Follow up to MIS-C post: Response from a Pediatric Immunologist

After publishing “[Did MIS-C Ever Really Exist?],” I received thoughtful comments As well as pushback from a pediatric immunologist who saw MIS-C cases firsthand during the pandemic. He told me there was no question MIS-C was a distinct clinical entity. He agreed that the definition of MIS-C lacked specificity, there was likely mislabeling of the disease and over counting, and that the wider community has not acknowledged the disappearance of MIS-C. He also agreed that the vast majority of kids that died were medically fragile.

So a lot of agreement to my post, but on to the points of contention. Here are his points, paraphrased, with my responses.

“MIS-C was definitely real. It was a distinct clinical entity.”

The post questions the diagnosis in the headline, in the form of a question. So the actual post doesn’t argue it doesn’t or couldn’t exist. The too long don’t read answer to the title is : Yes, MIS-C did exist, but…

What the post argues is that the 9,684 cases CDC counted were not all examples of that real entity, and that the surveillance system had no specific test to separate the real cases from the surrounding noise. Those two claims are compatible. A clinical entity can exist while the count of it is unreliable.

The question I’d push back on is not whether something real existed, but what the scope of the syndrome was. How many of the 9,684 represented it? If the answer is “I don’t know” — which I suspect is the honest answer — then the headline number was never doing the work it was being used to do. It was being cited to justify universal pediatric vaccination policy, to demonstrate a discrete pediatric COVID harm, to motivate continued surveillance funding. None of those uses survives the admission that the true count of the real entity is unknown.

“I can’t give you specific clinical criteria that diagnostically separate MIS-C from its mimics. But I know it when I see it.”

This is the entire methodological problem the post is about, and it’s not a problem unique to my immunologist friend. Pattern recognition by experienced clinicians is a legitimate diagnostic mode at the bedside. It does not generalize to a national surveillance count generated by thousands of pediatric hospitalists of varying experience, filing case reports based on a written case definition that does not capture what the experienced clinician is recognizing.

The 9,684 number was not generated by experienced pediatric immunologists making gestalt diagnoses on cases they personally evaluated. It was generated by the application of the CDC case definition to charts, by clinicians and reporters who had to follow written criteria because they didn’t have the pattern-recognition expertise to override them. The gap between what an experienced specialist sees at the bedside and what a national surveillance system captures is exactly where the manufactured part of the syndrome lives.

The Burns group at UCSD made the same point implicitly when they built KIDMATCH — a machine-learning algorithm to distinguish MIS-C from Kawasaki disease. They built it because clinical judgment at the bedside was insufficient. If the world’s leading Kawasaki research group needed an AI model to tell these diseases apart, the average pediatric hospitalist filing a CDC case report was unlikely to be doing much better.

“The volume of kids showing up in a shock state that were responsive to steroids was unusual. Pre-pandemic, we didn’t see that many.”

This is the strongest part of the pushback. It’s a real clinical observation from someone who was there. If the experience at major pediatric centers was that the volume of steroid-responsive pediatric shock cases visibly increased during 2020-2022, that signal is meaningful.

But three things need to be said about it.

First, the observation is compatible with my post. A real population of post-COVID Kawasaki-spectrum vasculitis cases — what the post referred to as Group A — clinically resembled Kawasaki disease, responded to Kawasaki treatment (IVIG plus steroids), and was almost certainly biologically real. The post doesn’t dispute their existence. It disputes the scale of the surveillance count.

Second, “steroid-responsive shock” describes a treatment response, not a specific etiology. Steroids are immunosuppressive and reduce inflammation across a wide range of inflammatory shock states — toxic shock syndrome, severe sepsis with hyperinflammatory features, macrophage activation syndrome, septic shock with adrenal insufficiency, and severe Kawasaki disease all show partial steroid responsiveness. The “steroid-responsive” criterion doesn’t separate a discrete MIS-C entity from a heterogeneous population of pediatric inflammatory shock cases. It tells you the immune system was a major driver of the illness, which was true in many of the conditions on the differential.

Third, the volume question is the question. If the pediatric immunologist’s true estimate of the discrete entity is, say, 500 cases per year nationally during 2020-2022, that’s a real signal worth paying attention to, but it’s also dramatically smaller than the 9,684 in the CDC count. The disagreement between his clinical experience and my post’s argument may not be about whether MIS-C existed but about what fraction of the count it represented.

