Medical Nihilism

by Jacob Stegenga

This book argues that if we consider the ubiquity of small effect sizes in medicine, the extent of misleading evidence in medical research, the thin theoretical basis of many interventions, and the malleability of empirical methods, and if we employ our best inductive framework, then our confidence in medical interventions ought to be low.

Genres: Nonfiction, Medicine, Science, Philosophy, Health, Medical, Skepticism

391 pages, Kindle Edition

Published March 8, 2018

Reviews

[Петър Стойков · Rating 4 out of 5]

Науката напоследък е много модерна и има много фенове. За съжаление голяма част от тях не са фенове на науката като метод, а на Нáуката™, демек "науката" като медиен феномен и култура към която да се причислят и да се имат за много умни. Съвсем логично, след като са възприели Нáуката™ за определяща характеристика на личността си, тия хора реагират доста остро на всяка критика към нея - печалните резултати от което могат да се наблюдават изобилно в редит, в групите на войнстващи атеисти (да бе, и аз бях от тях когато беше модерно, а аз млад, признавам си) и в популярни личности като Сам Харис.

Въпреки, че са дразнещи, ние няма да се занимаваме повече с тях, а със самите научни работници и изследователи - които за съжаление също доста често не използват научния метод по начините, за които той е създаден, и науката страда от това. Т.е. сега ще критикуваме не Нáуката™, а науката.

Като цяло, има няколко основни типа широкообхватна критика към науката:

1. Елементарна, религиозно мотивирана, обикновено насочена към отрицание на биологията (еволюцията), която най-често с интерес на посетители на зоологическата градина наблюдаваме сред евангелистите в САЩ, но понякога се промъква даже у нас.

2. Философски подход с вербалистичен уклон, най-често под формата на социален конструктивизъм, по-рядко научен антиреализъм или глобален епистемичен скептицизъм (за да видите, че и аз ги знам дългите думи). Това се среща най-вече във философията на теория, а на практика в хуманитарните дисциплини, които го използват за да обосноват, няма какво да се лъжем, резултатите от политическите си изкривявания - голяма част от идеите на цялата woke сбирщина попадат тук.

3. Конспиративни теории с шизофрениформен характер, често комбинирани с „Фарма-мафията“, не сме кацали на Луната, евреите са виновни за 11 септември и др. Интересното на конспиративните теории, разбира се е, че понякога те се оказват верни накрая, защото както знаем и счупеният часовник е верен два пъти на ден (тоя израз зуумърите не могат да го разберат, защото масово не могат да познават часовник със стрелки).

4. Научен скептицизъм, който изследва границите на науката и ограниченията на прилагането й от хора, чрез самата наука.

Горните са генерализирани антинаучни нагласи, които в различна степен твърдят, че (големи части от) науката не може да произвежда надеждно знание поради разнообразни ограничения.

В Medical Nihilism Якоб Стегенга, естествено (естествено е, защото ако не беше така, изобщо нямаше да се занимавам с него), се съсредоточава върху четвъртата точка, която обичайно се основава на метанаучни изследвания (науката за науката), подсилени с философски анализ. Книгата се занимава основно с ненаучната наука в медицината - нещо, което е отдавна забелязано като проблем както на теоретично ниво, така и в чистата медицинска практика и отбелязано от други автори.

Как се аргументира тезата, че голяма част от медицината е ненадеждна, или както авторът формулира: „трябва да имаме ниско доверие в ефективността на медицинските интервенции“?

От десетилетия е известно, че учените по правило не публикуват отрицателни/нулеви резултати, поради което публикуваната литература рисува прекалено розова картина. В някои случаи е възможно да се проследят всички проведени изследвания, най-вече благодарение на регулаторната документация. Тогава може да се сравнят публикуваните с непубликуваните резултати — и често се наблюдава именно описаната картина. Този тип изследвания, по мое мнение, предоставят най-силните доказателства срещу медицината в общ план.

За повечето терапии има само няколко, да речем пет, публикувани проучвания. Възможно е те да изглеждат убедително. Но остава упоритото опасение, че съществуват още, например, пет също толкова качествени изследвания, които са негативни, но ние просто не ги виждаме, защото не са публикувани. Няма начин да знаем дали такива проучвания съществуват. Следователно, дори когато разполагаме със сравнително приличен метаанализ, остава съмнение — понякога значително — относно дадено лечение.

