Scholarlink

Introduction <p>Primary 'predominantly antibody deficiencies' (PADs) are rare disorders characterized by increased susceptibility to infections, autoimmunity, allergies, and malignancies. Their low prevalence and heterogeneity often delay diagnosis, increasing morbidity and mortality. This study identifies infection patterns in PAD patients and analyzes predictors of bronchiectasis presence at PAD diagnosis (BPAD), aiming to facilitate early diagnosis and improve prognosis.</p> Methods <p>Using fully monitored data on PAD patients without identified genetic origin from the ESID Registry, infection types and pathogens across PADs were compared (chi-square analysis) and predictive factors for BPAD were identified (generalized mixed-effects logistic regression). Additionally, five machine learning classifiers (logistic regression, (bagged) decision tree, random forest and support vector machines) were trained and evaluated by area under the receiver operating characteristics curve (ROC-AUC), confusion matrices and F1-score.</p> Results <p>Recurrent respiratory infections were predominant in our cohort of 861 patients from 11 centers. Cultures were conducted in only 5.9–12.3% of infections. Encapsulated bacteria were most frequently isolated. The number of organ systems affected by recurrent infections, occurrence of specific serious bacterial infections, number of identified encapsulated bacteria and common variable immunodeficiency disorders (CVID) diagnosis were predictive of BPAD. Machine learning models achieved moderate discrimination (ROC-AUC range 0.679–0.746).</p> Discussion <p>This study highlights the predominance of recurrent and encapsulated bacterial respiratory tract infections in PAD and the underutilization of microbiological cultures. Generalized mixed-effects logistic regression best predicted BPAD. Clinicians should consider immunologic evaluation in patients presenting with serious bacterial infections, multi-system recurrent infections or the repeated isolation of encapsulated bacteria in unusually severe or recurrent infections.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploring the Pathogens in Primary Predominantly Antibody Deficiencies of Unknown Genetic Origin

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月12日

查看全文
  • Barbara Bergmans,
  • Lisanne Fontane Pennock-Janssen,
  • Eugène van Puijenbroek,
  • Roeland van Hout,
  • Jean-Luc Murk,
  • Esther de Vries,
  • Jaap ten Oever,
  • Stefanie Henriet,
  • Riet Strik-Albers,
  • Koen van Aerde,
  • Marina Garcia-Prat,
  • Miriam Gonzalez Amores,
  • Pere Soler-Palacin,
  • Lucia A. Baselli,
  • Rosa Maria Dellepiane,
  • Maria Carrabba,
  • Giovanna Fabio,
  • Leif Gunnar Hanitsch,
  • Renate Krüger,
  • Milos Jesenak,
  • Lenka Kapustova,
  • Otilia Petrovicova,
  • Judith Potjewijd,
  • Suzanne Bazen,
  • Efimia Papadopoulou-Alataki,
  • Kyriaki Chiona,
  • Bram Rutgers,
  • Annick van de Ven

摘要

Introduction

Primary 'predominantly antibody deficiencies' (PADs) are rare disorders characterized by increased susceptibility to infections, autoimmunity, allergies, and malignancies. Their low prevalence and heterogeneity often delay diagnosis, increasing morbidity and mortality. This study identifies infection patterns in PAD patients and analyzes predictors of bronchiectasis presence at PAD diagnosis (BPAD), aiming to facilitate early diagnosis and improve prognosis.

Methods

Using fully monitored data on PAD patients without identified genetic origin from the ESID Registry, infection types and pathogens across PADs were compared (chi-square analysis) and predictive factors for BPAD were identified (generalized mixed-effects logistic regression). Additionally, five machine learning classifiers (logistic regression, (bagged) decision tree, random forest and support vector machines) were trained and evaluated by area under the receiver operating characteristics curve (ROC-AUC), confusion matrices and F1-score.

Results

Recurrent respiratory infections were predominant in our cohort of 861 patients from 11 centers. Cultures were conducted in only 5.9–12.3% of infections. Encapsulated bacteria were most frequently isolated. The number of organ systems affected by recurrent infections, occurrence of specific serious bacterial infections, number of identified encapsulated bacteria and common variable immunodeficiency disorders (CVID) diagnosis were predictive of BPAD. Machine learning models achieved moderate discrimination (ROC-AUC range 0.679–0.746).

Discussion

This study highlights the predominance of recurrent and encapsulated bacterial respiratory tract infections in PAD and the underutilization of microbiological cultures. Generalized mixed-effects logistic regression best predicted BPAD. Clinicians should consider immunologic evaluation in patients presenting with serious bacterial infections, multi-system recurrent infections or the repeated isolation of encapsulated bacteria in unusually severe or recurrent infections.