Don’t Blame COVID for this Diabetes “Surge” in Kids

A new study of type 2 diabetes in kids makes a fundamental error

Source: Two papers that shouldn’t be combined (see below)

This week, you’ve probably been hearing about the “surge” of type 2 diabetes among children and adolescents in the US. An apparent four-fold increase in the prevalence of the disease from 2017 to 2024.

What is to blame? Fingers point to the COVID pandemic sitting right in that time frame and a nice little picture is painted. Type 2 diabetes, the kind associated with obesity and insulin resistance spikes after the COVID pandemic, either due to the structural changes the pandemic placed on kids (more sedentary time, less activities) or based on the tropism of the virus to the pancreas itself.

[Want to keep reading? Get a Medium subscription or here, have a friend link.]

And that’s all very scary.

There’s only one problem. I don’t think it is correct.

The hubbub this week was caused by this paper, appearing in JAMA Pediatrics. It is the first truly nationally-representative sample of the prevalence of type 2 diabetes in children and adolescents ever published.

It sources data from the National Survey of Children’s Health. This is a survey administered by the census bureau and it contains data on more than 160,000 children, which can be scaled up to represent the full population of the US. That’s powerful.

But it’s cross-sectional. It is a snapshot of one moment in time. And I may actually put this on my tombstone: you can’t make longitudinal inferences from cross-sectional data.

I’ll explain the issues in a second but let me give the study it’s due.

As part of this survey, parents answered questions about their children’s health including this one: “Has a doctor or other healthcare provider ever told you that this child has type 2 diabetes?”

OK. We have our first issue right there. Not because it depends on parental recall. I have a feeling most parents would get this right about their kids. But because they may mistake type 1 diabetes for type 2 diabetes.

The survey asks about type 1 diabetes as well, but in a different way. It asks “Has a doctor or other healthcare provider ever told you that this child has Autoimmune disease (such as Type 1 Diabetes, Celiac, or Juvenile Idiopathic Arthritis)?”.

You see the problem, right?

Can you see how this might be confusing for a parent? I am sure that there is going to be some mislabeling of type 1 diabetes as type 2 diabetes in this dataset. And this is a real problem because type 1 diabetes is more common in kids than type 2 diabetes and also the factors that cause those two diseases are completely different, making the public health implications of an increase in prevalence completely different.

The major headline here is that, in this dataset collected between 2022 and 2024 the prevalence of Type 2 Diabetes (defined in this awkward way) was 1.4 per 1,000 kids overall, and 2.6 among kids aged 10–17.

It’s that 2.6 that is generating all the buzz, because, if that is the real number, it represents a huge increase from prior estimates. A four-fold increase to be exact.

But those prior estimates do NOT come from the National Survey of Children’s Health. The authors of the JAMA Pediatrics paper, citing this “surge” in prevalence, link to a longitudinal study of diabetes in kids called the SEARCH study.

Running since the year 2000, the SEARCH study is NOT nationally-representative. It is run out of five health centers around the country. But it doesn’t rely on parental report either. SEARCH employs trained research staff who actively identify children with diabetes at those centers based on physician reports, medical record reviews, hospital and health plan data in addition to self-report.

In 2017, the SEARCH study, using it’s less-representative, but much better phenotyped data, reported that the prevalence of type 2 diabetes among 10 to 19 year-olds in the US was 0.67 per 1000. That’s one fourth what this JAMA Pediatrics study suggests.

But the methodology is entirely different. This is comparing apples to oranges. Wait I need a better simile. This is like comparing DKA to HHS. OK we’ll workshop that and get back to you.

The point is a cross-sectional datapoint from a totally different survey methodology can not be meaningfully compared to an established, longitudinal study.

And the SEARCH study, since it has been running for 20 years, has already shown us that Type 2 Diabetes in increasing in young people. Here’s that data. This was actually published in JAMA in 2021. And it’s concerning — there is nearly a doubling in the prevalence from 2001 to 2017, with girls and older kids at higher risk. There really is an issue here.

But, for context, here is where the new paper suggests the prevalence is. A pretty dramatic inflection, driven by the 14–17 year age group.

And who knows? Maybe when the SEARCH study reports out its post-covid data this is what we’ll see. But it seems disingenuous for the authors to use historical controls with data collected completely differently to leverage the narrative that there is this explosion in cases.

Cross-sectional studies can’t give you longitudinal inferences, for sure. But they are good for some things. The authors of the JAMA Pediatrics paper used their data to examine what groups were most at risk for Type 2 Diabetes, and those findings are genuinely interesting.

They show that risk increases with age, and decreases based on parental income and education. They show higher rates in those of Hispanic ethnicity that evaporate when controlling for socioeconomic factors, and dramatically higher rates among Non-Hispanic Blacks that do not. These are important findings.

They may help to guide the next set of studies trying to figure out how to prevent this disease that we really should never have to see in children. But instead, we hear about this surge with an unsupported hypothesis that we can blame COVID for the problem.

Look, I would love to see nationally-representative, longitudinal data with accurate assessments of Type 2 diabetes rates in kids that goes right through the pandemic. Right now, the literature doesn’t provide that. What we do have shows that rates are increasing as childhood obesity is increasing. That is not surprising but it is important. Because if we don’t address that problem, well, a surge is coming.

A version of this commentary first appeared on Medscape.com.