Ben Goldacre: Building RCTs into GP practice

[Sasha]

I know Ben Goldacre hasn't exactly covered himself in glory in relation to PACE but he does nevertheless say some interesting things. At the beginning of this talk he presents an interesting proposal about how to get a cheap and easy randomised trial across general practice whenever there is treatment uncertainty in medicine.

@Jonathan Edwards has mentioned wanting clinicians who treat PWME with unproven therapies to do trials - I wonder if something like this could be the mechanism.

(BTW, I'd like to keep this thread for the discussion of this idea, not for criticism of Goldacre. )

 

[Jonathan Edwards]

I know Ben Goldacre hasn't exactly covered himself in glory in relation to PACE but he does nevertheless say some interesting things. At the beginning of this talk he presents an interesting proposal about how to get a cheap and easy randomised trial across general practice whenever there is treatment uncertainty in medicine.

The basic idea that people should be randomised to different options during standard health care makes some sense. But there has to be a prospect of actually achieving your objective. Gold acre uses the example of not knowing which drug to choose for diabetes. That is either because they are about equally good so it does not matter, or about equally not very good, in which case it doesn't matter either, and moreover when a good drug comes along all the effort in choosing between two mediocre drugs will have been a waste of effort.

Goldacre also goes into hyperbole about 'standard academic papers' being useless. That is just wrong. The big steps forward are published in standard academic papers and get things right. They may be a tiny minority of papers but that is irrelevant.

I think the idea of randomisation in standard care would be very relevant to supportive care like activity advice and management in ME. It seems a good idea to have clinics where PWME get advice on how to manage their illness but we don't know what is the best advice. So maybe people should be randomised to some sensible options. The difficulty is that it is probably just as important that the carers do not know they are being randomised as the patients. That raises pretty huge political/ethical problems. It is the sort of thing probably only Scandinavian countries have ever been any good at.

 

[Sasha]

I think the idea of randomisation in standard care would be very relevant to supportive care like activity advice and management in ME. It seems a good idea to have clinics where PWME get advice on how to manage their illness but we don't know what is the best advice. So maybe people should be randomised to some sensible options. The difficulty is that it is probably just as important that the carers do not know they are being randomised as the patients. That raises pretty huge political/ethical problems. It is the sort of thing probably only Scandinavian countries have ever been any good at.

But shouldn't everyone know that they're being randomised? I don't think he's talking about hiding that aspect. People would still need to give consent.

 

[Jonathan Edwards]

But shouldn't everyone know that they're being randomised? I don't think he's talking about hiding that aspect. People would still need to give consent.

No you cannot offer people the option of consent, because that would take too long. He is making the point that there is no ethical reason not to randomise rather than do what doctors do now - pick what they fancy, so no need to ask for consent. There is also the problem of knowing you are being studied. It changes all sorts of things. It would be much better if nobody knew they were involved in studies -either patients or carers. But it is too much of a political hot potato.

What might be feasible is to have a blanket condition of using a heath service that you were aware that any treatment you had would be part of ongoing randomised comparisons of what seem equivalent options.

 

[wdb]

I could see these randomised GP trials working very well for well defined conditions with objectively measurable outcomes such as testing treatments for high cholesterol. For ME/CFS or pain disorders or mental health though we'd just have the same old problems with non-treatments that only produce subjective benefits or positive thinking therapies make ill people say that they are well.

 

[Mithriel]

I can't watch videos so this may not be relevant, but as a diabetic, I know what I want from the drugs I get from my GP and I certainly don't want to be randomised into a group taking a particular drug.

With the internet making so much knowledge available I can check what is available and discuss with my GP. I may be unusual because of having ME but I checked the research into the newer drugs and know what I do not want. (One paper said the drug was found to have few side effects except for the risk of sudden cardiac arrest, no thank you!)

We are moving on to the days of the expert patient, especially in diabetes where monitoring your own condition is becoming very common. Instead of medicines being up to the whim of the GP, they are starting to be a dialogue on what is available and what suits the individual.

 

[Adrian]

That is either because they are about equally good so it does not matter, or about equally not very good, in which case it doesn't matter either, and moreover when a good drug comes along all the effort in choosing between two mediocre drugs will have been a waste of effort.

I'm assuming there will also be something around different potential side effects or interactions with other drugs that patients take. If there is then it would make any such trials difficult.

I'm wondering if he has issues with just saying that two drugs are roughly equivalent. But I'm assuming that other factors may come into the choice (such as availability and side effects, interactions etc).

I think the idea of randomisation in standard care would be very relevant to supportive care like activity advice and management in ME. It seems a good idea to have clinics where PWME get advice on how to manage their illness but we don't know what is the best advice. So maybe people should be randomised to some sensible options.

The problem with this is that it is making care into a trial and hence reducing the patient and doctors ability to choose something appropriate to the patient (hence potential ethical choices).

I wonder if there are alternative ways to analyse the data from treatments in the field to look at the performance of care. I don't see why the notion of an RCT needs to be considered the only ideal. If we have larger volumes of data then are there ways to select randomly (perhaps with multiple runs a bit like with cross validation) to see which effects are better or even if there are given important sub-cases. We have very good data mining and analytic techniques these days with all the dangers that they can be biased. But perhaps we should go back to looking at the principles behind what an RCT is trying to achieve and project this into data mining on data collected in the field.