Autoimmunity Article 2: Immunoadsorption, IVIG, Rituximab, Daratumumab — Removing the Antibodies, and What It Means When That Doesn’t Fix ME/CFS
If you have ME/CFS and evidence of GPCR autoantibodies, the logical next question is: can we remove them? The treatments exist — immunoadsorption physically strips IgG from blood, IVIG dilutes pathogenic antibodies with pooled healthy immunoglobulin, rituximab kills the B-cells that make them, daratumumab goes a step further and kills the plasma cells. But the evidence for each tells a consistent story: open-label results are dramatic; controlled results are sobering.
This article explains the treatment evidence. For the conceptual background — what GPCR autoantibodies are, the Fc glycoprofile, and the measurement controversy — see the companion overview article.
1 First, a plain warning
Everything below describes hospital-based interventions with significant risks: immunoadsorption requires central venous access and carries infection, thrombosis, and hypotension risks. IVIG causes aseptic meningitis, thromboembolism, and anaphylaxis. Rituximab carries a risk of progressive multifocal leukoencephalopathy. Daratumumab causes infusion reactions and immunosuppression. These are treatments for the severely affected, discussed with an immunologist — not self-directed and not first-line.
2 The treatment evidence, from most to least mature
2.1 Immunoadsorption — the most direct mechanism, the most contested evidence
Immunoadsorption (IA) is akin to dialysis for antibodies. Blood is passed through a column lined with a ligand that binds IgG — tryptophan, protein A, or a synthetic peptide — stripping ~80% of circulating IgG in a single session. A typical course is 5 sessions over 10–14 days.
The advantage over plasma exchange is selectivity: IA removes IgG without removing albumin, clotting factors, or other plasma proteins. The disadvantage is that it removes all IgG — protective antibodies, vaccine responses, and autoantibodies alike — and the autoantibodies rebound within weeks to months if the B-cells and plasma cells producing them are not addressed.
The open-label evidence is strong. The Scheibenbogen group has published a consistent series: - 2018 pilot (n=10): 70% rapid improvement, 30% sustained 6–12 months (Scheibenbogen et al. 2018). - 2020 retreatment (n=5): 80–90% IgG and β2AR-AAb reduction, 4/5 sustained improvement (Tölle et al. 2020). - 2025 (n=20 post-COVID): IgG reduced 79%, AAbs 77%, 70% responder rate (SF-36 PF ≥10 points), durable at 6 months (Stein et al. 2025).
The controlled evidence is sobering. Two developments in 2025–2026 changed the picture: - Anft 2025 (independent centre, n=12): AAbs were eliminated, cytokines reduced, neuropsychological improvement measured — but ME/CFS symptom scores were not improved, and AAbs rebounded within 1 month (Anft et al. 2025). This is the discordant result that needs explanation: the antibody removal worked biochemically but did not translate to clinical benefit. - IA-PACS-CFS (Charité, NCT05710770, sham-controlled, ~65–66 patients): preliminary conference report (May 2026) indicated no statistically significant difference between immunoadsorption and sham on the Chalder Fatigue Scale [Preßler et al. (2024)](Rücker 2026). One caveat frames it: the trial did not pre-select for antibody-positive patients, so a negative overall result does not falsify the autoimmune hypothesis — it may simply mean you need to remove antibodies from someone who has them.
The lesson of the discordance. Immunoadsorption removes antibodies. If removing them does not improve symptoms, either (a) the antibodies were not pathogenic (wrong Fc glycoprofile), (b) the pathology they caused is already irreversible (receptor internalisation, structural damage), (c) the antibodies were not the dominant pathology in that patient, or (d) the antibody rebound is too fast for clinical benefit. All four are plausible, and the current data cannot distinguish them.
2.2 IVIG — immunomodulation, not just antibody dilution
Intravenous immunoglobulin (IVIG) at 0.4 g/kg daily for 5 days, then 0.4 g/kg every 3–4 weeks, does two things: it dilutes pathogenic autoantibodies (a smaller effect than IA), and — more importantly — it provides immunomodulatory Fc glycans from thousands of healthy donors. The sialylated IgG in IVIG engages DC-SIGN on regulatory macrophages, inducing an anti-inflammatory shift. This is why IVIG works in diseases (Kawasaki, CIDP, ITP) where simple antibody dilution would not be sufficient.
