Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

If the research in PFAS is at all like the article posted yesterday which did multiple statistical tests without doing the appropriate correction, I am not convinced that PFAS are all that harmful.

Because PFAS are so widespread and they have been around so long, if they were really bad we would be seeing massive issues. However, we are not.

Likely they do cause some harm if you have enough statistical power, but likely below the threshold that we live with (for example birth control, going to the beach, driving a car, red meat, and maybe even cooking itself).



>Likely they do cause some harm if you have enough statistical power, but likely below the threshold that we live with (for example birth control, going to the beach, driving a car, red meat, and maybe even cooking itself).

Isn't the entire reason these things are unsafe that they bioaccumulate over long periods of small dose exposure?


I am interested in where/if this bioaccumulates, like mercury does in marine life.


| and they have been around so long, if they were really bad we would be seeing massive issues. However, we are not.

1. They really haven't "been around so long". Not even a single human lifetime.

2. Plenty of other things cause cancer, endocrine disruption etc. Red meat and the beach are here to stay, but why should I be ok with some unaccountable corporation dumping additional pollutants into my environment?

But, by all means, continue to avoid a rational, cautious mode of decision making. It's probably not that harmful in my opinion.


Interesting thought, can you talk more about the methods and their flaws? Also what’s the article you’re referring to?


The discussion is

https://news.ycombinator.com/item?id=32438368#32442794

Here is my comment:

This is actually very poor statistics. Take a look at the actual study:

https://www.jhep-reports.eu/article/S2589-5559(22)00122-7/fu...

And look at table 3. They are testing 6 different types of PFAS, and only 1 is statistically significant. They need to be using the Bonferroni Correction because they are checking multiple hypothesis. To do that, you divide the required p-value (0.05) by the number of tests (6). If you do the Bonferroni Correction, none of the PFAS is statistically significant.

Relevant XKCD: https://xkcd.com/882/

Thus the real conclusion is that the study did not find any statistically significant link between 'forever chemicals' and liver cancer.


In the text they mention an FDR correction. In which case you typically don’t use 0.05 as the threshold. Bonferroni isn’t the only way to do that correction, but it a valid method (depending on the pvalue distribution). But with only 6 tests, it’s not really that critical. I usually only think of multiple testing correction in the context of thousands of tests.

Given your comment I was expecting something like 5 patients per group. But they had 50 cases and 50 controls. In this context, an uncorrected pvalue of 0.02 is pretty solid.




Consider applying for YC's Fall 2026 batch! Applications are open till July 27.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: