Are My Kids Eating Too Much Sugar?

One of the themes of this newsletter is “correlation is not causation.” And one of the places that it comes up most is when I’m writing about diet. I’ve argued repeatedly (see here, and here, and here) that most of our research on what foods are “healthy” is flawed. Because the diet that people eat is so wrapped up in the other behaviors they engage in (smoking, exercise) or their access to resources (education, income), any study that just compares people’s health based on their diet is largely useless. 

This criticism especially applies when it comes to individual foods, or food groups. We do have a bit of data on overall diets. There is one large randomized trial that supports the Mediterranean diet, and, very generally, we have a sense that a diet high in ultra-processed foods probably contributes to obesity and therefore to worse metabolic outcomes. When we try to dial down further than that, though — to look at individual foods (chia seeds?) or even categories (like sugar) — the evidence gets mushy and unconvincing. 

You might ask: Why? Why is this hard? The primary issue is that in order to evaluate the effect of food on health, you actually need to have some random variation. One way to do this is to run a randomized trial. To do this with, for example, sugar, you’d need to randomly allocate amounts of sugar to different people, follow them for a long period of time, and observe their outcomes. If you wanted to know the lifetime effect, you’d need to follow them from birth through their lifetime. At this point, you would be dead, making it challenging to write up the paper. 

That explicit randomization is hard to impossible. But it’s not the only way — the important thing for an evaluation is that there be some random variation in the diet. In some cases, that random variation can be delivered by the world. We call this a “natural experiment.” 

With this long windup: a paper has done just such an experiment, for sugar. The researchers found some (arguably) random variation in early childhood sugar consumption and used that to look at the impacts of childhood sugar exposure on adult outcomes. Here: a deep dive. What did they do? What did they find? Are there any flaws we want to be aware of? And finally, should this change how you think about sugar?

What does this study do?

The paper is here. Any paper relying on this type of natural experiment idea needs … a natural experiment. In this case, the natural experiment the authors rely on is World War II sugar rationing in the U.K.

During and after WWII, sugar in the U.K. was rationed (along with a number of other foods, such as meat and butter). The rationing began during the war and continued until the early 1950s. Under these rations, households were able to consume only a certain amount of those food products. Sugar rationing ended in September 1953. At that point — and this is key to the research — consumption of sugar increased immediately and sharply.  

We can see that in the data below, where I have replicated Figure 1 from the paper. When sugar rationing ends, there is a huge increase in sugar consumption. Notably, there are fairly minimal trends in consumption of other products over this period, even though some other foods (like butter) also came off rationing around the same time.

Importantly for later interpretation: the amount of sugar available to people during rationing was similar to current guidelines for a healthy exposure to sugar. The authors report that adult sugar consumption increased from about 40 grams per day to 80 grams per day over this short period; data for children shows a similar percentage increase. In comparison, current guidelines suggest limiting sugar consumption to 50 grams per day (for adults).

The sharp timing of the change in sugar consumption is an important part of the authors’ approach. The idea is to compare children who were conceived or born just before versus just after this cutoff. The children who were in utero or young during the period while sugar was rationed had less sugar exposure than those who came slightly later. Because the timing of the change was so sharp and so big, the authors argue that any differences they see are likely due to the sugar change and not something else (more on this below).

The data for the study comes from the UK Biobank, which collects a lot of information about the health of individuals. The authors focus on a sample of about 60,000 people born between October 1951 and March 1956, a group that ranged in age from 51 to 66 when they were surveyed. The key health outcomes were whether the individual had been diagnosed with diabetes or hypertension.  

What does the study find?

The headline finding of the study is that chronic disease in adulthood — specifically, hypertension and diabetes — is reduced for children who were exposed to less sugar in utero and in their early years of life.  

To demonstrate this, the authors divide the sample of people based on the amount of rationing they experienced. In their main graph, which I’ve included below for diabetes, they have three groups. The first group was rationed for their entire first three years — that is, by the time sugar rationing ended, they were already 3 years old. A second group was born sometime in the nine months after rationing ended. This group spent at least some of their in utero time during a period in which sugar for their mothers was limited. The third group was conceived after sugar rationing ended, and therefore spent all of their in utero and early childhood period in a higher-sugar environment. 

