Shared publicly  - 
 
This is sloppy, risible science. A new paper finds that Pill use in different countries correlates with incidence of prostate cancer. Here is an extended version of a rant I had on Twitter.

1) You may have noticed that people who take the Pill don't have prostates. Their suggestion is that Pill use -> higher oestrogen levels in the environment or drinking water -> higher prostate cancer risk. The problem is that both of those links are tenuous as hell and they haven't made any efforts to substantiate them. Did they actually measure oestrogen levels in the environment to see if they correlated with either of those things? No. Because that would be too close to actual work. They also say that this mechanism is “plausible” without providing a reference. Which means they pulled it out of the air. Lots of studies produce correlations that aren't necessarily causative, but if you don't have the first clue about a mechanism that could plausible explain your result, perhaps it's time to find one first?

2) There are all sorts of things that vary between countries that could affect these results and it's good to see that they accounted for all of them. They adjusted their stats for GDP and... wait, that's IT? Give me a break.

3) What the hell do national boundaries have to do with anything? If this mechanism is at all plausible, the variation would be at the local or regional level. Why compare national differences? Because it's easier.

4) They say study is "speculative" and “hypothesis-generating”. That's a common refrain with these sorts of "ecological studies". It is a fancy way of saying "so shallow as to be useless”. No, these studies can't prove causation on their own, but the evidence they provide is so pathetically weak that they could equally well have generated some hypotheses by shouting them to the winds.

5) Rubbish like this is a complete insult to scientists in fields who actually have to do significant work before they can publish a paper. Just think of how much effort goes into your average neuroscience or molecular biology or behavioural ecology paper - months, probably years, of experiments, pernickety reviewers, and more. And these folks just clumsily mash together two data sets, found a correlation that they can't explain, and publish it?? Thanks. Thanks a bunch.
50
27
Amy Woods's profile photoJoseph Zullo's profile photoMark Goodwin's profile photoPhạm Hải's profile photo
30 comments
 
Keep on ranting, Ed!

Seems to me like a clear case of peer review/editorial oversight fail. Unfortunately these type of studies appear to have become more frequent in bulk open access journals like BMJ Open and PLoS ONE, where peer review is often minimal.
Ed Yong
 
And you can actually see the reviewers' comments for yourself on the paper link
 
Unfortunately, there is a trend to mix causation and correlation especially in epidemiological studies and sociology. It's so easy. I made a "study" like that in less than 90 min, by taking some basic socio-historico-economic parameters and trying to correlate them with the perception whether a country is from "West" or no. I got results. But this has not the pretention of explaining. Just correlating.
Correlation should be the basis for more in-depth studies, they should help to highlight and guideline further studies.
Ed Yong
+
5
6
5
 
No, hang on. They're not making claims to causation and I haven't criticised them on that grounds. My point is that throwing up a correlation is completely useless as a guideline for future studies without having at least a smidgen of corroborative evidence.

And that in some fields, scientists are expected to find that evidence before they can publish.
 
Thanks Ed, I missed the link to the reviewers' comments (which are astonishingly uncritical: how did that happen?).
 
In addition to being insulting rubbish, this kind of studies are frighteningly good at generating false opinions in the general population's consensus. People eyeball through these studies (or just read the titles of short news stories about them) and just remember the "punchline" and take it as a scientifically proven fact. I'm sure there's already a bunch of people wanting a ban on the Pill because it causes prostate cancer. It's kinda sad. Actually, very sad.
 
Havent read the artical
Because one I'm
Tired and well yea but what I remember I'd that some scis believe the g spot Is the female prostate which could be the effect?
 
Omgoodness! This reminds me of the 'study' done on Austism Spectrum Disorder and vaccinations. >:(
 
The headlines in the press today: pill 'linked to' PCa, 'associated with', 'may have increased', 'behind', 'may be tied to', 'connection', 'ups', 'increases risk', etc....from some sources that I, as a non-scientist member of the public, would maybe have trusted.
So glad I noticed you criticisms on Twitter last night, +Ed Yong Thanks!
 
Amy Woods nailed it. It also sounds like the garbage I have ead about Multiple Sclerosis and. . .mercury, viruses, tanin(!), sun exposure (both lack of and excessive). . Lousy "studies" that produce junk science.
 
Blindly passing along questionable findings is irresponsible reporting. Sometimed I weep for the state of journalism. Thanks for reading the study before writing about it, Ed. 
 
They're just not spinning it the right way. New headline: Homeopathic Estrogen Linked to Prostate Cancer
 
Statistics is not a branch of epistemology. But apparently it has become not just a branch, but the trunk of science.
 
while statistics are evil and can be manipulated and massaged to find a "result" where none actually exists, it would be nearly impossible to get a paper published without them....
 
Another slight nit... You get those extra oestrogens even by skin contact. Which would suggest married men ought to have higher incidence of prostate cancer, when controlling for STD's and what not. That however is not the case, and it isn't as though that kind of data isn't available for cross-validation, in more than one form. So even if you stoop as low as to go with correlations alone, you could try doing it right.
 
To be fair, they do conclude that environmental factors associated with oestrogen levels (e.g in the drinking water) could be worth investigating.
 
Kind of like how "everything" seems to be "known to cause cancer in the state of California"...
 
+Ed Yong ... you say they pulled the correlation "out of the air." Reading that made me feel uncultured as I had just mentally concluded before that sentence that they had pulled the correlation out of their ass. Oh, by the way, thank you for your hard work.
 
Publishing standards practically don't exist anymore. I blame the interwebz (joking... sort of).
 
my first guess would be the correlation is due to availability of medical care: in places where medical care is sufficient for the pill to be widely available, it is also sufficient for most instances of prostate cancer to be diagnosed instead of overlooked.
 
Thanks, +Ed Yong, good analysis. These kind of studies are generally only good for the authors, who get to go on the Today show, lazy journalists who get something to fill their time/space and industry lobbyists who can use it to distort facts and sway public opinion.
Ed Yong
 
+Joanna Staebler-Kimmel That's exactly what I thought too. Plus those countries might also tend to have more accurate cancer registrations.
 
I've noticed more of my neighbors are unemployed. I've also read that the unemployment rate for the entire U.S. is pretty high. As a result, a plausible explanation is that when my neighbors are unemployed, it causes a domino effect across the U.S. So, if I can just help my neighbors get jobs, the entire U.S.'s economy will be restored.

OK, now I'm just awaiting publication.
 
This study reminds me of that great old graph which plots the change in numbers of pirates sailing the seas and average global temperatures over time. Naturally, there is an inverse correlation.
The caption?
"Save the planet. Become a pirate!" TROLOLOLOLOL
 
As usual, XKCD has the relevant joke: http://xkcd.com/925/

But yeah, when you have the following paragraph in a paper:
"Some may argue that our results only reflect screening and treatment patterns for prostate cancer, with the more developed countries having both a higher use of OCs and a higher incidence of prostate cancer. Unfortunately data on worldwide screening tendencies or prostate specific antigen (PSA) use is unavailable. However, we included a multivariable analysis... "
You're a SCIENTIST. If data does not exist, you go out and generate it. That's WHAT WE DO. I understand it's not practically possible given limited funds and time to gather all the data that would make your story complete. However, when you're essentially saying "There's a giant whole in this story, but plugging it would take too much effort, and only some of you are going to call us on it, so we'll just see if we can skate by with a half-assed " mulitvariable analysis' " it really does say that your paper shouldn't have been published to begin with.
Add a comment...