May 17, 2019

expert reaction to study looking at sugary drinks, fruit juice and risk of premature death in US adults

Research published in JAMA Network Open suggest higher consumption of sugary beverages, including fruit juice, is associated with increased mortality. 

Dr Alison Tedstone, Chief Nutritionist at PHE, said:

“This study is a reminder that consuming sugary drinks can contribute to dental caries, increased calories, weight gain, and associated ill health.

“Current advice is to swap sugary drinks for water, lower fat milks and lower sugar or diet drinks.  While fruit and vegetable juices can contribute to one of your 5 A Day, it’s important to limit juice and smoothies to a total of 150 mls each day as they can still contain high amounts of sugar.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This is a good study that is competent statistically, and I’m reasonably convinced by its findings that consumption of sugary drinks is associated with increased death rates, and that this applies whether the drinks are sweetened with sugar or are 100% fruit juice.  However, the results are rather complicated, and there are several important limitations that you should be aware of.

“I’ll begin with the limitations.  First, the study was observational.  That is, people weren’t allocated by the researcher to drink particular amounts of sugary drinks – they just drank what they would have drunk anyway, and the researchers asked them about that and recorded deaths that occurred.  So, inevitably, there were other differences between people who drank large and small amounts of sugary drinks, apart from what they drank – for example, differences in age, or in the other things they ate and drank, or in education, or in physical activity levels.  These other differences, called confounders, might be the cause of differences in death rates, rather than the differences in sugary drink consumption.  It’s possible to make statistical adjustments to allow for these differences, and the researchers did that quite comprehensively.  These adjustments made quite big changes in their detailed findings, which demonstrates the important role that confounders have in this study.  But you can’t adjust for differences that you have no data on, or that you just don’t know about.  So we can’t be completely sure about what causes mortality differences.

“Then there are issues about the sample that was used.  It was originally put together for a study of strokes, so participants were particularly drawn from areas in the US where strokes are particularly common (the so-called ‘stroke buckle’ and ‘stroke belt’).  Although half the participants came from other parts of the USA, there may still be doubt about how representative the sample is of the whole US population, let alone the populations of other countries.  This concern is heightened by the fact that nearly one third of the original participants did not provide data on their food and drink consumption, so could not be included in this study – perhaps those who did provide food and drink data weren’t even typical of the sample as a whole.

“But, most importantly for me, the sample wasn’t very big, in terms of the numbers that matter.  Yes, measurements were obtained from over 13,000 participants, which sounds like rather a lot, but the key numbers in a study of death rates, like this, are the numbers of participants who died.  Because the follow-up time wasn’t terribly long – 6 years on average – only 1000 of the participants died in all, and only 168 died of coronary heart disease (CHD).  As a result, the estimates of difference in death rates are not very precise, particularly for CHD deaths.  This is why the researchers are calling for new studies with more statistical power (so that the associations can be estimated more accurately) and longer follow-up.  This new study can’t be the last word.

“What about the detailed results?  One important point is that they are summarised rather strangely in the abstract of the research report, and (particularly) in the ‘Key points’.  Basically, the writers chose only to mention, in these summaries, results that were statistically significant, implying that other results either aren’t of any interest or that the lack of statistical significance means that there was definitely no association.  That just isn’t the case.  In particular, though the results for sugar-sweetened drinks taken on their own are not statistically significant, in statistical terms they are actually not much different from those for all sugary drinks, or for fruit juices on their own.  The statistical significance, or lack of it, isn’t the main issue here.  Also, despite the Key Points saying that “similar associations were not observed for sugary beverage consumption and coronary heart disease”, in fact the associations for deaths from CHD are of comparable size to those for deaths from all causes.  The reason why the associations are statistically significant for all-cause mortality and not for CHD could be because there really are no associations with CHD mortality, but it could also be simply because there were not many CHD deaths so that the sizes of these associations are not very precisely estimated.

“Finally, it might make more sense of the figures for relative risk increases (hazard ratios) in the research report if I convert them to changes in life expectancy.  The research reports that drinking an extra 12oz of sugary drinks daily (sweetened with sugar, or fruit juices, taken together) is associated with an 11% increase in all-cause mortality risk.  If someone drinks the extra 12oz every day throughout their adult life, that corresponds to a reduction in life expectancy of roughly one year – but again, that’s only if the relationship really is causal, and we can’t say that it is from the study.  But the 11% figure is subject to quite a bit of statistical uncertainty (because there were not many deaths).  Plausibly, it could be anywhere between an increase of 3% or and increase of 19% in risk.  A 3% increase corresponds to a reduction in life expectancy of about 4 months, and a 19% increase to a reduction of around 2 years.  So these aren’t huge changes compared to, say, the effects of cigarette smoking (where 20 a day throughout your adult life knocks something like 8 years off average life), but nor are they negligible.

“In thinking about this, though, it’s worth noting that 12oz daily of sugary drinks is quite a lot.  It corresponds to 355ml.  The current NHS recommendation (see https://www.nhs.uk/live-well/eat-well/water-drinks-nutrition/) is that we should drink a combined total of no more than 150ml of fruit juice, vegetable juice and smoothies each day.  150ml is one small glass.  The advice is also to keep consumption of sugar-sweetened beverages to a minimum, and that children should avoid them completely.”

Dr Gunter Kuhnle, Associate Professor in Nutrition and Health, University of Reading, said:

“This is a very important study, especially as fruit juices are often seen as a “healthy” alternative to sugar-sweetened beverages, even though they often contain much more sugar (especially smoothies).  Fruit juices can provide vitamins and even some fibre, but there is little health benefit beyond this: the amount of phytochemical found in juices is too low to have any further beneficial effect, and there is no beneficial health effect from so-called antioxidants.

“This observational study has shown that the consumption of sugary drinks – whether they are fruit juices or lemonades – is associated with a modest increase in premature death.  However, this is the first study to compare fruit juices with sugar sweetened beverages, and surprisingly it found very similar associations with mortality for both fruit juice and sugar sweetened beverage consumption (indeed, the risk was higher for fruit juice consumption, but this is difference is too small to be meaningful).

“If the association is shown to be causal (which we don’t know yet), this would have a number of implications: first of all, it would suggest that it does not matter whether sugary drinks are lemonades or fruit juices.  This is important, as fruit juices and smoothies are not commonly perceived as sugary drinks.  Secondly, it would suggest purported health benefits of fruit juices are not sufficient to counteract their sugar content.  Fruit juices are a poor replacement for actual fruit consumption, in particular as they can be much more easily over-consumed.  Indeed, a 150 mL glass of orange juice is made from about 2 oranges – but it takes much longer to eat two oranges than to drink the juice.

“In the UK, the general recommendation is that a 150 mL glass of fruit juice can provide one of the five-a-day, but not more.  This is less than half of the amount found in this study to result in a modest increase in mortality, so there is no suggestion from this study that one glass a day is problematic.”

‘Association of sugary beverage consumption with mortality risk in US adults: a secondary analysis of data from the REGARDS study’ by Lindsay J. Collin et al. was published in JAMA Network Open at 16:00 UK time on Friday 17 May 2019.

Declared interests

Prof Kevin McConway: “Prof McConway is a member of the SMC Advisory Committee, but his quote above is in his capacity as a professional statistician.”

None others received.