Failures of Occam's Razor In Medicine

[Dhondy]

William of Oakham, mispronounced as Occam, used to live not far from where I now stay....but around 700 years ago. His lasting legacy to philosophy, and indeed to problem solving, was the principle called "Occam's Razor", which essentially states that in trying to resolve to a problem, try and reach for the simplest solution that explains all parts of the problem.

 

Occam's razor implies that in reaching a solution, try and make as few assumptions as possible. I have outlined this in bold, as this is the key to interpreting Occam's razor. This is also called the "Principle of Parsimony".

 

Lest people use this blindly, let me quote something I read, "Occam's razor is no substitute for insight, logic and the scientific method".

 

In the words of Einstein, "Everything should be made as simple as possible, but not simpler."

 

Indeed, much of science draws on Occam's razor, and mostly it stands up to scrutiny. There are occasions though, when it lets you down badly, and one must always be aware of such pitfalls.

 

Let me give you three examples, one from medicine, one from genetics, and one from nature to illustrate this better.

 

Until very recently, upto six years ago, hormone replacement therapy (HRT) was considered the panacea for women who had recently reached menopause.

 

It made intuitive sense- it made them look younger, it stopped their hot flushes, it strengthened their bones, and above all, the oestregen component lowered the levels of cholesterol in the blood, thus holding out the promise of reducing heart attacks and strokes in the future.

 

Quite clearly, Occam's razor pointed to a win-win situation for HRT. Why would anybody be worried about using it?

 

Imagine the shock of the medical community therefore, when, a few years ago, a very large controlled study, called the Women's Health Initiative, showed that HRT in fact increased the risk of heart attacks and strokes in post-menopausal women significantly.

 

The finding led to a worldwide drop in prescriptions for HRT. It is now only used in the short term for alleviating troublesome symptoms such as hot flushes.

 

Another example.

 

High alcohol intake has been shown to be a risk factor for head and neck cancer in observational studies. This is thought to be due to acetaldehyde, a carcinogen that is obtained during oxidation of alcohol in the body. There are two allelles for the gene that oxidizes alcohol- one, let's call it B2, converts alcohol to acetaldehyde 100 times faster than the other, say B1. Scientists therefore justifiably postulated that persons who had two copies of B2, would be at a far higher risk of head and neck cancer than those with B1-B1 or B1-B2.

 

In reality though, it turned out that both B1-B2 and B1-B1 had a far higher risk of developing cancer than B2-B2. The reason for this was that acetaldehyde has very unpleasant effects on the body- it makes people feel unwell. The homozygotes for B2-B2 therefore cut down on their alcohol consumption to escape the dysphoric effects of acetaldehyde, while those with genotypes B1-B2 or B1-B1 continued to drink with impunity, and thus became more susceptible to head and neck cancer.

 

The principle of parsimony would have suggested a higher risk for B2-B2 based on the genotype, but it obviously could not account for confounding by alcohol consumption.

 

Finally, let me give you a topical example from nature. If you watch National Geographic, you will know that there is a huge concern about the dwindling number of cheetahs. There are only 10,000 of them left in the wild, they do not mate very easily, and the cubs they produce are often killed by other predators.

 

Scientists were concerned whether, firstly cheetah females had enough males to mate with, and secondly even if there were enough males, whether mating was taking place often enough, a view based on Occam's razor, due to the low density of cheetahs in general, and the behaviour of females studied on camera.

 

Imagine the surprise of nature scientists therefore, when DNA analysis of cheetah cubs recently revealed that female cheetahs were in fact shamelessly polyandrous, with up to half of their litter being sired by different males. That is to say, they were sleeping with every stranger that came along!

 

It is thought that this kind of behaviour offers a survival advantage by allowing genetically variable offsprings to adapt better to vicissitudes and by reducing infanticide by male cheetahs, who are often confused as to the parentage of the cubs!

 

Who would have thought, eh?

