When the sums don’t add up

January 15th, 2004 by Ben Goldacre in bad science, scare stories, statistics | 1 Comment »

When the sums don’t add up

Ben Goldacre
Thursday January 15, 2004
The Guardian

Like most people who know the first thing about science, I don’t usually bother to read scare stories in the media, as I know they’re not going to tell me anything useful. Parabens from deodorant in breast cancer cells? That sounds interesting. Oh, you haven’t measured it in normal cells yet. Thanks for wasting my time. Four minutes poorer, life goes on.

And the stories always fit that anti-science, anti-government, anti-industry, anti-medical agenda, as if they were the same thing, with shared blame. But here’s some real news: we’re not doing so badly. In the chemical industry, red-list discharges – releases of the most noxious substances – have fallen by 95% over the past decade.

So I’ll tell you what I want to know from a scare story, or any science story. If it’s about an experiment, I want to know what the experiment was, what the results were and where it was published. That’s not so much to ask. If it’s a story about risk, based on population data, I want to know the number of people looked at and the statistical significance of the data – the figure that tells me how reliable the data is. That’s really not unfair.

Look at the crimes of scare stories. They extrapolate from speculative laboratory data pretending the population data really exists. Or they leave out key information. So this week, eight newspapers reported a possible link between pancreatic cancer and aspirin in women. Most managed to report fairly accurately that it was a study of almost 90,000 women, that those who took two 325mg tablets of aspirin a week for 20 years had a 58% increased risk of pancreatic cancer, and that in women who had taken two tablets per day, the risk increased to 86%. Although only the Times and the Daily Mail reported that the total number of cancer cases was 161 over 18 years. Crucially, only the Times made it clear the 86% increase was only associated with high consumption for at least four years.

Perhaps the public have got their heads screwed on anyway. It’s been shown that scare tactics have little long-term effect on public behaviour: that’s why all those well-intentioned lies about illegal drugs never worked. If the media are to be believed, the day after the salmon story, fish were lying unsold in the ice of Smithfield market, but less than a week later sales are back to normal, just like beef and baby milk.

It looks like the public, unconsciously, apply what those who formulate public policy informed by science call “the precautionary principle”: act cautiously, over a speculative finding, until further data fails to back it up. We stop buying salmon for a day or two on the one in a thousand chance that we’re not being misled, shrug, and get on with our lives. As Bill Hicks once said: “I’ve got news for you: non-smokers die every day.” Two thousand of you will die this year from accidents at home; 200, I’m afraid, will be assaulted and then die; 500 will be accidentally poisoned. Tie your shoelaces, check the milk, and stop looking at my bird, if I were you.

Is there a link between breast cancer and antiperspirant?

June 19th, 2003 by Ben Goldacre in bad science, scare stories | 2 Comments »

Is there a link between breast cancer and antiperspirant?

Ben Goldacre
Thursday June 19, 2003
The Guardian

This scare started in 1999 with a circular email claiming that toxins were “purged” through perspiration, and that when the armpit sweat glands were blocked, toxins built up in the lymph nodes behind them, causing cancer in the upper outer quadrant of the breast.

While it’s true to say that an excess of cancers occur in this quadrant of breast tissue – the one closest to the lymph nodes – it is also by far the largest area of breast tissue.

Furthermore, sweating is certainly not the primary means by which the body rids itself of toxins. But the email caused so much concern at the time that the American Cancer Society and the National Cancer Institute issued statements to reassure people, and scientists began to study the issue.

If you really want to get to the bottom of whether an environmental factor is causing an illness, it’s good to be able to compare one group who were exposed to your toxin, antiperspirant for example, with a group who weren’t. In 2002 a large study was published looking at 1,600 women and found no link between antiperspirant use and breast cancer. It was testament to the influence of the original “hoax” email (as Harvard medics called it) that they also went out of their way to study its specific claim that shaving before using antiperspirant increased the chances of it causing cancer, because toxic chemicals could get in more easily, and again found no link.

This week the story rose again when Dr Philippa Darbre, a molecular biologist from Reading University, published a speculative hypothesis paper, in a small but respected academic journal, describing the effects of some of the ingredients of antiperspirants on non-living tissue in the laboratory, in other peoples’ experiments. Specifically, that aluminium has been reported to at least bind to DNA, and also to alter gene expression in some situations, and that parabens (preservatives used in some cosmetic creams) have been reported to work similarly to oestrogen in some experiments.

This has certainly generated more media interest than any of her other more eminent papers on, for example, IGF-II receptors and breast cancer, although she tells me that she never sent out a press release, but happened to mention it to a friend who worked for the Sunday Times.

Science is all about testing interesting hypotheses. In this case we have an interesting but half-finished laboratory-based story against a convincingly large study of real people that showed antiperspirants to be safe.

Unfortunately, it looks like the largest risk factors for breast cancer at the moment are the ones we can’t control such as age, sex, and family history; or the ones we might not want to such as smoking and age at childbirth.