Some researchers from Milan, Italy [it had to be from Italy] say that women with the most severe form of endometriosis are also the most attractive.
Their study involved three groups of 100 women--one group with rectovaginal endometriosis (RVE), the worst kind; a second group with peritoneal/ovarian endometriosis (POE), which is milder; and a third group who had no evidence of endometriosis.
Attractiveness was judged on a scale of 1 to 5 by raters blinded to the diagnoses of the women.
They found that 31/100 with RVE were attractive or very attractive compared to only 8/100 of the POE group and 9/100 in the No Endometriosis group. The difference was significant with p < 0.001.
Further proof of the attractiveness of the RVE women was that they first engaged in sexual intercourse at a significantly earlier age than those in the less attractive groups.
Ages of women in the three groups were similar at 32. Morphologic characteristics of the women, such as hair and eye color, BMI and waist-to-hip ratio, were similar for the three groups but RVE cohort had a significantly higher breast-to-underbreast ratio [a scientific term for big boobs].
The authors postulated that higher estrogen levels are linked with attractiveness and might also be responsible for the development of endometriosis.
The paper was published ahead of print last month in the journal Fertility and Sterility.
Some issues come to mind after reading this paper.
As they would have to be, the sexual histories of the subjects were self-reported. This sort of thing is always open to skepticism.
Four doctors, two men and two women, judged the subjects’ attractiveness. Is a panel of only four doctors sufficient to rate attractiveness? Who knows? In fairness, the inter-rater reliability of the four judges was good.
The criteria for attractiveness were not stated, so evidently “Beauty is in the eye of the beholder.”
The rating scale of 1 to 5 was converted to a 3-point scale as follows: > 3.5 very attractive or rather attractive; 3.5-2.5 averagely attractive; < 2.5 not very attractive or not at all attractive [What must the < 2.5 group have looked like?].
I’m not a statistician but that doesn’t seem kosher to me. 1) It’s not a 5-point scale if you have decimals. 2) I believe the conversion of the scale from 5 to 3 was done to provide better numbers for comparison. Otherwise, why not have the judges just rate them on a scale of 1 to 3 from the beginning? There must not have been a statistically significant difference when they crunched the numbers using the 5-point scale.
Since the group without endometriosis serves as a control, it represents a sample of all women of Milan. Is it possible that only 9% of women in Milan, the fashion capital of Italy [some say the world], are rather attractive or very attractive?
As usual, the media reported these findings uncritically. Along with its article, a site called Medical Daily posted a sort of NSFW photograph of celebrity Padma Lakshmi, who is very attractive and has endometriosis, severity unstated. She is not a resident of Milan but did some modeling there in the past. Also, she was once married to author Salman Rushdie, who has a son named Milan from a previous marriage.
The Medical Daily headline “Why Women Diagnosed With Rare Gynecological Condition Are "Unusually Attractive" is misleading because the condition is not rare, ‘rather attractive’ is not ‘unusually attractive,’ and the study only speculates but does not explain why the women are attractive.
Cosmopolitan handled the story with typical restraint using the title “This Painful Condition Makes You Attractive,” which of course is a complete misinterpretation of the results. The short write up is accompanied by some scathing comments from endometriosis sufferers wondering why research is focused on how women look rather than curing the disease.
There is one good thing about research like this. It keeps me occupied.
10 comments:
Footnote b of Table 4 tells you how the 5 point scale worked "A score was calculated as the mean of the judgments of four independent observers (two females and two males), who independently gave a score of 5 (very attractive), 4 (rather attractive), 3 (averagely attractive), 2 (not very attractive), and 1 (not at all attractive)."
So four people judged each woman on the 5 point scale, those scores were averaged and then those scores were split into three groups according to their >3.5, 3.5-2.5 and <2.5 system. I also am not a statistician but this doesn't seem like it's a statistical cheat.
Dear Seeking, thanks for commenting. I am having an exchange with @reedroberts about the same issue on Twitter. I guess I didn't state it well in the blog.
I should have said I don't see how you can convert a 5-point ranking scale into a 3-point ranking scale and then convert the 3-point ranking scale into continuous data with averages.
With the caveat that statistics is the least exact area of mathematics, I hope someone who is a statistician will comment.
Thanks for the reply.
