Low Fat Diet and Sunscreen: a Recipe for Disaster
people.csail.mit.edu
people.csail.mit.edu
There's nothing wrong with hypothesizing that (a) modern medical practices are leading to deficiencies in Vitamin D among certain populations and (b) Vitamin D deficiency is linked to autism. But the responsible next step is to get some good data, or at least link to some. It's not to write stuff like this:
Since the 1970's, our country has witnessed a 3000% increase in the sales of sunscreen products, alongside a 30% increase in deaths from melanoma[17]. Isn't it time to admit that sunscreen is not working?
This is a telling paragraph. It tells me that the author is prone to making mistakes straight out of Darrell Huff's How to Lie With Statistics: Clock A strikes midnight a few seconds before clock B; isn't it time to admit that A caused B to strike?
Autism seems to be one of those issues that really inspires the quack epidemiologist. When you combine it with the opportunity to lecture new mothers -- our culture's favorite vice -- the temptation becomes irresistable. But it's worth remembering that actual epidemiology is really hard. Diagnosing autism without bias is hard. Randomly sampling the population for autism is even harder. Even something relatively simple, like measuring Vitamin D levels in the blood, is tedious and expensive and difficult to do without bias. But just because anecdotes and handwaving are much easier doesn't make them more believable.
There is no fallacy in the statement you've focused on. I'm a bit confused. Sunscreen is sold and used to prevent cancer. There's no evidence it has, and sound reasons to suggest sunscreen has contributed to cancer. That's all it's saying.
> Diagnosing autism without bias is hard.
Are you suggesting the autism epidemic is a product of better diagnosis? If so, where are all the autistic 50 year olds?
The observation is that sunscreen sales and melanoma rates are correlated. That's probably true, but it doesn't tell us anything about causation, one way or the other. Perhaps sunscreen is ineffective against melanoma. Or perhaps melanoma rates are rising because sunscreen use causes melanoma. Or perhaps sunscreen sales are rising because melanoma rates are rising. (As someone who has just watched a close relative die of metastatic melanoma, I assure you that the experience really inspires you to wear sunscreen and go to the dermatologist. Every time I get a pimple, I have to fight the desire to visit the emergency room.)
Perhaps sunscreen sales and melanoma rate would be completely unrelated, if they weren't both driven by an external cause. Perhaps they're both caused by population increases in the Sun Belt, or by other demographic changes.
Perhaps sunscreen is highly effective against melanoma, but there are still lots of older patients who got all their sun in the 1940s, 50s, and 60s before sunscreen was invented, and those patients are still getting melanoma at large rates. Or perhaps sunscreen is completely ineffective, but we won't know until the people who used it as kids reach their sixties. Or perhaps sunscreen causes melanoma... but it takes more than fifty years to do so.
Perhaps sunscreen causes autism! I guess that's what this article is trying to suggest.
As you can see, pulling hypotheses out of one's ass is easy to do, and it's fun for the whole family. It's also uninformative. Scientific studies would be informative. I'm sure there are lots and lots and lots of published studies which address hypotheses like these. It's a pity that the article didn't cite any of those studies.
"While few epidemiologic studies have examined the relationship of sunscreen use and skin cancer, two studies suggest that sunscreens may not be effective in preventing skin cancer. A large case-control study showed higher risks of melanoma in men who used sunscreens, and a large prospective study showed a higher incidence of basal cell carcinoma in women who used sunscreens. The excess risks in the latter study persisted after multiple adjustment for differences in skin type and time spent outdoors."
-- "Could Sunscreens Increase Melanoma Risk?" http://www.ajph.org/cgi/reprint/82/4/614
This is from 1992 and earlier, though. One wonders if there has been any more progress in recent years.
And here is where I get to rant about my own pet peeve: Why should I be surprised that so few people know how to properly cite scientific literature? That literature is locked behind paywalls! I'm not going to even bother using PubMed on this, and it's only partly because I don't care. It's mostly because, even if I turn up a definitive set of review papers in Nature on this very topic, I won't be able to read them without paying $15 each, or whatever.
I, too, wonder if there aren't more recent studies. The two studies cited were from 1985 and 1990, though; if that's the most they could come up with in 1992, it sounds like the area is sparsely researched.
If sunscreen sales clearly negatively correlated to melanoma, that would be a point worthy of consideration in this discussion, no? It is reasonable to consider this lack of correlation as a piece of evidence in support of the discussed framework. Apply your own grain of salt: this is not a scientific publication and the audience is assumed to be a reasonable person.
There would be practically nothing to say in the "non-sciences" (medicine, economics, etc.) if one were not allowed to admit "circumstantial evidence."
I'm just finished with it. It doesn't have much explanatory power. It fits a big universe of hypotheses, some of which I have outlined above. It's provocative, which is not the same thing as conclusive. But consider me provoked. I'm curious. Bring on the science.
