Thursday, August 22, 2013

Chinese medicine is carcinogenic

The following insights look particularly ironic and make me think about the prescientific way of analyzing Nature.

One of the most omnipresent chemical compounds in Chinese herbal medicine have been aristolochic acids (AA) which are found in wild ginger, pipevines, and other long tubular flowers. AA I is almost omnipresent in Aristolochia species.

This European Asarum edition of wild ginger almost looks like a tumor itself.

The three hexagons plus a pentagon in the chemical formula don't look particularly healthy to me. ;-) But just to be sure, not all polycyclic compounds such as AAs are aromatic; aromatic compounds are only those whose ground state ket vector is a linear superposition of the formulae with several distributions of the single and double C-C/C=C bonds so that it's more accurate to indicate the alternating bounds by a circle inside the polygons. This allows the molecule to stregthen the binding relatively to the non-aromatic case and this also tends to make the compounds unhealthy. It's sort of remarkable that this class of organic compounds with an interesting two-state quantum mechanical effect were already named by August Wilhelm Hofmann in 1855. Some people are able to get very far in quantum mechanics without even knowing the symbol \(\ket\psi\). ;-)

For quite some time, the compound has been known to be carcinogenic – I didn't know about it – and probably more efficient in causing the disease than the UV radiation and smoking. As such, it's been banned in many Western countries. The tumors appear in the urinary tract, kidneys, and liver. The compound is probably behind the Balkan endemic nephopathy (nephro- means "related to kidneys").

A new paper
Genome-Wide Mutational Signatures of Aristolochic Acid and Its Application as a Screening Tool (Science)

Scientists Find Link Between Aristolochic Acid And Liver Cancer (Asian Scientist, review)
by Song Ling Poon and dozens of Taiwanese and Singapore-based co-authors was apparently able to identify some genetic fingerprints of the mutations caused by the compound which may turn out to be very useful.

The Chinese shamans and their consumers, much like many other pre-scientific cultures, must have been able to identify certain, sufficiently strong and speedy, influences of the plants. Without science as we know it, many cultures – and even animals that lived long before humans – must have been able to identify poisonous plants as well as those that are helping you quickly enough (at most days?) in certain conditions.

But I would say that whenever the impact was delayed by more than weeks, they must have been lost. Their memory wasn't good enough. I suspect that they also believed that no effect that took longer than two days (needed for the food absorbed through the mouth to get out of your body via another hole) could ever be relevant. They obviously didn't evaluate the consumers of their ancient drugs and other remedies statistically, and so on.

So I guess that many such herbal medicines may have determined – at least approximately – which spices and other plants have antimicrobial properties because this effect may be fast and visible. On the other hand, it's likely that carcinogens were completely invisible to the primitive methods of the ancient Chinese folks and other sufficiently primitive tribes.

In some cases, the plants may have been helpful by accident. In most others, they neither hurt you nor help you. And there are apparently also cases in which you may be recommended a plant that is downright harmful – and very harmful, in fact.

The latter situation is particularly unfortunate and you may blame it on the low statistical significance or speed of the impact of the plant. Science has learned to identify all these things systematically. More importantly, the ancient believers viewed a story told by another human (an older relative etc.) to be at least equally important as your direct empirical experience. Science has learned to totally ignore the stories.

So while the ancient people wouldn't notice (with the reliably correct sign) a plant unless it "immediately and visibly" helps you or hurts you, science got good in identifying various "signals" that are hidden in the noise. We could say that the modern medicine research got "too obsessed" with the asymmetries. This obsession could be described as the "opposite bug" than what the Chinese herbal medicine has suffered from.

What I mean is that we can see lots of medicine research papers in which some 2-sigma or so signal is spotted and despite this very low confidence level which makes it extremely likely that the apparent signal was just a fluctuation, such papers often penetrate to the general "scientifically literate" culture or even to further research papers.