There's no conversion though. Suppose we have the picture of a patient. Dr A rates them 4, Dr B rates them 5, Dr C rates them 3 and Dr D rates them 5. How can you use these four ratings to get some kind of measure of attractiveness? You could just sum them up. That would give you a measure of attractiveness. The advantage of taking the average and dividing by 4 is that it gives the numbers some sort of meaning on the original 5 point scale. A score of 4.5 is in the "more attractive" region of the scale while a score of 1.5 is in the "less attractive" region. It's not really continuous data, more like a squeezed 16 point scale (20 points being the highest possible score from the sum of the 4 Drs and 4 being the lowest).
You'd end up with a set of data of the form Patient A-3.25, B-2.5, C-4.75, D-1.25, E-4.00 etc etc. Next you want to bin them somehow. This would be no different to binning a data set like weights. It would be legitimate to group people with weights over X as heavy, under Y as light and between X and Y as average. Similarly you can do that here to say under 2.5 is less attractive, 2.5-3.5 is average attractiveness and 3.5 and above is attractive. You probably could bin them in such a way to skew results and, if there binning was unusual (5-4 is attractive, 4-1.5 is average and 1.5 and below is unattractive) then that might suggest cheating the statistics. But their binning is such that there are 6 possible values in the attractive bracket (5.0, 4.75, 4.5, 4.25, 4.0 and 3.75), 5 in the average (3.5, 3.25, 3.0, 2.75 and 2.5) and 6 in the unattractive bracket (2.25, 2.0, 1.75, 1.5, 1.25 and 1.0). Those bins are all equally sized (along the attractiveness spectrum) and they correspond to what they were measuring (>3.5 means more scores of 4's and 5's, <2.5 means more 1's and 2's and in the middle more average scores) so it seems pretty legit.
Of course the best thing to do, and what we should all be pushing for, is open data. If the data were all available then this sort of argument could quite easily be cleared up by simply running the numbers with different bin criteria!
Otherwise, a nice post! One other thing you could point out is that, like many diseases, endometriosis has some heritable component so if the women were from the same families we might expect to see them also looking alike. Could only really rule this out by repeating study in other countries.
Thanks for the detailed explanation. I understand what you are saying except for one thing. Since it's a ranked scale, a subject rated 4 is not twice as attractive as a subject rated 2. In that respect, it is not like weight where we can say that someone who weighs 200 lbs is twice as heavy as someone who weighs 100 lbs.
Since the morphological parts of the body were similar, I suppose they were just looking at the face.
I suspect the unattractiveness of any abdominal surgical scars from surgery for endometriosis was not considered.
Joking aside,
It is pretty well documented that humans find attractiveness by symmetry in facial features. Just ask any plastic surgeon.
Certainly a picture and a computer program could make a determination and give some real numbers.
I can tell you that my best buddy and I disagree on what is attractive all the time, so I don't know if I would agree with just 4 docs.
More wasted research
RuggerMD, Thanks. I agree with you. I've been to Italy but not Milan. I can tell you that in Rome, Florence, Venice and elsewhere, there are more attractive women than 8 or 9 of every 100.
Did a bit more research, the type of scale that they used is known as a Likert scale (http://en.wikipedia.org/wiki/Likert_scale). It's somewhere between an ordinal and interval type data so using a mean is pseudo-ok. Using the mean is a common thing to do in sociological and psychological studies and some studies have shown that it quite closely approximates a truly continuous scale. (http://statisticscafe.blogspot.co.uk/2011/05/how-to-use-likert-scale-in-statistical.html). Looks like they may have been better off using a VAS (http://en.wikipedia.org/wiki/Visual_analogue_scale) but, for ease and speed, of processing a Likert type scale is not a bad system and using the mean is not too much of a problem.
If they wanted to do the study correctly, they would have used the RuggerMD scale.
Attractiveness is based on how many beers I need to drink to find them attractive.
Sophia, that's why I said statistics is the least exact math discipline. Ask enough statisticians and they will tell you what you want to hear.
RuggerMD, they also should have said when it was that they rated the attractiveness of the women. Reference: Mickey Gilley's country song "Don't the girls all get prettier at closin' time." Here's a link to the song http://www.dailymotion.com/video/x2ec89_mickey-gilley-don-t-the-girls-all-g_news#.UMM_aWfNlQs.
That was my theme song all through college and med school.
I know it very well
Post a Comment
Note: Only a member of this blog may post a comment.