Sunday, November 05, 2006

Hirsutism and genetic diversity   posted by agnostic @ 11/05/2006 11:00:00 PM

On Comedy Central I just saw Russell Peters (admixed South Asian and British, superficially more of the former), and figured I'd watch since stand-up comic routines are a great source of data for HBD issues that are taboo elsewhere (Comic View on BET is always reliable). In the clip below, skip and watch the bit from 2:45 to 6:00 if you haven't already guessed where this is going.

Now, we've covered hairiness in humans before (here and here), but the focus was mostly on the difference between homo sapiens and other primates. But as Peters clearly lays out -- and as always comes up whenever we discuss which ethnic group's females are hotter than which other's -- this trait varies substantially enough between populations that it's obvious to anyone who isn't blind and who's had even a little experience with the relevant groups. But the reasons for this variation seem even less clear than those behind the chimp-human variation. Continue below the fold for data on hirsutism and some (hopefully) educated speculation on why it varies between groups. Leave your own (hopefully) educated speculation in the comments.

Beginning with definitions, hirsutism in women is measured by some version of the Ferriman-Gallwey score, which tests different areas of the body for varying degrees of hairiness from 0-4 in increasing hairiness (example combinations). It seems the consensus is that scoring 3 is sufficient to qualify as hirsute for women -- for example, light hair on the upper lip, lower back, and chest would get an FG score of 1*3 = 3. This trait is not normally distributed: most women score low (2 or less on a scale of 0-36, where 36 = 4 points max per 9 areas measured), so you'd have to model it with a log-normal, gamma, exponential, or other such distribution. If you wanted to apply La Griffe's method of thresholds, then, you'd have to change the quantile score from the skewed distribution -- say it's log-normal -- to the z-score of the normal distribution, according to:

Q = exp(m + qs)

Where Q is some quantile score for the log-normal and q the corresponding z-score from the normal distribution, with m and s the normal's mean and SD. Using the standard normal with m = 0 and s = 1, the above simplifies to:

Q = exp(q)

It might be simpler still to just apply the method of thresholds to the log-transformed Q values instead of the Q values themselves (to reduce effects of skew) in order to estimate the difference in means between groups. Not being a math wiz (relatively speaking), I simply note the choices to be made, present the data below, and leave it to those more skilled at modeling to resolve.

Now, to the rough topography of the hairiness map. What data I've found through PubMed suggests that the stereotypes about body hair are true, so in the absence of data on a particular population, going with your gut is probably safe assuming you've had minimal experience with said group. First, a recent study found no difference in hirsutism between Whites and Blacks in the southeastern US -- given the region, most likely the Whites are northern European and the Blacks sub-Saharan African, from the West most likely (cite):

MAIN OUTCOME MEASURES: Terminal body hair growth was assessed using the mFG scoring system; nine body areas were scored from 0-4 for terminal hair growth distribution. RESULTS: The mFG scores were not normally distributed; although cluster analysis failed to identify a natural cutoff value or clustering of the population, principal component and univariate analyses denoted two nearly distinct clusters that occurred above and below an mFG value of 2, with the bulk of the scores below. Overall, an mFG score of at least 3 was observed in 22.1% of all subjects (i.e. the upper quartile); of these subjects, 69.3% complained of being hirsute, compared with 15.8% of women with an mFG score below this value, and similar to the proportion of women with an mFG score of at least 8 who considered themselves to be hirsute (70.0%). Overall, there were no significant differences between Black and White women.

That fits with most people's experience that Blacks don't appear any more or less hairy than northern Europeans. But who's even less hairy? East Asians, of course! Even in patients who have elevated levels of hormones which tend to lead to greater body hair, the Japanese women were not hirsute, while the American and Italian women were (cite):

RESULTS: Women from Japan were less obese (p <>, although the percentage of cystic ovaries (68% to 80%) was comparable. Serum luteinizing hormone, testosterone, and estradiol were similar, but levels of 3 alpha-androstanediol glucuronide, which was elevated in women from the United States and Italy, was normal in women from Japan. The adrenal androgens, dehydroepiandrosterone sulfate and 11 beta-hydroxyandrostenedione were elevated in 48% to 64% of the patients and by a similar percentage in the three groups.

