Abstract
Background
Infants are at a high risk of food allergy (FA); however, owing to the lack of commonly recognized simple and noninvasive diagnostic tools, definitive diagnosis of infantile FA is challenging. In this study, we aimed to establish a diagnostic model comprising highly suggestive indicators to facilitate the early identification of infantile FA.
Methods
In this case-control study, we enrolled two groups of infants with suspected FA. FA diagnoses were confirmed through oral food challenges (OFCs). The training set, which included infants enrolled between 2022 and 2023, was used to develop a logistic regression diagnostic model and perform internal cross-validation. The testing set comprising previous cases between 2016 and 2021 was used to perform external validation. We assessed the discrimination and calibration of the diagnostic model using the area under the curve (AUC), Hosmer-Lemeshow goodness-of-fit test, calibration curve, and decision curve analysis (DCA).
Results
We identified variables for the diagnostic model, including eczema, stool form, hematochezia, failure to thrive (FTT), respiratory symptoms, and family history of allergic diseases. The AUCs for the training set, internal cross-validation, and external validation were 0.86 (0.83–0.90), 0.86 (0.76–0.95), and 0.87 (0.83–0.92), respectively, showing good diagnostic performance of the model. The Hosmer-Lemeshow goodness-of-fit test results and calibration curve showed that the model had good calibration. DCA results showed a high net benefit value in clinical decision-making.
Conclusions
The diagnostic model, constructed with the six aforementioned variables, can serve as a simple and noninvasive diagnostic tool for clinicians to effectively distinguish FA from similar diseases.
Keywords: Diagnostic model, eczema, food allergy (FA), hematochezia, infant
Highlight box.
Key findings
• This study developed and validated a diagnostic model for infantile food allergy (FA) using six non-invasive clinical variables: eczema, stool form, hematochezia, failure to thrive, respiratory symptoms, and family history of allergic diseases.
• The model in this study had good diagnostic performance and demonstrated effectiveness for clinical decision-making, providing a simple and convenient diagnostic tool for clinicians to effectively distinguish FA from similar diseases.
What is known and what is new?
• Current diagnosis of infantile FA remains challenging due to the impracticality of the gold-standard oral food challenge and the absence of reliable biomarkers, highlighting a critical unmet need for accessible diagnostic tools.
• This study established and initially validated a simple and noninvasive diagnostic model based on routinely assessed clinical symptoms and risk factors, enabling early and objective identification of FA in infants.
What is the implication, and what should change now?
• The diagnostic model is a noninvasive and standardized diagnostic tool for infantile FA beyond medical conditions and can be adapted by primary healthcare facilities. We intend to simplify the model into a clinician-friendly calculator to facilitate widespread adoption.
• Further prospective multicenter studies that incorporate additional predictive variables are needed to refine the diagnostic model.
Introduction
Food allergy (FA) is characterized by a specific abnormal immune response that recurs upon exposure to particular food protein antigens (1), presenting as multi-system manifestations. Infants are at a high risk of FA, with a high prevalence of 5–10% (2,3). Allergic reactions stemming from inadvertent allergen exposure are frequent because of unidentified allergens (4). Infancy constitutes a critical stage of growth and development, and excessive avoidance of suspected allergens can increase dietary restrictions, potentially leading to malnutrition (5), thereby aggravating the social and psychological burden of this disease (6). Accurate diagnosis is essential to help infants and their families to promptly and reasonably avoid allergens, enhance treatment compliance, and reduce adverse outcomes.
However, a definitive diagnosis of FA is challenging (7). The gold standard method, an oral food challenge (OFC), is complex, time-consuming, and sometimes potentially life-threatening owing to possible anaphylaxis and anaphylactic shock occurring during an OFC (8). OFCs are not widely used in primary healthcare facilities because of these limitations. Therefore, a simple and universal diagnostic tool is urgently needed. Current diagnostic tools for infantile FA, such as the Cow’s Milk Related Symptom Score (CoMiSS), show regional variability and suboptimal accuracy, which is a particular concern in Asian countries (9,10), while other models suffer from inconsistent diagnostic criteria or poor performance (11,12). To address this gap, we aimed to establish a diagnostic model for the early identification of FA in infants.
