Thursday, December 05, 2013

NIH discovers reproducibility

When someone first described to me the evidence-based medicine movement, I responded (shocked): "Is that like science-based science? What were they doing before?"
Nature News: In biomedical science, at least one thing is apparently reproducible: a steady stream of studies that show the irreproducibility of many important experiments.

In a 2011 internal survey, pharmaceutical firm Bayer HealthCare of Leverkusen, Germany, was unable to validate the relevant preclinical research for almost two-thirds of 67 in-house projects. Then, in 2012, scientists at Amgen, a drug company based in Thousand Oaks, California, reported their failure to replicate 89% of the findings from 53 landmark cancer papers. And in a study published in May, more than half of the respondents to a survey at the MD Anderson Cancer Center in Houston, Texas, reported failing at least once in attempts at reproducing published data (see 'Make believe').

The growing problem is threatening the reputation of the US National Institutes of Health (NIH) based in Bethesda, Maryland, which funds many of the studies in question. Senior NIH officials are now considering adding requirements to grant applications to make experimental validations routine for certain types of science, such as the foundational work that leads to costly clinical trials. ...
A comment from a researcher quoted in the article notes
... a broader need to shift biomedical research from categorical statements and simple schematics to quantifiable hypotheses backed up by modeling and computation, open access to data (itself requiring new approaches and infrastructure) and better application of probability theory and statistics.
See also Medical science? (are most research findings false?) and Bounded cognition.

16 comments:

SethTS said...

When I first got acquainted with some of the research being published by medical faculty many years ago, my main impression was "these people need to have their licenses to practice statistics revoked".

A great deal of social science publishing depends on vagueness about causation vs. correlation. Unfortunately, medical research often isn't much better. The human body is pretty d&*n complex, and controls can be difficult to construct or fraught with ethical issues.

Diogenes said...

YET americans spend more on healthcare than they do on food.



imho there's an exception to association isn't causation..when the odds ratio is very high. smoking and lung cancer is an example. drinking and throat cancer is another. odds ratios 12 to 1 and 3 to 1 respectively.

Diogenes said...

again physicists smart, everyone else dumb, but vaccines and antibiotics conway should have added to plumbing (for the zero people who read my last comment).


nobel laureate linda buck, AFTER her nobel, did what? tested 80k compounds for their effect on the lifespan of c. elegans. two tetracyclic anti-depressants were hits along with an anti-psychotic (anti-seizure drugs have been hits in the past). retested under different conditions by a german group, the worms got fat and died prematurely. these two anti-depressants are known to cause weight gain in humans.


so were buck's results wrong? steve should cut medical scientists some slack. they aren't mathletes, but what they're dealing with is so complex that even the description "complex" may not be apt. the description "complex" assumes that the phenomena ARE mechanistic.

Diogenes said...

on medical science:



even "in this day and age" people still die from causes unknown.

perhaps i can use this post inappropriately as a "requiem" for my uncle. phd in econ from yale could never find a professorship except in china. he was fluent in chinese. lost his speech, then lost his ability to walk, then died. a few days ago from an unknown organic brain disease. i hope it's not genetic.

BobSykes said...

It's not just medical research. Any research related to biology and the environment is suspect.


In part, the systems are so complex that research is very difficult conceptually and the required sample sizes are enormous. The most famous medical research program is the longitudinal Nurses Health Study, now some 20 years old or more. But its results are not really applicable to people in general. It's a study of women, not men, and the women are nurses, which is a subset of women, and the nurses have upper middle class incomes, often higher. My niece makes $120,000 per year at Brigham and Williams Hospital in Boston as a nursing administrator. [Her graduate degrees are from Phoenix U. Another interesting topic.]


Doctors and biologists and environmental scientists have the most minimal mathematical skills, essentially only high school algebra, and no understanding of statistics. I occasionally had a PhD candidate biologist in my mathematical ecology course, and they were hopeless with the simplest calculus. Much of the climatology controversy circles around utterly incompetent statistics by the more famous warmists, including even documented data input errors.


Then there is fraud. Fraud is very widespread in academic medicine, biology and environmental science. Academic researchers are left alone to do what they want as long as they bring in money. Not so in the corporate world, which is heavily scrutinized, so fraud in corporate research is rare, the tobacco companies not withstanding. Physicists and mathematicians never encounter fraud in their disciplines, but it is so widespread in the biology related fields that hardly anything can be trusted.


The fraud exists because it is easy to do and the financial rewards are enormous, as well as the fame. Highly successful academic scientists have fat salaries and benefits and live upper class (top 10% or higher) jet-set lifestyles funded by the government. When was the last time a major conference was held in Akron?


