Five days ago I reported on the unctuous Dr. Mehmet Oz and his new claim that various medical conditions and ailments were correlated with your astrological sign. I poo-pooed this since I couldn’t see any obvious connection between birth date and medical conditions. Well, I may have spoken a bit too fast, though, as I note below, there is still not a scintilla of evidence that the specific ailments Dr. Oz and his toadies ascribe to your “sign” are correlated with the zodiac.
Yet there may be a correlation between birth month (not astrological sign) and disease because of the environment in which you’re born, most notably the temperature and amount of sunlight. That, at least, is the tentative conclusion of the paper below, which I believe is in a reputable journal (J. American Medical Informatics Association is published by the Oxford University Press). Click on the screenshot to see the article (h/t to reader David for calling this to my attention):
In brief, the authors did a huge survey of nearly two million people—1,749,400, to be exact—born between 1900 and 2000; all records were from the New York-Presbyterian/Columbia University Medical Center. This means that all the subjects spent at least their infancy in New York, and experienced the climate of New York (though of course babies aren’t outside that often). The authors then correlated the birth month with conditions identified in the patients (median age 38 years) using a logistic regression and various methods to rule out “false positives” due to random variation. They assessed 1688 medical conditions from the records of all patients. To find correlations, at least 1000 patients had to show the condition.
Results:
- There were 55 diseases significantly associated with birth month, although the level of significance (“adjusted P”) was not huge for some of them, and, in general, the effects were small. In other words, there are highly significant associations for some conditions, but that reflects not a huge effect of birth month on disease but the very large sample size of people, which made it easier to detect small effects. Here’s a list of some of the diseases associated with birth month and the pattern over seasons; many are cardiovascular conditions.
Other diseases showing strong associations with birth month but not shown here include asthma, AHHD, reproductive output, and myopia.
- Cardiovascular conditions tend to peak in February-March, but aren’t all correlated (some are unimodal, others bimodal). They do tend to decrease over the year. Several studies from Scandinavia parallel this pattern. The authors postulate that this may be due to maternal infections in the winter that could damage their fetuses’ or children’s hearts, as well as to seasonal vitamin D deficiency.
Asthma peaks in the summer, and this has been shown in several studies (below; but note the y axis showing that the proportions of afflicted by month differ only slightly, though significantly. The authors posit that this might be because dust mites are highest when sunlight is at its peak, and the implied idea is that dust mites cause asthma that remains chronic. The months are offset, with Scandinavia showing a peak a few months later, corresponding to their their dust-mite season (summer) being a bit later. But in both New York and Denmark, asthma is correlated with peak sunshine in the countries where disease rates were assessed:
What about birth output, venereal disease, cancer, myopia, ADHD, and other conditions? It’s not clear. For ADHD, the authors posit that children born later in the year (who have a higher risk of ADHD) might simply have their ADHD detected more easily as children born up to December 31 go to school with all the kids born in that same year. As the authors say, “this occurs because more immature children (i.e., younger in age) face higher demands earl on in their school years making them more susceptible to ADHD diagnosis.” Well, that’s one theory, but it doesn’t sound that convincing. As for myopia, cancer, and other stuff, who the hell knows?
What we have, then, are suggestions, corroborated in part by studies in Europe, of correlations between 55 maladies in children and the month in which they were born. And there may well be something to this, though for many of the disease the causes (if this is corroborated elsewhere beyond New York City) are not obvious.
Does this vindicate Dr. Oz, then? Not on your life. Remember, zodiac signs don’t correspond to birth months, but to 2/3 of one month and 1/3 of another (see below). But given the peaks, there would almost certainly be significant correlations between zodiac signs and disease anyway, at least for the New York sample, and probably for the Danish samples for heart disease. The signs:
There are two other issues, though. First, the relative risks based on birth month are very small: so small that you don’t want to tell someone that they’re a lot more susceptible (or even appreciably more susceptible) to a disease because they are, say, a Pisces rather than a Leo.
Second, these samples are from one place, New York City. These associations need to be replicated in other places. This is particularly important south of the Equator, where the sunlight and temperature regimes are reversed, and so the associations should be much different by month and by Zodiac signs. And on the equator, there should be no associations, at least for those diseases associated with sunlight exposure.
Finally, the maladies that Dr. Oz says are associated with Zodiac signs aren’t any of the ones found in this study. The list of Dr. Oz’s astrology-sign diseases is here. Here are the 12 maladies supposedly associated with your astrology sign:
Migraines
Stiff neck
Pain or stiffness in the hands
Water retention, low energy, cysts
Upper back issues
Digestive or gastrointestinal issues
Lower back pain
Inability to let go of past relationships and experiences
Sciatica and hip problems
Buckling of knees or knee weakness
Ankle tension
Foot pain
If Dr. Oz wants to tell people that their “signs” predispose them to these conditions, let him show data supporting it. As for now, the maladies associated with birth month (and probably with “sign”) aren’t those touted by Dr. Oz and his flacks.
sub
I might add that there are likely to be important social effects here too. For example children born either side of school entry cutoffs may have better or worse educational attainment and therefore lifetime experience of health.