The evidence in ME/CFS is uncontrolled. The Eckey 2025 patient survey (n=3,925) reported IVIG among the more positively rated options in the multisystemic symptom cluster (a figure around 73% net-positive was reported for it) — but this is patient-reported, uncontrolled, and subject to selection bias, and the survey’s full breakdown should be read directly (Eckey et al. 2025). There is no placebo-controlled IVIG RCT in ME/CFS.
IVIG for PANDAS — a cautionary parallel. PANDAS (paediatric autoimmune neuropsychiatric disorder associated with streptococcal infection) is mechanistically analogous: post-infectious autoantibodies targeting the basal ganglia cause neuropsychiatric symptoms. A 2024 placebo-controlled IVIG RCT in PANDAS was null — IVIG was no better than placebo. The optimistic reading is that the trial was underpowered or the wrong patients were enrolled. The pessimistic reading — the one that should constrain expectations for ME/CFS — is that post-infectious autoantibody syndromes may not respond to antibody dilution alone (Loth 2026).
2.3 Rituximab — the B-cell depletion lesson
Rituximab (anti-CD20) depletes B-cells — the precursors of the plasma cells that produce antibodies. It spares the long-lived plasma cells (CD20-negative) that reside in bone marrow and continue to secrete antibodies for decades. This is the mechanistic explanation for why rituximab failed in ME/CFS.
The arc of the evidence is instructive. Fluge & Mella’s initial open-label studies (2009–2011) reported 67% response rates — dramatic, widely reported, and the basis for the phase III RituxME trial. RituxME (n=151, double-blind, placebo-controlled) found 26% rituximab vs 35% placebo at 24 months — the placebo group did better (Fluge et al. 2019). The open-label signal was noise, amplified by expectation and regression to the mean.
The lesson applies to every other open-label treatment signal in ME/CFS: uncontrolled response rates overestimate benefit. The RituxME null did not disprove the autoimmune hypothesis — it proved that depleting B-cells without depleting plasma cells does not work.
3 Expected results
- The best-case treatment response is partial. Even in the positive open-label series, “responder” meant clinically significant improvement — not recovery, not return to work, not normalisation of PEM. The treatments reduce the autoimmune contribution; they do not cure ME/CFS.
- IVIG response may take 3–6 months. The combination of antibody dilution and immunomodulatory Fc-glycan effects is gradual. A single cycle is not an adequate trial, but a null result at 6 months of adequate dosing is sufficient to stop.
- Immunoadsorption works biochemically — IgG drops by 80% — but this does not guarantee clinical improvement. The Anft 2025 discordance (AAbs gone, symptoms unchanged) is the cautionary data point. IA tells you something about the antibody; the clinical response tells you something about whether the antibody was pathogenic.
- Daratumumab responds with an 8–9 month latency, if the signal is real. This is the longest latency of any immune-targeted treatment, and it means a 3–6 month trial may be falsely negative. A 12-month trial is the minimum to judge the daratumumab signal — and such a trial has not been done.
- Treatment decisions in 2026 are made under genuine uncertainty. The IA-PACS-CFS result has been reported preliminarily (a May 2026 conference report suggested no significant benefit over sham) and the EXTINCT controlled result is still awaited (Rücker 2026). The glycoprofile hypothesis is mechanistically elegant but clinically untested. The honest approach is a time-limited, endpoint-defined trial — and the discipline to accept the answer it gives.
4 An example structured trial
This is an example to give your clinician something to work from — not a self-prescription.
The pre-requisite is documented GPCR autoantibodies on a functional assay (bioassay, not commercial ELISA), a post-infectious onset, and a POTS-dominant autonomic profile. If these are not met, the pre-test probability of a dominantly autoimmune mechanism is low and these treatments are unlikely to be appropriate.