Another problem is that when fraud is detected deans, chairs and administrators will usually cover it up, or at least try to. I saw this happen three times in the 37 years I worked in academia. The classic case is Robert Gallo at CDC. Fausey is still defending him.


No retired, I have no respect at all for academic science other than mathematics and the hard physical sciences like chemistry and physics. No science publication should be take at face value, and no science organization, including specifically the NSF, can be trusted.

Anonymous said...

Best wishes... If I may ask. Looking through this blog; your profile seems to be very political. Before I ask the question I must provide context. In the USA indian americans(from india) earn the most per house hold a year out of all ethnic groups($90000 a year for indians, $74000 for east asians, $60000 for whites and 3600 for hispanics or so and 24000 for blacks). Indians seem to be over represented in fields of medicine and just acedemic achievements. In India there is a space programe with 900million people on less than $5 a day. It seems the caste system has bred a genetic elite... How do you thing genetic engineering caste based hetrogenous countries like.america and india when genetic engineering is avaliable?

Diogenes said...

yes. american healthcare is overpriced almost ENTIRELY because healthcare workers are overpaid.

Hacienda said...

You see, there really are only two subjects in the world- physics and everything else, which we can call "philosophy".

Owen said...

"american healthcare is overpriced almost ENTIRELY because healthcare workers are overpaid."

American health care workers are overpaid because we refuse to train enough of them. Then we demand that they work extra long hours to get enough of them to meet demand and we end up paying steep overtime prices to get the health care done.



With respect to physicians,
* It's very hard to get a US visa and license for a foreign physician
* Almost thirty years went by without a new medical school permitted to open until the late 2000's
* Nurses and assistants are more restricted from performing independent services in the USA than any other nation; nearly every medical service, no matter how safe and trivial, requires physician supervision
* Half of new US physicians are women and women would like to work shorter careers and less hours, but we did not change the number of doctors trained while this transition happened
* US medical education usually makes incoming physicians study about 12 post-secondary years before they are qualified for independent practice; the OECD average is three or four years less schooling and internships.

JayMan said...

When someone first described to me the evidence-based medicine movement, I responded (shocked): "Is that like science-based science? What were they doing before?"

You sure you want to know? ;)

On that, actually:

Trans Fat Hysteria and the Mystery of Heart Disease | JayMan's Blog

So if dietary trans fat can’t explain heart disease, what does? The short answer is we don’t know. Contrary to what health experts might lead you to believe, fundamentally, we have no idea what causes heart disease. There are all sorts of hypotheses, but, as we see, all face significant trouble when subjected to scrutiny.

Kennon Gilson said...

That's a pretty silly comment, given Physics is a branch of Philosophy. I suggest you understand what they teach in Philosophy. Maybe take a course in Philosophy where they tear apart a lot of lousy work in current Physics.

Hacienda said...

By philosophy I meant literally everything else. To me working as a busboy, or having dinner with the familly, or whatever and any experience should be counted as a knowledge experience. Physics could be simply one of these experiences and not given any priority over any thing. But, it seems to be the one field that actually has "meaning". Unlike all others, even including math.

Iamexpert said...

How does what you wrote relate to the vaccines-autism debate?

barhahra said...

physics is a restricted and narrow subset of knowledge. Step outside it, you would be surprised to see the world as it really is. For instance, social interaction is a skillset that most scientists lack. But it is a fine and wondrous skill, a form of knowledge in itself. For any nerd who moves to a new city, it is a wonder to behold the socially savvy friend who establishes a wide network of partiers and dates, while the nerd himself hangs around at home fapping to porn.
For some reason we intellectuals don't value the man who can ruthlessly ascend the social ladder as much as a man who can solve a physics problem. But the fact is that the ability to weave together coalitions of allies, fight off challengers, and guard your winnings is an incredible task that requires a synthesis of intellect, will and courage.
The Ancient Greeks knew this, Aristotle and Plato wrote on natural and human sciences giving both equal importance.

5371 said...

It doesn't require intellect at all, and courage only in some environments.

Hacienda said...

Physics is quite possibly mankind's greatest friend or enemy. I mean that seriously, so it can't be considered simply a narrow subset of knowledge. There is no more disruptive discipline than physics. It wins wars, destroys civilizations, destroys biospheres (well at least ecologies and in concert with Western (okay Eurasian civilization's) greed).


And in terms of the ways physics changes opinions- Bell's experiments, the measurement of the speed of light, the splitting of the atom, Newton's evil calculus vs Leibnitz's friendly one . These surely are not narrow subsets of knowledge.

Blog Archive

Labels