Clinical heterogeneity in chronic inflammatory demyelinating polyneuropathy: data-driven subtypes challenge current classification.

BACKGROUND

Chronic inflammatory demyelinating polyneuropathy (CIDP) presents with substantial phenotypic heterogeneity, yet current classifications do not adequately capture clinically meaningful subgroups. Thus, we aimed to discover novel, clinically meaningful CIDP subgroups using unbiased clustering.

METHODS

In this observational, multicenter retrospective cohort study we screened 425 patients diagnosed with typical CIDP or CIDP variants from 10 German centers. Standardized data including demographic, clinical, electrophysiological, laboratory, comorbidity, and treatment information were collected. 325 patients were then randomly assigned to the discovery cohort and 100 to the validation cohort. We performed unbiased clustering and post-hoc analysis to identify clusters and define subtypes.

FINDINGS: Four distinct clusters showed marked differences: Cluster 1: Female patients with high symptom level, high disability scores, ANA positivity and autoimmune comorbidities Cluster 2: Moderate to low symptom level with low disability scores, demyelination-dominant and less comorbidities Cluster 3: High symptom level, axonal-demyelinating damage and cardiovascular comorbidities Cluster 4: Multifocal CIDP with high disability scores as well as frequent ataxia and areflexia Post hoc analyses showed marked differences in disease progression and therapeutic strategies, with clusters 1 and 4 showing particularly high rates of progression and disability.

INTERPRETATION: Unbiased clustering revealed four clinically meaningful CIDP subtypes, each exhibiting distinct presentations and trajectories. We identified two red flag subgroups suggesting that CIDP comprises biologically and clinically heterogeneous disease phenotypes. These findings may support more individualized monitoring and therapeutic strategies for high-risk patients and raise the possibility of distinct underlying pathomechanisms across CIDP subgroups.