Sunday, March 01, 2009

I'm going to live forever, cancer-free; part of a continuing series

via Pejman Yousefzadeh, the latest in the parade of good news about the health benefits of the nectar of the gods.
A cup of joe a day may help keep skin cancer away: A new study shows that caffeine helps kill off human cells damaged by ultraviolet light, one of the key triggers of several types of skin cancer.

The finding, detailed in Feb. 26 online issue of the Journal of Investigative Dermatology, could one day lead to the development of caffeine creams or ointments to help reverse the effects of UV damage in humans and prevent some skin cancers.

Nonmelanoma skin cancers, which rarely metastasize or cause death, are the most common form of cancer in humans, with more than 1 million new cases occurring each year in the United States alone. (Melanoma is, however, one of the deadlier cancers.)

Exposure to ultraviolet light is one of the most important factors in causing nonmelanoma cancers. The rays cause DNA damage to skin cells, which then mutate or become cancerous.

Several studies have shown that people who regularly drink coffee or tea seem to have lower incidences of nonmelanoma skin cancers. One recent study of more than 90,000 Caucasian women found that with each additional cup of caffeinated coffee consumed, there was an associated 5 percent decreased risk of developing one of these skin cancers (decaf coffee had no effect).


Todd Seavey said...

As you know -- but as it wouldn't hurt reminding less-skeptical readers -- you can find these sorts of subtle (i.e., meaningless) tiny-statistical-variation studies leaning one way, then the other nearly every day in the news about coffee, for the simple reason that it's a popular substance and attracts attention. They're meaningless. Would that all such studies had as little impact on behavior and regulation as coffee studies likely do, though.

Jacob T. Levy said...

I can never figure out what your attitude is toward peer-reviewed science. If something *hasn't* been shown by such science, then it's junk science to be disregarded entirely-- but you don't seem to give any consistent weight to the nonnegative results of scientific inquiry. "Subtle" and "tiny" and "meaningless" are none of them the same concept; results can be consistent and meaningful although small. A consistent 5% drop in cancer rates across multiple studies of tens of thousands of participants adds information.

What's your threshold for "meaningfulness" of a medical-science finding, and why should we prefer yours to those of peer-reviewed grants and scientific publications? Is there any positive relationship between any behavioral or environmental variable besides smoking and a health outcome that you *do* believe to be borne out by the evidence?

Anyway, as best as I can tell the coffee studies are pretty consistently in the same direction-- even more consistently than the moderate-red-wine studies. The findings on blood pressure are mixed, but the findings on skin cancer, dementia, diabetes, gout, and Parkinson's all seem to be uncontradicted-- and some of the effects are pretty large.

Todd Seavey said...

I'll refer people to my co-workers at for a lot of case-by-case examples of health scares, but a (very rough) rule of thumb to keep in mind here that helps explain why we so often say smoking/lung cancer and smoking/heart disease studies matter in a way that these coffee studies do not is that they are not even remotely comparable in the size of the statistical fluctuations.

Whereas coffee, fad nutrition, and chemical-scare stories routinely involve tiny 1 or 2% increases or decreases in disease rates -- easily produced by random fluctuations or confounders -- smoking is the poster child for epidemiological evidence because it increases lung cancer not by 1% or 2% but by something like _1200%_.

As another very rough rule, you'll sometimes hear epidemiologists say you should at least see a _doubling_ in incidence before thinking you're onto something, though that's a shorthand for a more complex calculation dependent on your sample size, of course.

And this is not "just us" saying this, as I fear you're suggesting -- this is what the best scientists in most fields will say, but more and more the peer-review journals will publish things _despite_ also (much more quietly) acknowledging that they don't mean much (in much the same way TV news lets itself off the hook morally by throwing in one line saying "though many experts disagree").

In any case, though, if you look at the numbers on smoking vs. coffee-type stuff, you'll see it's not some subtle, arcane difference but more like mountains vs. anthills.

(This is somewhat related to my reasons for not wanting to read too much into small-sample political generalizations, of the "no senator in an off-year election since 1968 from the south" etc. type.)

Long story short: I'd ignore most mainstream health reporting, as do most nutrition scientists, you'll find. The disjunct between what those in the know know and what goes on in the popular culture is far more vast than almost anyone realizes (and disturbing though it might be, I must assume the same is likely true in all those other fields about which I know little -- think of how many errors there are in local newspaper stories every time it happens to be a topic you know well...).