Some clinical features and risk factors for infantile FA have been suggested. First, the clinical manifestations of FA vary from infancy to adulthood; gastrointestinal symptoms are the main manifestations of FA in infants, and skin and respiratory manifestations may also be present (13). Second, some laboratory examinations (mainly those for total and food-specific IgE) indicate the presence of potential FAs (14). However, these tests have limited diagnostic value in infants. Moreover, their clinical utility is further constrained by the requirement for invasive sampling. Third, genetic and certain environmental factors may be associated with a high risk of developing FA (15,16). Thus, exploring a comprehensive noninvasive diagnostic model incorporating the above clinical data may aid in the early diagnosis of infantile FA.
In this study, we enrolled infants with suspected FA to compare clinical data between FA and non-FA groups, screen for suggestive indicators, and establish and validate a clinical diagnostic model for FA in infants, which may contribute to the early identification and timely treatment of infantile FA. We present this article in accordance with the TRIPOD reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-aw-783/rc).
Methods
Study design
This was a case-control study. Infants treated at Peking University Third Hospital between January 2022 and June 2023 were selected as the training set for the initial establishment of the FA diagnostic model. Infants admitted to Peking University Third Hospital between January 2016 and December 2021 were selected as the testing set to further evaluate the accuracy and goodness-of-fit of the model. The OFC is considered the gold standard method for diagnosing FA. Since the participants in this study were infants, who were less affected by psychological factors, we used the open OFC method for diagnosing FA. Following a 2–4-week elimination of suspected allergenic foods with symptom resolution observed, we conducted OFCs according to the standardized protocol established in 2020 (17). Physicians performing OFCs were blinded to participants’ clinical information to reduce potential assessment bias. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Peking University Third Hospital Medical Science Research Ethics Committee (approval No. M2020405, approval date: April 28, 2021), and the guardians of the enrolled infants signed an informed consent form.
The inclusion criteria for the FA group were as follows: (I) full-term infants aged <12 months; (II) presence of one or more symptoms, such as rashes, runny nose, nasal congestion, sneezing, coughing, wheezing, vomiting, reflux, diarrhea, hematochezia, constipation, mucous stool, failure to thrive (FTT), feeding difficulties (FD), crying, frequent night awakenings, and repeated eye rubbing; and (III) symptom improvement after avoiding suspicious allergens and a positive OFC result. The exclusion criteria were as follows: (I) family members refused to sign an informed consent form or (II) infants with complications such as hepatic and renal insufficiency, hematological diseases, autoimmune diseases, congenital diseases, inherited metabolic diseases, neuromuscular diseases, and immunodeficiency diseases.
The inclusion criteria for the non-FA group were as follows: (I) full-term infants aged <12 months; (II) presence of one or more of the aforementioned symptoms of suspected FA; and (III) no improvement in symptoms after avoiding suspicious allergens and a negative OFC result. The exclusion criteria for the non-FA group were similar to those for the FA group.
Collected data
We identified common clinical manifestations (13,18,19), risk factors (15,16,20,21), and laboratory examinations (22) for FA in infants through a literature review. Detailed data were obtained via inquiries, questionnaire surveys, medical record reviews, and regular follow-ups.
