The issue with the state of the art

At every moment in time throughout history, the state of the art treatments that the medical profession collectively holds is touted as the end-all, be-all of health - or at the very least, better than the existing alternatives.

In some cases, this is totally correct. Smallpox killed 400M+ people before it was eradicated, saving enough lives that any potential downside to the vaccine is clearly worth it from a societal perspective.

In other cases, the state of the art treatment is worse than something else. Bloodletting is worse than not having all your blood. Thalidomide is worse than morning sickness (at least for your baby).

Of the current treatments and lifestyle prescriptions that the medical profession collectively holds, there is no doubt that there are treatments that fall within each category. But how to know which is which?

Issues with determining good from bad

Attempting to determine good treatments vs. bad treatments by considering the reasoning that leads to those treatments is tricky.

On one hand, it seems necessary to see why we might think something is a good treatment before we can judge if it’s a good idea. How can we write off statins without first understanding 1) statins apparently lower cholesterol, and 2) cholesterol apparently causes heart attacks, and 3) heart attacks are unhealthy?

On the other hand, the fact that the medical collective makes the wrong decision so often clearly demonstrates that this sort of reasoning is very hard. Educated professionals who do their very best to make decisions in the best interest of their patients very often end up hurting them, for example by prescribing pain medication that just ruins their life. In fact, this sort of reasoning is hard enough that there is little hope of a non-knowledgable person like myself making the correct decisions; at best, I’d be fooling myself on my ability to asses such lines of reasoning.

So what do? How can I justify ignoring the reasoning that leads to the state of the art, if I’m rejecting the state of the art? On the other hand, how can I justify engaging with this state of the art reasoning given how hard it clearly is, and how uneducated I clearly am?

Where things go wrong

Our only hope for escape from this unpleasant situation is some sort of higher-level reasoning. If we can extract a general sort of structure for where the state of the art reasoning tends to be correct and where it tends to be wrong, then we may be able to make judgements about the state of the art treatment without needed to engage with the specific reasoning that leads to it.

Of course, this higher-level reasoning will not be perfect. We may reject some treatments that are in fact great, and take some treatments that are worse than the alternative. In some cases, the best treatment will be to do nothing at all.

But as we will explore in later sections, these rules of thumb (MAY) do better than the state of the art throughout most of history.

Personal rules of thumb for evaluating treatments

Ultimately, most health decisions are profoundly personal ones. Here, we propose a series of simple rules to help make these decisions.

Notably, these rules are designed to at most require an understanding a) the current prognosis of the ailment you’re looking to treat, and b) what humans have done through history. We avoid decision making based on complex models of biological cause and effect, due to their clear complexity and continual failure in many domains.

Note that these are rules of thumb, not absolute rules that should be followed exactly. As we show below, following these strategies would have led to you doing better than the medical establishment for

ROT 1: avoid treatment when the aliment is not serious. Seek aggressive treatment when the ailment is serious.

Serious, here, means fatal or so-bad-for-one’s-quality-of-life-that-death-sounds-like-a-alternative. So: not serious is a headache, muscle soreness, a wart, a cold. Serious is a cluster headache, loss of a limb, an STD, a novel pathogen, cancer.

Of course, one person’s cluster headache is just another persons hangover, so this (like all the rules here) are not objective. And this is not a hard rule - you should just be more hesitant to treat a sickness the less serious it is.

ROT 2: try and avoid treating symptoms that might evolution deemed protective.

The injury: a broken bone. The body handling it: swelling. The injury: a flu. The body handling it: a fever. Reducing swelling with ice, reducing a fever with advil: both might be comfortable, but should be avoided. Simply put, these responses are evolutionarily adaptive and likely give some benefit.

Of course, as any rule of thumb, this can be taken too far. You’re body might respond to a cancer with some response, but that doesn’t mean you should just let the body handle it. If you’re going to treat non-serious aliments, make sure you’re actually treating the ailment vs. something that is helping you get over the aliment. Evolutionary thinking helps here.

ROT 3: time-permitting, reversible or forwards-compatible treatments should be tried first.

Serious aliments usually are time-sensitive. If you’ve decided to treat a less serious aliment that is not particularly time sensitive (e.g. back pain), try treatments that are reversible or at least don’t stop you from seeking other treatments.

If you’ve got back pain, try strengthening the muscles that support your back before you try surgery. You can’t take surgery back after you’ve done it, but you can just stop strengthening your back if it’s not helping. If you’re is mentally ill, consider aggressively experimenting with your sleep and exercise schedule before scheduling yourself for a lobotomy (or going on an SSRI that is effectively impossible to get off of).

