CDC propagandized diabetes — as a pediatric health risk associated with COVID-19 — on the basis of a sloppily constructed and reasoned study

10 January 2022

Today, I'll pick on the Centers for Disease Control and Prevention (again)

They seem to have become one of the United States' preeminent taxpayer-funded sources for delivering fright-propaganda and little of medical or epidemiological value.

The CDC's most recent fear-mongering claim — arrived three days ago

Published in the MMWR — my additions in bracketed italics:

New diabetes diagnoses were 166% (IQVIA [database]) and 31% (HealthVerity [database]) more likely to occur among patients with COVID-19 than among those without COVID-19 during the pandemic and 116% more likely to occur among those with COVID-19 than among those with ARI [acute respiratory infection] during the prepandemic period.

© 2022 Catherine E. Barrett, Alain K. Koyama, Pablo Alvarez, Wilson Chow, Elizabeth A. Lundeen, Cria G. Perrine, Meda E. Pavkov, Deborah B. Rolka, Jennifer L. Wiltz, Lara Bull-Otterson, Simone Gray, Tegan K. Boehmer, Adi V. Gundlapalli, David A. Siegel, Lyudmyla Kompaniyets, Alyson B. Goodman, Barbara E. Mahon, Robert V. Tauxe, Karen Remley and Sharon Saydah, Risk for Newly Diagnosed Diabetes >30 Days After SARS-CoV-2 Infection Among Persons Aged <18 years — United States, March 1, 2020–June 28, 2021, Morbidity and Mortality Weekly Report (07 January 2022)

The CDC then further escalated this claim . . .

. . . to "2.5 times more likely" — in a tweet based on the above-cited report:

[C]hildren and teens 18 years & younger who have had #COVID19 are up to 2.5 times more likely to have a #diabetes diagnosis after infection.

Prevent COVID-19 by using tools like masks and #vaccines for those eligible.

Naturally

The alleged finding will scare the pants off any parent, who is reasonably familiar with juvenile diabetes' severe depredations upon a child's future.

Is the CDC's claim meaningfully correct?

No.

At least, not in any reasonably assessed statistical risk sense.

Here are the numbers (from the CDC's own data)

Point one — consider the CDC's only dubiously sound methodology:

The MMWR team pulled its data from diabetes-related diagnosis and billing codes from two different databases.

Those of you in the medical field, will recognize how inaccurate these codes can be, depending upon where the financial and medical competence incentives lie.

Point two — the absolute risk of allegedly developing diabetes (whether type 1 or 2) from a COVID infection turns out to be trivial — even according to data contained in the report's Table 1.

For instance, the calculations for the IQVIA database work out this way:

With respect to the two pooled databases, only 68 COVID kids — out of 80,893 such — were coded as having become newly diabetic.

Most people would consider the risk posed by 68 divided by 80,893 to be pretty insignificant.

This 'risk' works out to be only 0.08 percent.

That is 8 hundredths of one percent.

I doubt that anyone would build themselves a bomb or tornado bunker, under those odds.

CDC sensationalizes the COVID-diabetes relationship

The MMWR report compares the above COVID-related risk to supposedly normal background rate (drawn from the two databases) of developing diabetes in children.

That calculation works out this way (for the IQVIA data) — 132 newly diabetic kids out of the 404,465 that did not have COVID.

See, again, the report's Table 1.

This gives us 0.03 percent. Or 3 hundredths of a percent.

If we calculate the relative risk of developing diabetes between the COVID and not-COVID groups, we see that:

(a) a child is 2.67 times more likely to be diagnosed with diabetes-related issues from, during or after a COVID-19 illness

as compared to

(b) what he or she would have, while operating under more 'normal' (non-COVID) circumstances.

You can do this calculation by dividing 0.08 by 0.03.

The CDC 'humbly' reduces this 2.67 escalation to only 2.5 in a tweet. A reduction that I suspect somewhat hints at how bogus they recognize their own study (and its reasoning) are.

Does the increased diabetes risk amount to anything meaningful?

Probably not.

This is where understanding the difference between absolute risks and relative ones is important.

We can magnify fear, whenever we decide to investigate trivial risks by comparing them relatively.

For example, if our risk of developing something bad (meaning undesirable) is close to being 'absolute' zero — would we be concerned — if this trivial risk were multiplied by 3?

I don't think so.

At least not, if were rationally cognizant of the fact that there are many more statistically worrisome risks to think about.

