Lunacy

sesquiterpene said:

You are quite right - I am approaching this from a scientifiic viewpoint (retired chemist) and reading Science-Based Medicine blog, and attributing qualities to Evidenced-based Medicine which it doesn't possess - I apologise.


What biases would a prospective trial have that the current trial doesn't? And just as importantly, what biases does the current trial have that a prospective one wouldn't?



I think that most of the disagreement here is due to the difference between EBM and SBM. For instance, one of main criticism coming from SBM is that EBM suffers from an undue reliance on clinical studies, and (this criticism come from several sources, I believe) a slavish devotion to p<.05.

I can see the latter in your repeated descriptions of statistically significant findings as 'definitive' and 'conclusive', and lately 'pass[ing] Epidemiological muster'.

This simply isn't true. There are many different ways that even statistically significant data can be misleading or even flat out wrong. As Kimball Atwood remarked on the SBM blog:


There a several scenarios in which Wehr's statistically significant data might be false. He provides analyses for individual patients but doesn't pool the data from those 43 years. Is there any statistical significance left? You can't tell by reading the article.

You mention that the length of tidal periods are unusual in biology, but a 14 day period is ubiquitous in human society and could easily make their way into the data. Perhaps patients 1-8 just really don't like Mondays, and patients 9-17 have enormous antipathy to Nurse Ratched who fills in on alternate weekends. 14-day periods differ slightly from the tidal periods {edit: and they diverge over time}, but Wehr doesn't try see if the data is good enough to differentiate between the two. You can't tell by reading the article. In fact, Wehr doesn't discuss possible confounders or sources of error at all. ( that alone probably invalidates the 'pass[ing] Epidemiological muster' label ).

My initial comment was spurred by your [mis]use of of Statistically significant as 'definitive' and 'conclusive', and yes that is probably due to the differences between EBM and a more scientific approach. I'll stick with science.

Click to expand...

Luckily, I am a doctor and not a scientist; and this is Medicine we are dealing with. Medicine has traditionally been termed an Art, not a Science; and with the almost wholesale adoption of EBM which does not follow Scientific Method, those that wish to argue it a Science have an even weaker case (depending of course how you define Science, a quite protean term).

SBM is a bunch of folks in the outer darkness, with much wailing and gnashing of teeth, while EBM is literally taught in Medical schools as the standard. For SBM is really incoherent - it adopts EBM principles when it wants to, but when the data doesn't fit their induction-derived hypotheses, they yell foul. Essentially, you have your cake and eat it too, as whenever you wish, you can fall back on taking 'expert opinion' in lieu of evidence or in place of it. No wonder it has failed to make headway.

Some of its cricitisms aren't without validity, but it has failed to present a system to replace EBM. Essentially, the days of consensus conferences are behind us, and SBM has no coherent system to suggest Best Practice or Clinical Decision Making. They only make a stink about p values when it suits them, as the CHEST trial makes clear: SBM advocates did not raise a whimper against it, until EBM tried to digest it, and piece by piece its shortcomings were laid bare.

In Medicine we need to be sure about things, for moral and ethical reasons, which is why EBM has become so dominant. We can't go around just guessing, no matter how informed those guesses may be, as lives are at stake, and people can be highly litigious. In Chemistry we can mess around with hypotheses that our data doesn't support until we are satisfied it isn't so, or experiment with the unknown, and sometimes such diligence would reward with brilliant discoveries - Ethically we can't risk human life in the same manner, which is why Inductive reasoning in Medicine is largely frowned upon. So even if the data doesn't fit our preconceived notions; our pet theories, our framework of testing; that data if properly collected and valid, should be given paramountcy. Of course EBM is tempered by Clinical Medicine, but it should not be by a priori decisions and just opinion. In this case, we have essentially 43 years of data, with a 13.7 day cycle (which would not correlate well over 43 years with 14 day cycles, though one could conjecture that is where those cycles arose as 'half moons'). The data is clear, even if people don't want to believe it.

