Friday, 20 April 2018


In my last article I launched the CPL Nanny State Annual Awards for Fake News and Phoney Statistics. Several people contacted me and asked was this tongue-in-cheek, and did misleading statistics really matter, given the environment in which we live where information on social media sites like Twitter have a shelf-life attuned to the concentration span of a mouse. The answer is that whilst our awards night will be light-hearted and enjoyable, the impact of fake news and phoney statistics can be very serious indeed.

As an example, take the now-infamous quote from Professor Dame Sally Davies, England’s Chief Medical Officer of Health: “There is no safe level of alcohol consumption.” And to illustrate the point she asks women to consider: Do I want that glass of wine or do I want to raise my own risk of breast cancer?” Notwithstanding the fact that she meant to say: “Do I want that glass of wine or do I want to avoid raising my risk of breast cancer?” it is the alarmist and misleading nature of the warning that is so damaging.

I am very sensitive to the emotional nature of anything connected with cancer, because it touches the lives of almost every family in the country, including my own. But let me explain why Professor Davies’ quotes are so misleading and why they raise women drinkers’ concerns unjustifiably.

Firstly, the implication is that one glass of wine a day will raise a female drinkers risk of breast cancer, and to most people that means raising their risk of dying from it because in the popular imagination cancer = death. The good news is that If you are a middle-aged woman you have just a 0.3% chance of contracting and then dying from breast cancer over the next decade. And this chance is the same whether you drink alcohol or not. There is slightly more chance that as a light-to-moderate drinker your eventual and inevitable death will be from breast cancer, but only because light-to-moderate drinkers are much less likely to die prematurely from heart disease, and you must die from something, eventually.

An analysis of various causes of death of middle-aged and elderly Americans (Thun et al 1997) found that, of the 251,420 women in the study, 0.3% of the non-drinkers and very light drinkers died from breast cancer, over the 10 years of the study’s duration. And the same percentage applied to the moderate to heavy drinkers. These were women drinking 1 to 4 standard drinks a day. In America a ‘standard drink’ is about 14 grams of undiluted alcohol, so approximately 1.4 units. So, a woman drinking 1 to 4 standard drinks is drinking between 1.4 to 5.6 units of alcohol a day and has the same chance of contracting and dying from breast cancer as that of a non-drinker.

In a smaller mortality study (35,000 women, Fuchs et al 1995) the chance of death from breast cancer during the 12-year follow-up period was 0.4%, and this was identical for non-to-light drinkers and moderate-to-heavy drinkers. And again, crucially, there was a much higher chance of death from cardiovascular disease among the non-drinkers.

Reading all this confused me somewhat. I was sure that there were numerous population studies that showed drinkers had a one percentile higher chance of getting breast cancer (from just under 10% lifetime risk to just over 10%), and there are, but these are studies of “incidence” (how many women are diagnosed) not studies of mortality (how many women die). So, women who drink alcohol have a slightly higher chance of being diagnosed with breast cancer, but if they are, they have slightly less chance of dying from it. That didn’t seem to make sense. So, I did a bit more research. It turns out that around 90% of women who are diagnosed with breast cancer don’t die from it, partly because medical science cures many of them, but also because many of the diagnoses are incorrect. Research also shows that women who drink alcohol are more likely to screen more for breast cancer than non-drinkers (Mu and Mukamal 2016), so you would expect them to be disproportionately represented in the diagnostic figures arising from screening activity. So, the link between drinking and the incidence of breast cancer is a statistical correlation, not a causal link – it is, in fact a self-fulfilling prophecy.

So, here’s what Professor Dame Sally Davies should have said: “When asking ‘do I want that glass of wine, or should I not drink at all’, women should balance a slightly elevated risk of getting breast cancer, which may be explained by the fact that drinkers are more likely to screen for it, with the actual elevated risk of cardiovascular disease associated with not drinking at all.”

This is why straight statistics matter.

No comments:

Post a Comment