Assessing Cardiac Amyloidosis Subtypes by Unsupervised Phenotype Clustering Analysis.

Background: Cardiac amyloidosis (CA) is a set of amyloid diseases with usually predominant cardiac symptoms, including light-chain amyloidosis (AL), hereditary variant transthyretin amyloidosis (ATTRv), and wild-type transthyretin amyloidosis (ATTRwt). CA are characterized by high heterogeneity in phenotypes leading to diagnosis delay and worsened outcomes.

Objectives: The authors used clustering analysis to identify typical clinical profiles in a large population of patients with suspected CA.

Methods: Data were collected from the French Referral Center for Cardiac Amyloidosis database (Hôpital Henri Mondor, Créteil), including 1,394 patients with suspected CA between 2010 and 2018: 345 (25%) had a diagnosis of AL, 263 (19%) ATTRv, 402 (29%) ATTRwt, and 384 (28%) no amyloidosis. Based on comprehensive clinicobiological phenotyping, unsupervised clustering analyses were performed by artificial neural network-based self-organizing maps to identify patient profiles (clusters) with similar characteristics, independent of the final diagnosis and prognosis.

Results: Mean age and left ventricular ejection fraction were 72 ± 13 years and 52% ± 13%, respectively. The authors identified 7 clusters of patients with contrasting profiles and prognosis. AL patients were distinctively located within a typical cluster; ATTRv patients were distributed across 4 clusters with varying clinical presentations, 1 of which overlapped with patients without amyloidosis; interestingly, ATTRwt patients spread across 3 distinct clusters with contrasting risk factors, biological profiles, and prognosis.

Conclusions: Clustering analysis identified 7 clinical profiles with varying characteristics, prognosis, and associations with diagnosis. Especially in patients with ATTRwt, these results suggest key areas to improve amyloidosis diagnosis and stratify prognosis depending on associated risk factors.