“The cases clustered 4-6 weeks after each COVID wave. That’s specific to MIS-C.”

It’s specific to post-infectious hyperinflammatory syndromes generally. Kawasaki disease itself has been observed to follow respiratory virus waves with a 2-6 week lag — the original epidemiologic descriptions of KD in Japan in the 1960s and 1970s noted clustering after seasonal respiratory virus seasons. The temporal lag tells you the syndrome is post-infectious. It doesn’t tell you the syndrome is a novel discrete entity distinct from Kawasaki spectrum vasculitis triggered by SARS-CoV-2.

If the clinical phenotype is Kawasaki-like, the treatment is the same as Kawasaki, the temporal pattern matches Kawasaki, and there is no specific biomarker that distinguishes it from Kawasaki — the parsimonious interpretation is that we observed Kawasaki disease triggered by a new pathogen, not that we observed a new disease. The pediatric rheumatology literature contains a real debate about this. It is not a fringe position to suggest that MIS-C is the COVID-triggered presentation of Kawasaki-spectrum disease.

“The disappearance was because of Omicron. The variant just produces less MIS-C.”

Perhaps. But MIS-C is supposed to be a discrete first encounter post-infectious syndrome. Children born in 2023 and 2024 are SARS-CoV-2-naive at birth. Their immune systems have not encountered the virus before. By age 1-2, virtually all of them will have their first SARS-CoV-2 infection — by definition a first encounter with a novel pathogen by a naive pediatric immune system. That is exactly the immunological setup that was claimed to produce MIS-C, and yet this population doesn’t develop the disease anymore.

The cleanest explanation is that MIS-C as a discrete first-encounter post-infectious syndrome was always a smaller phenomenon than the pandemic throes surveillance count suggested, and the apparent epidemic was generated by the same factors that generated other pandemic-era surveillance artifacts — universal exposure criterion meeting nonspecific clinical definitions that intersected with a lot of institutional attention. When the attention faded, the count faded, even though the biological substrate (naive children encountering a novel virus) continues to be regenerated constantly.

If MIS-C is a discrete syndrome of first SARS-CoV-2 exposure in a naive pediatric immune system, where are the cases in the 2024 toddlers who just had their first infection? If the answer is that the virus has changed, what specifically about Omicron and its descendants makes first-encounter infection in a 2-year-old fail to produce the syndrome that first-encounter infection produced in a 8-year-old in 2021? These are answerable questions. The pediatric immunology literature has not engaged with them seriously, preferring vague variant-related explanations to alternatives that would require revisiting the original framing of the syndrome.

Take the fact that MIS-C was first recognized and reported in 2020, but in much smaller numbers than 2021. Maybe MIS-C is extremely variant specific, but my point is that there are other explanations that are just as, or more plausible.

The Immunity Debt Story

The COVID response during 2020-2021 (masking, school closures, social distancing, hygiene theater) suppressed the circulation of pediatric respiratory pathogens — RSV, influenza, parainfluenza, adenovirus, common cold coronaviruses, mycoplasma, pertussis. Children who would normally have been exposed to these pathogens in their first few years of life were not. When NPIs ended in 2021-2022, the pediatric population had an accumulated cohort of immunologically naive children encountering pathogens later than they otherwise would have, and at higher initial inoculum doses, in larger waves than normal. So when it comes to trying to discern the actual numbers of this clinically distinct entity that was related to COVID, the 2020 exposure, that represents a very clean test of what happens to a pediatric population exposed to COVID in the background of the usual petri dish of infectious pathogens kids swim in, generated the smallest count of MIS-C.

That 2020 count — modest, geographically concentrated, occurring in immunologically normal children encountering the virus in a normal pediatric infectious-disease ecosystem — is probably the cleanest available estimate of MIS-C as a real biological entity. It suggests the actual MIS-C count is dramatically smaller than the 9,684 total. The larger peaks of 2021 and 2022 occurred against a background of suppressed pediatric pathogen circulation, rebounding into a population of immunologically unusual children, all of whom now had universal SARS-CoV-2 antibodies and could be channeled into the MIS-C surveillance bucket. It is possible the real entity never got much bigger than its 2020 baseline. The surveillance count did, and that gap is what the original post was about.