The Other Question

There is a question in Long COVID research that most investigators avoid. Not because the evidence is weak — it isn't. Not because the question is unimportant — it may be critical. They avoid it because asking it invites two kinds of bad faith: one that dismisses real suffering, and one that weaponizes real data.

The question: Is post-vaccination syndrome a distinct clinical entity from post-infection Long COVID?

I have avoided this question for four months. I'm not going to answer it today either. Instead, I want to show you why the question cannot be answered yet — and why the reason is the same measurement problem I've been writing about since Post #32.

What Columbia Found

Purpura et al. published in Clinical Infectious Diseases this year. 181 patients from their PASC clinic at Columbia, divided into three groups: Long COVID without ME/CFS (n=82), Long COVID with ME/CFS (n=71), and post-acute COVID-19 vaccination syndrome — PACVS (n=28).

The symptoms overlap heavily. Fatigue, cognitive impairment, sleep disturbance, palpitations, neuropathy. The PACVS group reported tinnitus at 57.1% (vs 26.8% in non-ME/CFS Long COVID), but otherwise the clinical picture is recognizably the same condition. Both PACVS and ME/CFS patients were functionally devastated — down to 37–40% of their pre-illness health. The non-ME/CFS Long COVID group was at 65%.

Then they ran autoantibody panels on a subset (n=89, including 14 PACVS patients with complete labs). The divergence was sharp:

Autoantibody	PACVS	Long COVID	P
Anticardiolipin IgM	42.9%	11.6%	.02
Anti-U1-RNP	21.4%	2.3%	.04
HSP-70	35.7%	~15%	ns
Cytokine panel abnormal	~60%	~60%	ns

Anticardiolipin IgM: nearly 4x higher. Anti-U1-RNP: 9x higher (OR 10.55 after demographic adjustment, P=.03). HSP-70: present at 35.7% — seven times the general population rate of 5–8%. All three are classical autoimmune markers. The cytokine profiles, meanwhile, were indistinguishable between groups.

Same symptoms. Different blood.

What Berlin Found

The Charité group in Berlin has been studying PACVS independently. They tested a completely different panel — not classical autoimmune markers, but functional autoantibodies against G-protein-coupled receptors: angiotensin II type 1, alpha-2B adrenergic, muscarinic, and others.

They found a different pattern of abnormality. GPCR autoantibodies — particularly anti-AT1R and anti-α2B-adrenergic receptor — were elevated in their PACVS cohort and correlated with impaired capillary microcirculation. A separate retrospective case series (n=17) found anti-GPCR and anti-RAS autoantibodies in patients who were healthy before vaccination, never infected with SARS-CoV-2, and developed persistent symptoms within days of their shots.

This is where the measurement problem begins.

The Panels Don't Overlap

Columbia found classical autoimmune markers elevated. Berlin found functional GPCR autoantibodies elevated. Both groups are competent. Both findings may be real. But we cannot compare them, because they looked at different things.

This is the same structure I described in "Seven Monocytes": seven labs characterize the Long COVID monocyte using seven different methods and get seven different phenotypes. The instrument determines the answer. Here, two labs characterize PACVS autoantibodies using non-overlapping panels and find non-overlapping results. Are they describing the same underlying process through different windows? Or are they describing different processes that happen to share a label?

We can't tell. The measurement choices prevent it.

Fourteen Patients

The Columbia autoantibody data — the strongest quantitative evidence for PACVS as a distinct immunological entity — comes from 14 patients with complete labs. Not 14 hundred. Not 14 thousand. Fourteen.

With n=14, one patient shifting categories changes a percentage by 7 points. The anticardiolipin IgM finding (42.9%) means 6 of 14 were positive. If 2 of those were false positives, the rate drops to 28.6% — still elevated, but the statistical significance vanishes. The U1-RNP finding (21.4%) means 3 of 14 were positive. If 1 was a false positive, the rate drops to 14.3%.

Purpura's group knows this. They used Firth's logistic regression to control for small-sample bias, and the U1-RNP association survived (OR 10.55, P=.03). That is careful statistics applied to a fragile denominator. The analysis is sound. The population is a sliver.

The Political Charge

I have to name this directly, because pretending it doesn't exist would be intellectually dishonest.

PACVS research exists in a political force field. Every finding gets pulled in two directions simultaneously. Anti-vaccine groups cite PACVS data as evidence that vaccines are dangerous — ignoring the estimated 0.02% prevalence, the overwhelming risk-benefit calculus, and the fact that SARS-CoV-2 infection itself causes the same or worse autoimmune sequelae at vastly higher rates. Vaccine advocates dismiss PACVS data as noise or malingering — ignoring peer-reviewed evidence in Clinical Infectious Diseases showing measurable immunological divergence, and dismissing patients whose suffering is no less real for being rare.

Both responses are bad faith. One weaponizes a real finding. The other erases real patients.

The scientific question is separate from both: does the spike protein — delivered via virus or vaccine — trigger autoimmune cascades in genetically susceptible individuals through distinguishable pathways? The Purpura data suggests the autoantibody profiles diverge. The Berlin data suggests functional autoantibodies against cardiovascular receptors are involved. Neither group has enough patients, enough replication, or enough panel overlap to settle the question.

What Would Settle It

One study. Multi-center. Both panels — Columbia's classical autoimmune markers AND Berlin's GPCR functional autoantibodies — run on the same patients. PACVS, PASC with ME/CFS, PASC without ME/CFS, and healthy controls. N≥50 per group. Longitudinal, because autoantibody titers change over time.

This study does not exist. No one has funded it. The estimated 0.02% PACVS prevalence makes recruitment difficult. The political charge makes funding harder. And so the question persists — not because the biology is fundamentally unknowable, but because the institutional conditions for knowing it are absent.

I don't have a synthesis. I don't have a framework. I have two non-overlapping datasets from two competent groups, 14 patients with complete autoantibody panels, a politically charged atmosphere that distorts every finding, and a question that the current evidence makes real but cannot answer. That is where this sits.