That’s an exceedingly tenuous assumption.
Tenuous it may be (it’s an SEP), but it this is a claim that is made a lot. It being an SEP, I don’t know if being older than average for your school cohort is an advantage or a disadvantage ; I wouldn’t be at all surprised if there were studies pointing in both directions, being cited by people with (genetic) bets going in either direction.
Sub
Sub
Without going into detail, those propensities closely match my own health issues and lack thereof.
Yo already raised the issue of the Southern Hemisphere, where seasons are reversed re the North. Also, if this is an environmental effect, could it be operating on the individual, not (only) on the mother? (Say, born & raised in NY vs born in NY and raised from age 1 in Australia?)
BTW if the month/season effect holds for Southern Hemisphere births, then it’s a big “poof!” For any astrological arguments…
I would have thought that some Southern Hemisphere data might be derived from the Dunedin Multidisciplinary Health & Development Study which has now been following a group of 1000odd children born in 1972-73 for several decades.
So, according to Dr. Oz (and Vitruvian Man?) as a piscine, I will have foot problems (maybe frogfish, family Antennariidae, have these issues too?) for reasons that escape me and all logical thought. The study, however, suggests a possible link to a lung condition I’ve never heard of. I can at least see how there might be a slight causal link to lung conditions in babies born in the winter, stuck inside during their first few months of life, covered with a blanket, as parents do, whenever going outside… however, I’m willing to bet good money that being born into a family of smokers, in the 70’s when damn near everyone smoked, everywhere and all the time, would have a great deal more impact on my lung health than pretty much anything else.
My feet are fine, by the way.
I just browsed through it and it looks quite well done (the permutation analysis of literature reported effects gives some relief to the idea that much of the medical literature is not reproducible) and Dr PPCe is correct – the effects are truly tiny – one could not have noticed these effects without such a large data set.
But one sentence struck me as funny in a “boy is that an understatement” kind of way. In their limitations of the study section they state (emphasis added);
Yep, health records from hospitals do tend to have a lot of sick people in them. Oh, and the Grand Canyon is a hole in the ground.
Christ, now ya got me fingering my carotid lookin’ for coronary issues. Think I’ll just go back to readin’ Jeane Dixon.
So, all in all, it seems it is best to be born in the fall.
Except you might vomit a lot.
“Yet there may be a correlation between birth month (not astrological sign) and disease because of the environment in which you’re born, most notably the temperature and amount of sunlight.”
Not having studied the literature, I imagine there are any number of other factors such as: genetic predispositions; age, diet and health of the parents; weather conditions in addition to temperature and sunlight; environmental factors such as air quality, dust (and dust mites), water contamination and chemicals in the soil, etc.
There is a genetic predisposition to asthma in my family. I was born with asthma in southwestern Missouri in January to older parents. My asthma attacks were triggered by weather. My parents said they could tell when a storm was coming by how I breathed. They left Missouri for the west coast when I was
2 or 3 attempting to keep me alive. That, or other factors worked, because I “outgrew” the asthma by age 7 in a better climate and am still here.
Both of my grandsons had/have asthma. There was a genetic predisposition on both sides of the family. One had a mild case and outgrew it. One had a severe case and, although he is still alive, he almost died from this disease
several times (was without oxygen long enough in one of those events to suffer some brain damage.) An additional factor in his case is that as a baby and toddler he lived near a military base in Japan where the air quality was terrible due to trash/garbage being burned (remember those huge old upside down cone shaped structures there used to be?)
I’m all for scientific and rational investigations into the causes of human illness. The pseudoscience of Dr. Oz deserves to go into that trash burner I talked about
above.
It is always interesting to know about these correlations, but the emphasis on p-values instead of effect size (with error bars) is a common malady of medical research.
The asthma graph, which does provide information about effect size (though without error bars) has been rescaled to show only the variable part. That’s perfectly reasonable for analyzing possible causal factors. But we shouldn’t lose sight of the fact that the proportion of people affected per month varies by less than 3% between worst and best months.
I was taking the proportion on the Y axis as percentage, with the highest number being ~ 0.0620 or 6.2%. The then low value would be 0.0595, or 5.95%. Thus a difference of around 0.25%….as always, I may be wrong. But seems that’s what a massive sample size will give you, though to me an increased risk of a fraction of a percent = meh.
Glad you directed me to that graphic, as I’d skipped it. The authors committed one of the cardinal sins of data presentation: never use two different y-axes on the same graph! They actually did this several times and the axes’ windows and scales have (of course) been chosen to make the trends they found match up with the trends found in the literature.
Other than that though, I don’t have much to complain about regarding their analysis, with the understanding that this type of work is purely exploratory. I like Jerry’s suggestions about seeing if their flagged items replicate in other parts of the world where the seasons are reversed or the amount of sunlight differs greatly. It would have been great if they did this kind of external validation rather than the largely uninformative internal validation that they reported on instead.
