Luca no background

Hi! I’m Luca. How can I help?

Email me I reply within 24h.

Luca no background

Hi! I’m Luca. How can I help?
Email me. I reply within 24h.

skip to Main Content

Today I read this tweet, written by @calmasarock and retweeted by @naval:

“Health isn’t found in parts of the system but in the system as a whole. It’s not the vitamin C that makes you healthy – it’s the Orange. It’s not the Iron – it’s the Spinach. And so on.”

Too often, medicine follows a process similar as the below:

  1. We compare physiological data from a group of healthy people and people with a given disease or disorder.
  2. We look for a physiological marker which can predict which group does a randomly chosen patient belong.
  3. We infer causality and think that correcting such indicator might make patients healthy.

Sometimes the indicator found do causes to the disease or disorder (e.g., patients with the flu virus in their blood have flu, and removing the flu virus will cure them), but sometimes it doesn’t (e.g., patients with flu have fever, but healing the fever alone doesn’t cure the patient of the flu).
Western medicine has been incredibly successful in the fields where an indicator which coincides with the root cause of the disease/disorder has been found (e.g., infective diseases, bone fractures), but has been incredibly unsuccessful in the fields where it could not find any yet (e.g., chronic pain, migraines).

I suspect that, often, medicine is playing a game of whack-a-mole with indicators which do not coincide with the root cause of a disease or disorder. Curing a symptom without curing the cause, will only cause another symptom to pop up.

Let’s take the example of the tweet that inspired this post.

  1. Some people show a given symptom.
  2. Researchers test those patients by extracting blood samples and compare the results with those of healthy people.
  3. Researchers find a correlation between an indicator (low vitamin C) and the symptom.
  4. They ask a subset of patients to eat more Vitamin C.
  5. Those patients eat more oranges.
  6. The patients who ate oranges become healthy.
  7. The researchers claim the study is a success and start prescribing Vitamin C to new patients with that disease.
  8. Pharmaceutical companies start producing Vitamin C integrators.
  9. Patients buy Vitamin C integrators, but their overall health does not seem to increase.
  10. Researchers make new studies: how strange, now some patients have high level of Vitamin C in their blood but are still not in good health!
  11. The researchers are puzzled; some even suspect a new disease or disorder.

What might have happened?
Perhaps, the symptoms were not caused by not having enough Vitamin C, but by not eating enough oranges. In this case, eating more Vitamin C would not solve the problem. Also, before Vitamin C was manufactured in pills, fruits were the primary source of it. The correlation between high fruits intake and high levels of Vitamin C in the blood *was* valid. Now that new sources of Vitamin C have been introduced, however, the correlation might not be valid anymore.
(Perhaps, low vitamin C was a byproduct of the disease or disorder, and fruits, not Vitamin C, were the cure.)
(I do not know if this specific example is correct but, please, observe the process in the example, not its specific content. This essay is not about Vitamin C diets, but about the consequences of curing proxy indicators.)

This is Goodhart’s law in action: finding a metric correlated with a disease or disorder caused humans to try to game it (by producing ad-hoc cures), which ultimately caused the metric to become ineffective.

When medicine fails over time to treat a disease or disorder, it did not understand its root cause. In such cases, curing a symptom is never a substitute for curing the cause. Curing symptomatic indicators is only as effective as a game of whack-a-mole, and sometimes it is even dangerous (because of side effects which are not offset by real, long-term benefits).

Tomorrow I will expand on why curing symptomatic indicators is dangerous. Subscribe to my mailing list to get the article.

Secured By miniOrange