Има и много други проблеми - дори да се абстрахираме от откровените лъжи и измами (да, учени и лекари също си служат с тях, човешкото его е неукротимо), остава огромният проблем с кризата на повторяемостта на изследванията в цялата наука и особено в по-меките й варианти (хуманитарни, икономически и медицински дисциплини), както и прозаичното невежество на много "изследователи" (боли ме сърцето и не мога да кажа учени) относно основни научни принципи - водещо до практики в медицината основани на изводи базирани на статистически асоциации, вместо на причинно-следствени връзки.

Последното е толкова преобладаващо в изследванията на храните и различните вещества и хранителни добавки и връзката им със здравето, че аз лично приемам всяко твърдение в тази сфера като по дефиниция невярно. Достатъчно е да слуша човек какви ги плещят популярни лица като Андрю Хюберман и Ронда Партик и на какви мижави и съмнителни изследвания базират твърденията си за ефектите на това или онова върху човешкото здраве...

Тези проблеми са разрешими, например чрез регулации, които изискват задължителна публикация (какво разумно правителство би финансирало изследвания, които никой никога не може да прочете и да извлече поука от тях?) както и с употреба на изкуствен интелект за автоматично изследване на методите на изследванията и връзката им с абстрактите на публикациите (защото почти никой не чете друго освен абстракти, а там са най-големите извъртания, преувеличения и откровени лъжи).

[Zahwil Dossa · Rating 3 out of 5]

Before you take statins to lower cholesterol, or blood pressure drugs, or anti-depressants, you might want to read this book.

Stegenga crafts an argument, methodically if somewhat repetitively, that most medical interventions are unlikely to bring about the purported benefits.

He is critical of evidence-based medicine (EBM) which prizes the Randomized Control Trial (RCT) and meta-analyses that aggregate the results of many such trials.

Stegenga gives many reasons for his skepticism which has engendered outright nihilism, meaning he disbelieves in the value of medicine altogether.

He does not disbelieve in science, oppose vaccinations, or urge us to employ holistic medicine. Rather, he is led to medical nihilism by his study of RCT’s, the way they are designed, their extrapolation from clinical settings to the general public, and the way the results are reported by researchers and corporate backers.

He concedes the value of penicillin and insulin but argues such "magic bullets" are very uncommon in medicine, have not been replicated with other like discoveries, and due to the physiology of disease and the profit motive of the pharmaceutical industry, are unlikely to turn up in the future.

Many diseases do not have necessary and sufficient causes which can be intervened on with the specificity and effectiveness of penicillin or insulin. Meanwhile, the pharmaceutical industry is motivated to create drugs that treat widespread, mild diseases, often launching educational campaigns for conditions like shyness, obesity and anxiety, with a treatment plan to follow.

Stegenga’s self-described Master Argument for medical nihilism is based on Bayes’ theorem.

P(H|E) = P(E|H) x P(H)/P(E)

The left side of the equation is the probability (likelihood) of a hypothesis that a medical intervention will be effective given the evidence E.

On the right side of the equation, Stegenga argues that the former two terms ought to be low, and the latter term ought to be high, and thus P(H|E) ought to be low.

The prior probability P(H) is influenced by the proportion of past interventions that have been effective, and as there are few magic bullets in medicine, it ought to be low.

The evidence for the hypothesis about a drug's efficacy is invariably based on two RCTs, as this is the bar set by the FDA for a drug's approval.

An RCT is a trial in which subjects are allocated randomly, usually without their knowledge, to either an experimental group which receives the intervention under evaluation or a control group which receives a placebo.

RCTs have a plurality of flaws that should weaken their evidentiary value:

1. RCTs are biased because they are asymmetrically designed to detect benefits at the expense of detecting harms
o Subjects in the experimental group may withdraw from the trial due to harms which then go unreported
o The trial runs just long enough to detect the benefits, while harms often manifest years later
o Harms are often observational or anecdotal which are discounted by medicine as weak forms of evidence compared with the RCT

2. RCTs can fall prey to selection bias, whereby a third factor (a confounding factor) introduces relevant differences between the groups.
o Randomizing the allocation of subjects is intended to mitigate selection bias, but even so, subjects are not drawn from a random sample of the broader population who have the disease. Certain people are excluded such as the elderly, patients on other drugs and patients with other diseases; yet these are the very patients who are likely to consume the drug in a real-world setting.

