E-IJD® - ORIGINAL ARTICLE
|Year : 2022 | Volume
| Issue : 3 | Page : 311
|Forecasting of atopic dermatitis in newborns
Nurangiz Hajiyeva1, Ismail Gafarov2, Adelya Hajiyeva3, Nailya Sultanova1, Tahira Panahova1
1 Department II of Children's Diseases, Azerbaijan Medical University, Baku, Republic of Azerbaijan
2 Department of Medical Physics and Informatics, Azerbaijan Medical University, Baku, Republic of Azerbaijan
3 Department of Premature Infants, Scientific-Research Institute of Pediatrics Named After K. Faradjeva, Baku, Republic of Azerbaijan
|Date of Web Publication||22-Sep-2022|
Department II of Children's Diseases, Azerbaijan Medical University, AZ1022, 167 Samed Vurgun Str., Baku
Republic of Azerbaijan
Source of Support: None, Conflict of Interest: None
| Abstract|| |
Background: Early forecasting of any pathological process is of great significance from both medical and economic point of view. An illness requires much more attention in the light of exhaustion of resources of the body, and a doctor should be maximally aware of the near and far future of a patient. In this regard, the preparation of forecasting programs on a mathematical basis would be a rational and, most probably, the only true approach to the solution of forecasting. Aims and Objectives: The aim of the article is to study the forecasting of atopic dermatitis (AD) in newborns. Methodology: The authors studied 109 clinical and laboratory indicators in children without and with AD. Discriminant analysis was used as an algorithm for the resolution of diagnostic issues. Results: The main indicators acceptable as a forecasting criterion in the formation of AD in children were defined. The sensitivity, specificity, and general diagnostic value of statistically valid differing factors in the formation of AD were studied. Key rules of the forecast were formed after processing all indicators through the KU–Kruskal–Wallis discriminant criterion, a universal computer method. Conclusion: It was concluded that the power of influence of rhinitis, cluster of differentiation 31, mucin 2, and intestinal trefoil factor 3 are higher in the AD model.
Keywords: Correlation analysis, disease, medicine, theory of probability
|How to cite this article:|
Hajiyeva N, Gafarov I, Hajiyeva A, Sultanova N, Panahova T. Forecasting of atopic dermatitis in newborns. Indian J Dermatol 2022;67:311
|How to cite this URL:|
Hajiyeva N, Gafarov I, Hajiyeva A, Sultanova N, Panahova T. Forecasting of atopic dermatitis in newborns. Indian J Dermatol [serial online] 2022 [cited 2022 Sep 30];67:311. Available from: https://www.e-ijd.org/text.asp?2022/67/3/311/356776
| Introduction|| |
Early forecasting of any pathological processes is of great significance from both medical and economic points of view.,, An illness requires much more attention in the light of exhaustion of resources of the body, and a doctor should be maximally aware of the near and far future of a patient., The knowledge about the result of each individual case would ensure the inclusion of the patient in a more or less vulnerable group. Thus, this study discusses the choice of necessary therapeutic approach as well as the assessment of the consequences of various methods of treatment.,,,,
The analysis of literary information indicates that the assessment of illness forecasting based on individual or isolated indicators is not promising.,, In this regard, the preparation of forecasting programs on a mathematical basis would be a rational and, most probably, the only true approach to the solution of forecasting.,,,,,
Any event in clinical medicine can be currently assessed as a probable process and studied by the theory of probability. These conditions would ensure the application of mathematical methods based on the principles of probability with the possibility of positive results in the forecasting of the course of the illness. Recently, more attention is attached to the application of exact sciences in medicine and the establishment of mathematical expert systems on computers. It will help intellectualize the decisions of doctors and prevent unpleasant results based on algorithms, including the facilitation of individual preventive measures.,,
| Materials and Methods|| |
Discriminant analysis was used as an algorithm for the resolution of diagnostic issues. Key rules of the forecast were formed after processing all indicators through KU–Kruskal–Wallis discriminant criterion, a universal computer method. At this stage, 63 forecasting criteria were selected for further analyses. After thorough processing of these 63 indicators by means of discriminant (Pearson's Chi-square) and disperse (analysis of variance test – F-Fischer) methods, statistically and specifically different 28 clinical and 6 laboratory indicators were kept for the formation of the model. Thus, certain indicators have less rate of spread, and we decreased clinical factors to 19 by combining certain indicators under the title of “other antenatal factors” and “other neonatal factors.” Receiver operating characteristic (ROC) analysis was conducted for the evaluation of laboratory indicators (forecasting markers), the cut-off point was determined based on the coordinates of ROC curves, and the sensitivity and specificity of the markers were calculated.