The collected data included the following: (I) demographic data: age, sex, gestational age, delivery mode, body length, and weight at different time points (birth, onset of symptoms, and outpatient visit); (II) symptoms and their severity—(i) skin: eczema and Scoring Atopic Dermatitis (SCORAD) index; (ii) digestive system: vomiting, reflux, stool form, defecation frequency, constipation, hematochezia, mucous stool, FD, and FTT; (iii) respiratory system: runny nose, stuffy nose, sneezing, coughing, and wheezing; and (iv) behaviors that may be associated with discomfort: eye rubbing, crying, and frequent night awakenings; (III) possible risk factors: genetic factors, such as family history of allergic diseases and autoimmune diseases, and environmental factors, including infection history, antibiotic exposure, and the aforementioned demographic characteristics (21); and (IV) laboratory examinations: total IgE, food-specific IgE, and fecal occult blood test.
Definition and assessment of symptoms
Eczema is common in infants and refers to a clinical feature shared by a group of heterogeneous diseases, including FA (23). Skin changes caused by eczema are characterized by erythema, edema, dryness, desquamation, and hyperkeratosis, accompanied by itching, burning, or tingling sensations (24). The severity of eczema was categorized as absent (0 point), mild (＞0 to < 25 points), moderate (25–50 points), or severe (>50 points), based on the SCORAD index (25).
Digestive symptoms and assessment were defined as follows. (I) Vomiting is the forceful expulsion of gastric contents from the oral cavity, regulated by the autonomic nervous system. It frequently accompanies nausea and retching owing to a nervous system emetic reflex and retrograde contractions in the upper gastrointestinal tract (26). (II) Reflux is the return of gastric contents into the esophagus and beyond (27,28). (III) Constipation is characterized by a decreased defecation frequency and dry stools. The diagnosis was made based on the Rome IV criteria (29), and dry stools were evaluated according to the Brussels Infant and Toddler Stool Scale (BITSS) [The Bristol Stool Form Scale (BSFS) types 1–3]. In addition, we described the stool forms as formed or soft (BSFS type 4), loose (BSFS types 5 and 6), and watery (BSFS type 7) (30). (IV) Hematochezia included gross bloody stools and positive occult fecal blood samples. (V) FD (31) involved age-inappropriate impaired oral intake and are associated with FA, gastrointestinal diseases (32), feeding skills, and psychosocial dysfunction. The diagnosis is based on international standards (33,34). (VI) FTT may be a multifactorial disorder associated with food-induced gastrointestinal FA. The diagnosis of FTT requires meeting one or more of the following criteria: (i) weight-for-age or weight-for-length below the third percentile; (ii) multiple consecutive assessments of weight-for-age or weight-for-length below the fifth percentile; or (iii) weight loss exceeding two percentiles on the growth charts for age. Growth charts and Z-scores were obtained from the official website of the World Health Organization (35).
Night awakenings, possibly related to infant colic and discomfort, are common among infants. In this study, night spanned from 6:00 PM to 6:00 AM the next day, and frequent night awakenings were defined as at least 14 awakenings per week (36). Repeated eye rubbing and unexplained crying may indicate factors causing discomfort in infants. The above symptoms caused by infection and accidental factors usually last <1 week; therefore, we defined discomfort symptoms as the persistence of at least one of the above symptoms for >1 week to avoid the inclusion of infection and accidental factors.
Sustained coughing, wheezing, sneezing, and runny or stuffy nose may suggest an allergic disease. The presence of at least one of the abovementioned respiratory symptoms that persisted for >1 week was recorded to avoid the inclusion of common non-FA-related factors (37). All assessments of clinical symptoms were conducted by blinded clinicians using standardized objective criteria, with no access to participant information or modeling details.
