I don’t know about you, but news headlines with the phrase “early death” tend to capture my attention much more than, say, those with the phrase “second quarter economic forecast”. Here’s one from the Washington Post, warning us against all sodas–with and without sugar. Either way, they lead to our premature demise.
Salon, in its coverage of the study, decided that a picture was worth a thousand words. I think we get it:
You may be wondering:
I think the answer is “no, not really”. But of course that’s not the whole story. Why?
The real and longer answer is “we don’t know, based on the results of the study”. So, let’s take a look at that study and the results.
Here’s the Washington post summary:
The study, one of the largest of its kind, tracked 451,743 men and women from 10 countries in Europe. It found that consumption of two or more glasses of artificially sweetened soft drinks a day was positively associated with deaths from circulatory diseases. For sugar-sweetened soft drinks, one or more glasses a day were associated with deaths from digestive diseases, including diseases of the liver, appendix, pancreas and intestines.
The researchers recruited people from Britain, Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain and Sweden between 1992 and 2000, surveying them on their food and drink consumption. Participants were excluded if they reported incidents of cancer, heart disease, stroke or diabetes. Mean age was 50.8, and 71.1 percent of participants were women.
This sounds fairly clearly true, as reported by the Post. So why am I skeptical?
Any big population epidemiology study on nutrition and mortality risk that states some categorical claim is automatically going to raise my antennae. It’s notoriously hard to do this kind of research and get clear results. Why? Think about it for a moment: trying to isolate the contribution to better or worse health over time of the occasional intake of some substance for a whole population is incredibly difficult. For some substances, like tobacco, it turns out it’s pretty easy to establish that it’s harmful in just about any amount. But for, say, alcohol intake, when we graph the risks for drinking it, we get what’s called a J-curve. It looks like this:
What we see here is that, as alcohol consumption increases from zero to about one drink a day, the risk of death drops, and then starts to increase slowly. Looking at it without any other information, you would think that drinking up to say, 5 drinks a day is actually beneficial for your health. Pro health tip: it’s not. But there’s a lot more needed than just this data on this graph in order to understand alcohol consumption and health. If you’re interested in more about J curves, public health and what to do when you get one, look at this nice JAMA article (which is where the graph came from).
Back to the soda study: when we look carefully at the data presented here in the article, what we see is another J-curve. That is, the people in the study who drank between 1 glass a month and 6 glasses a week of soda (either sugar-sweetened or artificially sweetened) had a LOWER risk of all-cause mortality than those who drank less than one soda per month. The risk was a bit higher for those who drank 1–2 glasses of soda (either kind) per day and then higher again for the group who drank more than 2 glasses per day.
When you look at their disease-specific mortality risk data, results are mixed, but there are still J-curves. For women who drank between 1 glass a month and 6 glasses a week of soda, their cancer risks were lower. For men, the risks were mixed, but not appreciably higher. Ditto (almost) for heart-related disease risks (it’s a little more mixed, but a similar pattern).
I would like to thank both Samantha for pointing me to sources and Health Nerd on Twitter for his very detailed analysis of the study (which saved me some time and directed me to particularly interesting data patterns). Here’s the J-curve we see in this study, compliments of Health Nerd):
Does this mean we should all start drinking a little soda a day, to keep the grim reaper away? NO. What this means is that we need to look at the data a lot more closely. Health Nerd points out that the researchers didn’t control for income or ethnicity. We know from various studies that soda intake is higher among Latinx and African-American populations than among White populations, and it’s also higher among lower-income populations. We also know that there are major health disparities in the US both for lower-income and for non-white populations. So, is the soda playing some role in the health of these populations, or not? This is a very important question, and researchers are working on it. But it’s not easy to find a clear answer. And this JAMA study is not, in my view, giving us one.
Lots of people (including me) believe that sodas (either kind) are unhealthy-for-them (according to their views of what healthy-for-them means). There is some scientific evidence that supports this view. What headlines like this do, though, is provide an excuse to scare people and to demonize or blame those who drink soda for whatever happens to their health over time. Looking more closely at what drinking soda means to people seems a good place for public health to start, if the goal is to help people make their own lives healthier-to-them.
Here’s to your health!
Readers– what do you think about soda? Is it a daily part of your life? An occasional drink? Totally off your radar? Do you think it’s the devil, or the pause that refreshes? I’d love to hear from you.
I wish that your post and the JAMA article included a reminder that a statistical connection is not the same as a cause/effect connection. While a study like this one is very interesting (and I agree, not at all a good basis for public policy), and may suggest that soda is the benefactor/culprit, all it really does is suggest that a closer look may be worthwhile.
Both you and JAMA point out that certain major other possible factors are not controlled for. But even controlling for them would not be enough. To isolate the cause of different outcomes, one must also isolate the factor being studied. So if the study began with its large group (with, on average, varying ordinary levels of all readily identifiable characteristics including income levels, general health, dietary habits including soda consumption) and had defined subgroups drink from zero to lots of either diet or regular soda, and results showed significant differences between the groups, we would know why. Now, J curve or not, we just plain don’t know.
It worries me that we are bombarded with this sort of statistical connection constantly, and are often lured into making behavioral choices from its sometimes impressive numbers.
We could publish a study showing a 100% link between aging and greater chance of death. The older you get, the more likely…Ok, that one is patently absurd, but many – probably most – of the statistics we read are equally unlikely to give us a reason to believe that a very specific change in our behavior will lead to the outcome suggested by those statistics.
HI Winnie and thanks for your comments. Yes, you are right– this was a retrospective observational study, finding only associations. We’d need a completely different study design to get much stronger evidence to justify potential causal claims. And even randomized controlled trials can get it wrong with respect to causal connections. I’ve written about this in the past on the blog, but it’s certainly a good idea to keep making the point, as we are indeed bombarded with false causal claims to try to scare us into trying to change our behaviors (and then blaming ourselves when those don’t stick).
What about soda soda, i.e. completely unsweetened soda water / carbonated water (with or without flavouring)? Does not even seem to be on the research radar….