So I'll stick with Medicine - proven, supported, Evidence-based and limiting bias. The tyranny of Consensus Conferences driven by those with the strongest opinion, highest credentials, and support base, has had its day, thank you very much. Such quackery has no place in a discipline that requires as much precision and confidence as possible, seeing that human biology is so variable and our therapeutic indices so narrow. I'll trust my health and that of my patients to Evidence, not conjecture (even if dressed up as Hypothesis).

sesquiterpene said:

Wait, what? What ethical and moral reasons do you need to "sure about things"? Despite your assertion about how you are not doing science, but some art of medicine, the use of p values necessarily includes uncertainty - and mostly , a quite arbitrary level of uncertainty. Do you think being sure is useful for projecting a your personal level of confidence, for convincing your patients to undergo a treatment, or merely to protect yourself from lawsuits? Please choose some other option.

Click to expand...

Let me give you an example: Cricoid Pressure.

This is a technique to try and limit passive regurgitation during a rapid sequence induction of Anaesthesia, by pressing on the cricoid cartilage and thus hopefully obstructing the Oesophagus. Now, the studies upon which this practice is based were done on Obstetric patients back in the 60s, which showed a marked reduction of Aspiration. Subsequently it was universally adopted and extrapolated into other patient groups that have a similar risk for aspiration - due to raised intra-abdominal pressure or delayed gastric emptying, say. However, it has a downside: It potentially distorts the airway, making intubation more difficult, the anaesthetic assistant occupies one of their hands to give it, with active wretching there is marked risk for oesophageal tears, etc.

To top this off, we can never repeat these studies to test them - those pesky Moral and Ethical reasons. Once it was shown beneficial, we cannot morally not provide it to at risk patients, so no study with controls not given cricoid pressure can ethically be repeated in Obstetric GAs. From the principle of Beneficence, we cannot withold it. Even if patients agreed, it would be Medical personnel vs layman, so our reassurances or such, would be from a position of implicit trust and thus not 'freely chosen', but an assumed safety by the patient.

So these initial studies are it. We have to provide care that has been shown beneficial. We cannot withhold it, we cannot falsify these studies via repeating them. By doing so, we would not have those patients' interest at heart, and are thus in contravention of medical ethical principles of non-maleficence and Beneficence.

This holds true for anything that shows strong benefit (or conversely strong negative effect) - such as giving anti-hypertensives, providing emergency blood, discontinueing drugs that give dangerous effects, etc. We cannot repeat such studies, so Repeatability becomes an issue in Scientific Method, and why it cannot be properly applied. So we must be sure of ourselves in these things.

That is where EBM comes in so brilliantly, as it treats the data as non-falsifiable information. By utilising p values and the like, we can minimise the practices that are thus elevated to sacrosant Best Practice - the type that would have you scrapped from the medical rolls, or taken to court for malpractice, if you did not do them; and by doing them, potentially reduces morbidity and mortality. Everyone expects the best possible care for their own health, and barring lapses in Medical ethics like the Holocaust doctors, we are stuck here.

Of course, we can never be hundred percent sure. That is one of those epistemologic pitfalls of life. You can always doubt a result, back down to the axiomatic if need be. Mathematics has opted for those Fisherian variables of 5% statistical significance - as you said, arbitrarily (to facilitate calculations with a slide-rule if I recall), but all the safeguards and plotting and graphing that Math developed, was built around this mark. So do you have a better alternative? Not that I am aware of, nor that SBM supporters have preposed - So with p values and the ilk, we are just dealing with whining that the world isn't perfect, akin to why can't I eat as much carbs as I want and not get fat.

So it is theoretically imperfect, but by experience, the sensitivity is more than adequate. Retrospectively we can confirm these values, by seeing what had happened in practice (so again, not denigrating lower tier evidence, which can help to be confirmatory). At times, we adopt things even with imperfect p values, such as the widespread use of Anti-depressants, so when something does pass muster in this regard, we of course take notice.

sesquiterpene said:

In the invidious comparisons category for $200, I'd like to suggest that this resembles creationist misunderstandings of science. There are no proofs in science, and that really extends to medicine too.

Click to expand...

Really? These are Mathematically proven values by statistical analysis. Of course, you can always argue Math hasn't been proven itself, which is true seeing that Russel failed to prove its axiomatic grounding and the Incompleteness theorum, but that is really just being abtuse.

So according to our best indicator for 'proof', cold hard numbers, these are valid findings. Again EBM is deductive - meaning its findings are necessarily valid and sound based on the supplied data; which differs from Induction systems where we cannot affirm it necessarily valid. Ever heard of Hume's Swans?

