National and Kapodistrian University of Athens
Artificial intelligence methods in clinical practice and research/development of drugs
Abstract
dc:descriptionBackground: Allergic contact dermatitis (ACD) is a delayed hypersensitivity skin reaction that occurs as immune response to contact haptens and only in previously sensitized individuals. In particular, contact sensitization is the pro-inflammatory skin condition, which is provoked due to hapten exposure without any clinical sign, while ACD lesions are developed after subsequent exposures to the same haptens. Of note, sensitized patients to one allergen have been shown to be at high risk of developing multiple contact allergies. The term multiple sensitizations or polysensitization is, generally, defined as having positive patch test reactions to three or more non-related haptens. Furthermore, occupational contact dermatitis (OCD) is the most frequent type of occupational skin disorders. In industrialized European nations, notifications for suspected OCD accounts for 30-40 % of all occupational diseases, while ACD-related to occupation accounts for 80 % of all occupational dermatoses’ cases. Common high-risk occupations for ACD are healthcare doctors, nurses, dental technicians, pharmaceutical workers, cleaners, domestic workers, resins-plastics workers, construction workers, painters, rubber workers, textile workers, leather tanners, beauticians, nail technicians, and hairdressers. Overall, ACD affects around 5.7 million adults and 4.4 million children annually. So, ACD is a significant global health problem, that affects patients’ physical, emotional, social and mental health, as well as has a great impact on industry productivity and healthcare costs. The prevalence of contact allergy is estimated to be influenced by environmental and genetic factors. However, the etiopathogenesis and mechanisms of contact sensitization are still largely unknown. Consequently, the early diagnosis of ACD and the detection of contact sensitization sources are crucial steps for management of ACD patients. Currently, the in vivo method of epicutaneous patch testing remains the diagnostic gold standard for ACD. The European Baseline Series (EBS) is the core set of contact allergens for patch testing. In particular includes the most prevalent allergen categories that elicit ACD, such as topical medicines, excipients, preservatives, metals, fragrances, dyes-colorants, acrylates and epoxy resins. The allergen inclusion criteria are formed and updated by the European Society of Contact Dermatitis (ESCD) based on national, social, and labor norms. This means that new allergens can be added to the EBS, while allergens with no longer clinical relevance can be deleted from EBS. Therefore, allergen trends vary over time and between geographical regions. Notably, topical drug products have the potential to cause ACD. In particular, drug-induced ACD can occur with topical application of local anesthetics, topical antibiotics, topical corticosteroids, antimycotics, non-steroidal anti-inflammatory drugs (NSAIDs), estrogens and testosterone, as well as their formulations contain excipients and preservatives, that can also act as allergenic factors. Nevertheless, safety data regarding localized skin reactions are not required in formulations for topical drugs, if the ingredients are in line with the Ministry of Health standards. Recently, FDA has recommended the use of diagnostic patch testing with the individual ingredients (i.e. active substances and excipients) as well as the finished product. Overall, the use of patch testing during R&D of topical drugs could be a further proof of their safety. In the big data era, the artificial intelligence (AI) applications are outlined for a variety of medical specialties, such as dermatology, neurology, cardiology, pediatrics, surgery and other More specifically, machine learning (ML) algorithms have the ability to uncover data patterns from extensive repositories of patient information. So, ML models may contribute to ACD early diagnosis and enhance its precision, as well as the treatment strategy. Aims: The present dissertation aims to investigate the sensitization patterns in order to understand the linkage among hypersensitivity to medicines and excipients/preservatives, ACD, patients’ clinical and demographic characteristics using artificial intelligence (AI) methods. Specifically, this study highlights the potential of AI in unveiling unique structures within dermatological data, improving patient monitoring and paving the way for fundamentally healthcare system transformation. The whole investigation was conducted under the prism of how a reproducible approach like patch tech could be applied in R&D of topical drugs, thus enhance their safety profile and minimize ACD incidence. Methods: Retrospective and prospective observational clinical studies were performed investigating ACD prevalence. Overall, clinical, demographic and patch test data from 1200 ACD patients were collected at the National Reference Center for Occupational Dermatoses "Andreas Syggros" Hospital in Athens, Greece. The entire data management was conducted under supervision of the treating physicians of the ACD patients, as well as the Scientific Review Board of the hospital reviewed and approved each study protocol separately. All ethical aspects of the study were fully in line with the Helsinki Declaration (1975, review 2000). The whole study was conducted with respect to medical data confidentiality. Contact sensitization was tested with a panel of 30 contact haptens (allergens) of an adapted EBS. Eligible patients were adults diagnosed with ACD and sensitized to either medicines, or excipients/ preservatives. As exclusion criteria for patch testing the high UV exposure and the chronic use of corticosteroids, immunomodulators, and