Correct application of probability methods requires correlative non-dependence of forecasting criteria. Thus, overlapping of unilateral results of correlative indicators could lead to a strong disruption of the truth in forecasting. We selected 15 non-correlative indicators during the correlation analysis (p-Spearman) between the results.
First, we used to distribute these indicators under the principle of available/not available, i.e., there is atopic dermatitis (AD) (should not be AD). In this case, it is advisable to apply the Bayes formula. To this end, there are two theories: H1 – Hypothesis 1: this value of the indicator is characteristic of AD positive (AD+) children; H2 – opposite hypothesis 2: this value of the indicator is characteristic of AD negative (AD-) children. Obviously, the sum of the hypotheses is (Eq. 1):
Then, Bayes formula is as follows (Eq. 2):
Whereas P (AD+/H1) is the conditional probability of the formation of AD in a child within the H1 hypothesis; P (AD+) is the unconditional probability of the formation of AD in a child; P (H1/AD+) is the conditional probability of the value of the indicator being in conformity with H1 hypothesis in a child with AD; P (AD−) – 1 − p (AD+) is the unconditional probability of the absence of AD in a child; P (H1/AD+) is the conditional probability of the value of the indicator being in conformity with H1 hypothesis in a child without AD.
Before calculations, it is admitted in the vaguest way that AD+ and AD− probabilities are equal at first glance (Eq. 3):
Taking into account every next factor, a posteriori probabilities are calculated under the Bayes formula: p1, p2,…. It is obvious from the above that the sequence of the calculation for every specific case is conducted under the Bayes formula.
The next stage of the attempt to establish a forecasting program exceeded the borders of the theory of probability and the science of mathematical statistics. Thus, it was impossible to determine a one-digit probability of AD formation on a specific unit of laboratory indicators. Therefore, we had to refer to elements of the theory of “fuzzy” logic.
To this end, we gathered the laboratory results of children with and without AD into a “fuzzy” multitude in the form of an n-sized ellipsoid with focuses being medians of AD+ and AD− multitudes. Then, we divided this ellipsoid into the multitude of interconnected ellipsoids through percentile evaluation. This allowed us to determine the point of connection of the ellipsoid in the form of the one-digit (given any set of laboratory results) on every specific occasion. As the focuses of the ellipsoid are obvious, the distances to the focuses of each point became a metric for us. These figures were normalized and recognized as an unconditional probability, which allowed us to conclude calculations on a forecasting model again under the Bayes formula. According to the requirements of “fuzzy logic” theory, we rejected absolute “0” and absolute “1” for determining the distance to avoid any indefinite situations and made use of 0.01 or 0.05 instead of “0” and 0.99 or 0.95 instead of “1” – figures close to former ones, depending on the shape of the ellipse on each specific occasion.
A special “ADYR-2019” program was worked out in Visual Basic algorithm language based on MS Excel-2013 component for modeling the above. The program management only requires a doctor to enter patient data in necessary boxes within seconds with the help of capabilities at the user level. The program controls the information entered as well, thus a doctor is called to be careful by marking the box with false data as “False.” It should be noted that cases in the absence of any information about patients are also taken into account. In this regard, the result is calculated even in the absence of any information in that box. However, the inclusion of all data selected in the program ensures a more accurate result; thus, it is advisable to mark all boxes. The program calculates the probability of AD in an examined child and submits the result on a chart [Figure 1].
Such information would play a vital role in the selection of future treatment tactics by a doctor. The program is protected from accidental amendments and requires standard MS Office software on a simple configuration computer. The program is at a capacity of < 50 Kbyte. For the sake of friendly use, the ADYR-2019 program is placed on the website www/[email protected]_, and an extra page is added to the program for suggestions and comments. The program is planned for improvement in the future for the purpose of more accurate results based on these suggestions. All indicators studied in the groups with and without AD were taken into consideration for the evaluation of results. According to the quantity of real positive and real negative results, “specificity” and “sensitivity” were evaluated in the ADYR-2019 program.
| Results|| |
In the current research, we sought to define the main indicators acceptable as a forecasting criterion in the formation of AD in children. From this point of view, we studied 109 clinical and laboratory indicators in children without AD (I group – n = 260) and with AD (II group – n = 268). The sensitivity, specificity, and general diagnostic value (GDV) of statistically valid differing factors in the formation of AD were studied [Table 1].