ROT 4: if something has been a non-issue for most of human history, it’s probably not an issue now, unless there’s been a major context change.

Early humans likely took care of their teeth, but certainly didn’t use fluorinated toothpaste. And yet they mostly had great teeth. However, we eat a lot more carbs and sugars now, so perhaps there is good reason we needed to update the treatments for our rotting teeth.

But if there is not a relevant context change, then treating something that appears to be an issue will likely do more harm than good. If you mostly walk and run on trails, then consider ditching your running shoes with massive soles. If evolution didn’t think your Achilles was enough as a shock absorber, you’d have a bigger one.

Looking @ you, skin care!

ROT 5: reduce the number of products that you rely on, and prefer to rely on naturally derived ingredients.

There was a voluntary recall by sunscreen brands Neutrogena and Aveeno this year for many of their sunscreen brands, specifically because they contained dangerously high levels of benzine, which “increases the risk of developing leukemia and other blood disorders.” So by attempting the relatively visible and easily treatable skin cancer, we gave ourselves blood cancer.

Every additional product you apply to your body is another chance for a large corporation to forget to notify you about some horrible feature of their product. This happens with cars, baby formulas, sunscreen, and much more.

On natural ingredients: of course, not everything natural is good. Arsenic can be found in nature. But given that our bodies were created in an evolutionary environment where we were not even distinct from nature, we’re much more likely to have some tools for dealing with natural ingredients than artificial ones.

Poisonous things created by nature usually taste bad to us (something something alkaloids), but poisonous things created by humans can be distinctly yummy (trans fats, the smell of gasoline).

ROT 6: consider attacking base-sources of unhealthiness before treating specific ailments.

One can treat back pain in a lot of ways. One of the more surprising ways is to just loose weight, so you’re carrying less weight on your back.

In general, when optimizing a system, if you can start by addressing “base” issues that contribute to the other specific issues that exist, you’ll get much bang for your buck (and likely have more success). This is true when optimize your code, or optimizing your health.

What specifically are base issues that are worth starting with?

1. Get enough good sleep. This effects both mental and physical health in massive ways.
2. Maintain a healthy weight. Healthy means healthy, don’t take this to far.
3. Exercise, without injuring yourself. Cardio, strength, and stability are all important. Not exercising has correlative risk factors that make it look worse than smoking cigarettes (p.s. don’t smoke cigarettes).
4. Eat fresh food first. Avoid the standard American diet of processed foods. They are clearly worse than the alternative

Group-based rules of thumb for those who affect decisions over multiple people

ROT 1: if your aliment evolves, respect the evolutionary game you’re playing.

Viruses, cancer cells, and bacterial infections reproduce and evolve. As such, changing the environment that they exist in will select them to adapt for this environment.

Heavy antibiotic use will leave us humans worse off than before, as we breed antibiotic resistant bacteria. Likely, cancers would be much better treated with respect for the evolutionary adaptations of cancer cells (look up pest evolutionary dynamics!).

Respect these dynamics, and do you’re best to think about the second order effect that this game will have!

TODO

On steady states, and non-linearity

I used to wash my face with face wash (yes, it was gentle), and moisturize to replace the oils I had stripped off (yes, it was dermatologist recommended). When I moved to just washing my face with water, I was a real oily MF for a few weeks.

This, of course, went away. Your body is generally pretty good at adapting to the situations that you put it in. My skin stopped being oily quickly enough.

But there is still a question of how to transition to a different steady state: if I run in cushioned shoes, how should I start running in more minimalist shoes?

It might seem like the answer is to gradually transition, but in some cases this appears to be worse. Switching to minimalist shoes might force you to adapt a new running style, due to explicit pain; gradually transitioning to minimalist shoes might allow you to get away with bad form while not having enough padding, and so injuring yourself.

On the other hand, moving from being on SSRIs to not being on SSRIs (something one should only do if they talk with their doctor, I’m just a random fucker, don’t listen to me) seems like it would fuck your shit up, and actually that tapering works better here.

So, I’m not sure. I also don’t have any good rules of thumb for figuring out which is which!

Non-linearity

TODO: add a section on fasting, and taking days off from exercise. Areas of no stress are good, and areas of stress are good too.

Is this always true? Should we sometimes fast water or air? Of course, you can take it to far, but breath holding, or some form of breathing exercise seems is often touted as having positive benefits.

Evaluating these rules of thumb

It’s easy enough to just lay our rules of thumb; it’s important that we actually argue that these result in better livin’.

