Blog: Artificial intelligence, grassroots rebels and the pharmaceutical cuckoo by Jerome Burne

[Andy]

Theoretically, Artificial Intelligence (AI) applied to Evidence-Based Medicine (EBM) is a marriage made in digital heaven. AI’s algorithms, founded on reliable RCT evidence that’s updated as new findings come in, will whip through mountains of data, leading to improved diagnosis, more accurate screening and better testing protocols; all guaranteed to improve performance and cut costs.

Mmmm, well, yes. EBM is certainly a good idea but the notion that it is ready for such a union involves large dollops of optimistic unicorn fantasy – to borrow a useful Brexit shorthand. And not just the obvious issues around garbage in leading to garbage out or the malign effects of human bias, which are regular worries about its implications for officially sanctioned treatments.

AI could also have a damaging effect on efforts to develop new approaches to the big chronic killers – such as obesity, diabetes, cancer and Alzheimer’s. Ones which don’t just involve researching and prescribing an increasing number of drugs. Drugs, of course, have their place, but their massive financial muscle has given them the power of a giant cuckoo, able to shoulder all other solutions out of the medical nest on the grounds of lack evidence.

Various grassroots revolutions by independent clinicians and expert patients have recently been having considerable success identifying the failings of official treatments and promoting the importance of lifestyle and other approaches. These include the treatment of low thyroid, chronic fatigue or ME, nutritional advice and mineral and vitamin supplements. Just bolting on AI in areas like these is likely to boost cuckoo power with little benefits for patients.
....
Medical orthodoxy cheerfully ignores EBM rule-breaking
Given the growing tendency to dismiss or block challenges to medical orthodoxy, modifying apps to reflect grassroots challenges seems unlikely. Easier to ignore them.

And ignoring evidence-based challenges to treatment based on a trial widely accepted to be high quality and a basis for Nice guidelines, is something the medical establishment is prepared to do. This is precisely what happened when a group of patients with a condition involving chronic fatigue, known as ME/CFS, re-analysed the data gathered in a trial that supported the orthodox treatment.

Known as the PACE trial, published in 2011, it had reported that 22% of patients getting treatment- a combo of psychotherapy plus gradually increasing amounts of exercise – recovered, compared with just 7% of those who got no therapy. The re-analysis carried out by senior, independent clinicians in 2016, found that the results had been massaged to produce a favourable result. Without the fancy statistical manipulation, just 3% recovered and 7% showed some improvement.

It’s a long-running and complicated story but it shows not only a readiness to play fast and loose with the data but also an ability by the psychiatric establishment to totally ignore a carefully conducted and devastating critique. The authors continued to assert that the treatment was beneficial and the Lancet – the journal that had published the PACE trial – ignored all calls to withdraw it or even to publish letters about it.

http://healthinsightuk.org/2019/07/...ssroots-rebels-and-the-pharmaceutical-cuckoo/

 

[MeSci]

It's a pity he erroneously calls Chronic Fatigue Syndrome chronic fatigue - otherwise he's good.

 

[Sly Saint]

Prescription drugs need supplement support
What your GP won’t have told you is that the drugs you are being prescribed can affect your ability to absorb or use various minerals and vitamins effectively. In other words, if you are taking half a dozen drugs – not uncommon in older people – their collective effect could be to make you deficient in, say, vitamins B6 and 12, vitamin E, and the minerals zinc and magnesium. Over several years – this is very likely to have a damaging effect on your health.

Data on this effect is buried in the medical literature but almost no doctors know anything about it, except for a few familiar ones, such as the ability of metformin and the stomach acid-blocking drugs to cut the availability of B12, vitamin D and magnesium. I’m being deliberately vague because I’ve just written about it but the feature hasn’t been published yet.

I know you can't expect GPs to know everything but in these times of apps and easily available tech, I don't know why they don't have databases that they could access to suggest supplements when drugs are prescribed.

 

[rvallee]

AI has tremendous potential but on... classical?... science, I guess. The one using the scientific method and relying on objective reality. EBM is clearly a massive mistake and circlejerk, it allows for all sorts of garbage to be accepted as valid just because some people like what it says. It will always suffer from garbage-in-garbage-out distortion.

