But is EBM really all that it is cracked up to be? As it has evolved, EBM refers mainly to statistical evidence of a treatment's effectiveness in trials.
What was a means to an end - statistical significance as a way to understand the huge empirical data gathered in trials - has become an end in itself.
In EBM today, the best evidence is assumed to be that from randomised controlled trials (RCTs) - and from meta-analyses, which are systematic reviews of a series of RCTs.
In an RCT, subjects are randomly allocated one treatment or another, including a placebo. This is the best way to overcome researcher and subject bias.
The problem is that these days, any treatment that achieves some degree of statistical significance in an RCT for its effectiveness, however little, is regarded as effective - even if the drug's known chemistry and the biology of the disease concerned both say it is scientifically implausible for the drug to work in that disease condition.
Though EBM was first mooted in 1992, its roots are traceable to 18th-century Britain, according to Ulrich Tröhler's To Improve The Evidence Of Medicine (2000). Before that, gathering empirical data was frowned upon when proving the effectiveness of a treatment.
This was because, until EBM, it was all "eminence-based medicine" - the more senior the doctor, the less empirical evidence was needed. It was held as a matter of doctrinaire faith that experience trumped evidence.
The British Navy was one of the first sites for medical trials, as the ship surgeon held his study subjects captive for months at sea. Thus James Lind's famous experiments in 1747 when he showed fresh lime prevented scurvy or vitamin C deficiency in seamen who rarely had fresh fruits and vegetables for months.
But 42 years would pass before eminent physicians of the day accepted it because, as Dr Trohler explained, no one knew about vitamins. The lesson from 1747 is that the basic science behind any hypothesis an RCT seeks to test matters: If the science reasoning says it is implausible that a treatment could work, testing it in humans is not justifiable.
There was no knowledge of vitamins at all in Dr Lind's time, so no one knew why lime juice would work and sauerkraut would not. Thus doctors of the day were not convinced. If the research hypothesis had been that lime juice would work because it is full of vitamin C whereas sauerkraut would not because its vitamin C has been destroyed, then there would be reason to try out lime juice against sauerkraut. In that case, lime juice did not kill so an "RCT" on it did not matter.
But today, RCTs see dangerous drugs being tested and RCTs cost lots of money. RCTs ought not be done unless the research hypothesis is grounded in science.
In summary, the hypothesis to be studied must be scientifically plausible, that is, a drug's "prior probability" of effectiveness must be reasonable given what is known about the biology of the disease and the chemistry of the drug concerned.
Take, for example, homeopathy, which asserts that small doses of a substance can cure symptoms that large doses of it are known to cause. So a cure consists of that substance in dilute solution because it can supposedly imprint a mirror image and thus the inverse of itself on the water used.
As the inverse, the image's effects are supposedly the reverse of the substance. Hence its curative powers. To achieve these minute doses, ultra-dilute solutions are made, which after many iterations of dilutions, must inevitably be plain water again.
But there is no scientific mechanism for water to carry information or have memory. None has ever been shown in the laboratory and there is no reason to think water cures illnesses.
Now we see a hint of why Britain's Department of Health, when queried by Parliament in 2009, said that "no scientific evidence was examined in drawing up the National Rules Scheme" for homeopathic treatment provided under its National Health Service. A homeopathy RCT is unjustifiable on the science. Yet EBM would take it on.
By the late 20th century, large RCTs - and meta-analyses of large RCTs - had become the gold standard for judging whether a treatment worked or not. But this unwavering faith in the statistical massaging of numerical data RCTs collect as the sure-fire way to find "truth" is unwarranted.
And because the method has transmuted into an end in itself, conversely, any treatment not "proven" by RCTs is considered non-EBM or "unproven".
In 2003, a spoof appeared in the British Medical Journal to mock this rejection of treatments not proven by RCTs.
The article's title said it all: "Parachute use to prevent death and major trauma related to gravitational challenge: A systematic review of RCTs". You don't need RCTs to show chutes work - elementary physics would say so - and who wants to be assigned to a control group with no chute?
Without insisting on "prior probabilities", EBM research has become the remorseless, idolatrous pursuit of statistical significance.
Even treatments that would work only if established laws of biology, chemistry and physics were utterly suspended are deemed worthy of RCTs.
But as Dr John Ioannidis, an epidemiologist at Greece's University of Ioannina, showed in his 2005 PLoS Medicine study, the more scientifically implausible a hypothesis, the more likely RCTs will produce equivocal results, with lots of false positives. This is because, among others, the placebo effect is very marked particularly for culturally resonant sham treatments.
In a 1999 Annals Of Internal Medicine study, Dr Steven Goodman, a Johns Hopkins epidemiologist, noted that "biological understanding plays little formal role in the interpretation of quantitative results... in modern medical research data".
The medical community idol worships "this statistical approach", oblivious, he lamented, to the fact that statisticians themselves have been vigorously disputing whether this is a valid approach to truth since the 1930s.
The only way out of this research cul-de-sac is to insist on prior scientific plausibility.
But it is dogma in EBM, as it stands today, that established basic science knowledge, even when it refutes a hypothesis, is the least important kind of evidence, or only as important as expert experience, according to the authoritative Oxford Centre for Evidence-Based Medicine.
Since this structural weakness is intrinsic to RCT design today, EBM research is fraught with false positives. And the more of them there were, the more improbable the hypothesis that was tested.