The highest sensitivity (67.9 ± 2.9%) and GDV (64.2%) among the above-mentioned factors are noted in allergic factor on parents; the highest specificity (94.2 ± 1.4%) and the effect of evaluation under positive predictive value (71.2 ± 6.3%) and likelihood ratio of the positive result (2.39 – satisfactory) include seasonal factor. [Table 2] presents the specificity, sensitivity, and GDV of neonatal pathologies and symptoms.
|Table 2: Informative value of some neonatal symptoms and pathologies in the formation of AD|
Click here to view
The highest sensitivity and GDV among the above-mentioned factors is rhinitis; the highest specificity is food allergy and constipation; the highest effect of evaluation under positive predictive value is food allergy and wheezing, and the highest effect of evaluation under negative predictive value is rhinitis. The likelihood ratio of positive results (LR+) is observed in the frequency of symptoms of rhinitis (1.77 – satisfactory) and food allergy (2.08 – satisfactory), and it is of great importance in forecasting AD. A more sensitive ROC analysis was held on the markers of genetic and mucous membrane among laboratory indicators where statistically valid differences are monitored [Figure 2] and [Table 3].
|Figure 2: Results of ROC analysis on CD31 marker also known as platelet endothelial cell adhesion molecule (PECAM-1) and HRH4 (human) markers|
Click here to view
Apparently, a great part of the ROC curve for the cluster of differentiation 31 (CD31) marker is located above the standing line. In other words, the CD31 immunological marker has high specificity and sensitivity in the formation of AD. The area of the ROC curve amounts to 0.685 ± 0.56 (95% energy intake (EI) – 0.578–0.791), and the likelihood ratio was calculated as P = 0.002. As it is obvious from the ROC curve, the area of specificity in the reliability interval at 95% of histamine receptor H4 (HRH4) allergic marker comprises 0.604 ± 0.56 (P = 0.52), and referential indicators vary between 0.494 and 0.714. According to the results, this marker may be considered as the one with high specificity and sensitivity in newborns. [Figure 3] and [Table 4] describe the results of the ROC analysis in the classification of the markers of the mucous membrane.
|Figure 3: Results of ROC analysis in the classification of the markers of the mucous membrane|
Click here to view
The area of specificity of the ROC curve of the G1 marker amounts to 0.551 ± 0.50 (P = 0.307). Referential indicators of this marker in reliability interval at 95% vary between 0.454 and 0.649. According to the ROC curves, mucin 2 (MUC2) and intestinal trefoil factor 3 (ITF3) genetic markers have high specificity and sensitivity in children with AD. Thus, the area of specificity of MUC2 calculated under the ROC curve is equal to 0.692 ± 0.046 (P < 0.001), and its upper and lower limits in reliability interval at 95% are accordingly 0.601 and 0.783. The area of specificity of ITF3 constitutes 0.740 ± 0.048, and referential indicators in reliability interval at 95% vary between 0.645 and 0.834. The ROC analysis of the two indicators, which do not differ in a common group, but statistically validly differ, at the same time, which are noted in medical literature were held in the population studied previously [Figure 4] and [Table 5].
|Table 5: Results of ROC analysis of the statistically valid different indicators|
Click here to view
According to the ROC curve, immunoglobulin E (IgE) and Vitamin D (VitD) indicators are considered markers with low specificity and low informative value. Thus, the area of specificity of IgE is 0.534 ± 0.063, and its referential indicators at 95% EI are 0.411 and 0.658. The upper and lower limit of VitD at 95% EI is defined accordingly as 0.441 and 0.565; its area of specificity is 0.503 ± 0.032. Apparently, neither IgE nor VitD indicators can be recognized as sensitive and specific indicators in the formation of AD. The next stage envisages finding cut-off points – the farthest point from the standing line among interval figures at a variation interval of the indicators in the result of ROC analysis. The calculation was conducted on statistically valid differing indicators in the ROC analysis [Table 6].
It should be noted that a positive solution direction in forecasting AD in children was not observed in pediatric practice. The power of influence of the factors present in the model prepared based on the results was calculated with the help of the Fischer–Snedecor method at the next stage, and the final results are described in [Table 7].
|Table 7: Informative value of factors participating in AD forecasting model in newborns|
Click here to view
Based on the factors studied, ITF has the highest informative value, thus the fact is the quantity of this factor (n = 101 persons) is higher than 25.0; its specificity, informative value, and GDV are, respectively, 50.8 ± 6.4%, 95.0 ± 3.4%, and 68.3 ± 4.6%; the effect of evaluation under positive and negative predictive value is accordingly 93.9 ± 4.2 and 55.9 ± 6.0, which proves that this indicator is of great importance for forecasting of AD. According to the results of the correlation analysis, direct proportionality was found between the zone factor and allergy on the father, allergy on the mother, and food allergy; and inverse proportionality was found with the constipation factor.
