Earlier this week there was a big media hoo-haa over a PLoS One published paper on placebos.
The BBC lead with “’Most family doctors’ have given a patient a placebo drug” while the Daily Mail ran with “Nearly ALL doctors have given patients a placebo – either to keep them happy or reassure them” and both mentioned the attention grabbing result that 97% of doctors have knowingly prescribed a placebo. Both the BBC and the Daily Mail specifically give antibiotics being prescribed for viral infections as an example of placebo.
The study itself was titled ‘Placebo Use in the United Kingdom: Results from a National Survey of Primary Care Practitioners’ and, as PLoS One is an Open Access journal, is free for anyone to read. The result, as reported by the mainstream press, certainly do make interesting reading, but I wonder if these figures hold up on slightly closer scrutiny?
The survey gained responses from 783 general practitioners, via a web based survey. This is said to be representative and comes from random sampling from Doctors.net.uk registrations. According to the 2012 NHS Workforce Survey, there are 40,265 GPs in the UK which leads these 783 GPs sampled to be around 2% of the available population, chosen from a pool of just 71% of available GPs.
The really interesting detail in this paper though is the classification of placebo.
Here, placebo is divided into ‘pure’ and ‘impure’. A pure placebo, according to the paper, is a sugar pill or saline injection. There is a rather impressive list of what could pass as an impure placebo:
- Positive suggestions
- Nutritional supplements for conditions unlikely to benefit from this therapy (such as vitamin C for cancer)
- Probiotics for diarrhea
- Peppermint pills for pharyngitis
- Antibiotics for suspected viral infections
- Sub-clinical doses of otherwise effective therapies
- Off-label uses of potentially effective therapies
- Complementary and Alternative medicine (CAM) whose effectiveness is not evidence-based
- Conventional medicine whose effectiveness is not evidence-based
- Diagnostic practices based on the patient’s request or to calm the patient such as
○ Non-essential physical examinations
○ Non-essential technical examinations of the patient (blood tests, X-rays)
A few interesting ones there, especially ‘Positive Suggestion’ and ‘Complimentary and Alternative medicine whose effectiveness is not evidence based’ (pretty much all of it, then). It’s safe to say that a number of the intervention on this list would not normally be classified as placebo.
In adding Positive Suggestion to this list of placebos, they have essentially included almost every GP consultation in this group. If a GP tells you that this treatment ‘will make you feel better’ or similar, they are using positive suggestion. If they neglect to do this, they could be accused of having poor ‘bedside manner’. Even if a GP gave a sugar pill and told a patient that it’d make them feel better, this would count as impure placebo.
The only feasible way that a pure placebo could be registered here is if a GP gave a patient a sugar pill/saline injection without any guidance – ‘Here’s some pills. There you go…’
Unsurprisingly, the 97% figure given is for impure placebo use, where the pure placebo result was reported at 12%. This sounds incredibly high for the pure placebo group, until you take into consideration the fact the numbers are also grouped into frequency.
The 12% pure placebo use figure is an ‘at least once in career’ statistic. When we look at the ‘frequently’ numbers this drops dramatically to just 0.9%. This type of placebo prescribing is very rarely used in real life.
So, looking again at the impure placebos and concentrating on the frequently used (daily or weekly) the overall incidence drops to 77% – quite a drop from the headline grabbing 97%. Of this 77% the most frequently used are ‘Non essential physical examinations’ and ‘Positive reassurance’.
The ‘non-essential’ exams sound a little disconcerting, but don’t forget that a doctor often has to carry out a number of different tests to reach a diagnosis. Does a test that comes back negative constitute as being unnecessary?
When we remove the slightly obscure definitions for impure placebos, three placebos stand out. They are ‘Off-label uses of potentially effective therapy’, ‘Antibiotics for suspected viral infections’ and ‘Conventional medicine whose efficacy is not evidence based’.
I’m going to let off-label use slide. Off-label prescribing isn’t necessarily a bad thing and many drugs are used for multiple ailments. For example, the tricyclic antidepressant amitriptyline is prescribed off-label for neuropathic pain. The GMC understands the need for and use of off-label prescription, and although it can be beset with problems, including it here with the placebos is at best questionable.
The figure for antibiotics being prescribed for suspected viral infections is 25.2%. This, it has to be said, is pretty high. I would suspect that it’s more to do with doctors giving patients what they expect and demand than doctors willingly giving a placebo. If this is the case, it’s pretty poor practice.
So finally we reach ‘conventional medicine whose use is not evidence based’. This group of ‘impure placebo’ is frequently prescribed by 26.2% of GPs surveys. Interesting? Well, that depends on the context in which you look at the study.
This study was part funded by the Southampton Complementary Medical Research Trust and co-authored by George Lewith of the Southampton Complementary and Integrated Medicine Research Unit.
Lewith is particularly involved in the research of homeopathy and acupuncture (as well as other CAM modalities) and has published research backing the idea that the homeopathic consultation, not the sugar pill, is what is responsible for any effect seen (i.e. homeopathy is placebo). He has been criticised for continuing to prescribe homeopathy despite this.
Could it be that this study – partially funded by a CAM group, partially designed and carried out by a CAM practitioner/researcher to show that placebo use is widespread – could be in turn cited as an argument for CAM use? After all, if CAM medications are placebo, is it better to prescribe them for ailments than, for example, antibiotics for viral infections? If medicines that have no evidence base are prescribed regularly, then why shouldn’t evidence-free CAM be an option?
That’s certainly one way of looking at the results as presented, and especially as reported by the press. But digging a little deeper, and being a little more careful as to selecting what counts as placebo, this survey actually shows that placebo is much less widespread in it’s use, and that CAM really isn’t an alternative to anything. The real story here is how this paper has been presented to and by the mainstream press and what fallout there may be from that.
The press release from the University of Southampton only contains the 97% and 12% figures for impure and pure placebos. The vital frequency information has been stripped out completely. What’s happened here (and the researchers will have known this would happen) is that the press release has been churned into an article without any one of the journalists looking into this having digested the original PLoS One study.
Running this through Churnalism.com shows that both the Daily Mail and the Independant have respectively taken 51% and 63% of the press release verbatim, unsurprisingly taking out the details of funding and not bothering to go into detail about the paper itself. This is not only a clear show of poor journalistic standards, but an example of dishonesty in presentation of research results. Moreover, the press release itself has been specifically designed to use cherry-picked data that will grab headlines rather than show representative data.
What is already a badly designed study, which could only ever show a high proportion of placebo use, has been deliberately misrepresented to gain maximum exposure. All this shows everyone involved in a bad light, starting with the researchers, through PLoS One and ending with bad journalism.
The annoying thing is that this could have been so different. The study could have chosen its differentiation of placebo types better, possibly splitting them into 3 or more types; pure (as before), reinforced (placebo with psychological reinforcement) and active (off-label prescribing, etc) to start with. There could have been a larger sample size to reduce the possibility of noise. More importantly, though, the results of the survey could have been presented more honestly, without the spin and headline-grabbing tactics. Sure, this wouldn’t have caught the media’s attention quite so much, but it would have been a more valid representation of placebo use and more useful for future research. Unless of course, that was the intention all along?
Scott Gavura has written a similar piece on the Science-Based Medicine blog. It’s worth a read.