For instance, would most of us get excited about an escalated risk of falling into a bottomless pit, if the absolute risk of doing so was tripled from 0.033 to 0.099 on any given trip to an NFL playoff game?

Let's take the alleged COVID sting . . .

. . . out of the CDC's deceptively presented data.

The MMWR report compared COVID's diabetes number (in kids) with the diabetes volume that "acute respiratory infections" generated.

The latter turned out to be a reported 0.06 percent.

Now, compare the 0.08 percent (diabetes from COVID) to the 0.06 percent of newly acquired diabetes (due to what are, for most kids, essentially 'colds').

The relative difference there is 1.33.

This means that a child is 33 percent more likely to develop diabetes from COVID than they are, presumably, from a cold.

Or looking at the issue in the other direction, according to the CDC's own data — a child is twice as likely to develop diabetes from having a cold, as they are from not having a cold.

To calculate that relative rate — divide 0.06 percent (diabetes related to acute respiratory infections) by 0.03 percent ('normal' diabetes background incidence).

Is the CDC's ridiculousness beginning to sink in?

One can do this relative risk thing all day long — when one wants to make trivial issues and associations look both important and causative.

All told — crappy MMWR methodology

We have already addressed the problems inherent in doing research associated with diagnosis and billing codes. Let's address other shortcomings in the MMWR report.

For example, though the research team probably had some of the desirable data, they did break the allegedly diabetic (and not diabetic) children down into medically appropriate demographics:

Who had comorbid conditions, including obesity?

Who, especially, already had vascular issues, which we know are likely to be subject to SARS-CoV-2 attack?

And what about all the kids, who never went to the doctor and/or were never COVID tested?

All told, we do not have a clue, as to how many of those unreported children had COVID — and (a) had no symptoms and no diabetic indications — or (b) had symptoms and no diabetes — or (c) had symptoms and did have unreported or undetected diabetic signs.

In sum, the MMWR 'kids and COVID diabetes' report is a Big Vat of Nothing.

The research team partially admits as much . . .

. . . but buries what should have been their up-front confession in the report's second-to-last paragraph:

The findings in this report are subject to at least four limitations.

First, the definition of diabetes might have low specificity because it used a single ICD-10-CM code, did not include laboratory data at the time of diagnosis, and could not reliably distinguish between type 1 and type 2 diabetes.

Second, patients infected with SARS-CoV-2 without a COVID-19 diagnosis or documented positive test result might be misclassified as not having COVID-19.

Third, the present analyses lacked information on covariates that could have affected the association between COVID-19 and incident diabetes, including prediabetes, race/ethnicity, and obesity status.

Finally, estimated associations are only representative of persons aged <18 years seeking care included in these commercial claims databases and not of pediatric populations with SARS-CoV-2 infection without commercial health insurance or who do not seek health care.

Consider a more realistic perspective — drawn from the same CDC data

If you read Table 1, did you notice that the relative risk of being hospitalized for COVID — among the IQVIA database of kids being discussed — was 4.75 times the lesser risk of developing diabetic indicators?

You can do this calculation by dividing 6473 hospitalized children by 1,698,753 total (in the database). That results in 0.00381.

Multiply 0.00381 by 100 to get a percentage — 0.38 percent.

Now, obtain the relative risk — of being hospitalized by COVID — versus getting diabetes from COVID.

To do that, divide 0.38 percent (hospitalization) by 0.08 percent (diabetes indicators).

The relative risk equals 4.75.

The CDC blithered on about the COVID-associated risk of developing diabetes in kids — when the risk of COVID hospitalization (in the IQVIA database) was close to five times that.

For those of you, who have never been in a hospital, I can assure you that going there brings all sorts of frequently enough nasty, occasionally lethal, issues with it.

In my parental view, the CDC loses all traces of common sense in its MMWR report's breakdown of potential issues to legitimately worry about.

The moral? — CDC's COVID and diabetes report is medically and epidemiologically useless

As I have hinted in the past — see here and here, for instance — the MMWR is not noted for the solidity of its research and reasoning.

And today, the CDC has further deteriorated its ability to generate legitimately delivered respect by having turned itself into a (near-pure) propaganda organ for Big Pharma and tyrannically oriented government.

PeteFree.com

CDC propagandized diabetes — as a pediatric health risk associated with COVID-19 — on the basis of a sloppily constructed and reasoned study