3. RCTs are subject to confirmation bias which is our tendency to overweight evidence that confirms our prior beliefs, and underweight evidence that disconfirms our prior beliefs
o When we consume a pill, we become more sensitive to perceiving improvements in our states, and are likely to infer causation when there is not
o Many illnesses, such as influenza, improve on their own as part of the natural course of the illness
o People with psychiatric diseases are especially responsive to the placebo effect
o The subjects may guess which group they are in, increasing the likelihood that the people receiving the treatment perceive a benefit (this is known as detection bias or blind-breaking)
o Social media exacerbates confirmation bias by promulgating personal anecdotes and creating feeds that reinforce our prior beliefs

4. The designers of RCTs have strong incentives to overestimate the effectiveness of the drugs.
o Researchers and scientists face incentives to publish findings that are meaningful to raise their status
o Pharmaceutical companies that sponsor the research are motivated to develop profitable products
o Three quarters of trials only report relative outcome measures, which lead physicians and patients to overestimate the effectiveness of drugs. Relative differences promote the base rate fallacy which is making an inference about the probability of an event without considering its prior probability. Say the prior probability of a good outcome for a person with a hip fracture is only 1%. An RCT may report a 50% relative difference between the experimental and control groups, but that only represents an absolute difference of 1%.
o A more informative metric to report is the Risk Difference (RD), or the difference in probability of having an outcome with and without use of the intervention. Or the inverse of the RD which is the Number Needed to Treat (NNT), which is how many people would have to use an intervention to achieve one of the outcomes of interest
o “When outcomes have low base rates, a high relative effect size masks a small absolute effect size”

5. RCTs are subject to publication bias:
o Trials that suggest a medical intervention is ineffective or harmful are unlikely to be published
o Extrapolation from a trial can be misleading because the published trials are only a fraction of the total trials performed. This is known as the filing cabinet effect, as we don’t see the results of all the trials that are stashed away because there was no meaningful effect.

6. The researchers and government agents within the FDA have massive conflicts of interest
o Medical researchers routinely have financial ties to the pharmaceutical industry as consultants, shareholders, patent holders, etc.
o The FDA receives fees from the pharmaceutical industry for the applications submitted on prospective drugs, and these fees are used to pay their salaries

7. RCT’s are vulnerable to p-hacking (spurious correlations):
o P-hacking is the likelihood of a statistically significant correlation arising by chance as the complexity of a dataset increases, and as more and more analyses are performed

With both the opportunity and the incentive to design and report RCTs to overestimate benefits, we expect many drugs will be later withdrawn from the market, and this is in fact what we see.

Stegenga’s book thus leaves us with a bleak picture of medicine.

“The vast majority of medical interventions introduced in the last several decades have very small effect sizes and have a plethora of harmful side effects.”

He puts forward the notion of “gentle medicine” or pursuits like eating better and walking to minister to our health.

One of the quotes that expresses this idea is from Shakespeare:

“When Macbeth asks a doctor if he can cure a diseased mind, a troubled brain, or a heart weighted with sorrow, the doctor claims that ‘the patient must minister to himself’”.

[Lee · Rating 4 out of 5]

Everyone knows that in the old days medicine mostly didn't work. But nowadays many believe that medicine mostly does work. This book by a philosopher of science argues that actually that is wrong: medicine mostly is still bunk. There are a few exceptions: antibiotics, insulin, a few other "magic bullets." But otherwise, it's trash. Here's why.

The drugs on the market in America passed FDA review. To pass FDA review, a drug company has to present results of studies showing their product is effective at treating a disease. There are several good reasons to think those studies overestimate the effectiveness of a drug, and there are independent reasons for thinking the true effectiveness is usually zilch or worse.