The same appears true for Southeast Asians such as Thai women (cite):

RESULTS: Five hundred and thirty-one women underwent a physical exam. The women who had the total hair-growth score of 0, 1 and 2 by mF-G-L method accounted for 97.8% of all the subjects. All of the 11 subjects with a total score of 3 or more considered themselves to have excessive growth of hair.

So, only 2.2% of representative Thai women were hirsute compared to the 22.1% of northern Europeans and sub-Saharan Africans mentioned above -- an order of magnitude less! Who's more hairy than northern Europeans and s-S Africans? The Greeks, of course! Take the following with a grain of salt, since it was taken from a representative sample of a Greek island (Lesbos), and it's well known that island people may turn out weird due to founder effects, inbreeding, and what-have-you. But Lesbos is very close to other Greek islands and to the heavily populated part of Turkey that borders the Aegean Sea and the Sea of Marmara. Bearing that in mind (cite):

To determine the prevalence of PCOS [Polycystic Ovary Syndrome -- agnostic] in the Greek population as well as the metabolic parameters, we performed a cross-sectional study of 192 women of reproductive age (17-45 yr), living on the Greek island of Lesbos. They were divided into 4 groups according to the presence of hirsutism (defined as a Ferriman-Gallwey score > or = 6) and O/M: group N (n = 108), regular menses and absence of hirsutism; group 1 (n = 56), regular menses and hirsutism; group 2 (n = 10), O/M and absence of hirsutism; and group 3 (n = 18), O/M and hirsutism.

Then (56+18)/192 = 38.5% of the sampled women were hirsute. This is actually an underestimate since in this study the threshold was not an FG score of 3 or greater but a more stringent 6 or greater! Again given the skew of the hirsutism distribution, a more realistic estimate might be somewhere around 50% scoring 3 or greater. As for Spaniards (cite):

Hirsutism was defined by a modified Ferriman-Gallwey score of 8 or more . . . PCOS was present in 10 (6.5%), hirsutism was present in 11 (7.1%), and acne was present in 19 (12.3%) of the 154 women.

Without a clear understanding of the exact shape of the distribution, we can't know what percent of Spanish women scored at just 3 or greater on the FG scale in order to compare them with the above studies, since the Spaniards were judged hirsute only if they measured 8 or greater. It seems reasonable to assume that on average the Spanish are equally or somewhat less hairy than the Greeks, but hairier than northern Europeans, s-S Africans, and NE and SE Asians. The only data I could find on New Worlders was a not representative sample of Mexican-American women who'd been pre-selected for a family history of coronary artery disease. Thus the following is certainly an overestimate, especially since the hirsutism data were self-reported rather than measured on the FG scale (cite):

Using the questionnaire to diagnose PCOS, we found that 20 (13%) of the 156 women studied met criteria for PCOS (self-reported irregular menses and clinical signs of hyperandrogenism). In the remaining women, 17 (11%) reported menstrual abnormality only, 71 (46%) reported hyperandrogenism only, and 41 (31%) reported no abnormalities.

Then, for what it's worth, 0.13 + 0.87*0.46 = 53.0% of this selected sample were hirsute. As we've mentioned before, Mexican-Americans are not Amerindians -- they are admixed, so it could well be that even correcting for the bias in the above sample, much of the rest may simply reflect admixture from hairy Mediterranean populations. At least from what I've seen from National Geographic or The Discovery Channel, the indigenous of the Americas are more like East Asians in hairiness.

And to bring it back to Russell Peters' original remark, the only study that looked at hirsutism among South Asians and Caucasians (most likely Britons) did so as part of a larger study on PCOS (cite). Among the 47 South Asian and 40 Caucasian women who had PCOS, the median FG score was 18 for the former and 7.5 for the latter (P less than 0.0001); and even among the 11 South Asian and 22 Caucasian controls, the median FG score was 8 for the former and 1.5 for the latter (P less than 0.04). So, South Asians are clearly much hairier than northern Europeans, regardless of having PCOS or not. It should be noted, though, that 40 of 47 South Asians with PCOS and 9 of 11 South Asian controls were Pakistani, and consequently 29 of 47 of those with PCOS and 4 of 11 of the controls came from consanguinous backgrounds (which is more prevalent in Pakistan). None of the Caucasians with PCOS or controls were from consanguinous backgrounds.