Sample size calculation and statistical analysis
The sample size of the diagnostic prediction model is typically at least 10 times the number of study variables. In this study, we included 23 variables and calculated that at least 230 cases of FA were required. The ratio of samples in the training set to those in the testing set is usually 8:2–6:4. We used EpiData 3.1 (https://www.epidata.dk/download.php) for data entry and accuracy testing; additionally, we used IBM SPSS for Windows version 23.0 (IBM Corp., Armonk, NY, USA) and R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria) for statistical analyses and the generation of accurate charts, respectively. Count data are expressed as constituent ratios or percentages (%) and non-normally distributed data as medians with interquartile ranges. A P value of <0.05 was considered statistically significant. Initially, for the training set, we established a diagnostic model for FA in infants using logistic regression. A nomogram was used to visualize the model. Subsequently, to test the overall fit and efficacy of the diagnostic model, we performed internal ten-fold cross-validation using the training set and external validation using the testing set. We assessed the predictive performance of the model using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. Additionally, we used the Hosmer-Lemeshow goodness-of-fit test, a calibration curve, and a Brier Score to evaluate the calibration of the model. We conducted a decision curve analysis (DCA) to assess the utility of the diagnostic model in clinical decision-making. Variables with >30% missing data were excluded, and multiple imputation was performed for all remaining covariates prior to statistical analysis.
Results
Participant enrollment
Overall, 572 infants were included in the training set, with 286 in the FA group and 286 in the non-FA group (Figure 1). Baseline demographic characteristics are presented in Table 1. In total, 260 infants were included in the testing set, with 137 and 123 in the FA and non-FA groups, respectively (Figure 1). The role of the two sets, model establishment, and validation process described below are visually presented in Figure 1. During data analysis, IgE data were removed because >50% of the infants did not undergo IgE testing. No statistically significant differences were observed in baseline clinical characteristics between the training set and the testing set.
Figure 1.
Flow diagram of the enrollment and process for the establishment and validation of the model. The training set was used to train variables within the model. Internal validation was used to assess the status and convergence of the model during the training. The testing set (external validation) was used to evaluate the generalizability of the model, that is, its practicability. FA, food allergy; OFC, oral food challenge; OFC+, positive OFC result; OFC−, negative OFC result.
Table 1. Baseline demographic characteristics and univariate logistic regression analysis results in the training set.
| Variable | FA (n=286) | No-FA (n=286) | P | OR | 95% CI |
|---|---|---|---|---|---|
| Age, months | 3.93 [2.13, 5.60] | 4.1 [2.17, 6.47] | 0.28 | 0.97 | 0.92, 1.03 |
| Male | 147 (51.40) | 147 (51.40) | >0.99 | 1.00 | 0.72, 1.39 |
| Gestational age, weeks | 39.43 [38.43, 40.00] | 39.3 [38.4, 40.0] | 0.63 | 1.03 | 0.90, 1.18 |
| Birth body length, cm | 50.0 [49.0, 51.0] | 50.0 [48.0, 50.0] | 0.02 | 1.10 | 1.01, 1.21 |
| Birth weight, kg | 3.25 [3.03, 3.60] | 3.10 [2.90, 3.50] | 0.003 | 1.85 | 1.23, 2.78 |
| Cesarean section | 128 (44.76) | 126 (44.06) | 0.87 | 0.97 | 0.70, 1.35 |
| Eczema | |||||
| Absent | 91 (31.82) | 145 (50.70) | – | – | – |
| Mild | 105 (36.71) | 88 (30.77) | 0.001 | 1.90 | 1.29, 2.80 |
| Moderate | 64 (22.38) | 38 (13.29) | <0.001 | 2.68 | 1.66, 4.33 |
| Severe | 26 (9.09) | 15 (5.24) | 0.004 | 2.76 | 1.39, 5.49 |