[ecoli]

When doing a controlled experiment, occam's razor is most often the correct one. In practice, there are lot of variables that we do not know about. Occam's razor still holds true... scientists just don't know the whole story. It's difficult to make an accurate hypothesis is you're not aware of all the variables.

[YT2095]

agreed, it seems all 3 were down to simply Not being in possession of all the facts, not a failing of the principal behind Occams razor.

[lucaspa]

His lasting legacy to philosophy, and indeed to problem solving, was the principle called "Occam's Razor", which essentially states that in trying to resolve to a problem, try and reach for the simplest solution that explains all parts of the problem.

 

Actually, that's NOT the Razor. William of Ockham actually meant that, when describing a phenomenon, include no unnecessary entities.

 

In Ockham's time, people said "an object moves because of an impetus". Ockham argued that "moves" means simply locations in space at different times. Therefore, "because of an impetus" is unnecessary. All we should say is "an object moves".

 

The idea that the simplest solution is the correct one is actually the position William of Ockham argued against. People of the time noted that in reflection, the angle of reflection = the angle of incidence. They reasoned from that that, in refraction, the angle of refraction should = 1/2 the angle of incidence, because that was the next "simplest". Ockham thought that reasoning was stupid. There was no a priori reason to think that. The angle of refraction, instead, had to be established by measurement. See John Losee's Introduction to the Philosophy of Science for a fuller discussion of this.

 

The idea that the simplest solution is the correct one comes from Newton. And it's wrong.

 

There are many better examples of how wrong the Razor or Parsimony is in biology. Basically signal transduction and transcription control have thrown out Parsimony and they invent more complicated systems all the time. Why? Because the "less simple" systems eventually turn out to be correct.

 

I'll give you another real world scenario, this one involving controlled experiments. In 1962 Marshall Urist discovered that implantation of demineralized bone matrix (DBM) under the skin of a rat or rabbit would cause bone to form at that site -- a site where bone would normally NEVER form. He established, by controlled experiments, that the responsible agent was a protein.

 

The Razor would say that there was only ONE protein -- that would be the simplest solution. Therefore over a dozen laboratories over the next 35 years tried to purify THE protein. All failed. They always ended up with at least 3 peptides. Finally, Wang and colleagues at Genetics Institute threw out the Razor and stopped trying to purify "the" protein. Instead, they cloned cDNA from every protein in the partially purified mixture. A brute force, non-simple approach. Turns out that there were five proteins, each of which would induce bone formation. The proteins were dimers of 2 identical peptide chains held together by disulfide bonds. They were called bone morphogentic proteins (BMPs) and were given numbers. It turns out that the naturally occuring BMP in bones is a heterodimer of BMP-2 and 7. Again, not the simplest.

 

E.coli, if you try to modify parsimony like you did: "Occam's razor still holds true... scientists just don't know the whole story. It's difficult to make an accurate hypothesis is you're not aware of all the variables." you basically reduce it to a tautology: The simplest answer is correct because, however complicated it is, it won't work if it is simpler. That may "save" Ockham's Razor, but it does so by making it useless.

[Dhondy]

Actually, that's NOT the Razor. William of Ockham actually meant that, when describing a phenomenon, include no unnecessary entities.

 

I did say this in the next line.

 

Occam's razor implies that in reaching a solution, try and make as few assumptions as possible. This is the key to interpreting Occam's razor. This is also called the "Principle of Parsimony".

 

Good example, anyway.

[Dhondy]

OK, here's another example of failure of Occam's razor in medicine.

 

We all grew up reading that bugs couldn't survive for long in the acid environment of the stomach, right? Surely, that was one of the reasons why the colon was chockful of the little buggers, while the stomach had none.

 

The idea therefore that little rod shaped bacteria called H. pylori could cause gastritis and peptic ulcer was completely counter-intuitive to the scientific community. When Barry Marshall, Aussie medic extraordinaire, in conjunction with the older, seasoned Robert Warren mooted this idea, he was subject of much ridicule from his peers.

 

In fact, Barry had to drink a solution teeming with the bugs, and develop an endoscopy proven ulcer, before people woke up and realized that he was onto something.

