Two stories appeared in the news yesterday – one in The Mail and the other, apparently exclusively, in the Express – claiming that scientists had found that drinking coffee during pregnancy raised the child’s risk of leukaemia.
Whilst it’s always good to see solid investigations into potential factors that contribute to the development of blood cancers, I was concerned reading these that the alarmist reporting may be upsetting for parents whose child has had one of these terrible diseases and for many who’ve been affected in the past.
The stories, as reported, do not give proper attention to the complexities of the issue and seem to miss some crucial details.
What do we mean by risk?
There are no links to the actual research (what look like hyperlinks in the Express article are actually commercial advertisements). It was actually a study published online last September (2013) and appeared in print in the American Journal of Obstetrics & Gynecology in February 2014. The study was as a well-conducted ‘meta-analysis’, which is where researchers take data from a few different studies, mesh them together, then re-interrogate to see whether the original conclusions stand.
The Express reports that childhood leukaemia risk “increases by up to 72% for those drinking “high levels” of coffee compared to those who drink little or none”. Yet, unhelpfully, the news report never mentions how “high levels” are defined. The Mail goes into a bit more detail and reports: “babies of women who drank coffee during pregnancy were 20 per cent more at risk of developing leukaemia. But if they drank more than two cups a day the risk rose to 60 per cent...And for women who had four or more cups a day it rose to 72 per cent.”
But crucially, these figures give the relative risk, and not the actual absolute risk. The overall risk of a child developing leukaemia is a fraction of a percent, so even a 72% increase is still a fraction of a percent (see more here). In fact, another confusion that creeps in to the Express article is the suggestions that “about 1,600 children are diagnosed with cancer every year in the UK” – but this research concerns acute leukaemias, which account for 400–500 childhood cancers a year in the UK, so the Express seem to be trying to inflate the problem.
Repeat: "correlation is not causation"
One of the major limitations of the study, which does not get discussed in either news report, is that despite being thorough and well-conducted, these types of studies can only tell us about associations.
An association between two factors – in this case childhood leukaemia risk and coffee drinking during pregnancy – does not say that one caused the other. It could be that a third factor, such as socioeconomic status, type of job or diet, drives both an increase in cancer risk and coffee drinking (for the record, there is no evidence for these – just speculation to illustrate a point). These types of studies try to account for these potentially ‘confounding’ factors, but it does complicate the analysis and strength of the conclusions.
The authors of the research study were well-aware of this – in the last line of the article summary, they acknowledge the “limited studies” in this area and call for further, more definitive studies to explore this possible effect. One important approach to take would be to follow groups of people over time, so that researchers can collect and control for all potentially confounding variables at the start. This also minimises errors that that can creep in when people report themselves what they remember their drinking habits to be or what they drank at some point in the past.
Other questions that sprang to my mind were whether, if the effect turns out to be real, it is coffee drinking or, more specifically, caffeine consumption? How did the studies control for how strong the cups of coffee were, or how they were drunk (with milk, as espressos, etc.)?
These nuances are absent in both the Mail’s and Express’ reporting. That’s not to say the study isn’t not worthwhile, but to draw firm conclusions as these newspaper have done is very premature.
Won't someone please think of the evidence
The Express extensively quotes Professor Denis Henshaw of Bristol University but neglect to mention, unlike the Mail, that he was not involved in the research. There was no input from the authors or experts in epidemiology.
Prof Henshaw claims that incidence of childhood leukaemia has “increased steadily in recent decades”, so infers that there must be environmental causes. But it is not that straightforward. The University of Oxford’s Childhood Cancer Research Group that houses the National Registry of Childhood Tumours point out that the apparent rise in incidence may more reflect our increasing ability to detect and record new cases than a genuine rise in risk.
Prof Henshaw also mentions power lines and magnetic fields, but neither of these has been reliably shown to be linked to the development of leukaemia, despite being the attention of much study (see here and here and here). There is some evidence of effects of natural background radiation, such as radon from rocks and soil, on childhood leukaemia risk, but these nevertheless remain very small and may be affected by bias or confounding factors. In any event, this seems to be a way to tangentially slip in Prof Henshaw’s long-standing advocacy topic. For the Mail to claim that “academics” (note, plural) are calling for a change in health advice is, I think, disingenuous.
The Express reports that … “significantly increases”. This appears to be a misrepresentation of what researchers mean by ‘significantly’ and what the everyday meaning is taken to be. “Significant”, in a statistical sense, tells us whether the observed differences between two groups are "real" or are simply due to chance. It doesn’t necessarily mean that the finding is important or substantial, or even that it is meaningful in the real world.
What does this mean for patients?
Cancer Research UK has written before about the problem of media coverage making people “feel guilty” about their disease and how wrong it is for anybody to feel at fault. It is also important to realise that lifestyle choices are about stacking the odds in your favour, rather than a simplistic ‘this causes that’.
Causes of cancer are typically a complex mix of genetics, environmental influences and bad luck. Blood cancers in particular, by and large, do not have clear, single factors – from either nature or nurture – that substantially raise the risk of disease. To suggest otherwise, especially when the evidence is not clear, is not a useful contribution to informed decision making.
This article is published under a Creative Commons Attribution-NonCommercial-ShareAlike License.
Image by Flickr user:Mortefot [Latte art CC BY-SA 2.0]