Traditional Chinese medicine syndrome patterns and associated factors in adults with type 2 diabetes and metabolic syndrome: a data-driven analysis

Background: Syndrome differentiation is fundamental to traditional Chinese medicine (TCM) diagnosis and treatment, yet its application is complicated by population heterogeneity and disease phenotypes. Data-driven methods to patient stratification offer a pathway to refine syndromic classification beyond expert consensus. This study aimed to identify and characterize TCM syndrome-based patient subgroups and their associated factors in a Hong Kong cohort with comorbid type 2 diabetes mellitus (T2DM) and metabolic syndrome (MetS).

Methods: This cross-sectional study included 505 adults with comorbid T2DM and MetS in Hong Kong. Data on TCM symptoms, clinical profiles, and patient-reported outcomes were collected. To identify patient subgroups, we applied principal component analysis (PCA) and cluster analysis to derive syndrome constructs, and latent class analysis (LCA) to identify latent patient subgroups based on symptom patterns. Multinomial logistic regression was used to explore factors associated with the derived subgroups.

Results: Complementary data-driven approaches identified distinct patient subgroups based on TCM syndrome patterns. PCA derived 5 symptom-based syndromes: Spleen and Kidney Yang Deficiency (17.62%), Qi and Yin Deficiency (22.97%), Kidney Essence Deficiency (10.89%), Yin and Yang Deficiency (22.97%), and Phlegm and Blood Stasis (25.54%). LCA identified 5 latent patient subgroups: Liver Depression and Spleen Deficiency (14.46%), Liver Depression and Spleen Deficiency with Qi and Yin Deficiency (14.46%), Liver and Kidney Yin Deficiency (36.24%), Qi and Yin Deficiency with Phlegm-Blood Stasis (13.27%), and Phlegm and Blood Stasis (21.58%). Multinomial regression indicated that syndrome patterns were significantly associated with multiple factors (all P < 0.05), including sociodemographic characteristics (age, gender, monthly household income), clinical parameters (hypertension, triglyceride levels, basal metabolic rate), lifestyle behaviors (alcohol and caffeine consumption), and patient-reported outcomes (fatigue severity, Pittsburgh Sleep Quality Index score, Audit of diabetes-dependent quality of Life weighted score, and energy intake).

Conclusion: This study demonstrates that complementary data-driven methods, specifically PCA and LCA, can effectively map the heterogeneous landscape of TCM syndromes in patients with comorbid T2DM and MetS. The analysis validates core constructs, including Phlegm and Blood Stasis, and links deficiency syndromes to severe fatigue and poor sleep. Future TCM syndrome research may benefit from prioritizing these empirically derived, multidimensional classifications to inform the development of personalized management strategies.

Clinical trial registration: ClinicalTrials.gov, identifier NCT06703684.