So yes, proven by the best means currently available to humanity, backed up by extensive statistical analysis. While not 'proven' in an absolute epistemologic sense, frankly nothing is; and nor did I ever make that claim. You are just erecting Strawmen here once more.

True, these deductive values only apply internally, so we must show external validity of a study in a population, but again we must show this mathematically usually, by investigating the variables. This is where Bias becomes so important. Recently they did a mock-serious study of testing Parachutes when jumping from planes, to help illustrate this concept; where the study found parachutes offered no benefit, but could of course only be conducted on a stationary plane on the ground, so external validity to flying planes is lacking. With extensively studied things like Cardiovascular risk, we can usually achieve external validity though. In this OP's findings, a prospective study would not be possible for instance, as any enrolled subjects would need to have their mental illness treated ethically, thus biasing results and ruining any validity it might have had.

Frankly, I don't know what you are trying to achieve here. You clearly are out of your depth where Medical Studies are concerned, as you even try and conflate EBM with SBM or 'science' (whatever you mean by that) in general. You keep casting aspersions on p-values, but what would you have us replace it with? Nor have you shown in what way the q-value in this study was flawed, as the study itself addressed the low number of cases it investigated, and supplies justification why it remains valid regardless.

So, um, what are you trying to achieve? What is your point? For at the moment we have Evidence, EBM, Math, on the one side; and solely your opinion on the other. As I said before, I feel akin to arguing with an Antivaxxer - which is not meant as an insult, but a juxtaposition to maybe help you see the silliness thereof. In my opinion, Evidence of something that can be supported by ancillary means, always trumps mere anecdote and personal opinions on it. You are free to disagree with the study's findings, but you are really not talking from any position of strength.

essentialsaltes said:

No, it hasn't. That 5% is an arbitrary value for some fields, not determined or chosen by mathematicians.

Click to expand...

Chosen by the mathematician Ronald Fisher in the 1920s to indicate statistical significance. Look it up.

essentialsaltes said:

Yes, barring mistakes, it is mathematically proven that there is a 5% chance the results are entirely mythical.

Click to expand...

This is untrue. The p value is that if the null hypothesis is true, the chance this would be a different result. It is about probability such results could be coincidence, not a 5% chance the results would be false.

Ophiolite said:

In the best dramas, intense scenes are followed with light relief. This is such a moment.

Since you have been discussing statistics, this is a statistical interlude. A friend of mine had been raised in Kenya. He offered me the following defence of big game hunting. "The lion has a fifty-fifty chance. Either you hit it, or you miss it. That's fifty-fifty."

Click to expand...

My wife says the same about being in a falling aircraft, somewhat but not completely in jest. You either are, or you aren't.

essentialsaltes said:

Just because it has become established in the medical community (apparently to the point of dogma)

Click to expand...

My friend, you have no idea. I spend much effort defending EBM to the non-medical, and limiting it to the Medical. People forget it is heuristic, so many slavishly follow Best Practice recommendations as if gospel, and the amount of people that don't critically evaluate studies for their external validity before applying them, is shocking. There are even vocal groups that complain how Totalitarian EBM is, as more and more gets subjected to it, and position or grant is dependant on applying its principles. Even medical studies have a sort of 'newspeak' in which the jargon of EBM appears: You'll see medical journals seldom mention words like Scientific; but Evidence, Best Practice, gold standard, and other such narrowly defined EBM words are legion, and if you aren't familiar with EBM, you miss the implications such terms carry.

Ideally though, EBM should not be dogma, but help to weight clinical decision making appropriately as a method to evaluate how much confidence should be placed in results. At heart, EBM is a sceptical system, that knows everything may be wrong or biased, but tries to assist the doctor to navigate through the countless studies and developments in medicine. It is a golden mean, between trusting results and distrusting them, that we are aiming for.

As I said, you can believe this specific study, or you don't. I think much of the opposition is related to the traditional bias against assigning lunar influence (though that is not what this study indicates, but correlation). One should always carefully evaluate things, but if you start out rejecting standards of evidence widely held, you can reject anything you want (especially if you apply double standards, as SBM is wont to do). Maybe this study is flawed, time will tell; but a p value is usually a good indicator it isn't, though not foolproof, or completely devoid of bias. We are simply trying the best we can, in a world of inherent uncertainty about everything.