To summarize the picture so far, it looks like hairiness is most frequent in the area from roughly Greece through the Middle East and into South Asia, is perhaps somewhat lower in the rest of Mediterranean Europe, is lower still in northern Europe and s-S Africa, and lowest of all in NE and SE Asia (and likely among the Amerindians as well). This inter-group variation clearly reflects "recent" evolution, as in after the major continental races went their separate ways; and given the northern European - Mediterranean difference, this may reflect even more recent evolution, as in the past 10,000 years. Neutral drift is out of the question. Though hairiness may not be at fixation in South Asia, by all accounts it's close enough for government work. Since the time in generations for the fixation of a single mutation = 4Ne, where Ne is the effective population size, let's make the generous assumptions that this neutral drift began 10,000 years ago rather than more recently, and that 1 generation = 20 years rather than some longer estimate like 25 years. Then, under these assumptions, the effective population size from which this emerged would only have been 125! That's not the same as census size, for sure, but it's still off by at least an order of magnitude. As usual, neutral drift is unlikely to be informative about evolution in large, non-founder populations over a "recent" timescale for functional traits (and vice versa).

We rehashed a lot of adaptationist accounts of why humans are hairless for primates in the threads linked to at the beginning of the post -- adaptation to climate, protection against ectoparasites, etc. But the split between relatively more and less hairy populations doesn't mirror the two big splits of the past 10,000 years, both a result of the agricultural transition: 1) increasingly complex societies, which likely selected for higher IQ; and 2) a shift in diet and pathogen stress. Hairiness clearly doesn't track social complexity or current mean IQ of different groups, and it's not clear how hairiness would help or harm integration into more or less complex societies. Neither does hairiness appear to track dietary changes, though there may be some subtler difference in one particular foodstuff I'm unaware of that does carve up the world the same way hairiness does. And if it reflected recent changes in pathogen stress, you'd expect it to slice up the world into one group containing s-S Africans, Middle Easterners and South Asians, and Southeast Asians; and another containing northern Europeans, Amerindians, and NE Asians. This mapping seems the least erroneous of those examined -- perhaps pathogen stress might account for some substantial fraction of the variance, even if not a majority of it. However, this still seems unsatisfying since most recent changes in pathogen stress involve really nasty fuckers like malaria, small pox, measles, plague, and so on -- so we'd expect a response that kicked into gear earlier than puberty.

Which brings us to the next area of explanations: sexual selection. The fact that women of various parts of the world vary significantly between racial and ethnic groups, but are apparently close to the levels of their co-ethnic males, would fit the pattern of slow dimorphism. Since body hair sets in at puberty and declines somewhat with old age, it seems likely that this is a secondary sex characteristic even among females. Clearly no male would prefer this trait, so it's conceivable that females were sexually selecting males who had masculinely hairy bodies, and that their children -- sons and daughters alike -- inherited alleles that predisposed to hairy bodies. Presumably, over a long stretch of time this might become more dimorphic, so that South Asian females would come to resemble Swedish females in body hair, but as Russell Peters notes, we've arrived at a cultural point where hair-removal is relatively cheap and commonplace. Thus, the alleles may persist at their high frequencies among the hairier populations since their effects (at least as regards male choice) will be largely mitigated by cultural innovation.

We already know that females from more pathogen-stressed environments emphasize "good looks" in a male mate more than do females from less-stressed environments (Gangestad & Buss 1993). And this study suggests that at least some degree of body hair is attractive on males. So it could be that, for whatever other reason, South Asians and nearby populations were slightly hairier than s-S Africans (who also select for good looks due to pathogen stress), but that this difference was magnified by sexual selection in the former areas but not the latter, if we assume there was less variability in hairiness in older s-S African populations relative to older South Asian populations. Perhaps due to small differences in local ecology, hairlessness is under tighter functional constraint in s-S Africa and SE Asia, while it's not in other germ hotspots like South Asia and the Middle East. Evolution is stochastic, after all. Well, I took a shot at it anyway. What better ideas are there that are consistent with the data outlined here?