Clinical-cognitive profiles in long COVID and their associations with inflammatory-endothelial biomarkers and vascular measures

Abstract

Long COVID is clinically heterogeneous, and reproducible biomarker patterns remain limited. This cross-sectional study aimed to identify clinical-cognitive profiles in 304 patients with long COVID and examine their associations with inflammatory-endothelial biomarkers and intermediate vascular measures. Latent profile analysis was performed using cognitive, subjective memory, fatigue, sleep, pain, affective, frailty, dyspnea, functional and health-related quality-of-life indicators. Associations between retained profiles and biomarkers or vascular variables were examined using linear models adjusted for age, sex, educational level and body mass index. A four-profile solution was retained, comprising a low global burden profile (42.1%), an intermediate affective-mental profile (22.1%), a high multisystem burden profile (17.9%) and an intermediate physical-functional profile (17.9%), with mean posterior probabilities from 0.924 to 0.965. The two intermediate profiles showed a cross-over SF-36 pattern, with lower MCS and relatively preserved PCS in the affective-mental profile and the reverse pattern in the physical-functional profile. Profiles were associated with ICAM-1, F(3,267) = 5.84, BH-adjusted p  = .008, partial ηp 2  = .065, but not with vascular measures. ICAM-1 remained significant across analyses accounting for classification uncertainty, whereas the endothelin-1 finding was not consistently reproduced. These findings support latent phenotyping as a descriptive strategy for organising long COVID heterogeneity and identify ICAM-1 as the principal hypothesis-generating inflammatory-endothelial correlate of the clinical profiles.