Until someone brings some more studies to the table there's really not much more to say. I'd rather spend my afternoon getting some carefully modulated sun exposure. ;)
OTOH, some 10 years ago, some sunscreen makes were forbidden (at least in Spain) because a component in them (I can't recall the name) had been proven to cause melanoma.
Circumstantial evidence is court admissible.
The autism "epidemic" is largely being caused by an expansion of the definition of autism. Hence, lots of young people with Asperger's, and other mild autisms, but not many 50-year-olds. They're not being diagnosed. (Whether Asperger's is an actual syndrome or just a product of upbringing and personality type: that's a different argument. I tend toward the latter -- I am skeptical about its existence at all.)
Basically, consider the sumptoms of deafness, impotence and tingling in the hands and feet (D, I and T). Suppose that deafness, impotence and tingling are all independent of each other (P(D | I) = P(D)). In this case, there is no syndrome. If, however, they are strongly correlated(P(D | I) >> P(D)), we have a syndrome.
Basically, "deaf and impotent provides no information about tingling" == no syndrome.
"Deaf and impotent implies a high probability of tingling" == syndrome.
Anyway, the main point remains: be careful with "common sense" conclusions.
Unfortunately adult autism care in the US is fairly deplorable.
As I read it, I quickly categorized it as an interesting persuasive essay citing supporting facts. I have no problem reading papers like this and gaining what I can from them. There is a place for persuasive essays and a place for research papers. As long as the author is not trying to misrepresent - no problems.
I read up on the author a bit and she seems to be closer to an expect on these issues than a vast majority of op-ed diet and autism write-ups that I've seen. Plus she's got an undergrad biophysics degree to boot.
Anybody can put forward a testable hypothesis. That doesn't make you a good scientist.
A good scientist is one who puts forward a testable hypothesis and then tests it before publishing.
A great scientist is one who creates a testable hypothesis, tests it, and then -- when the data fail to prove the hypothesis -- shrugs, publishes the definitive refutation of her own hypothesis, and moves on to something else. It took me years to realize just how rare a skill this is, even for scientists. Part of the problem is that the career system for scientists actively discourages this -- once you've finally managed to sell a line of research to a funding agency, there are very strong incentives to keep that research going, even if it isn't getting anywhere.
But as you say - it's a testable hypothesis, so I don't care how it's been written up, it would be good to test this.
1. If someone feels the need to point out, at the end of an essay, that they have a PhD, they're probably a kook.
2. That goes doubly for anyone who feels the need to tell you which prestigious institution their PhD is from.
I'm sorry, but the format of this site trips off my whackjob detector and no one should be willing to take random medical advice without at least the appearance of support from the medical literature.
This is doubly true when it comes to subjects of fertility or diet.
Their study:
http://www.johnson.cornell.edu/faculty/profiles/Waldman/AUTI...
If you sample 100 uncorrelated variables looking for significance at 95% confidence, you should find about 5 which are significant.
Data mining to produce hypotheses is fraught with danger. You should declare your hypothesis and then go to the data, otherwise your confidence measurement is bogus.
High pasta diets are not exactly on the menu in Mexico.
ORLY?
Seriously, is there study that show a low-fat/type I diabetes connection?
When mom eats french fries and rice the insulin spikes permanently screw up the kid's endocrine system, leaving a propensity to obesity if not diabetes.
Japan is in love with fat in all its forms (fatty marbled beef, fatty tuna, fatty pork) and there is almost no culture of sun-screen usage, but they also are much better at producing melanin than europeans, making for pretty effective natural protection.
They also have one of (if not the) largest populations of centenarians on earth.
I think you meant "Type II".
When "verifying things on your own" you run into all kinds of unfounded and delusional ravings on the internet. Doctors may err on the side of conventional wisdom, but that's not a call to disregard them and err on any nonsense that you come aross.
Now granted, those articles could have been biased, but I had no idea there was any sort of left wing vs right wing thing going on. Stupid politrix.
The entire movement reminded me of Moneyball by Michael Lewis.
However, it's extremely common for doctors to prescribe medicine off-label, before the full level three trials have completed for that use (which can take a decade). There is often compelling evidence a medicine is effective (in the Bayesian sense) far before there is official "Evidence" which counts for "Evidence Based Medicine".
The very fact that it's called "Evidence Based Medicine" should set off alarm bells. Being against it sounds like being against science in medicine, which would just be stupid. In fact, it's code for "restrict public care facilities from being able to offer a full range of medical benefits to conserve money".
P.S. We just fully implemented Twisted at TicketStumbler :) (I heard you guys use it, and love it, at Justin.TV).
Twisted really is awesome; we've been super happy with it. A little sad they wouldn't take our memcache get_multi patch (they said we hadn't written enough tests for it).
What on earth is "right-wing" about that?
Everyone should of course be for applying scientific evidence to medical practice.