This paper came up on my news feed today:

According to the research, daily intake of glucosamine/chondroitin supplements for at least a year is associated with a 39% reduction in all-cause mortality.

This is a cohort study, not a randomized trial, so the evidence is not conclusive. Individuals taking supplements were self-selected. Perhaps there is another factor not controlled for in the paper. The researchers are calling for a more thorough study.

However a 39% reduction appears to be a pretty large effect. And glucosamine/condroitin supplements are fairly inexpensive and available over-the-counter, and seem likely to be relatively harmless even if the effect is illusory. Maybe the kind of person who takes this kind of bet comes out ahead, even if they're wrong sometimes.

I'm not looking for mere social permission or a gut-feel answer here. I want to understand your reasoning. How should a rationalist reason about this? Do we plug numbers into a Bayesian net? Should I take the bet and start supplementing daily? Is the effect too small to bother with? Should I wait for more evidence?

New Answer
New Comment

7 Answers sorted by

This is not just some random data fishing result. Even before the results of the two glucosamine papers this year, longevity researcher James Clement wrote in "The Switch", that glucosamine is "an autophagy inducer in a pathway separate fron mTOR-inhibiting." That is big, since autophagy seems to be responsible for most of the benefit from fasting and CR, but almost all autophagy that we know is activated via mTOR.

That means that you have a clear prior that it should increase your healthspan (but as with all autophagy inducers, like resveratrol or cold showers, don't take it after exercise, since it dulls the response). The prior together with the study results working out as expected is a very good sign for a rationalist IMHO.

Unfortunately glucosamine has at least one significant negative effect as well (see e.g It is similar to glucose, so can disrupt blood sugar stability.

Our biology is complex. Don't take substances just because of correlations. Strive to know why they work. If you are going to take glucosamine, do it when you want to be catabolic for a while (it has 15h half life), take it together with fat (I have fish oil and quercetin in my notes for reasons I forget - probably blood sugar results from the detailed notes on and know if you are in a risk group for diabetes 2, because it might mess slightly with your blood sugar.

The above is from my personal research notes. I don't take glucosamine yet, but believe I will start. I am not a health professional, just a hobbyist.

A much larger cohort study was published just before the Dana King paper:

They have adjusted for a wide variety of factors and concluded there was a 15% overall reduction in mortality associated with glucosamine supplementation.


Daniel Lemire covers medical developments in his blog, it is well worth following.

As you said, it's a cohort study. The cohorts are pretty different (gluco/condro users were more white, educated, non-smorkers, physically active than non-users, though pretty older in comparisons) and the adjustment could be skewed by the small number of users. But, it's mostly harmless and the effect size looks promising. Personally I'd make a quick review of the literature to decide whether to commit to taking the supplement and set up an alert on Pubmed.

As an aside, though it addresses the actual question, I'm developping a model to do precisely what you ask: providing health guidance and dashboards tailored to the desire to know yourself, propensity to risk (or better risk aversion) and data you can generate. I'm focusing on the PoC and business side for now, so can't discuss the specifics here.

Glucosamine is used by people with arthritis to reduce joint pain with activity. It would not be used by a group to ill to be active i.e. if you are healthy enough to worry about taking glucosamine you probably aren’t going to die this year. This result is purely selection bias obtained by data mining.

They said they controlled for that:

Multivariable-adjusted HR showed that the association was maintained after adjustment for age, sex, race, education, smoking status, and physical activity

Is there something in particular you saw that they did incorrectly in their analysis here?

Those factors kind of cover health, but I'd guess there's a big enough residual variation in health-consciousness to be responsible for the results.

Correlation does not imply causation. There is a clear correlation between the decline in piracy and global warming. Subsidizing pirates won't solve the climate crisis however. On a more serious note, glucosamine is usually a treatment for mild to moderate arthrosis. People who use it are likely the kind of people that take care of themselves and thinks good health is important.

Look for other differences between the two groups and you'll likely find something else there that is more likely to increase your lifespan.

I take glucosamine at least 4 times each year to sustain my high weightlifting.

All cause mortality? I very much doubt it'll protect you from car accidents or bear attacks.

Far more likely that people that people that take such supplements do so as they're health conscious and any perceived benefits are halo effect.

Eat a balanced diet, include skin, bone and shellfish and you'll get more than enough of both plus lots of other nutrients.

All cause mortality? I very much doubt it'll protect you from car accidents or bear attacks.

Improvements in all-cause mortality mean that if you sum up all causes of mortality, the risk goes down, not that all causes decrease.

For example, if the only ways to die are car accidents or bear attacks, with equal probability, a self-driving car that perfectly avoids all accidents would decrease all-cause mortality by 50% even though it doesn't do anything to prevent bear attacks.

An improvement in all-cause mortality is a useful metric because it avoids mistakes where you take an action that decreased one cause of mortality while raising another. Like how drinking poison would lower your risk from car accidents and bear attacks by 100% but would presumably raise all-cause mortality.