Valerie Beral: evidence is the challenge in medicine

‘Evidence-based medicine’: it’s been a holy grail in healthcare for many years. It doesn’t always stop doctors preferring their own instincts over boring tests and, say, sticking with a diagnosis of migraine until the real problem, a brain tumour, becomes untreatable, as happened to a friend of mine recently.

Generally, though, there is a growing sense that things ought to be done differently. In particular, public health medicine seems convinced of the need to collect and analyse data rigorously. So I was fascinated to listen to the BBC Radio 4 programme The Life Scientific this morning, because it centred around Valerie Beral, now Head of the Cancer Epidemiology Unit at the University of Oxford and Cancer Research UK.

Valerie Beral:
intelligent use of evidence to challenge thinking in medicine

Beral is best known for having been the main voice linking Hormone Replacement Therapy with increased probability of breast cancer. Her position earned ferocious criticism but over a million women have come off HRT over the last ten years, while regulatory body guidelines continue to recommend keeping the dose of HRT down to the minimum and taking it for the shortest time possible. 

In Beral’s view this has avoided 10,000 additional cases of breast cancer over that time.

What fascinated me was her attitude towards healthcare data analysis.

First of all, she believes it matters. She first made her name working on the contraceptive pill, and she got into the subject while working at a Family Planning clinic. Women would constantly ask her what long-term effects the pill would have; she could only answer, truthfully, that she didn’t know. So she decided to find out, not just about the pill, but about women’s health and, indeed, about public health generally.

Secondly, she’s clear about the need for certain standards in analysing data. Years ago a colleague of mine published tables comparing cardiac surgeons. His tables showed that the worst-performing had a mortality rate of 33.33%. Sadly this figure was based on a sample of three patients, one of whom died. Converting that kind of figure into a percentage is meaningless. The word ‘meaningless’ was the least offensive used by the surgeon in question.

It isn’t enough just to be rigorous about sample sizes, though. Even if you’re working on meaningful numbers (and Beral set up the ‘million woman study’), you need to be careful about any correlations you think you’ve found.

A classic example came from a University of Pennsylvania study published in no less prestigious a publication than Nature in 1999. It found that children sleeping with their light on are more likely to be myopic. Only a review of the data established that myopic parents are more likely to leave the light on in the child’s bedroom, and children of myopic parents are more likely to be myopic themselves. So the two phenomena weren’t directly related but independently linked to a separate common cause, the myopia of the parents.

The third aspect that I found refreshing follows on from the previous one: Beral also points out that you don’t actually have to propose a mechanism for the cause behind a correlation. She gives the example of the link between smoking and lung cancer: no-one can explain how smoking increases the risk of lung cancer, but the evidence for the link is undeniable, and barely anyone denies it any longer.

Interestingly, the programme mentions that when that link was first shown, the British government tried to hide it, so as not to cause concern. Clearly, it isn’t enough only to collect evidence, you have to be prepared to do something about it.

That can often be uncomfortable. One of the cases mentioned by Beral, that I found most interesting was that of the far higher incidence of breast cancer in the richer nations than in the developing world. She points to evidence that the risk of contracting breast cancer falls by 10% if a woman gives birth at around 20 and breast feeds her baby; it falls by a further 10% for a second child.

Does that mean that our societies have to switch to encouraging much younger childbirth and much longer breast feeding? Not at all. As Beral makes clear, childbirth and breast feeding must be generating hormones that are giving these young women their relative immunity to breast cancer. Research should be identifying which hormones are in play and a means of delivering them without necessarily going through the pregnancy.

The problem is that the research needed to produce this result doesn’t fit with the pattern of medical research funding we have adopted in developed nations, and which is focused on three or five year programmes. The research Beral is calling for would take more like ten years.

This is where listening to the evidence becomes particularly powerful: when it challenges some of our fundamental habits and assumptions. And that’s when it becomes most exciting.