Hacker News Reader

Comment by serverholic | Hacker News Reader

Epidemiology strikes again! I watch a lot of videos on nutrition and researchers often say that randomized control trials are too expensive so we need to rely on epidemiology despite its flaws.

What I find crazy is that nations around the world collaborate and spend billions of dollars building a particle accelerator and yet we can't seem to find money to research something that affects every human every day of their lives.

View on HN

I guess that is because particles in the accelerator are not alive, so you can accelerate anyway you want. Do you really want to round up a group of seniors and gives them different doses of alcohol and see how much you can accelerate their demise?

That's a strawman, a randomized control trial does not have to be dangerous. In fact, we already do randomized control trials, they are just expensive and uncommon.

An experiment is not required, as the behavior (acceleration towards demise) may be observed and tracked at scale in the field

I know a few seniors who'd be down for that!

It's not just a money issue. It would be impractical and unethical to confine study subjects in controlled environments for long periods of time. That's what we would have to do in order to gather reliable data and eliminate confounding factors.

Most human nutrition studies are only observational and barely a step above junk science. Don't take them too seriously.

I'm not so confident RCTs of nutrition are unethical. The data are too damn valuable for society to ignore. Perhaps participants could be informed and willing to partake in the challenge trials, which although these participants might not be reflective of the overall population, we can get a better sense than correlational studies.

Humans are incredible explorers, self-sacrificing for the greater (perceived) good. We should consider continuing our legacy by understanding ourselves better through rigorous science. (And hopefully end the historical legacy of making thousands of other species extinct via our exploration.)

I believe the Journal of Controversial Idea (https://journalofcontroversialideas.org/) will explore some of these topics, shortly.

There are ways to at least mitigate some of those challenges and achieve a meaningful RCT without locking people up. Encouragement designs, for example: https://link.springer.com/article/10.1023/A:1008045618501

Randomized control trials are done all the time. They can and are done ethically. I don't know where you people are getting the idea that RCTs are unethical.

Proper large scale, long term RCTs have almost never been done in the human nutrition field. In order to really get the "controlled" part of an RCT the researchers have to confine people somewhere that all of their food and beverage intake can be accurately counted. Obviously this is very expensive and researchers can't ethically confine people in controlled environments for years. At most those studies have been done with very small subject counts for a few months at a time: hardly sufficient to obtain reliable results.

So as a practical matter most studies tend to rely on subject reported data, which is notoriously inaccurate. People forget to record little snacks, or don't weigh their portions, or lie about their behavior because they're ashamed.

Where the did you get the idea that I'm suggesting we lock people away for years?

Yes of course it's expensive which is why I said we should have larger scale studies with more test subjects funded by the government.

Also, what does subject reported data have to do with RTCs? What are you even arguing about?

You're totally missing the point. If the study relies on subject reported data then there's no proper control regardless of number of subjects. More funding won't fix that quality problem.

Why are you assuming this hypothetical study will rely on subject reported data?

What other method is there to gather nutritional data? The researchers either have to confine and monitor the subjects in a controlled environment, or leave them loose and rely on subject reported data. Maybe someday we'll have better technical solutions but for now those are the only practical options.

If you have a better approach for study design then I'm sure nutrition researchers would love to hear it.

You're making so many assumptions in your posts.

Yes, to do a proper RCT you need to confine and monitor subjects in a controlled environment. YOU used the word "years" and nobody ever suggested a study that long. If you're implying that you need a study that long to gather meaningful results then that's just simply not true.

Let me put this plainly, you can design RCTs that are:

A) Safe for participants.

B) Aren't too long.

C) Don't rely on subject reported data.

D) Still give meaningful results.

These are not mutually exclusive and they are done all the time. These are gold-standard experiments in the nutrition field.

Like you said, they are expensive to do and usually only have a small number of subjects. That's why I'm suggesting that wealthy governments should collaborate to fund massive-scale trials at maximum quality.

In the general case "B" and "D" are mutually exclusive. For the important questions in human nutrition like whether a diet high in saturated fat increases the risk of all cause mortality it takes many years for significant effects to appear.

There are many useful things we have learned and can learn from shorter studies.

Best of luck to you. I'm out.

Deliver food for the family to their door every morning?

Yes, it is relatively expensive, but not prohibitively so.

That has been done in a few studies. But delivering food doesn't mean that study subjects actually ate it, or didn't eat other food. So it doesn't solve the data quality problem.

To be fair, particle physics is a lot easier. Your argument holds in terms of investment though, if not in terms of expected outcome.

Because it involves humans? I am only half-joking, but I think it's much harder to track and study large numbers of humans than it is to do esoteric particle physics experiments.

We can't use maths to describe our digestive systems adequately, but we do it for particles.

I think your last paragraph is spot on!

Reply on news.ycombinator.com