The upshot is that, even if your doctor says she has RCT support for the latest aesthetic treatment, keep her away with a long pole, for these treatments are, by and large, scientifically implausible to begin with. Or else wealthy old women won't look a day over 21.
andyho@sph.com.sg
What was a means to an end - statistical significance as a way to understand the huge empirical data gathered in trials - has become an end in itself.
In EBM today, the best evidence is assumed to be that from randomised controlled trials (RCTs) - and from meta-analyses, which are systematic reviews of a series of RCTs.
In an RCT, subjects are randomly allocated one treatment or another, including a placebo. This is the best way to overcome researcher and subject bias.
The problem is that these days, any treatment that achieves some degree of statistical significance in an RCT for its effectiveness, however little, is regarded as effective - even if the drug's known chemistry and the biology of the disease concerned both say it is scientifically implausible for the drug to work in that disease condition.
Though EBM was first mooted in 1992, its roots are traceable to 18th-century Britain, according to Ulrich Tröhler's To Improve The Evidence Of Medicine (2000). Before that, gathering empirical data was frowned upon when proving the effectiveness of a treatment.
This was because, until EBM, it was all "eminence-based medicine" - the more senior the doctor, the less empirical evidence was needed. It was held as a matter of doctrinaire faith that experience trumped evidence.
The British Navy was one of the first sites for medical trials, as the ship surgeon held his study subjects captive for months at sea. Thus James Lind's famous experiments in 1747 when he showed fresh lime prevented scurvy or vitamin C deficiency in seamen who rarely had fresh fruits and vegetables for months.
But 42 years would pass before eminent physicians of the day accepted it because, as Dr Trohler explained, no one knew about vitamins. The lesson from 1747 is that the basic science behind any hypothesis an RCT seeks to test matters: If the science reasoning says it is implausible that a treatment could work, testing it in humans is not justifiable.
There was no knowledge of vitamins at all in Dr Lind's time, so no one knew why lime juice would work and sauerkraut would not. Thus doctors of the day were not convinced. If the research hypothesis had been that lime juice would work because it is full of vitamin C whereas sauerkraut would not because its vitamin C has been destroyed, then there would be reason to try out lime juice against sauerkraut. In that case, lime juice did not kill so an "RCT" on it did not matter.
But today, RCTs see dangerous drugs being tested and RCTs cost lots of money. RCTs ought not be done unless the research hypothesis is grounded in science.
In summary, the hypothesis to be studied must be scientifically plausible, that is, a drug's "prior probability" of effectiveness must be reasonable given what is known about the biology of the disease and the chemistry of the drug concerned.
Take, for example, homeopathy, which asserts that small doses of a substance can cure symptoms that large doses of it are known to cause. So a cure consists of that substance in dilute solution because it can supposedly imprint a mirror image and thus the inverse of itself on the water used.
As the inverse, the image's effects are supposedly the reverse of the substance. Hence its curative powers. To achieve these minute doses, ultra-dilute solutions are made, which after many iterations of dilutions, must inevitably be plain water again.
But there is no scientific mechanism for water to carry information or have memory. None has ever been shown in the laboratory and there is no reason to think water cures illnesses.
Now we see a hint of why Britain's Department of Health, when queried by Parliament in 2009, said that "no scientific evidence was examined in drawing up the National Rules Scheme" for homeopathic treatment provided under its National Health Service. A homeopathy RCT is unjustifiable on the science. Yet EBM would take it on.
By the late 20th century, large RCTs - and meta-analyses of large RCTs - had become the gold standard for judging whether a treatment worked or not. But this unwavering faith in the statistical massaging of numerical data RCTs collect as the sure-fire way to find "truth" is unwarranted.
And because the method has transmuted into an end in itself, conversely, any treatment not "proven" by RCTs is considered non-EBM or "unproven".
In 2003, a spoof appeared in the British Medical Journal to mock this rejection of treatments not proven by RCTs.
The article's title said it all: "Parachute use to prevent death and major trauma related to gravitational challenge: A systematic review of RCTs". You don't need RCTs to show chutes work - elementary physics would say so - and who wants to be assigned to a control group with no chute?
Without insisting on "prior probabilities", EBM research has become the remorseless, idolatrous pursuit of statistical significance.
Even treatments that would work only if established laws of biology, chemistry and physics were utterly suspended are deemed worthy of RCTs.
But as Dr John Ioannidis, an epidemiologist at Greece's University of Ioannina, showed in his 2005 PLoS Medicine study, the more scientifically implausible a hypothesis, the more likely RCTs will produce equivocal results, with lots of false positives. This is because, among others, the placebo effect is very marked particularly for culturally resonant sham treatments.
In a 1999 Annals Of Internal Medicine study, Dr Steven Goodman, a Johns Hopkins epidemiologist, noted that "biological understanding plays little formal role in the interpretation of quantitative results... in modern medical research data".
The medical community idol worships "this statistical approach", oblivious, he lamented, to the fact that statisticians themselves have been vigorously disputing whether this is a valid approach to truth since the 1930s.
The only way out of this research cul-de-sac is to insist on prior scientific plausibility.
But it is dogma in EBM, as it stands today, that established basic science knowledge, even when it refutes a hypothesis, is the least important kind of evidence, or only as important as expert experience, according to the authoritative Oxford Centre for Evidence-Based Medicine.
Since this structural weakness is intrinsic to RCT design today, EBM research is fraught with false positives. And the more of them there were, the more improbable the hypothesis that was tested.
The upshot is that, even if your doctor says she has RCT support for the latest aesthetic treatment, keep her away with a long pole, for these treatments are, by and large, scientifically implausible to begin with. Or else wealthy old women won't look a day over 21.
andyho@sph.com.sg