To do so, we shall consider a random sampling of treatments in medical documents from history, and evaluate them with the above rules of thumb. We shall also judge them on their “correctness” vs. the alternative treatment of “nothing”

On evaluating the past from the current

TODO: talk about problems with judging the past with the current, because we don’t htink it’s good (but it’s at least better?). Hmmm.

The savvy among you will notice an immediate problem: selections from different groups of treatments can easily lead to different conclusions. If we look at a book of basic remedies for illness, many of them may be correct, while more advanced illnesses (like cancer) may be more off. Or something like that.

To fix this problem, we shall randomly select books from the Wikipedia list of medical textbooks using the following process:

Process for treatment selection:

1. There are 31 historical medical books listed.

2. We will look at the Bitcoin block #500,000, and take it’s merkle root:

   **31951c69428a95a46b517ffb0de12fec1bd0b2392aec07b64573e03ded31621f**
   

3. We use this to randomly select 5 book from the set with the Python code:

   import random
   random.seed("**31951c69428a95a46b517ffb0de12fec1bd0b2392aec07b64573e03ded31621f"**)
   print(random.sample(range(31), 5)) # 22, 27, 3, 23, 14, which we match to book indexes
   

4. Then, we went to find a translation of each of these documents

   1. London medical papyrus - Column 1 - 29 lines, Column 2 - 30 lines, Column 3 - 28 lines, for a total of 87 lines.
   2. Shennong Ben Cao Jing - no longer exists
   3. Adab al-Tabib - no online version found
   4. *Book of the Ten Treatises of the Eye - there are 125 pages in this book*
   5. Commentary on Anatomy in Avicenna's Canon - no online version found

5. Ok, we add another 5 to get more (increase sample code to 10), and get 30, 25, 6, 12, 9

   1. Brugsch Papyrus - no online version found
   2. *De materia medica -* 926 pages in the book
   3. Wushi'er Bingfang - no online version found
   4. Book of Optics - this does not give perscriptions
   5. Rosa Anglica - there are 508 pages

6. We have 4 now, and we’re gonna use a very similar process for selecting random pages or lines from them.

   1. We hash each of these depending on which index they were from the random selection.

7. As we are considering n books, for book i (0 indexed), we will take the hash (shaof this merkle root i times to generate a unique random number.

   • Code

     original_root = "31951c69428a95a46b517ffb0de12fec1bd0b2392aec07b64573e03ded31621f".encode('utf8')
     books = [
     	'Edwin Smith Papayrus',
     	'Cannon of Medicine',
     	'1900 book',
     	'current book',
     ]
     for book_index, book in enumerate(books):
     	random_number = original_root
     	for i in range(book_index):
     		random_number = hashlib.sha256(random_number).hexdigest().encode('utf8')
     	
     	
     	
     hashlib.sha256(b"Nobody inspects the spammish repetition").hexdigest()
     

8. Then, we will look at how many total treatments are presented in this book. If there are subsections in the book, we will count these (or do our best, per the book). Then, we will select 10 random numbers, using the randomly generated number above.

The other problem is that

Here, we consider a random sample of 10 pages from a variety of medical documents from history.

We start with the Edwin Smith Papayrus (written nearly 5000 years ago), the earliest known medical text. We then skip forward a few millenia, and look at the cannon of medicine, written around 1000 AD and remaining the most influential medical text for the next 600 years.

TODO: find one from 1900

TODO: find a handbook from the current day.

On evaluating the past with the current

The savy among you probably have two immediate issues with my proposal for evaluating these rules of thumb.

Selecting random treatements

First, before looking at these texts, we commit to a process of random selection for the treatments.

Edwin Smith Papyrus

TODO:

The Cannon of Medicine

TODO:

Here, we consider the earliest ever medical text, t. We also consider a random sampling

ROT 2:

ROT 2: time-permitting, the best treatment is the most gentle one.

The first medical textbook

A recent medical textbook

When to consider a treatment

Want to get over a sickness better? Bloodletting. Want to

Of course, sometimes

Throughout all of human history, humanities best effort at modeling the human for the purpose of health have been a failure. Specifically, many of the common best medical practices of the time have been explicitly harmful. This includes: blood letting, radiation use for non-serious causes, creating trans fats and avoiding animal fats, and many more.

Of course, there have been places where medicine has totally succeeded; specifically, small pox and chicken pox area areas where we’ve done wonders and saved millions of lives with a downside that is so small it is not worth worrying about.

Where and why do we succede and fail, continually, over time?

We succeed in areas that have a few conditions. First, the thing being treated must be acute, and quite harmful.

The main question here: is this

The goal here, specifically, is to make myself robust to model error.