The headline result of the graph below (which shows the data for diabetes, but the results for hypertension look similar) is that the increase in diabetes diagnosis with age is larger for the groups exposed to more sugar in childhood. The authors argue that the group with the least sugar exposure is about 40% less likely to be diagnosed with diabetes and 20% less likely to be diagnosed with hypertension than the group with the most sugar exposure.

The authors show (in other figures) that the effects grow based on the length of exposure to rationing. Notably, the largest impacts seem to be a result of having more sugar access after six months of age, when the child would be eating on their own. 

Diabetes and hypertension are the primary outcomes here. The authors (in their supplementary appendix) also show some impacts on obesity and waist-to-hip ratio. They do not find impacts on blood pressure. 

This study isn’t able to directly identify the mechanisms by which these effects occur, and there are a number of possibilities. One explanation is that, because we know from other work that tastes are informed by childhood experiences, children who are exposed to more sugar may become adults who like more sugar. This could contribute to metabolic disease. For the in utero impacts, in particular, it may be that more sugar exposure when in utero changes insulin resistance. 

From a causality standpoint, mechanisms here matter less. What the paper demonstrates is that something about this early-life sugar exposure seems to increase later health risks. 

Couldn’t something else be going on?

The paper isn’t a randomized experiment, and the authors therefore do not control everything. This leads to the question: could something else have changed at the same time as the end of rationing that drives these effects?

It is impossible to fully address this concern, but frankly, the authors do a very good job of tamping it down. They show that despite other changes in rationing over this general period, it is sugar consumption that really changes. They also show, for their primary outcomes, that there are no trends in the outcome before the change in sugar rationing. 

Nothing is perfect! If I were a referee on this paper, I would have given them a hard time because I see a bit of a pre-trend in Appendix Figure S9 and I think they could have used some more recent techniques to think about its importance. But at the end of the day, the results are compelling. 

Should we throw away all sugar? I’m panicking. 

Science (where the paper is forthcoming) released an early version on Halloween. I think this was mean! I had just gotten done explaining to people that it’s okay to let your kid have Halloween candy, and then this. 

In a sense, though, that juxtaposition is good, because both things can be true.  

On one hand: this adds a very convincing brick to a pile we already have that suggests that a lot of added sugar is linked to worse health. It is consistent with what we know about the importance of childhood in developing tastes. And yes, it does suggest that we want to be cautious about overexposure to sugar. The average American child consumes about 68 grams of sugar a day, per the CDC. That’s similar to the higher levels in this study. Working on ways to change that — as individuals, but probably largely through policy — could have benefits (if you’re interested in the policy angle, we have an episode of Raising Parents about this). 

On the other hand: this isn’t to say that we should never allow any sugar, and, indeed, making sugar a totally forbidden food may backfire (kids are drawn to foods they cannot have). There is a balance. It can be true both that Halloween should be a fun time with unlimited candy and joy and that on a regular basis kids’ exposure to soda should be limited. 

It’s also true that responsibility for monitoring the sugar in things shouldn’t lie exclusively in the hands of individuals.

Often when I talk about “panic headlines” (which this was for sure, for some people), the answer is that the research is bad and that’s why you should not only not panic but actually do nothing. Here, the research is not bad — it’s good — and it could suggest some behavior change. But there is still absolutely no reason to panic. 

The bottom line

• A compelling new research paper argues that chronic diseases in adulthood are reduced for children who were exposed to less sugar in utero and in the early years of their life.  
• Approach sugar consumption for your children with balance. There’s no need to be overly restrictive, but consider finding ways to cut back if possible.

Get more from ParentData when you sign up

This article is part of your free preview. By signing up, you can get:

Access to ParentData

Newsletters delivered to your inbox

Community discussions with other data-driven parents

Sign up

Community Guidelines