Yet again the public is being subjected to warnings about the potential health risks of consuming a common food item based upon insufficient evidence. Last month it was oat products, and now it’s sweetened drinks.
The study is a prospective cohort study, which means it is observational. The researchers looked at over 200 thousand participants in the UK biobank. At the start of the study they did not have a condition known as atrial fibrillation (A-fib), and by the end of the study there were 9,362 cases. They found that the risk of developing A-fib was 20% higher among those who drank 2 liters or more of artificially sweetened drinks, and 10% higher among those who drank a similar amount of sugar sweetened drinks. They also found that risk was 8% lower among those who drank unsweetened pure fruit juice, like orange juice.
One author states:
“Our study’s findings cannot definitively conclude that one beverage poses more health risk than another due to the complexity of our diets and because some people may drink more than one type of beverage,” said lead study author Ningjian Wang, M.D., Ph.D., a researcher at the Shanghai Ninth People’s Hospital and Shanghai Jiao Tong University School of Medicine in Shanghai, China. “However, based on these findings, we recommend that people reduce or even avoid artificially sweetened and sugar-sweetened beverages whenever possible. Do not take it for granted that drinking low-sugar and low-calorie artificially sweetened beverages is healthy, it may pose potential health risks.”
And this is where I have to strenuously disagree. I don’t think this one study is sufficient to make sweeping recommendations about diet. The quote above misses a major reason why we cannot make conclusions about actual risks from this data – it is not controlled. This is observational data, so we cannot make conclusions about cause and effect (something the authors do acknowledge, although then go on to make sweeping diet recommendations).
All the caveats of uncontrolled, self-reported, observational data apply. There are many social and health factors that influence our diets. People in the different groups were not comparable, and it was not really possible to adjust for all the potential confounding factors. They did adjust for genetic risk, but that’s it. They noted, for example, that those who drank more sweetened drinks tended to be more overweight, have other cardiac risk factors, and be of lower socioeconomic status.
That is all the generic limitations of large observational studies – you cannot control for all confounding factors. As the authors note, the confounding factors with something like diet are massive.
But I also had some specific questions and concerns about this study. First, it strikes me as odd that there would be similar risks, as least in the same direction and similar magnitude, for both artificially sweetened low calorie drinks and sugar sweetened drinks. Other than the fact that they are sweet, what would be the similar mechanism here? Sugar and low-calorie sweeteners are very different things. This suggests to me that, if there is an actual risk factor for A-fib here, it’s not the sweetener. It’s something else in the beverage, something not present in fruit juice.
The first thing that occurred to me is so obvious I thought – they must have controlled for this. Then I got to the line, “It is also unknown if the sugar- and artificially sweetened drinks contained caffeine.” To me that makes these study results difficult to interpret, to say the least.
The question of the relationship between caffeine intake and A-fib risk is a complex and interesting one. There is no simple answer at this time. Most of the research involves coffee intake, and coffee contains about 1000 compounds some of which are potentially cardioprotective. Low levels of coffee intake may even have a protective effect on A-fib risk, while higher coffee intake is neutral or may increase risk, depending on the methodology of the study. In the aggregate it’s a wash, without any apparent risk.
However, again, these are mostly from observational studies with lots of confounding variables. What I could not find are any studies looking specifically at caffeine intake from soft drinks. A-fib has more clearly been linked to energy drinks, which generally have higher caffeine and sugar content than soft drinks.
This study could have been an opportunity to look specifically at caffeine intake from soft drinks and A-fib risk, but it did not.
There is also another huge potential confounding factor. Whenever studying diet and trying to link a health risk to the consumption of a specific food, it must be considered that food intake is mostly a zero-sum game. We can only eat and drink so much, and every time we choose to consume one thing we are also choosing to some extent not to consume another. Therefore we need to ask – for those who were consuming large amounts of sweetened drinks, what were they not drinking? Fruit juice appears to be protective in this study, so is it, perhaps, that people who are drinking lots of soft drinks are just missing some of the nutritional benefits of consuming fruit juice?
In the end, this study, which is still useful data, at best can be used to generate hypotheses. But this kind of data – observational with massive potential for confounding factors – cannot be used as the basis for making any conclusions about cause and effect, and therefore should be viewed with caution. I do not think it should be used to make bottom-line public health advice.
Ultimately I find the results difficult to interpret and apply to daily life. But this can be used to design some follow up studies that are better controlled.
There is also the potential here for significant unintended consequences. Making big diet recommendations based on observational or preliminary data causes “alarm fatigue” in the public. If every week they are hearing about another thing they are supposed to avoid it all becomes impossible to manage. As I have written before, just about every food and ingredient has been linked to some health risk or benefit.
The public also loses faith in the medical establishment when such recommendations are constantly shifting or appear to be conflicting. This means they are more likely to ignore solid health advise when it is available.
The obvious solution here is to significantly raise the threshold for what constitutes sufficient evidence to put out a press release reporting health studies, and not to trumpet preliminary or ambiguous health evidence. Also, researchers and journalists need to be careful about how they frame such reports. You should not blare warnings in the headlines and opening paragraph, then bury the caveats deep in the article, where almost no one will read.
The signal to noise ratio is really low in mainstream health reporting. We need to significantly reduce the noise, so that the important signals can get through.