For example, my reading of the literature indicates that there exists no really solid evidence (at least 5 sigma, for example) that the second-hand smoking is harmful (in particular carcinogenic). If the original 2-sigma-like papers claiming that the passive smoking should be harmful were right, we would already be surrounded by totally convincing papers with much higher confidence levels because the research and surveys have grown by a huge factor from the first "hints" that the passive smoking should have been harmful. Five-sigma papers should be almost trivial to encounter these days. But they seem to be absent, reinforcing the opinion that the original claims were bogus and the health influence, if one exists, is far weaker and more subtle than originally thought.

Correct me if I have missed some important developments.

Of course, the second-hand smoking is sort of a "politically interesting" topic because everyone is either a smoker or a non-smoker and cares about the social interactions with the other group and the groups' "human rights". But there are arguably many "politically neutral" wisdoms that don't irritate anyone but that are still incorrect even though many people still believe it.

It could be interesting if someone were able and willing to cut through the vast medical literature and simply pick the results that have only been achieved with robust methods and at a confidence level that is appropriate in hard sciences. Otherwise while we've learned many things that were misunderstood by the ancient Chinese healers, we may still be believing lots of superstitions that will only be corrected by a more intelligent mankind in the future.


  1. Some years ago I attended a public health seminar, and the speaker casually mentioned that people who smoked a pack a day for most of their lives had a 15% chance of dying of lung cancer.

    Deaths from other smoking-caused diseases were not mentioned, and the total smoking-related death rate must be somewhat higher.

    However, the lung cancer death rate among smokers is striking low considering all the propaganda. And it strongly suggests that the second hand smoke claims are false.

    Nowadays all US agencies are heavily politicized and routinely produce nonsensical propaganda. No Federal report can be believed.

    Lysenko actually won the debate.,

  2. A very fair politically uncomfortable perspective, Lubos.
    You keep being my favorite 'scientific philosopher'!

  3. Secondhand smoking? Ha. This guy has moved beyond secondhand smoking to thirdhand smoking!


    p.81 and 82 summarize the research about the association between second-hand smoking and coronary heart disease. I would just add that these studies are not strictly medicine, it is rather epidemiology

    I believe that in similar studies it is very difficult to collect reliable data, i.e. you need a cohort study and follow the people over decades which is very demanding. How many people would you need to get a 5-sigma confidence interval? If the following formula is to be believed, you need a high sample size becase the SNR is very low. Unfortunately it is much more difficult to collect data about people than count collision events on a detector

  5. Dear Mephisto, as I have tied to teach you about 10 times but so far you failed, you only need (5/2) = 6.25 times greater sample to extend the 2-sigma confidence level to 5-sigma confidence level. Everyone claiming that getting 5 sigma is impossible in practice is either a crackpot unfamiliar with basic statistics or a fraudster.

  6. I am not that good in statistics, but could you point me somewhere where the sample size calculation in relation to increase of sigma is explained.

    I found

    the following

    On page two, it is explained that the power depends on the statistical test you are doing, chosen significance level, sample size, true difference between means (delta) and the two populations SDs.

    Let us say we choose a two sample t-test for equal size and equal variance groups and let us say that the true difference between means is very small (low signal). We increase the chosen significance of the t score (t-distribution) from 2 sigma to 5 sigma.

    I guess the formula from the chaotic reference above is explained in detail on page 5. (z score is a converted t score). The increase is power also depends on the signal (delta), if it is low you need a large sample.

    PS: I haven't read your post about the 5-sigma in particle physics yet, but I am going to do so now. I am actually curious what kind of statistical tests the physicists use to test test hypotheses.

  7. I blogged about the use of the 2-sigma 'standard' in medicine recently.
    The post includes a downloadable but rather scruffy spreadsheet that
    shows how sample sizes relate to sigma levels.

    It applies to the Normal distribution but that is far and away the most
    common because of the central limit theorem. I'm not a professional
    statistician but hope it is correct and you can see all the maths and check it for yourself:

  8. Try the Khan Academy videos. They are quite good for people at your level and they can be replayed until they are understood.

