Unlocking the potential of nailfold videocapillaroscopy in diagnosing and staging wild-type transthyretin amyloidosis: A preliminary approach

Abstract

Background: Wild-type transthyretin amyloidosis (ATTRwt) is a serious condition. At early stages, symptoms resemble those of heart failure with preserved ejection fraction (HFpEF). Our aim was to perform software-supported nailfold videocapillaroscopy (NVC) analysis to identify hallmarks useful for diagnosis and build machine learning (ML)-based models to assess severity.

Methods: Thirty-two ATTRwt patients underwent NVC. Nineteen initiated TTR-stabilizing therapy and had a new NVC 12 months afterwards. Forty-one capillary-related variables were analyzed. Thirty NVCs were randomly chosen to train models to discriminate between poorer or less poor prognosis according to N-terminal pro-B-type natriuretic peptide (NT-proBNP) or Cheng score (cut-offs: 2000pg/mL and 4 points, respectively). The remaining 21 NVCs were used for validation purposes. A control population of 99 patients with heart failure with preserved ejection fraction (HFpEF) but without signs of amyloidosis was included.

Results: A profound disorganization in the nailfold capillary architecture was generally observed. The models achieved accuracies of 0.81 and 0.90, respectively, in predicting disease severity. An additional model designed to distinguish a profile suggestive of amyloidosis (vs. HFpEF controls) achieved an accuracy of 0.73.

Conclusions: NVC-based ML models may contribute to early diagnosis and staging of ATTRwt.

Keywords: Amiloidosis por transtirretina wild type; Cheng score; Escala de Cheng; Machine learning-based model; Modelo basado en aprendizaje automático; N-terminal pro-B-type natriuretic peptide; Nailfold videocapillaroscopy; Péptido natriurético tipo B N-terminal; Videocapilaroscopia del lecho ungueal; Wild-type transthyretin amyloidosis.