Immunoadsorption (IA) is a therapeutic apheresis technique designed to remove pathogenic antibodies from circulation with high specificity (1). It has proven efficacy in immune-mediated neurological diseases such as autoimmune encephalitis (AE), myasthenia gravis, neuromyelitis optica spectrum disorders (NMOSDs), and Guillain–Barré syndrome (GBS) (1). In less critically ill patients, IA is generally well tolerated and associated with favorable clinical outcomes (2). However, its use in critically ill patients remains underreported, limited by concerns about safety, immunosuppression, and procedural risk. This retrospective study evaluates the feasibility, tolerability, and clinical outcomes of intermittent IA in critically ill patients with refractory immune-mediated neurological diseases, aiming to address this therapeutic gap.

Thirteen patients (9 women, 4 men, median age 46 years, IQR: 14–58, median body weight 70 kg, IQR: 40–80) met the inclusion criteria. Diagnoses include AE (n = 5), GBS (n = 4), neuromyelitis optica spectrum disorders (NMOSDs; n = 3), and chronic inflammatory demyelinating polyneuropathy (CIDP; n = 1). Although the median mRS score at both admission and discharge was 4 (IQR: 2–5 at admission and 1–5 at discharge), the distribution of individual changes showed a statistically significant improvement (p = 0.02) (Figure 1A). Specifically, 7 out of 13 patients demonstrated a ≥ 1 point reduction in mRS by discharge (Table 1), indicating meaningful functional recovery. The median APACHE II score was 10 (IQR: 5–25) on admission and 9 (IQR: 5–21) on discharge (p = 0.95), indicating a stable systemic condition (Figure 1B). The median duration of mechanical ventilation was 30 days (IQR: 0–62), and the median NCU stay was 41 (IQR: 23–65) days.

The median number of IA sessions per treatment course was 7 (IQR: 4–10), with inter-session intervals (ISIs) ranging from 1 to 30 days. Interruptions occurred in 92% of patients due to infection, hypotension, leukemoid reaction, or procedural complications (e.g., cardiac arrest and pelvic hematoma). Median IgG level decreased from 8.56 g/L (IQR: 6.17–14) to 4.18 g/L (IQR: 1.78–7.14), representing a median 55.6% reduction (Figure 1C, Table 2). In some cases, at our center, IVIG was administered during ISIs when serum IgG levels fell below 4 g/L, with or without concomitantly low IgA levels (~0.8 g/L), to support immunity (5). Pathogenic antibody analysis revealed negative conversion in 64.7% (plasma) and 37.5% (CSF), partial reduction in 17.65% (plasma) and 37.5% (CSF), and no change in 17.65% (plasma) and 25% (CSF) (Figure 1D).

Our findings suggest that intermittent IA treatment is both feasible and effective in critically ill patients with immune-mediated neurological disorders who fail standard first-line therapies. The flexible schedule allows clinicians to pause IA during critical complications and restart it when stable. In our center, decisions to suspend or prolong ISIs were based on multiple factors, including overall patient tolerance and the occurrence of serious adverse events (e.g., hypotension requiring vasopressor support, sepsis, or catheter-related infection). Among these, infectious complications typically result in substantially longer ISIs. This adaptive model may improve safety without compromising efficacy.

Notably, neurological function, as assessed by mRS, improved in most patients, whereas systemic disease severity, as measured by the APACHE II score, remained stable despite clinical complications. Given that the APACHE II score incorporates parameters from multiple organ systems, it provides an objective measure of overall physiological derangement (4). These findings suggest that intermittent IA treatment is a safe treatment option for critically ill patients with autoimmune neurological disorders.

Minor complications commonly associated with apheresis and vascular access (e.g., transient hypotension and hematoma) were observed in our cohort, although absent in some prior reports (6), and were clinically manageable. Leukemoid reactions were transient and self-limiting, while catheter-related infections and sepsis were successfully treated and did not necessitate permanent discontinuation of IA therapy. One patient experienced cardiac arrest during IA but was successfully resuscitated with cardiopulmonary resuscitation and completed the planned treatment course. In contrast, the reported adverse events in the intensive care unit during plasma exchange included anaphylactoid reactions, severe hypotension, catheter-related infections, pneumothorax, local bleeding, hypocalcemia, and paresthesia (7). These complications, largely related to plasma substitution and citrate anticoagulation, were uncommon in our IA cohort (Table 1), suggesting that IA may offer a favorable safety profile in critically ill neurologic patients, although confirmation in larger prospective studies is warranted.

While our observed IgG reduction was slightly lower than previously reported in general populations (55.6% vs. 62–93%) (1), this may reflect lower starting IgG levels or treatment interruptions. Despite concerns about IA-related immunosuppression, no patient in our cohort experienced fatal infection or irreversible deterioration. This supports the safe use of IA in selected patients with marginal IgG values, provided monitoring and IVIG support are available.

IA remains underutilized, partly due to high costs and limited guidelines for ICU settings. In Germany, a session costs ~€ 1,200 (8); in the U. S., five sessions may cost ~$ 58,952 (9). At our center, a typical course costs ~30,000 RMB (~3,000 per session). Costs vary widely based on location, condition, and institutional procurement, but early use in high-risk patients may reduce long-term ICU burden.

Our study has several limitations. First, its retrospective design introduces inherent biases, including a small sample size and the absence of a control group. Additionally, the study was not powered to detect differences across subgroups of immune-mediated neurological diseases. The absence of certain severe conditions, such as myasthenic crisis, further limits the generalizability of our findings. Future research should focus on prospective, controlled studies with multiple centers and large cohorts to validate our observations and explore the optimal IA treatment protocols for diverse neuroimmunological conditions.