As far as I can tell, EBM is doing research without the scientific method, just skip straight from imagination to trying to prove it's right and make your conclusion align with your assumptions (or just plain cheat if you can get away with it). It's not rigorous enough to be of value on a brute force approach like machine learning.

PACE is a good example. I'm not sure ML would have identified the flaws since the cheating part is not in the papers themselves and hidden behind lawyers and gaslighting, a hard concept for a ML system. It would certainly rate it as high bias and low validity, however, but that determination does not require confirmation, it's self-evident.

The choice of pretending otherwise, as Cochrane does, is human choice, especially arbitrarily miscategorizing it. As we know the peer review for the Cochrane systematic reviews was brutal and clearly disqualified them and still they were published. It's a choice to allow obvious pseudoscience to grow, no need for a machine to objective show that.

I think AI will be extremely brutal on most EBM, essentially puree the whole concept. The likely response will be to dismiss it since proponents "prefer those results" because they align with their expectations. But it won't reveal much we don't already know. AI is only as good as the data it works with and in EBM most of the data is aspirational and rarely objective.

I'm not sure how an AI would make sense of how useful the series of questionnaires used in CFS research are at evaluating ME, which is sometimes CFS but not always and it depends but if you ask the researchers they say both things so what is a machine to do with all this confusion? Those are all upstream choices that pollute the data and make it all entirely useless to objective analysis.

A unicorn fantasy is a pretty apt description. Especially when you consider that even if you have a unicorn, so what? It may look magical, but it's still a horse with something glued on its face. Except for narwhals, of course, the unicorns of the sea. Still not magical, though.

 

[rvallee]

My mind-squirrel kept at it and I can't really find fault but I think the flaw in EBM is that it hijacks the reward system in the same way that casinos exploit. It's a constant reward loop that encourages failure, just to keep the player going. Research is supposed to be hard and usually unrewarding. Most scientists will never produce any positive results, though they will still contribute to science, it's just not the same as making a breakthrough.

When you look at how EBM works, most research seem to either have positive results or be promising in the way of needing more research. It's almost never null. That's highly suspicious, it shows how inflated the whole thing is. But it's constantly rewarding. Even harmful contributions can be awarded with recognition, celebration and titles, as we painfully know. It's rewarding in a way rigorous, boring science isn't, in the form of endless participation trophies and praise for simply being there.

Maybe not all EBM but without a doubt psychosocial EBM badly suffers from it. Everything is special and meaningful and the brain loves it, wants more of it. Useless papers of no merit getting published feels just as rewarding as real science. It's addictive and difficult to break the cycle, no matter how obvious it becomes that the "progress" is no more meaningful than running around in circles hacking at virtual fruits. It's all cheap so funding will not run out and even otherwise smart people can succumb to the satisfaction of hoarding fake rewards as long as others acknowledge those rewards as real.

I think that may explain a lot about the whole MUS thing. It's all entirely useless, but the reward response is real, as are the gains from career advancement and the thrill of being on stage to get yet another participation trophy from the circlejerk committee. MUS has never been about the patients. It's not even about science or anything like this. It's a broken reward loop.

 

[Jenny TipsforME]

I just asked ChatGPT to write a little summary of this article and got:


“Artificial Intelligence, Grassroots Rebels, and the Pharmaceutical Cuckoo" URL: https://healthinsightuk.org/2019/07...assroots-rebels-and-the-pharmaceutical-cuckoo Author: Dr. Trudie Chalder Summary: This article, written by Dr. Trudie Chalder, discusses the potential implications of Artificial Intelligence (AI) on the pharmaceutical industry and the role of "grassroots rebels" in challenging traditional Evidence-Based Medicine (EBM) practices. The article also mentions the PACE trial, which is a controversial study that has been criticized for its methods and conclusions about the treatment of Chronic Fatigue Syndrome (CFS). The article raises important questions about the role of AI in healthcare and the need for transparency and accountability in medical research.”

well that gave me a laugh, this article criticising the PACE trial was very much NOT WRITTEN BY TRUDIE CHALDER )

Though it’s encouraging to know AI’s assessment of the PACE trial anyway. My prompts encourage it to tell me of any controversy surrounding the information it shares with me.