Blood bile acid profiles in chronic inflammatory demyelinating polyneuropathy

Abstract

Background: Chronic inflammatory demyelinating polyneuropathy (CIDP) is an acquired immune-mediated neuropathy with demyelinating features like Guillain-Barréé syndrome (GBS). Despite established diagnostic criteria, the lack of specific blood biomarkers highlights the need for novel markers to improve early diagnosis and disease monitoring. Bile acids (BA), cholesterol-derived molecules with immune modulatory properties, have been explored as biomarkers in immune-mediated diseases. This study investigates BA profiles in CIDP and evaluates their potential for diagnosing CIDP.

Methods: Patients with treatment-naïve immune-mediated polyneuropathies (CIDP and GBS) and age-matched healthy controls (HCs) were recruited from a tertiary referral hospital. Their plasma BA profiles were analyzed using liquid chromatography-mass spectrometry. A supervised machine learning model was employed to assess the BA profiles, and a simplified tree-based algorithm was developed based on the feature importance to differentiate CIDP, GBS, and HC.

Results: This study included 36 CIDP patients, 70 GBS patients, and 41 HCs. Compared with HCs, CIDP patients showed elevated levels of glycochenodeoxycholic acid (GCDCA, 1,306.64 vs. 614.16 nM, P < 0.001) and glycohyocholic acid (GHCA, 16.95 vs. 6.04 nM, P < 0.001). CIDP patients also exhibited higher levels of GCDCA (1,306.64 vs. 734.44 nM, P < 0.001) and cholenic acid (8.25 vs. 4.41 nM, P < 0.001) compared to GBS patients. The support vector machine model incorporating BA profiles demonstrated strong discriminatory power (AUROC: 0.878), while a simplified tree-based algorithm using four key features achieved better performance (AUROC: 0.929).

Conclusion: BA profiles have potential as diagnostic biomarkers for CIDP, enabling precise and timely patient management.

Keywords: Bile acid; Chronic inflammatory demyelinating polyneuropathy; Machine learning.