A practical caveat about the pre-requisite: the functional GPCR bioassay is mostly available in research settings, not routine clinical labs — so for most patients the antibody gate cannot actually be met before a trial can even be discussed. Where a functional assay is unobtainable, the practical pre-requisite collapses to the clinical picture alone: clear post-infectious onset plus a POTS-dominant autonomic profile with documented orthostatic symptoms. That is a weaker gate than the antibody test, and it should be read as such — it lowers confidence that the patient is antibody-positive but does not make the treatment inappropriate to raise with a specialist.
If the profile is right, start with the least invasive option: IVIG 0.4 g/kg daily for 5 days, then 0.4 g/kg every 3–4 weeks. Pick a single target symptom — the autonomic symptom that limits you most (standing time, presyncope frequency, heart-rate variability on orthostatic challenge). Reassess at 6 months. Stop if no improvement — a 6-month null IVIG trial is sufficient evidence that immunoglobulin therapy is not the lever. Be aware that IVIG is expensive (often several thousand dollars per infusion cycle) and, being off-label in ME/CFS, is frequently subject to prior authorisation and often denied — so access, not just clinical fit, is a real barrier that should be discussed up front.
If IVIG produces partial improvement but AAbs are still elevated, and the patient is severely affected (housebound or bedbound), discuss immunoadsorption with a specialist centre. A 5-session IA course (10–14 days) should reduce IgG by ~80%. Reassess at 1 month post-treatment. If symptoms have not improved despite adequate IgG reduction, the “antibodies are the dominant pathology” reading is unsupported — this could mean the antibodies were not pathogenic, that the damage they caused is irreversible, or that they were not the main driver; the data as this series discusses do not cleanly separate those. Either way the actionable clinical conclusion is the same: stop pursuing antibody removal and shift to downstream-amplifier management.
Only escalate to daratumumab (where available and under specialist supervision) if IVIG and IA have both failed in the AAb-positive post-infectious subgroup — and only within a controlled research protocol, given the early evidence stage and the on-target NK-cell depletion. Daratumumab is an expensive oncology biologic with no approved ME/CFS indication; in practice it is effectively confined to specialist/research centres that can obtain it, so for most patients it is not a realistically reachable option.
5 What it means if the treatment does not help
One: a failed immunoadsorption trial is informative. If IgG was reduced by 80% and symptoms did not improve, either the antibodies were not pathogenic (Fc glycoprofile mismatch), the pathology they caused is irreversible (receptor internalisation), or the antibody rebound was too fast. In the first case, further antibody-targeted treatment is unlikely to help. In the second and third, it might — but with a different strategy (maintenance with daratumumab, not pulse with IA). Two: a failed IVIG trial says IVIG does not work for you — nothing more. It does not rule out an autoimmune mechanism. IVIG is immunomodulatory, not antibody-depleting, and the dose, brand, and infusion rate all affect tolerability and possibly efficacy. But if IVIG at standard doses over 6 months produces no benefit, continuing indefinitely is neither evidence-based nor risk-free. Three: the RituxME null was a trial failure, not a hypothesis failure. Rituximab spares plasma cells. The plasma-cell factory model — supported by the daratumumab signal and the 8–9 month latency — is still on the table. The lesson of RituxME is not “autoimmunity is not the mechanism.” It is “do not deplete B-cells and expect to deplete antibodies.” Four: a sequence of failures — immunoadsorption removes the antibodies but symptoms persist, IVIG produces no benefit, daratumumab is unavailable — is genuine evidence that autoimmunity is not the dominant pathology, or that it was dominant once but the damage it caused is now self-sustaining. Either way, it is useful information. Stop pursuing antibody-targeted treatments and focus on the downstream amplifiers (POTS, MCAS, SFN) that the autoimmunity may have triggered.
6 The falsifiable predictions
- If the glycoprofile predicts IA response — G0F-dominant (pro-inflammatory) profiles respond; sialylated-dominant profiles do not — the Fc-mechanism model is clinically actionable and should be added to pre-treatment testing (Loth 2026).
- If daratumumab responders show a decline in pathogenic antibody titre over 6–9 months with clinical improvement lagging, the plasma-cell-factory model is supported and the latency is a real biological phenomenon, not a placebo effect.