 

The Nobel Prize in 2005 brought some consolation, I should think.

 

You may call this a failure of Occam's razor, you may chose to call this a triumph of cussedness over ignorance based on existing knowledge, or lack thereof, but it doesn't really matter, it's just a matter of semantics.

 

The old fogeys told off Marshall & Warren because they were making unproven "assumptions" in their view (Marshall observed the bugs on microscopy of gastric biopsies and then took a leap of faith in linking them to dyspepsia and ulceration). The point I am trying to make is that Occam's razor often involves taking the path of least resistance based on existing knowledge, and people use that to create and perpetuate scientific dogma, as in this instance.

[Sayonara]

You may call this a failure of Occam's razor,

 

Not the way you have described it, no.

[Dhondy]

Then how would you describe it, pray, dear mademoiselle/sir?

 

Do edify this poor yokel.

[Sayonara]

It just seems to me that if you are going to show how Occam's Razor fails to help explain gastritis and peptic ulcers, it might be useful to state how the principle was applied and why this was inadequate.

 

The way you have described it, it doesn't read as if there was much razoring going on at all.

[Sepiraph]

I think you understand the principle behind Occam's Razor, but you are applying it incorrectly. The examples you provided has to do with incorrect conclusion (possibly due to incorrect understanding) rather than using extraneous assumptions to reach a conclusion.

[PhDP]

The Occam's razor only means that, all things being equal, the most parsimonious hypothesis should be prefered.

 

Most people forget the "all things being equal" part.

[lucaspa]

Actually, that's NOT the Razor. William of Ockham actually meant that, when describing a phenomenon, include no unnecessary entities.

 

I did say this in the next line.

 

Occam's razor implies that in reaching a solution, try and make as few assumptions as possible. This is the key to interpreting Occam's razor. This is also called the "Principle of Parsimony".

 

Ockham did not talk about solutions. He talked about phenomenon! The example Ockham used most often (listed in Losee's book) was:

 

"A body moves because of an impetus." This is not a "solution", but a description of a phenomenon -- the motion of something. The phrase "because of an impetus" is a proposed solution to the question "why does the body move?" BUT, this is the part Ockham said to remove.

 

Ockham stated that the statement should simply be "a body moves". No solution attached.

[lucaspa]

It just seems to me that if you are going to show how Occam's Razor fails to help explain gastritis and peptic ulcers, it might be useful to state how the principle was applied and why this was inadequate.

 

The way you have described it, it doesn't read as if there was much razoring going on at all.

 

Sayonara, look at my post. It describes an actual case where the Razor (as stated as "parsimony") was used and failed. I was there. I was in one of the labs trying to purify BMP in the 1980s. The simplest explanation for having a protein that caused bone induction was that there was ONE protein that caused bone induction. So that was what everyone tried to purify.

[lucaspa]

The Occam's razor only means that, all things being equal, the most parsimonious hypothesis should be prefered.

 

Most people forget the "all things being equal" part.

 

Then you have made the Razor a tautology and useless. If you excuse all exceptions because "all things were not equal", then you have the tautology "the explanation found by the data is the simplest because the data does not show any simpler explanation."

 

But the Razor is used when there are competing explanations and insufficient data to decide between them. In that case, the simplest explanation of the possible explanations is supposed to be the correct one. Right?

 

But that is what we are saying doesn't work. The simplest explanation is very often NOT the correct one.

[Sayonara]

Sayonara, look at my post. It describes an actual case where the Razor (as stated as "parsimony") was used and failed. I was there. I was in one of the labs trying to purify BMP in the 1980s. The simplest explanation for having a protein that caused bone induction was that there was ONE protein that caused bone induction. So that was what everyone tried to purify.

Yes, but I was not replying to you. I was replying to Dhondy. Specifically to his "gastritis and peptic ulcers" example.

 

If he wants to use your example as evidence for his argument, fine, but he did not do this. Instead he chose to give an example that failed to support his argument (because of the way it was presented, rather than because of the specifics of the scenario), and I am attempting to encourage clarification of that.

