Work

Sex-Specific Phenotypic Architecture in Cardiac Amyloidosis Identified by Explainable Machine Learning

Year: 2026
Type: article
Study design: Observational Study
Abstract: BACKGROUND: Females are under-represented in cardiac amyloidosis (CA), but whether this reflects sex-associated phenotypic heterogeneity or differences in disease prevalence remains unclear. OBJECTIVES: The objectives of the study was to characterize sex-associated differences in presentation and prognosis across CA subtypes using unsupervised clustering. METHODS: We retrospectively analyzed 2,216 patients with confirmed CA enrolled between 2010 and 2022: light-chain amyloidosis (AL) (n = 583), variant transthyretin amyloidosis (ATTRv) (n = 432), and wild-type transthyretin amyloidosis (ATTRwt) (n = 1,201). Self-organizing maps with k-means partitioning were applied separately to each subtype, and conditional inference trees identified key phenotypic discriminators. RESULTS: Females represented 40.8%, 33.8%, and 15.9% of AL, ATTRv, and ATTRwt cohorts, respectively. Clustering identified 4 AL, 6 ATTRv, and 6 ATTRwt phenotypes. Across subtypes, female-predominant clusters generally showed preserved left ventricular ejection fraction, lower cardiac remodeling burden, and less pronounced biomarker elevation. Female-predominant AL clusters included preserved-function and multisystem profiles, whereas ATTRv clusters included neurologic-dominant and older cardiometabolic profiles. In ATTRwt, females were predominantly represented in a phenotype characterized by preserved systolic function and less structural remodeling. Conditional inference trees identified sex as the first discriminator in ATTRwt and a secondary discriminator in AL, whereas ATTRv clusters were primarily discriminated by cardiac variables and sex was not selected. CONCLUSIONS: Unsupervised clustering identified female-predominant phenotypes with less structural cardiac remodeling but distinct multisystem characteristics across CA subtypes. These findings support sex-associated phenotypic heterogeneity in CA and generate hypotheses regarding female underrepresentation in contemporary CA cohorts.
Language: English
Field: Medicine
Open Access status: gold
Estimated APC paid: $2,936
OpenAlex ID: W7236020533
FWCI:
Cites: 41
Cited by:
Related to: 10