  9. Dear Lubos, do you think exposure to low levels of radiation, like at Fukushima, might be in the same category?

  10. As an organic chemist by training with undergraduate lab work in a genetics lab, though I would have no apprehensiveness about whipping up pounds of such cute little polycyclics, including actual work I did extending this sort of three benzine ring boomerang all the way around to six rings with the last two not even being linked into a flat overall loop but actually physically overlapping to form helicines, when I look at the structure of such a small segment and consider it biochemically, I cringe indeed since it has the exact shape to slide right in (intercalate) between two otherwise happy base pairs of DNA and lock in by charge transfer physical stimulation tightly enough to fool finely tuned enzymes into making mistakes as they try to read, reproduce and repair it.

    As I informally recall, a polycyclic carcinogen is also formed when red meat is blackened due to a specific unstable amino acid that oxidizes, but at least one study suggested that merely rubbing the contents of a vitamin E capsule on the surface prior to cooking was enough to avoid it, and this may be an important finding since uptight women who prefer well cooked meat are said to have a 5X higher rate of breast cancer.

  11. It is no wonder that there is so much junk science around!

    A part of the problem is the publish or perish culture at universities. You need to publish to finish your Ph.D., to get a Postdoc job, to get a tenure. I doesn't matter what you publish or what is the quality of it. I see it all around. Young researchers are forced to write papers, so they are constantly searching for something that they could publish (publishable results are scarce). The result is a lot of papers of dubious quality - junk science. It is science not done for love of truth, but science done to get a job and inflate your ego.

    But how do you want to change it? You would need to pass some law that would force the research in medicine to use 5 sigma. All studies would need to be larger and the consequence for you would be for example and increase in the price of medicaments.

  12. It seems you are a smoker and you need to find some "rational" excuses for youreself to continue smoking. It is fine with me. It is your health and your problem. Google smoking and life expectancy. Smoking shortens life by about 10 years. If you die sooner, the society needs to pay less for your pension and the treatment of other diseases (if you live in Europe) and you pay the state a lot in taxes (cigaretes are highly taxed). If you live in the US, nobody cares because you pay these things for yourself. Your smoking is economically beneficial to the society because in old age you wont be a burden. Lung cancer is just the most well-known health benefit of smoking. COPD and atherosclerosis with its complications are other benefits.

  13. The record holder for longevity was a French woman Jeanne Calment, who died in 1997 aged 122 years and 164 days. She gave up smoking at 117 ;-)

  14. That's true Lucretius ! Also, just for the funny story, she sold her house in return for a life annuity to a couple when she was 90. The couple died before her which means that they paid twice the price of Jeanne's house!!

  15. Mephisto, I'm beginning to understand why it is always the same universities who reach the top of rankings: it's all about the number of publications!

  16. That's a good example of misjudging "life expectancy".

  17. Why should second-hand smoke be less carcinogenic than first-hand smoke (or rather first-mouth smoke)? Are there any indications that the carcinogenic substances involved have a very short decay time or something?

    It's obviously a lot harder to collect the essential data in the case of second-hand smoking, namely concentration of carcinogenic substances in the intake of breath and the length of exposure to those substances, than in the case of first-hand smoking, where you have the ersatz variable of cigarettes smoked per day, which should at least correlate with those factors.

    The dangers of second-hand smoking are very dependent on individual circumstances. If non-smokers sit in a hermetically sealed room with 1 smoker per square meter for hours the exposure of smokers and non smokers to carcinogenic substances won't differ by too much, while a whiff of smoke outside of buildings and amidst fresh air won't hurt anyone.

    5-sigma results are good and all, but sometimes the world is too complex to get data of good quality and that's when we must use inference and logic to assess the problem. Studies which do that are needed in the case of second-hand smoking in my opinion, and not studies which try to squeeze 5 sigma out of bad data.

  18. Here's a quick way of getting a feel for the thing by taking a simple example (and a little arm waving so as not to let a precision's concerns fog things up).

    Prelim: if the following simple points/terms are not clear then google them.
    1. A lot of actual variables have a distribution similar to the normal (Gaussian).
    2. The normal distribution is a very good approximation to the binomial distribution, especially for large n. (The further p gets away from 0.5 the larger n needs to be for the approximation to work — but this is just 'fog'.)
    3. In other words the binomial distribution can model a lot of actual variables. [Lots of arm waving on that one. :) ]
    4. The binomial distribution has a mean of np and a variance of np(1-p), hence a standard deviation of √[np(1-p)].