[PhDP]

Then you have made the Razor a tautology and useless.

 

What ? My definition is fine, I'm 100% certain I could find similar definitions in peer-reviewed journals. "All things being equal" means the competing theories have the same predictive/descriptive power, if theory A and B have the same predictice/descriptive power, but theory B is less parsimonious, then A should be prefered.

[lucaspa]

Yes, but I was not replying to you. I was replying to Dhondy. Specifically to his "gastritis and peptic ulcers" example.

 

If he wants to use your example as evidence for his argument, fine, but he did not do this. Instead he chose to give an example that failed to support his argument (because of the way it was presented, rather than because of the specifics of the scenario), and I am attempting to encourage clarification of that.

 

My apologies. I stand corrected.

[lucaspa]

What ? My definition is fine, I'm 100% certain I could find similar definitions in peer-reviewed journals.

 

You can find all manner of definitions for Ockham's Razor -- all of them incorrect.

 

"All things being equal" means the competing theories have the same predictive/descriptive power, if theory A and B have the same predictice/descriptive power, but theory B is less parsimonious, then A should be prefered.

 

And that is wrong. This is where 1) it isn't what Ockham stated and 2) it doesn't work. B should not be preferred because, quite "simply" , the simplest answer is NOT correct!

 

The only way to decide between competing theories is DATA! If you don't have the data to decide, then both theories stay on the table.

 

Let me give you an example from cosmology. There are many competing theories for First Cause. Two of them are A) God and B) ekpyrotic. Both have the same predictive/descriptive power in that each gives us the universe we observe. However, B is less "parsimonious" because it has 4 entities: a 5 dimensional universe, two 4-dimensional 'branes, and a 'brane released from one of the 4D 'branes that collides with the other -- making a "Big Splat" that wipes out that universe and starts another.

1. C Seife, Big bangs's new rival debuts with a splash. Science 292: 189-190, Apr 13, 2001. http://www.arXiv.org/abs/hep-th/0103239

http://www.sciencemag.org/cgi/content/full/292/5515/189

2. Turok on ekpyrotic http://online.itp.ucsb.edu/online/colloq/turok2/

 

Now, according to you, we should choose A because it is more parsimonious -- only 1 entity instead of 4. But I'd be willing to bet a large sum of money that you won't do this. So even you don't do this.

 

As one example of how the Razor is mistakenly presented -- by scientists -- here is the following:

 

“Consider for example the following two theories aimed at describing the motion of the planets around the sun:

The planets move around the sun in ellipses because there is a force between any of them and the sun which decreases as the square of the distance.

The planets move around the sun in ellipses because there is a force between any of them and the sun which decreases as the square of the distance. This force is generated by the will of some powerful aliens.” http://phyun5.ucr.edu/~wudka/Physics7/Notes_www/node10.html

 

Notice that this is identical to Ockham's example of "a body moves". It's just that, in this case, the body is "planets". The author wants to use the Razor to eliminate "aliens". However, according to Ockham, everything after "because" should be eliminated. The Ockham statement is "The planets move around the sun in ellipses."

[Sayonara]

My apologies. I stand corrected.

 

Not entirely your fault - I should really have made it clear who I was talking to.

[Dhondy]

If you accept that The Priniciple of Parsimony is, of necessity, based on existing knowledge, then you will understand what I meant when I cited the example of the role of H.pylori in peptic ulcers.

 

When Warren and Marshall made their groundbreaking discovery, the following were the existing beliefs:

 

1. Ulcers are caused by an excess of acid or by a breakdown in the protective mechanism of the gastric mucosa, caused by a reduction in prostaglandin synthesis due to drugs such as NSAIDs.

 

2. Bacterial infection could have no role to play in this, as bactreria simply could not survive in the acid environment of the stomach.

 

3. Reduce the acid, and the ulcers would heal, and stay healed.

 

Now, statement 3 was implicitly based on statement 1. If you felt that acid caused ulcers, William stipulated that you followed that up by deducing that reducing acid would heal ulcers.