anti-inflammatory drugs were considered, that might produce false-positive or negative results. In this dissertation, positive patch test results to budesonide 0.01%, caine mix III 7%, and neomycin sulphate 20% (medicines), ethylenediamine dihydrochloride 1% (excipient), as well as formaldehyde 2%, quaternium-15 1%, KATHON 0.02%, thimerosal 0.1%, methyldibromo-glutaronitrile (MDBGN) 0.5%, and paraben mix 16% (preservatives) were studied further. All allergen concentrations refer to % w/w in petrolatum (pet.) except for formaldehyde and KATHON cases in aqua (aq.).Finally, the clinical evaluation took place 48 and 72 hours after the first exposure to allergens and was according to the International Contact Dermatitis Research Group (ICDRG) criteria. Based on an extended MOAHLFA index, information was, also, collected for features, such as M (male), O (occupational dermatitis), A (atopic dermatitis), H (hand dermatitis), L (leg dermatitis), F (face dermatitis), A (age > 40), trunk, neck, and head dermatitis. Moreover, the occupation class was assessed according the International Standard Classification of Occupations (ISCO) guidelines. In particular, patients were divided into five major occupational classes: cleaning service, beauty industry, industry workers, healthcare workers, and office workers. Using the Antera® 3D camera, skin reactions to patch tests were, also, evaluated before patch testing (i.e., 0 h) and 72 h after the first exposure to the contact allergens. The collected image types included redness (haemoglobin), texture, and volume (elevation), which were utilized to calculate these skin parameters. Following data collection, descriptive statistics and classical methods were applied. In this dissertation, different unsupervised machine learning (ML) methods were performed for investigation of data patterns, as well as for data interpretability and efficient illustration of patients’ characteristics in a plan of two dimensions. In particular, multiple correspondence analysis (MCA), categorical principal component analysis (CATPCA), and factor analysis of mixed data (FAMD), which are dimensionality reduction methods, as well as Hierarchical and K-Means Clustering were used in this analysis. It is worth to mention that the selected ML algorithms fitted the goals of this study and were suitable to analyze the investigated variables (numeric, nominal, and ordinal). Data management and analysis were performed with the R software package (https://www.r-project.org), (version 4.3.3). ResultsML algorithms revealed numerous associations among several patients’ features. Regarding MOAHLFA index, MCA models revealed positive correlations among ACD anatomical regions, occupation class and eczema triggers, as well as CATPCA unveiled strong associations between eczema triggers and hand dermatitis. Also, hand dermatitis was primarily associated with disinfectants/cleaning products and industrial materials, followed by cosmetics and topical medications. Moreover, significant relationships were identified between the allergen group and all the affected body sites. In particular, CATPCA showed that thimerosal, formaldehyde and MDBGN patients seemed to develop occupational-related ACD primarily in hand, while KATHON patients developed primarily facial dermatitis. As far as, the occupation class, MCA found positive correlation between occupation class, and allergen group. Moreover, CATPCA revealed that industry and beauty industry workers were found to be primarily sensitized to formaldehyde 2%, followed by KATHON 0.02%, while cleaning service to formaldehyde 2% and healthcare workers to MDBGN 0.5%, followed by thimerosal 0.1%. Furthermore, MCA and CATPCA confirmed classic statistic results, showing that there were strong associations between allergen group and ICDRG 72h. MCA plots revealed additional associations among clinical characteristics, age group, gender, BMI class and smoking, as well as between atopic dermatitis (AD) and family atopy history. In the same direction, CATPCA joint representations revealed further associations between AD - phototype (II, III), as well as AD - age group (> 40). In addition, FAMD revealed associations between ICDRG 72 h and Antera® 3D haemoglobin 72 h showing that Antera® 3D haemoglobin at 72 h might be used as an additional diagnostic tool in patch testing. Finally, cluster analysis of patients’ biometrics and Antera® 3D measurements contributed to understanding of contact sensitization. In particular, HCA reduced the 8‐part data to a 4‐part subcomposition of dissimilar variables, and in agreement with K-means clustering, two clusters were created. Conclusions: The application of ML methods on demographic, clinical and patch test data significantly contributed to the investigation of patients’ profile unveiling unique patterns within the data. ACD interventions should primarily aim at clinical care, education, R&D of drugs and legislation, which can only be successful by the synergy of healthcare system, international scientific committee, and pharmaceutical industry. Overall, AI has the potential to fundamentally transform the healthcare system and improve patient monitoring.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Kyritsi, Aikaterini
- Κυρίτση, Αικατερίνη
Subjects
dc:subject × 19- Αλλεργική δερματίτιδα εξ επαφής
- Τεχνητή νοημοσύνη
- Μηχανική μάθηση
- Patch test
- Τοπικά φάρμακα
- Allergic contact dermatitis
- Artificial intelligence
- Machine learning
- Patch testing
- Multiple correspondence analysis - MCA
- Categorical principal component analysis
- Factor analysis of mixed data - FAMD
- Topical drugs
- Επιστήμες Υγείας
- Ιατρική και Επιστήμες Υγείας
- Ιατρική Πληροφορική
- Health Sciences
- Medical and Health Sciences
- Medical Informatics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/57183