The main reason to think the studies overestimate effectiveness is publication bias. This is a well-known criticism, so I had the vague impression that smart people somewhere must have done something about it. But according to Stegenda, no. Pull quote from footnote: "In a 1990 meeting this concern about the influence of publication bias on the FDA standard was articulated by a professor of statistics. The director of the FDA’s Division of Neuropharmacological Drug Products at the time stated that by law the drug under discussion, sertraline (Zoloft), must be approved as long as there were two positive trials—regardless of the number of negative studies." Here is a quote from a survey of the FDA's own evaluators: ""...when FDA approves a drug, it usually has no evidence that the drug will provide a meaningful benefit to patients."

This point alone is enough to carry the argument through, I thought.

There are other reasons to believe that the studies overestimate effectiveness, but they are more conceptually interesting than important to the main argument. For example, with the me-too pills that extend patents on an existing drug, there can be "disease-mongering" by industry, as in the infamous case of inventing female sexual dysfunction. Some subtler biases include the base rate fallacy introduced by the use of relative risk measures (eg, hip fractures reduced 30%) and inappropriate extrapolation from the effects on a proxy rather than the underlying target (eg, cholesterol reduced rather than heart disease prevented). Also, we should probably think of effectiveness as benefits net of harms, and if we do so, there is a serious problem with the design of the trials, which are insensitive to potential harms.

How do we know these biases, especially from publication bias, are serious problem, erring in the direction of overestimating effectiveness? The main reason to think this is that occasionally researchers do get ahold of the full data from all studies via a lawsuit or such, and when they do, the tendency is for the total evidence to point to near zero effectiveness. Stegenga argues there are independent reasons for expecting this to do with the complexity of the microphysical processes that we are often trying to manipulate with drugs.

That's the thrust of the main argument, and I mostly buy it.

Philosophically what I found most interesting in this book are some of the arguments denigrating meta-analyses. A meta-analysis pools data from many studies, usually RCTs, to summarize what we have reason to believe in light of all the evidence available.

Stegenga points out that these meta-analyses can inherit the biases in the individual studies, especially publication bias. Additionally, he argues that it would be better if these meta-analyses moved away from dichotomous decisions to include/exclude individual studies based on their category (RCTs, observational, etc). Instead he urges that they include all evidence available by weighting the studies using a quality assessment tool (QAT) that assigns a grade to each based on a checklist of criteria. I hadn't known about QATs, and I found these arguments convincing: don't throw out any information.

But he keeps going, and presses an argument that is unsettlingly pessimistic as it is interesting. Stegenga points out that there are multiple QATs available, with different criteria and scoring rules, and none is clearly superior to any other, and each has a fairly low inter-tool reliability score. He draws a pessimistic conclusion from this (likening it to the inverse of the Duhem-Quine thesis): the best evidence is underdetermined by theory.

I take the point--and it's a great one--but I found myself wondering whether he left unaddressed the original problem: having multiple QATs is bad in part because researchers can choose among them, introducing a degree of freedom that makes the conclusions of even the best meta-analyses malleable. If that is the problem, then it seems to me a merely political solution (akin to preregistration) would work: hold a vote among experts at some conference and just pick a single QAT for all to use. I have no idea if that would work, but I wish it had been considered even if dismissed as naive. There are such things as standards bodies, and they are useful for solving exactly this kind of malleability problem.

Highly recommend this book if you are interested in any of these topics and have a philosophical bent.

[Hoda · Rating 4 out of 5]

I'll start by saying that the clarity of Stegenga's writing is extremely refreshing. His argument for decreased confidence in medical interventions is also a compelling one. Probably the most understandable piece of evidence he provides for this claim is that of publication bias in medical research. Only studies with favorable outcomes are generally published and thus made available to the public, leaving one to think that most of the progress made in medicine is effective when in reality the majority of interventions are either ineffective or indeed harmful. The malleability of evidence-based medicine, grossly insufficient regulatory standards of the FDA, and lack of resources dedicated to studying the harms of proposed interventions are some additional reasons given for the overestimation of the effectiveness of medicine. I think it's important for people in healthcare to become familiar with this argument so that medical treatment become more closely aligned with the care people really wish to receive.
All that being said, I never quite became comfortable with Stegenga's use of the term "nihilism." It has been proven that patients who are administered some sort of treatment, despite knowing that the treatment is in fact a placebo, report better health outcomes. What this tells me is that it's important for patients to feel like they are being cared for and have a degree of confidence in the care they receive in order for it to be more effective. If that confidence were to be gained from a pleasant experience with a physician rather than a drug (which it should be if physicians are actually giving the "care" they trained to provide), then that would align well with Stegenga's suggestions for improvement. The challenge now is to shift the confidence from the drugs to the providers, and I don't know if the term "nihilism" will necessarily get us there.