 

Agreed?

 

William would have been wrong. In 75% of cases of gastric ulcers and >90% of duodenal ulcers, acid was not sufficient to cause ulcers, and reducing acid, while healing some ulcers, would inevitably be followed by a recurrence of ulceration at some stage, unless you got rid of H.pylori by triple thearpy.

 

Anyway, let's move on. The purpose of this thread was not to have a philosophical discussion about the correct application of Occam's razor, but to inspire ourselves as a community, with tales of scientific breakthroughs that went completely against the perceived wisdom of that age. Discoveries so unintuitive, they were considered almost heretical at first.

 

I'll try and think of some more.

 

Can I start with Barbara McClintock?

 

This lady has always been one of my biggest heroines because her work was so ahead of her times.

 

In the early 1940s, working on heritability in maize on Long Island Sound (between Connecticut & Long Island), she found that colours of maize kernels changed rapidly, with variegation is single grains of maize, indicating rapid transitions in genetic material. This mottling effect defied Mendel's basic principles of genetics because individual grains were multicoloured rather than a single colour.

 

From this simple discovery, McClintock thought of the concept of transposons, which are genes that move from one location to the other on the chromosome, and either inhibit or stimulate an active gene. In the case of maize, the position of transposons might inhibit or block pigment production in some cells. For example, if the transposon moved to a position adjacent to a pigment-producing gene, the cells were unable to produce the purple pigment. This resulted in white streaks or mottling rather than a solid purple grain.

 

Now consider that this discovery was made before the genetic code & the structure of the DNA double helix was known, and you begin to realize the enormity of her genius.

 

The importance of her discovery was not realized till the advent of the concept of onco-genes (cancer causing genes) & genetic engineering in the late 70s and early 80s.

 

Oncogenes may be activated by the random reshuffling of transposons to a position adjacent to the oncogene.

 

Similarly, transposons may be useful in genetic engineering with eukaryotic cells, by positioning transposons to activate certain genes. The most common example of this is found in genetically modified crops.

 

So it was that in 1983, fully 40 years after Barbara McClintock first described "jumping genes", she was awarded the Nobel Prize for Physiology & Medicine in recognition of her genius.

[Dhondy]

Time for more proximate stuff.

 

The human genetic code is stored in DNA, packed as a double helix. The purpose of genes is to carry the information for the synthesis of proteins. This information is transferred from the nucleus to the cytoplasm through the agency of messenger RNA (mRNA), which is a single helix, and is a complementary copy of the helix of DNA from which it is synthesized.

 

In the 1990s, scientists tried to get petunias to display the colour violet, by introducing mRNA for a protein that produced violet colour in other flowers. To their surprise, these petunias turned white. Not only were these flowers destroying the mRNA that were introduced, but they were also inactivating their own colour producing genes.

 

The problem was cracked by Craig Mello, working at Mass. University with a roundworm called C.elegans. Mello made several attempts to silence a certain roundworm gene by introducing mRNAs into the roundworm. No luck.

 

Then suddenly, with one batch of mRNA, it happened. The targeted gene was silenced, and stopped producing the putative protein. The effect spread remarkably fast through the worm, from tissue to tissue.

 

Mello was stunned. He discussed the finding with friend and colleague Andrew Fire at Stanford, who noticed that Mello had inadvertently included some double stranded RNA (dsRNA) into the batch that he had introduced into the roundworm. The pair observed to their amazement that ds RNA were 10 to 100 times effective in silencing genes than single stranded RNA.

 

The two published their paper on RNA interference in a 1998 issue of the journal Nature. It explained the petunia phenomenon, but many other things, such as the ability of human cells to neutralize RNA viruses.

 

It is now thought that this is a pretty ancient characteristic of most living cells, including humans, plants and animals, and plays a role in silencing invading viruses, but also jumping genes or transposons, which can often cause cancer.

 

Mello and Fire's discovery paved the way for further scientific work on such RNA, called small interfering RNA (siRNA), which are being used to silence certain genes in human and animal cells, and thereby elicit their function. It shouldn't be long before such research translates into therapeutic applications.