- If IA-PACS-CFS and EXTINCT (Hannover, NCT05954325, n=63) both return null in antibody-positive subgroups, the autoimmune hypothesis “takes a blow comparable to the rituximab-to-RituxME collapse” — the primary document’s own framing.
- If passive transfer of ME/CFS IgG reproduces autonomic dysfunction in mice — the experiment has been done in fibromyalgia and Long COVID but never in ME/CFS — the autoimmune mechanism moves from association to causation.
7 What you can actually do
- Get the right test, but read it critically. Functional GPCR autoantibody testing (bioassay, not ELISA) is the most meaningful — but it is available only in research settings. If you are offered commercial CellTrend ELISA testing, know that the 2022 replication study found 100% of healthy controls α1-AAb positive, and the test’s clinical utility is disputed. A positive result is ambiguous. A negative result on a high-quality functional assay is more informative than a positive result on a non-specific ELISA.
- The most honest predictor of treatment response is not the antibody titre — it is whether you have a clear post-infectious onset and a POTS-dominant autonomic profile. The β2-AAb signal correlates with autonomic severity; the post-infectious history places you in the subgroup where molecular mimicry is the plausible trigger. If your ME/CFS was gradual, non-post-infectious, and pain-dominant rather than POTS-dominant, the pre-test probability of an autoimmune mechanism being the dominant driver is lower.
- Immunoadsorption is a specialist decision for the severely affected with documented GPCR autoantibodies and a POTS-dominant profile. It is not a first-line treatment and not a generic recommendation. It requires a centre with IA experience and a prescriber who understands the evidence limitations — including the pending IA-PACS-CFS result and the Anft 2025 discordance. It is also among the most expensive and least available interventions described in this series (specialist centre, central venous access, a ~5-session course), so realistic access should be assumed to be a barrier in most systems.
- IVIG is a higher-risk, lower-evidence option than its patient-community reputation suggests. The PANDAS IVIG null is a cautionary parallel. The Eckey 2025 survey data are encouraging but uncontrolled. A 6-month IVIG trial with defined endpoints and a stopping rule is a rational conversation with an immunologist if GPCR autoantibodies are documented and less invasive options have failed — but indefinite IVIG without evidence of benefit is not good medicine. Standard pre-IVIG safety checks — serum IgA level (to exclude IgA deficiency, where anaphylaxis risk is higher), renal function, and a thromboembolic-risk assessment — are part of a responsible workup and should be discussed with the prescriber.
- Accept that the field is unsettled. The daratumumab signal is real but uncontrolled. The IA-PACS-CFS result is now being weighed in its preliminary form (suggesting no clear benefit over sham) and the EXTINCT result is still pending (Rücker 2026). The glycoprofile hypothesis is mechanistically elegant but clinically untested. A treatment decision made in 2026 is made under genuine uncertainty — and the honest clinical approach is to acknowledge that, make a time-limited trial with a stopping rule, and accept the answer the trial gives, whichever direction it points.
8 The bottom line
Autoimmune-targeted treatments for ME/CFS exist — immunoadsorption, IVIG, rituximab, daratumumab — but the evidence is uneven, the open-label signals exceed the controlled results in a pattern that RituxME made impossible to ignore, and the measurement tools that would identify who will respond (functional bioassays, Fc glycoprofiles) are not clinically available. The daratumumab signal — 60% marked improvement in an uncontrolled pilot — is the most encouraging recent development, but the RituxME precedent demands controlled replication before it changes practice [Fluge et al. (2025)](Loth 2026).
The honest clinical approach is a time-limited, endpoint-defined trial in the antibody-positive, post-infectious, POTS-dominant subgroup — and the discipline to stop if the trial fails and shift attention to the downstream amplifiers that the autoimmunity, if present, had activated.
Next in this series: “Chained together” — the synthesis capstone. How the amplifiers become one disease, and what that means for treatment.
For the comprehensive, fully-cited picture of how autoimmunity is weighed among the many candidate mechanisms in ME/CFS, see (Loth 2026).