| Discussion|| |
There are no specific diagnostic tests for AD. Diagnosis of the disorder is based on specific criteria that take into account the patient's medical history and clinical manifestations., AD occurs as a result of complex interactions between genetic factors, the environment, infectious agents, defects in the barrier function of the skin, and impaired immune response. Knowledge of etiological factors allows for adequate therapy, as well as primary, secondary, and tertiary prevention of this disease in children. According to research, the most important in the etiology of hypertension is burdened heredity of allergic pathology in families of children, as well as environmental conditions with factors of their influence, on which the child grows, develops, and stays. New information on hypertension indicates that both structural abnormalities of the skin and immune dysregulation play an important role in the pathophysiology of the disease. Therefore, the optimal treatment of hypertension requires a multifaceted approach aimed at healing and protecting the skin barrier, as well as the impact on the complex immunopathogenesis of the disease.
Direct proportionality is observed between allergy factor on the father and wheezing, conjunctivitis, rhinitis, food allergy, sensitive skin, trouble sleeping, MUC2, and ITF3. It shows that in the case of an allergy factor on the father, a newborn is at high risk of formation of such neonatal and antenatal symptoms. Direct proportionality exists between allergy factor on the mother and autoimmune illnesses, wheezing, rhinitis, food allergy, and CD31 factors. This dependence shows that in the case of an allergy factor on the mother, there is a high risk of autoimmune illnesses, at the same time, symptoms of wheezing, rhinitis, and food allergy may be noted in a newborn.
Children born via cesarean section demonstrate a higher quantity of HRH4 and MUC2 markers, thus their thickness is directly proportional to this factor. Wheezing observed in newborns has direct proportionality to conjunctivitis, rhinitis, food allergy, sensitive skin, trouble sleeping, constipation, CD31, HRH4, and MUC2. According to correlation results, damage to mucous membrane is observed through conjunctivitis, rhinitis, food allergy, sensitive skin, and trouble sleeping symptoms in newborns; and the increase of markers of mucous membrane proves it once again.
In a clinical setting, the proposed models can be used in the development of treatment tactics for newborns with AD.
| Conclusions|| |
To sum up, the power of influence of rhinitis, CD31, MUC2, and ITF3 factors are higher in the AD model. However, since the informative value of all factors used in forecasting is accompanied by statistical validity, the use of all factors in the model is advisable for the sake of more accurate results. The most sensitive factors are rhinitis; the highest specificity is food allergy and constipation; the highest effect of evaluation under positive predictive value is food allergy and wheezing; and the highest effect of evaluation under negative predictive value is rhinitis.
Thus, damage to intestinal mucosa with the influence of perinatal risk factors is an initiating and significant factor of atopic allergy. The identification of early and significant risk factors in the formation of AD would allow classifying a risk group with postnatal allergic pathology and preparing a set of treatment and preventive measures for the prevention of the formation of AD. The results obtained can be used in the selection of future treatment tactics by a doctor.
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. A study was approved by National Ethics Commission of the Ministry of Health of the Republic of Azerbaijan, October 23, 2021, No. 795-O.
Declaration of patient consent
The authors certify that they have obtained all appropriate patient consent forms. In the form, the patients have given their consent for their clinical information to be reported in the journal. The patients understand that their names and initials will not be published and due efforts will be made to conceal their identity, but anonymity cannot be guaranteed.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Venter C, Agostoni C, Arshad SH, Ben-Abdallah M, Du Toit G. Dietary factors during pregnancy and atopic outcomes in childhood. Pediatr Allergy Immunol 2020;31:889-912.
Gabes M, Chamlin S, Lai S, Cella D, Mancini AJ, Apfelbacher CJ. Development of a validated short-form of the childhood atopic dermatitis impact scale, the CADIS-SF15. J Eur Acad Dermatol Venereol 2020;34:1773-8.
van der Leek AP, Bahreinian S, Chartier M, Dahl ME. Maternal distress during pregnancy and recurrence in early childhood predicts atopic dermatitis and asthma in childhood. Chest 2020;158:57-67.
Kang C-M, Chiang B-L, Wang L-C. Maternal nutritional status and development of atopic dermatitis in their offspring. Clin Rev Allergy Immunol 2021;61:128-55.
Sánchez-García S, Rial MJ, Domínguez-Ortega J. Long and winding road: From infant wheeze to adult asthma. Cur Opin Pulm Med 2020;26:3-9.