[Philip · Rating 5 out of 5]

A fine slap in the face to current research practices. It's all very common sense but the author feels the need to make it rather complex at times. However, I get it. We are indeed up against Goliath.
I love when someone says," I love my doctor." They like his personality and style and with that unscientific data point use it to completely trust their doctor for any and all interventions. No one trusts their doctor in fact. What one trusts is pharma and their many nefarious practices. Pharma dictates what drugs are made and doctors use them. The real science that doctors learn in medical school: biology , microbiology, histology, anatomy, physiology etc are true enough. However, it's a big leap to go from that to pharmacological interventions and many biases ( up to 53) have been found when looking at trials in detail. All research should be done by public agencies and not by the manufacturers of said drugs.

[Nisarg · Rating 2 out of 5]

The content is good and worth knowing about, but I can't recommend the book because of the repetitiveness and poor quality of the writing. The author seems to have taken a 20 page paper and expanded it out to book length by inserting an irrelevant philosophical digression into the definition of disease, followed by endless forward and backward references to the content of his arguments.

[t herman · Rating 5 out of 5]

Scary

Detailed account of why we should put less faith in medical interventions. Well argued. It convinced me, and I'm a sceptic's sceptic.

[Jordan Papineau · Rating 5 out of 5]

An approachable though scholarly work that should be required reading for students of medicine, philosophers of science, and employees of regulatory bodies. Stegenga manages to home in on the methodological shortcomings of medical studies and financial conflicts of medicine to reduce the confidence one ought to have about any given disease intervention.

It is important to recognize on the outset that Stegenga is neither advocating for medical stasis (i.e., non-treatment), nor is he claiming true medical progress hasn’t occurred. Rather, Stegenga wants practitioners, scholars, and laymen alike to attenuate their confidence in the probability that drugs are often effective treatments outside of diseases like deficiency or of infectious etiology, i.e., for the diseases that are most commonly diagnosed and pharmaceutically intervened on, and encourage systematic adoption of “gentle medicine”.

There are portions of the book that emphasize the formalism of his logical argument. One need not have a background in formal logic or probability to understand many of the key arguments made in support of his overarching thesis.

Medical Nihilism is the first book in a while that has fundamentally shifted my perspective on a broad subject, namely, pharmaceutical interventions and medical interventions broadly construed.

[Dio Mavroyannis · Rating 5 out of 5]

It took awhile for me to finish this book but it was a great ride. Apart from learning a great deal about the kinds of tricks that go on in the medical profession I find the fundamental bayesian argument very convincing. This is one of those books that sounds absurd on its face but once you get your head around the argument, it becomes trivially true. 5/5 highly recommended.

[funsizedshark · Rating 5 out of 5]

"My hope is that a love of humanity can motivate us to improve the art of medicine-in all of its manifestations, including clinical practice, scientific research, and regulation-and conversely, rethinking the art of medicine could contribute to improving the condition of humanity."

rtc!

[Bruno · Rating 4 out of 5]

A challenging book … don’t be mistaken by the title: this is not a nihilist book about medicine, it is a forward-looking and optimistic book about medicine, while trashing current practices. Worth reading

[tom · Rating 5 out of 5]

crystal clear arguments laid out in a logical structure. very convincing!

[Sisyphus · Rating 5 out of 5]

5 stars because I want to promote medical nihilism. The book is a slog to read tho because its a thesis.

[Damian · Rating 2 out of 5]

It was too methodical for me. I probably should review the Bayes theorem part as that might be the most impactful.

[Navya · Rating 4 out of 5]

great introduction to the subject of medical nihilism. To those who are not already interested in the topic, it can seem dry or too technical. Important information and well-written.

[Gaby Chapman · Rating 5 out of 5]

An extremely technical, scientific, philosophical version of the best medicine is the least medicine.