It’s my own question, this time, and throws up lots of annoying little things.
The problem is straightforward: I’m a paediatric oncologist in my spare time, and was asked about the use of LifeMel honey to prevent infections.
Now, being both an EBMer and a Physician, I said I didn’t know, but didn’t think it could do, but might cause problems. (Remember we’re not supposed to give honey to infants ’cause of botulinism? Same concerns.)
Running a really simple PubMed search (“honey” and “infection” and “cancer” – to get rid of the leg ulcers) give 7 hits which are irrelevant. I tried again, ’cause I’d been told the website cited a real paper published somewhere … “life mel honey” gets a single hit. Looking at the Medline citation, where the index terms are listed, made me think that “honey” and “neutropenia” might be better – but still the single study.
The study is a cohort of 30 adult patients who’d previously had neutropenia. They then got given 5ml honey/day with the next course of chemo, and about 12 (40%) didn’t get very badly neutropenic this time, and 8 (27% – not 36% as quoted in the abstract) thought their quality of life was better.
Now ask yourselves a question – do you think the honey made a difference?
If it hadn’t been honey, but GCSF that had been given, would your answer be the same?
(I think most people would have a different answer if it’s honey or G, and the rest of this presumes that as a fact…)
The interesting points:
Firstly, I think that despite the claim we make to believe RCTs and p-values, most of us are inherently Bayesian: we move forward with new information not from an unbiased position but one of knowledge (or at least assumed knowledge). I come to the question with an assumption that honey might at the best have a 5% chance of doing something. And so I am not shifted at all. But is that reasonable, right and scientific?
Second, I am on the whole a “shruggie” when it comes to complementary therapies (as expounded by Val Jones). And I have no hard data for the botulinism argument. But at £37.50 for 120ml it seems a lot for a family who struggle to afford the transport to hospital to be spending (or guilty about not spending).
Before I go on, I’d like to note that the UK distributors of Life Mel make no claims of medicinal value for their product on their website, in keeping with UK and EU laws, and I’m not claiming in any way that their actions are illegal. The references to anything active – feeling better etc – are carefully silo’d into a well-linked sister website.
But back to the point – the questions I’d like to hear your answers on:
- Can we take a Bayesian approach to treatment like we do with diagnostics?
- Should poor quality benefit and harm data cancel each other out? Or how should they be balanced?
- Do adult data and customer feedback have any relevance to paediatric oncology patients?
Ben Goldacre’s BadScience mini-blog for the link to SBM