Machine learning in the diagnosis of asthma phenotypes during coronavirus disease 2019 pandemic.
COVID‐19 pandemic
machine learning
nonsteroidal anti‐inflammatory drug (NSAID)–exacerbated respiratory disease (NERD)
nonsteroidal anti‐inflammatory drug tolerant asthma (NTA)
oral aspirin challenge
Journal
Clinical and translational allergy
ISSN: 2045-7022
Titre abrégé: Clin Transl Allergy
Pays: England
ID NLM: 101576043
Informations de publication
Date de publication:
Oct 2022
Oct 2022
Historique:
received:
18
05
2022
accepted:
08
09
2022
entrez:
21
10
2022
pubmed:
22
10
2022
medline:
22
10
2022
Statut:
epublish
Résumé
During the coronavirus disease 2019 (COVID-19) pandemic, it has become a pressing need to be able to diagnose aspirin hypersensitivity in patients with asthma without the need to use oral aspirin challenge (OAC) testing. OAC is time consuming and is associated with the risk of severe hypersensitive reactions. In this study, we sought to investigate whether machine learning (ML) based on some clinical and laboratory procedures performed during the pandemic might be used for discriminating between patients with aspirin hypersensitivity and those with aspirin-tolerant asthma. We used a prospective database of 135 patients with non-steroidal anti-inflammatory drug (NSAID)-exacerbated respiratory disease (NERD) and 81 NSAID-tolerant (NTA) patients with asthma who underwent OAC. Clinical characteristics, inflammatory phenotypes based on sputum cells, as well as eicosanoid levels in induced sputum supernatant and urine were extracted for the purpose of applying ML techniques. The overall best ML model, neural network (NN), trained on a set of best features, achieved a sensitivity of 95% and a specificity of 76% for diagnosing NERD. The 3 promising models (i.e., multiple logistic regression, support vector machine, and NN) trained on a set of easy-to-obtain features including only clinical characteristics and laboratory data achieved a sensitivity of 97% and a specificity of 67%. ML techniques are becoming a promising tool for discriminating between patients with NERD and NTA. The models are easy to use, safe, and achieve very good results, which is particularly important during the COVID-19 pandemic.
Sections du résumé
Background
UNASSIGNED
During the coronavirus disease 2019 (COVID-19) pandemic, it has become a pressing need to be able to diagnose aspirin hypersensitivity in patients with asthma without the need to use oral aspirin challenge (OAC) testing. OAC is time consuming and is associated with the risk of severe hypersensitive reactions. In this study, we sought to investigate whether machine learning (ML) based on some clinical and laboratory procedures performed during the pandemic might be used for discriminating between patients with aspirin hypersensitivity and those with aspirin-tolerant asthma.
Methods
UNASSIGNED
We used a prospective database of 135 patients with non-steroidal anti-inflammatory drug (NSAID)-exacerbated respiratory disease (NERD) and 81 NSAID-tolerant (NTA) patients with asthma who underwent OAC. Clinical characteristics, inflammatory phenotypes based on sputum cells, as well as eicosanoid levels in induced sputum supernatant and urine were extracted for the purpose of applying ML techniques.
Results
UNASSIGNED
The overall best ML model, neural network (NN), trained on a set of best features, achieved a sensitivity of 95% and a specificity of 76% for diagnosing NERD. The 3 promising models (i.e., multiple logistic regression, support vector machine, and NN) trained on a set of easy-to-obtain features including only clinical characteristics and laboratory data achieved a sensitivity of 97% and a specificity of 67%.
Conclusions
UNASSIGNED
ML techniques are becoming a promising tool for discriminating between patients with NERD and NTA. The models are easy to use, safe, and achieve very good results, which is particularly important during the COVID-19 pandemic.
Identifiants
pubmed: 36267429
doi: 10.1002/clt2.12201
pii: CLT212201
pmc: PMC9579891
doi:
Types de publication
Journal Article
Langues
eng
Pagination
e12201Informations de copyright
© 2022 The Authors. Clinical and Translational Allergy published by John Wiley & Sons Ltd on behalf of European Academy of Allergy and Clinical Immunology.
Déclaration de conflit d'intérêts
The authors declare that they have no competing interests.
Références
Allergy. 2021 Aug;76(8):2407-2419
pubmed: 33314245
J Allergy Clin Immunol. 2012 Jun;129(6):1636-46
pubmed: 22418066
Allergy. 2020 Jul;75(7):1649-1658
pubmed: 32012310
Allergy. 2019 May;74(5):922-932
pubmed: 30446997
Allergy. 2020 Feb;75(2):311-325
pubmed: 31309578
Int J Med Inform. 2017 Dec;108:1-8
pubmed: 29132615
J Allergy Clin Immunol. 2017 Mar;139(3):819-825.e6
pubmed: 27567328
Respir Med. 2019 Jun;152:51-59
pubmed: 31128610
J Allergy Clin Immunol Pract. 2021 Jun;9(6):2255-2261
pubmed: 33618053
Eur Respir J Suppl. 2002 Sep;37:1s-2s
pubmed: 12361359
J Allergy Clin Immunol. 2006 Feb;117(2):312-8
pubmed: 16461132
Respirology. 2006 Jan;11(1):54-61
pubmed: 16423202
J Allergy Clin Immunol Pract. 2021 Mar;9(3):1133-1141
pubmed: 33065369
Allergy. 2007 Oct;62(10):1111-8
pubmed: 17521312
J Allergy Clin Immunol. 2015 Mar;135(3):676-81.e1
pubmed: 25282015
Clin Exp Allergy. 2015 Dec;45(12):1779-89
pubmed: 26449970
Ann N Y Acad Sci. 2017 Jan;1387(1):153-165
pubmed: 27627195
Allergy. 2019 Jan;74(1):28-39
pubmed: 30216468