When it comes to testing whether a therapy works, the randomised clinical trial (RCT) is usually the methodology of choice. Its principle could not be simpler: one has a group of patients, subdivides them at random into typically two subgroups, treats one with the experimental and the other with a control treatment and eventually compares the outcomes.
The RCT beats all other methods for one simple reason: it most effectively minimises the risk of bias. It is the study design most likely to deliver a true and reliable answer as to whether it was truly the treatment or some other factors that caused the observed result.
This is good news for all who want to know the truth, but it can be an annoyance to those who prefer to fish in muddy waters. In alternative medicine, for instance, there are many who see research as a means of promoting their pet therapy. Therefore, they don’t like stringent tests because they run a high risk of failing to generate the desired result. They prefer studies which appear to be rigorous and, at the same time, maximise the chances of producing the findings they want or need to promote their business. Consequently, many researchers of alternative medicine consistently fail to produce negative findings (a list of such pseudo-scientists can be found on my blog), and journals of alternative medicine publish largely positive results.
How do they do it? The answer, I think, is to follow relatively simple tips on how to mislead the public with seemingly rigorous clinical trials.
The most brutal method for misleading people is simply to cheat. Anyone can put anything on paper, and, in the realm of alternative medicine, we currently have plenty of journals which publish almost any rubbish.
The process of ‘peer-review’ is one of several mechanisms supposed to minimise the risk of scientific fraud. Yet some journals have introduced a system of peer-review that rarely involves independent and critical scientists; often you can even ask for your best friend to get invited to do the peer-review, and the journal will follow your wish. Consequently, the door is wide open to cheating. Once your fraudulent paper has been published, it is difficult to tell that something is fundamentally wrong.
Obviously, cheating is neither ethical nor is it always as simple as we imagine. Researchers cannot easily claim to have conducted a clinical trial without a complex infrastructure which invariably involves other people. And they are likely to want to have some control over what is happening. This means that the fabrication of an entire data set may not always be possible. What might still be feasible, however, is the ‘prettification’ of the results.
By just ‘re-adjusting’ a few data points that failed to live up to your expectations, you might be able to turn a negative into a positive trial. Proper research governance is supposed to prevent this type of scientific misconduct but, in alternative medicine, such mechanisms are rarely adequately implemented.
Another method is the omission of aspects of a trial which turned out to disagree with the desired overall result. Most studies include a myriad of outcome measures. Once the statistics of a trial have been calculated, it is likely that by pure coincidence some of them yield the wanted positive results, while others do not.
Consider, for instance, a clinical trial where 20 parameters are being measured to quantify the outcome. Let’s say these outcome measures are taken repeatedly, for example, once before the start of the therapy and then every week for nine weeks. This means we have 20 x 10 = 200 sets of results. In most trials, a result is considered to be ‘significant’ if the probability of a false-positive finding is 5 per cent. For our study, this would mean that we will have 10 positive results purely by chance, even if we test one placebo against another.
By simply omitting any mention of some of the negative results, a researcher can easily turn such a negative study into a seemingly positive one. Normally, investigators rely on a pre-specified protocol which defines a primary outcome measure. But in the absence of proper governance, it might be possible to publish a report which obscures such detail and thus misleads the public (on my blog, there are several examples where this sort of thing seems to have happened).
4. Studies that cannot generate a negative result
All the above tricks are fraudulent, of course. And, if found out, fraud is not well regarded by the scientific community. Therefore, researchers who are bent on publishing positive findings but fear tarnishing their record need a more legitimate method for misleading the public.
The optimal trial design in this respect is the ‘A+B versus B’ design which is currently particularly popular in alternative medicine research. In such a study, patients receive either a treatment (A) together with usual care (B), or usual care (B) alone. This looks rigorous, can be published as a ‘pragmatic’ RCT addressing a real-life problem. It has the advantage of never failing to produce a positive result: A+B is always more than B alone, even if A is a pure placebo. Consequently, we know the result even before the research has started (recent studies using this approach can also be found on my blog).
Alternative medicine is an area that is now beginning to attract a reasonable amount of research. If we cast a critical eye on what emerges from it, we are bound to be harshly disappointed. The journals of alternative medicine are full of studies apparently showing that this or that alternative therapy is efficacious. Yet there are several compelling reasons to be sceptical about such findings. Much of what is being published in this area is hardly worth the paper it is printed on, and many of the researchers of alternative medicine are pseudo-scientists who seem about as trustworthy as a £4 note.
Edzard Ernst, emeritus professor at the University of Exeter, is the author of Homeopathy: The Undiluted Facts and the awardee of the John Maddox Prize 2015 for standing up for science. He blogs at edzardernst.com.