 

Mello describes the morning of 13th October in 2006, when his phone started flashing at 4:30 AM. His wife asked him not to pick it up, as minutes earlier, somebody has made a "prank call" to the efffect that her husband had won the Nobel Prize. When Mello told his missus that it was the day that the Nobel Prize in Medicine was traditionally announced every year, she was completely gobsmacked.

 

For their work on RNA interference, The Nobel prize for Physiology and Medicine was awarded to Andrew Fire and Craig Mello in 2006.

[SkepticLance]

I have always loved the principle of Occams Razor. However, it is related to probability, not certainty. If various possible hypotheses compete as explanation for a particular phenomenon, then the hypothesis that is simplest, with the fewest new assumptions is the most likely to be correct. Not the one certainly correct. Just the most probable.

 

Having said that, I find Occams Razor most useful in dealing with human stupidity. Someone shows a photo of a ghost. Two possible explanations.

1. A hoax using photoshop.

2. A genuine phenomenon new to science.

Get the point?

[foodchain]

I have always loved the principle of Occams Razor. However, it is related to probability, not certainty. If various possible hypotheses compete as explanation for a particular phenomenon, then the hypothesis that is simplest, with the fewest new assumptions is the most likely to be correct. Not the one certainly correct. Just the most probable.

 

Having said that, I find Occams Razor most useful in dealing with human stupidity. Someone shows a photo of a ghost. Two possible explanations.

1. A hoax using photoshop.

2. A genuine phenomenon new to science.

Get the point?

 

 

 

Good point.

 

I like to think of QM here. Lets for instance look at the probability cloud with the uncertainty principal. In short, it says we don’t know.

 

Its probable but uncertain, at least that’s how I read it. Its like will it rain tomorrow, probable, but uncertain. What do you do with something like that with the razor? Follow the probable but uncertain? Basically I think to understand the razor then applies some basic scientific common sense? Or is it based on observation and experimentation with such? Like in geology, supposedly rivers always follow the shortest path to the ocean right? Well looking on maps I don’t see how this applies across the board, so what do you say then, form a hypothesis on what’s common or sort of known or looks probable but is uncertain? Then again maybe the rivers are all trying to do the same thing its just that the H2O and stuff does not move through the same material with the same levels of energy or all kinds of other variables that really quickly make it not a so easy to solve problem really, but its a principle that sounds easy and sort of makes sense and could probably be the answer but its probable and uncertain.

 

Personal its just philosophy to me, some realm of imaginary numbers or what not. I will probably never really use it in all actuality save for time critical situations really, I will always try to get some more absolute representation of whatever the phenomena happens to be present for study, such as what makes a good tasting cheeseburger.

[lucaspa]

I have always loved the principle of Occams Razor. However, it is related to probability, not certainty.

 

No, it's not. No one has ever done probability calculations based on it.

 

If various possible hypotheses compete as explanation for a particular phenomenon, then the hypothesis that is simplest, with the fewest new assumptions is the most likely to be correct. Not the one certainly correct. Just the most probable.

 

Then show us the calculations for that! This is an assertion.

 

Beyond that, look at what you are doing! What does "most likely to be correct" mean in science? Nothing! What we want to know is whether the hypothesis is correct or not. And saying "most likely correct" doesn't cut it.

 

The only way you can tell whether a hypothesis is correct or not is to gather the data to either falsify the hypothesis or falsify the alternative hypotheses we can think of.

 

Having said that, I find Occams Razor most useful in dealing with human stupidity. Someone shows a photo of a ghost. Two possible explanations.

1. A hoax using photoshop.

2. A genuine phenomenon new to science.

Get the point?

 

Yes. You are using it to dismiss entities you don't believe in. IOW, this isn't something for science and hypothesis evaluation, but a rationalization for your own prejudices.

[lucaspa]

Time for more proximate stuff.

 

Nice story, but how does it illustrate that Occam's Razor doesn't work. Where was the Razor being used here?