Yes, when I’ve done that sort of analysis I’ve reported parameter values and standard errors. Error bars and p-values then follow almost automatically, if you want them.
In this case you’d probably want the relative risks, with error bars, for each month. That’s a lot of output, though.
The p-values (corrected for multiple comparisons) are OK as a summary of whether a variable has an effect, though nowadays I would expect that sort of analysis to be done by comparing model deviances, probably using Akaike or similar.
Yes, the p-values, if properly corrected for multiple comparisons, do tell us how probable the observed results would be if there were no effect. But in these kinds of studies we never really care about WHETHER there was en effect. What we care about is HOW BIG is the effect. That’s usually the question that matters.
A low p-value can always be produced, in almost any biological experiment, if sample size is large enough. If the authors had a larger sample, they would have gotten many more low p-values. It is exceedingly unlikely that there would be exactly zero effect, so if the sample size were huge, almost all of the p-values could be made as significant as one wanted, even 0.00001, even if the actual effect sizes were a tiny fraction of a percent. It makes no sense to use p-values here.
Indeed.
Help, help, help, I’ma wishy washy Pisces, What am I goin’ to do?
In our small family this does not match up. My wife’s cardio problem is specifically mitral valve and she was born in Sept. My cardio problem is from birth defect so I assume it would not be part of the study at all.
I’m glad you are pointing out the self-serving lies of that dodgy huckster Oz. It’s a shame he gets air time to sell his BS.
Just to point out the obvious- and before we all get excited about birth month (magnet for headlines)- there’s the whole 280 days (average, from first day of last menstual period) of gestation before that – going back through all those prior months/seasons – that is of enormous significance on health of baby.
I.e. I can see getting excited about explaining some disease based on a certain birth month, without explaining why all the other months are NOT relevant.
Just sayin’
Yes, mechanisms would be nice.
“Yet there may be a correlation between birth month (not astrological sign) and disease because of the environment in which you’re born, most notably the temperature and amount of sunlight.”
I had read something similar a while back, but the speculation in what I read was that the causes had to do with the climatic conditions at the time of conception rather than at the time of birth.
L
I like the “non venomous insect bite” one.
That one makes me suspicious. How could your risk of being bitten by a non venomous insect be related to the month in which you were born?
I know. I think it is just something they observed so they felt they should put it down but it just makes the whole thing look suspicious and for some reason it kept making me laugh like the infant’s mother was Achilles’ mother and she put that kid down while she dipped Achilles in the river and the other kid got bitten by a gnat or something. Yes, just roll with it, it’s how my mind works.
It’s not the risk of being bitten, it’s the risk of having a reaction to a bite. Usually a type 1 immune response; hives, rashes, hypersensitivity, that kind of thing. Any immune response to an insect bite that is not related to venom.
Shhhhh, you’re ruining my hilarious story in front of Jeremy!
D’oh!
Maybe non-venemous insects are only alive or active during certain months.
According to NASA the 12 zodiac signs were a bodge to fit the calendar and there are actually 13 constellations (or more). Plus as https://spaceplace.nasa.gov/starfinder2/en/
says “Now, 3,000 years later, the sky has shifted because Earth’s axis (North Pole) doesn’t point in quite the same direction.”
So the ‘link’ between zodiac sign and disease is not as strong as some would claim.
Part of the environment of an infant of any species with parental care is the bacterial load that comes into contact with the infant from the parent(s). So, when it’s flu’ season, then the infants are also at risk of flu’, no matter how well swaddled they are (literally and figuratively).
“Diseases are correlated with birth months”
That is entirely credible. I assume ‘diseases’ there means chronic afflictions and their after-effects. Insofar as the time of year can influence the surrounding environment, up to and after the birth, it would be surprising if it had no effect whatever.
(One would similarly expect the location – country and region of the world – to have such an effect).
Of course this has nothing whatsoever to do with the alignment of the planets, the phases of the moon, or the age of your grandmother’s cat.
cr
Or with astrological ‘signs’, of course.
I remember reading a long time ago that people born during high pollen seasons are more prone to allergies. I was born during ragweed season and I’m certainly quite allergic!
I was born in December – low pollen season – and I have few allergies. Proof positive that people writing a long time ago were right.
1. Oz types do not show data, but only suggest correlations.
2. A lot of people are not aware of the need for data and are strongly influenced by suggested correlations.
3. A lot of money can be made by taking advantage of 1 and 2.
55 out of 1688 possible medical conditions they searched for in the data, eh?
What’s 5% of 1688?
It’s 84, which is more than 55. However, I thought this thing where you search for loads of correlations and cherry pick the few that appear to be significant was a well known statistical trap and that this study must be correcting for that.
Perhaps that’s a naive view.
Was a Bonferroni correction made?
Isn’t this just the well know phenomena that when you look for correlations in a large enough set of variables you will always find some? https://www.di.ens.fr/users/longo/files/BigData-Calude-LongoAug21.pdf
No, according to some that could never happen! See here for conclusive evidence that it is impossible:
http://www.tylervigen.com/spurious-correlations
sub