Rehbinder EM, Endre KMA, Carlsen KCL, Asarnoj A. Predicting skin barrier dysfunction and atopic dermatitis in early infancy. J Allergy Clin Immunol 2020;8:664-73.
Gerner T, Halling AS, Rasmussen RM, Haarup RN, Hjorslev KM, Menné BC, et al
. Barrier dysfunction in atopic newborns study (BABY): Protocol of a Danish prospective birth cohort study. BMJ Open 2020;12:1-7.
Hernández CD, Casanello P, Harris PR, Castro-Rodríguez JA, Iturriaga C, Perez-Mateluna G, et al
. Early origins of allergy and asthma (ARIES): Study protocol for a prospective prenatal birth cohort in Chile. BMC Pediatr 2020;15:1-9.
Alfonso J, Pérez S, Bou R, Amat A, Ruiz I, Mora A, et al
. Asthma prevalence and risk factors in school children: The RESPIR longitudinal study. Allergol Immunopathol 2020;48:223-31.
Looman KIM, van Meel ER, Grosserichter-Wagener C, Vissers FJM, Klingenberg JH, de Jong NW, et al
. Associations of Th2, Th17, Treg Cells, and IgA(+) memory B cells with atopic disease in children: The generation R study. Allergy 2020;75:178-87.
Kumar T, Pandey R, Singh Chauhan N. Hypoxia inducible factor-1α: The curator of gut homeostasis. Front Cell Infect Microbiol 2020;10:1-8.
Joneja JM. Infant food allergy: Where are we now? J Parenter Enteral Nutr 2012;36:49-55.
Oddy WH. Breastfeeding, childhood asthma, and allergic disease. Ann Nutr Metab 2017;70:26-36.
Hong S, Choi WJ, Kwon HJ, Cho YH, Yum HY, Son DK. Effect of prolonged breast-feeding on risk of atopic dermatitis in early childhood. Allergy Asthma Proc 2014;35:66-70.
Mazzocchi A, Venter C, Maslin K, Agostoni C. The role of nutritional aspects in food allergy: Prevention and management. Nutr 2017;9:1-12.
Smejda K, Polanska K, Merecz-Kot D, Krol A. Maternal stress during pregnancy and allergic diseases in children during the first year of life. Resp Care 2018;63:70-6.
Schoch JJ, Monir RL, Satcher KG, Harris J, Triplett E, Neu J. The infantile cutaneous microbiome: A review. Pediatr Dermatol 2019;36:574-80.
Underwood MA. Should we treat every infant with a probiotic? Minerva Pediatr 2019;71:253-62.
Yang G, Han YY, Forno E, Acosta-Pérez E, Colón-Semidey A, Alvarez M, et al
. Under-diagnosis of atopic dermatitis in Puerto Rican children. World Allergy Organ J 2019;12:1-5.
Ku MS. Neonatal phototherapy: A novel therapy to prevent allergic skin disease for at least 5 years. Neonatol 2018;114:235-41.
Kabashima K, Matsumura T, Komazaki H, Kawashima M. Trial of nemolizumab and topical agents for atopic dermatitis with pruritus. N Eng J Med 2020;383:141-50.
Ahadov RF, Gafarov IA. Modern ideas about the influence of risk factors on mortality of patients receiving hemodialysis treatment with concomitant metabolic syndrome. Eur Res 2016;4:156-8.
Safarova IA, Kaziyev AY, Gafarov IA, Jafarova GA. Study of informative value and prognostic significance of some cytokines and antimicrobial peptides for early detection of metastases in patients with cervical cancer. Theor Appl Sci (Austr) 2018;6:201-6.
Ahmadi H, Gholamzadeh M, Shahmoradi L, Nilashi M, Rashvand P. Diseases diagnosis using fuzzy logic methods: A systematic and meta-analysis review. Comput Methods Programs Biomed 2018;161:145-72.
Belyalov FI. Forecasting and Scales in Medicine. Moscow: Medproesineform; 2016.
Zhumalina AK, Tusupkaliev BT, Zame YA, Voloshina LV, Darzhanova KB. Clinical and immunological aspects of newborn adaptation born from mothers with intrauterine infection. Period Tche Quimica 2020;17:656-666.
Gritsenko DA, Orlova OA, Linkova NS, Khavinson VK. Transcription Factor p53 and Skin Aging. Adv Gerontol 2017;7:114-119.
[Figure 1], [Figure 2], [Figure 3], [Figure 4]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7]
| Article Access Statistics|
| Viewed||192 |
| Printed||8 |
| Emailed||0 |
| PDF Downloaded||0 |
| Comments ||[Add] |