To investigate performance of a clinical-electrophysiological model for distinguishing chronic inflammatory demyelinating neuropathy (CIDP) in an Australian cohort.
A probability calculator for CIDP, based on statistical modelling of clinical-electrophysiologic variables, was developed from a US center1. This clinical decision support tool uses four clinical (progression over 8 weeks, absent autonomic involvement, absent muscle atrophy, proximal limb weakness) and two electrodiagnostic variables (ulnar motor conduction velocity slowing <35.7 m/s and ulnar motor conduction block).
Twenty-nine patients were identified; mean age 43.4 ± 15.2 years, 69% male, mean disease duration 12.5 ± 11.3 years. CIDP phenotypes included typical (34%), multifocal/focal (24%), distal (17%), sensory/sensory-predominant (14%), and motor-predominant (7%). Mean number of nerve conductions assessed were motor n=5 (range 4-7) and sensory n=5 (range 1-7). At least one supportive diagnostic criteria was present in 86% of patients, including CSF protein elevation, imaging (ultrasound/MRI), nerve biopsy, and/or objective treatment response. Using 92% probability cut-off, 27/29 (93%) of patients were correctly identified. One multifocal and one distal CIDP patient were not identified, due to the presence of muscle atrophy and lack of proximal weakness. At 98% probability cut-off, sensitivity was 86% (25/29).
Findings in an Australian CIDP cohort support 92% probability cut-off in the model to minimize Type II error and indicates good sensitivity for CIDP, maximal in typical phenotypes. False negatives were likely contributed to by model application in a stable cohort with longer disease duration. Prospective validation studies in patients at initial evaluation for CIDP are required.