'9.2% of UK bladder cancer caused by tap water'

A story in many news sources and spreading on social media claims that drinking water disinfected with chlorine is responsible for up to 1 in 20 cases of bladder cancer across Europe. The authors claim the number is as high as 1 in 5 in some countries.

Really? How did we not know this?

Using chlorine to ensure water is free from Our Microbial Overlords and safe to drink is commonplace but not ubiquitous. And it leads to ‘disinfection by-products’ (DBPs) – among the most common of which are trihalomethanes (THMs).

The institute behind this study has tried to sell this link in the past in another paper that tried but failed to link trihalomethane exposure with bladder cancer. That paper also generated some scary headlines but didn’t tell us anything about bladder cancer risk; in fact all that it told us was that different people might behave differently with respect to how they use water.

So far so crap.

The New Study

So, the authors gathered data on trihalomethane levels in drinking water across 26 EU countries. They found that the average exposure was far below the maximum permissible limit, but in nine countries the maximum limit was exceeded.

In the paper they attribute just under 5% of all bladder cancer across the European Union to trihalomethane exposure, and of those cases they suggest 44% could be avoided if trihalomethane levels were reduced in all the 13 countries that exceeded the limit.

They came up with this figure by investigating total and individual trihalomethane concentrations in drinking water in all 28 EU countries and then, well, pulling some major assumptions out of their arses.

And as usual the media have reported this without even cursory scrutiny or challenge.

Are their conclusions correct?

There are many flaws in this study. For a start it estimates the number of bladder cancer cases that trace chemicals in drinking water might cause, but it does nothing to show whether chemicals such as trihalomethanes can actually cause bladder cancer in the first place.

Studies going back over 40 years have failed to establish a causal link between trihalomethanes in drinking water and bladder cancer but the data is messy. There’s a lot of noise and not much signal.

Yes, it has been shown that disinfection by-products can cause cancer in vitro (doesn’t everything) but whole animal studies have not shown any carcinogenic effect. And yes, the IARC lists trihalomethanes as possible carcinogens – just like it does everything else.

Back to the study. The lowest levels of THMs in drinking water were found in Denmark (0.02µg/l), the Netherlands (0.2µg/l), and Germany (0.5µg/l). The highest levels were found in Cyprus (66.2µg/l), Malta (49.4µg/l), and Ireland (47.3µg/l).

One would therefore expect to see a correlation with actual incidence of bladder cancer in these countries – unless this paper is Utter, Utter Bollocks (µ²β).

The problem is Denmark, the Netherlands and Germany – with negligible THM levels in drinking water – are 3rd, 5th and 9th respectively in the global ranking for bladder cancer incidence, and those with the highest THM levels are waaaaay down the list for bladder cancer ranking.

Which kind of flies in the face of the trihalomethanes / cancer causal link the authors imply and the media seize upon, obviously.

Hang on, buried in the discussion at the back of the paper is a very telling phrase:

“However, assuming a causal association, current levels in certain countries could still lead to a considerable burden of bladder cancer that could potentially be avoided by optimizing water treatment, disinfection, and distribution practices, amongst other possible measures.”

…”assuming a causal association“. That’s a pretty fucking massive assumption that pretty much contradicts everything else we know about trihalomethanes and the epidemiology of bladder cancer.

This smacks of Tooth Fairy Science. One could mount a massive study of how much money the Tooth Fairy leaves, the impact of sibling succession, tooth succession, socioeconomic and geographic factors – and come up with a highly-powered statistical model for Tooth Fairy behaviours.

But none of that would make the Tooth Fairy any more real.

How do we know this is likely very wrong?

In this paper they have assembled a large dataset. And if you interrogate the data hard enough, it will admit to anything you want – which it certainly seems to have done here.

The thing is, trihalomethanes were first detected in drinking water in the early 1970s. A discovery that resulted in many studies, some of which found correlations, some didn’t – but no causation. As I say, it’s messy. The data has a lot of noise and not much signal, which probably goes a long way to explaining the conflicting studies.

Also, due to increased scrutiny and regulation, THM and other disinfection by-product concentrations have declined over the decades, but bladder cancer incidence in many countries hasn’t. Certainly not substantially.

So, the risks from trihalomethanes might not be zero, but if risks exist they must be pretty small compared to the many other known bladder cancer risk factors.

Known risk factors

Let’s look at some other numbers. And the more established causes of bladder cancer.

  • Average US trihalomethane levels have declined from about 67μg/l in 1975 to about 30μg/l in 2015. Bladder cancer rates for both men and women – while different – were essentially unchanged during that 40-year period. Canadian results are similar.

  • In the UK, THMs have declined, as has male smoking, but female smoking less so. Male lung cancers have declined while female lung cancers have increased. Bladder cancer has declined, more so for males than females. Spooky.

  • The Netherlands – which stopped chlorinating drinking water in 1993 – shows a small bladder cancer decline for males and slight increase for women. Smoking rates and lung cancers declined for men but increased for women.

Do trihalomethanes cause cancer?

  • The idea that disinfection by-products are responsible for up to 20% of bladder cancers would mean we’ve been totally wrong about smoking and other proven risk factors for decades.

  • It’s inconvenient – to say the least – for the authors that their hypothesis does not tally with real-world incidence statistics for bladder cancers.

It has been shown trihalomethanes in drinking water are not carcinogens in whole animal tests – but it is still a concern. Also the aetiology of bladder cancer is complex. What we do know is incidence correlates with factors including age, smoking, Type 2 diabetes, gender and ethnicity. But smoking is the predominant risk and that link is well-established.

[There was a meta-analysis a few years back that showed a correlation between high doses of THMs – higher than found in potable water – and bladder cancer in men – but not women. But it didn’t control well for other factors. As I say, there’s lots of noise but not much signal.]

So, any link between THMs in tap water and bladder cancer is questionable – and if there is an effect it is likely to be vanishingly small, both in terms of absolute risk and relative risk compared to the other well-established risk factors.

It is plausible that THMs and other disinfection by-products could be harmful – and so it is entirely sensible to control them in potable water – provided it’s not at the expense of protection against Our Microbial Overlords. Cholera, anyone? Yes, quite.

I hated doing statistical analyses as an undergraduate – they made my head hurt – and I’ll cheerfully admit the fine detail of some of the statistical techniques in this paper are beyond me. But that’s not the point.

Basically this statistical analysis doesn’t stack up with what we already know about disinfection by-products and bladder cancer – and simply waving away these major flaws with:

“…assuming a causal link…”

…smacks of hubris or delusion. Their whole premise – and all the lazy, credulous reporting – is holed below the waterline by those three words. Their analysis and recommendations rely on a causal link where none has been shown to exist – and even then the observable data does not tally with their premise.

I don’t think for a moment the authors approached this with malfeasance in their hearts, no conflicts are declared and they don’t appear to be funded by Big Mineral Water – they just seem to have started by assuming their conclusion. Classic begging the question.

And yes, I know the link between smoking and lung cancer was proved before any known mechanism was elucidated and John Snow sussed out cholera with pure epidemiology before germ theory even existed – but I don’t think this paper will be judged by history as being in that league. Sorry.