If Stegenga is right about medical nihilism, an important implication is that we should repeal laws that lead to such excessive over-insurance. I would have liked the book to point out this implication clearly.

Aside from the policy conclusions, I am also skeptical about the ability of randomized controlled trials to find "the effect" of medicine. We like to think of medicine as hard science. Do X to the body and Y will happen because of biochemical pathway Z. But obviously we are quite heterogeneous in our responses to medicines. Our bodies are all different. Some differences in our responses to medicine happen for known reasons. Some people are fast metabolizers of opiates, for example, and some people are slow- or non-metabolizers. There are known genes that cause people to be fast- or slow-metabolizers, and having zero, one, or two copies of these genes can make someone more or less susceptible to an overdose, or can affect the dosage necessary to get pain relief. And this doesn't even get into the issue of subjective well-being, which might be different even for genetically similar individuals taking the same medicine. Medicine is not so much a chemistry lab with known quantities of well-behaved chemicals, but rather more like an exercise in fine-tuning. Finding "the right" medicine for a person might be more like figuring out their optimum shopping cart at the grocery store. Randomized controlled trials necessarily treat everyone like they're an "average" person rather than fine-tuning their shopping cart. Gwern said it better than I could when they wrote:

The point is making your life better, for which scientific certainty is not necessary: imagine you are choosing between equally priced sleep pills and equal safety; the first sleep pill will make you go to sleep faster by 1 minute and has been validated in countless scientific trials, and while the second sleep pill has in the past week has ended the sweaty nightmares that have plagued you every few days since childhood but alas has only a few small trials in its favor - which would you choose? I would choose the second pill!
 To put it in more economic/statistical terms, what we want from a self-experiment is for it to give us a confidence just good enough to tell whether the expected value of our idea is more than the idea will cost. But we don’t need more confidence unless we want to persuade other people!

It's conceivable that clinical trials will "average out" to a near-zero effect, even for a medicine for which the patients themselves can detect a benefit (or nasty side-effect) much larger than the average. That makes me hesitant about pulling drugs off the market because RTCs (which some people apparently think are synonymous with pure science) say they don't work. See also the point I made above about "selection effects" in clinical trials, which seems to imply that somebody has a sorting mechanism for who will/won't benefit from a drug, and that this knowledge exists even before the trial is conducted. Heterogeneity in responses to medicine is built into Stegenga's argument. He might have spent more time reflecting on this point.

Of course I am nit-picking. Medical Nihilism is a great book. Naturally I'm writing this post to respond to the areas of disagreement, because that's where I think I have something interesting to add. The rest of the book speaks for itself.