    Unreal example:
    You enter a coin-tossing competition but want to make sure the coin is fair, which you define as having a probability of it coming up heads of 0.5±0.01. Anything outside that band and the game's off. So you test it by tossing it a number of times, say n, and look at the resulting proportion of heads = (heads count)/n. You assume the coin is fair, i.e. that p = 0.5, and want to see if the test supports this. Also you want to be sure at the k-sigma level, i.e. you want |-p| < 0.01 = k√[np(1-p)]/n.

    This gives n = k²p(1-p)/(0.01)² [= 2500k²]. So, as Luboš says, going from 2-sigma to 5-sigma increases the sample size by (5/2)² = 6.25, with the squaring aspect here coming from the properties of the standard deviation of the binomial distribution — loosely speaking.

  19. Dear Barbara, I personally fully agree with your suggestion that the passive smokers qualitatively inhale similar compounds as the active smokers.

    It's really about the concentration, a quantitative difference. A passive smoker exposed to all the smoke from a particular active smoker simply gets less of the harmful stuff at least by an order of magnitude. Moreover, a smoker, as I imagine him, is really trying to avoid smoke so he will spend much less time in a smoked through environment than his colleague smoker. I would even dare to say that a non-smoker who doesn't avoid the smoke in this sense may be partially classified as a partial active smoker simply because it must be sort of pleasant to him.

    Because "full-steam" smoking shortens life expectancy at most by 10 years, "full-steam" passive smoking shortens it at most by 1 year and I guess that at that level, it become unmeasurable. For most passive smokers, the exposure is so low that it's equivalent to at most months of life expectancy and the change to the overall exposure you can make by changing the laws are at least weeks.

    Moreover, because one doesn't have direct data, one may conjecture all types of things such as the minimal threshold beneath which the body is able to eliminate the harm completely or even hormesis that says that low toxin levels are actually improving the health. I don't see a good enough theoretical way to eliminate any of these things. For this reason, it seems inappropriate to me to pretend that we really know how the low exposures etc. affect human health. And because it's at most weeks of life expectancy that the laws may add, assuming some not-really-established assumptions, it seems wrong to me to restrict the freedom of the smokers in a really brutal way - e.g. banning their smoking in all restaurants in a country or something like that.

    One must understand that the numbers of smokers and non-smokers are comparable. A smoker is OK to shorten his life expectancy by up to 10 years which really means that the smoking is this valuable to him. He shortens the life of a typical passive smoker, assuming typical social interactions in a country that normally tolerates smoking at places where smoking was common, by weeks relatively to the "banned public smoking" scenario.

    So one is really facing the following dilemma: is it better to allow the smokers to shorten the life of the non-smokers who sometimes become passive smokers by weeks; or is it better for non-smokers to steal almost 10 years-of-joy equivalent of life expectancy from the smokers' ilfe? In this way, I really converted both "harms" to the same units and it's clear that the smokers' freedom wins.

  20. That Georg E Matt looks remarkably like Mikey The Hokey Schtick Mann!

    The similarity suggests an hypothesis: bollocks hawking is genetic. I leave it as an exercise for others to dig out the full 5-sigma evidence of course. :)

  21. Medicine is a semi-science at best.
    a 5 sigma confidence level is impractical for things like drug research since if you have more people in your trial you increase the probability of something going wrong with someone (whether from your drug or other natural factors) and then having its safety questioned. (so maybe they do need to improve) There is also the issue of cost and availability of candidates.

    As far as protecting people from second hand smoke there is plenty of 5 sigma evidence to show the presence of mutagenic compounds and therefor it must be banned since smoking does not have any real benefits. The cost/benefit just doesn't add up even if you can't find a signal against the background noise. You can argue for its banning on a chain of evidence. It may not be scientifically valid but it is legally valid! The only other mutagen that is in common use is nitrite used in meat preservation. This is allowed because it prevents botulism poisoning!

  22. You said that second hand smoking is most probably harmless. This is plain wrong . I'ma medical doctor . Also there are many things here that are wrong .

  23. Second hand Smoking is even more dangerous in some circumstances !