| Vomiting | 56 (19.58) | 37 (12.94) | 0.03 | 1.64 | 1.04, 2.58 |
| Reflux | 33 (11.54) | 9 (3.15) | <0.001 | 4.01 | 1.88, 8.55 |
| Constipation | 7 (2.45) | 6 (2.10) | 0.78 | 1.17 | 0.39, 3.53 |
| Defecation frequency (times/day) | |||||
| ≤2 | 65 (22.73) | 165 (57.69) | – | – | – |
| 3–5 | 64 (22.38) | 37 (12.94) | <0.001 | 4.39 | 2.67, 7.21 |
| ≥6 | 157 (54.90) | 84 (29.37) | <0.001 | 4.75 | 3.21, 7.01 |
| Stool form | |||||
| Formed or soft | 68 (23.78) | 181 (63.29) | – | – | – |
| Loose | 58 (20.28) | 68 (23.78) | <0.001 | 2.27 | 1.45, 3.55 |
| Watery | 160 (55.94) | 37 (12.94) | <0.001 | 11.51 | 7.32, 18.11 |
| Hematochezia | 198 (69.23) | 70 (24.48) | <0.001 | 6.94 | 4.80, 10.04 |
| Mucous stool | 198 (69.23) | 83 (29.02) | <0.001 | 5.50 | 3.85, 7.87 |
| FTT | 91 (31.81) | 42 (14.69) | <0.001 | 2.71 | 1.80, 4.09 |
| FD | 57 (19.93) | 47 (16.43) | 0.28 | 1.27 | 0.83, 1.94 |
| Respiratory symptoms | 104 (36.36) | 30 (10.49) | <0.001 | 4.88 | 3.11, 7.64 |
| Discomfort symptoms | 154 (53.85) | 67 (23.43) | <0.001 | 3.81 | 2.66, 5.46 |
| Family history of allergic diseases | 210 (73.43) | 105 (36.71) | <0.001 | 4.76 | 3.34, 6.80 |
| Family history of autoimmune diseases | 17 (5.94) | 6 (2.10) | 0.03 | 2.95 | 1.21, 8.73 |
| Infection history | 33 (11.54) | 27 (9.44) | 0.26 | 2.59 | 0.55, 18.23 |
| Antibiotic exposure | 29 (10.14) | 26 (9.09) | 0.35 | 0.45 | 0.06, 2.22 |
Data are presented as number (%) or median [interquartile range]. CI, confidence interval; FA, food allergy; FD, feeding difficulties; FTT, failure to thrive; OR, odds ratio.
Univariate logistic analysis of fa occurrence in the training set
The results of the univariate analysis revealed differences in birth length, birth weight, eczema, vomiting, reflux, diarrhea, hematochezia, mucous stool, FTT, respiratory symptoms, discomfort symptoms, family history of allergic diseases, and autoimmune diseases between the FA and non-FA groups, indicating that they may be associated with FA in infants (P<0.05; Table 1). However, no significant differences were observed in age, sex, gestational age, cesarean section, constipation, infection history, or antibiotic exposure between the two groups.
Multivariate logistic regression analysis in the training set and preliminary establishment of a diagnostic model
We considered the occurrence of FA as the dependent variable and used the significant variables obtained from the univariate analysis as independent variables. Subsequently, we employed backward stepwise regression in the multivariate logistic regression analysis to obtain variables for FA diagnosis. This process resulted in the development of a diagnostic model for FA in infants comprising the following variables: eczema, stool form, hematochezia, FTT, respiratory symptoms, and a family history of allergic diseases. The result of multicollinearity analysis showed that the variance inflation factor for the six predictor variables was 1.054–1.725, indicating that there was no obvious multicollinearity among these variables. The variables incorporated in the model are presented in Table 2, and the weights for each clinical feature were derived from multivariate logistic regression coefficients (B values). Since the constant term is fixed, we omitted it for ease of calculation. A diagnostic model with its formula for infant FA was obtained: the scores of the diagnostic model = 0.23 × eczema (mild) + 1.02 × eczema (moderate) + 1.35 × eczema (severe) + 0.47 × stool form (loose) + 1.42 × stool form (watery) + 1.60 × hematochezia + 0.78 × FTT + 1.19 × respiratory symptoms + 1.36 × family history of allergic diseases. The presence of a variable was denoted as 1 and the absence was denoted as 0. The scores were calculated by inserting the results of all variables after assessment into the above formula. FA was diagnosed in infants with scores of ≥3.19. Based on the logistic regression analysis results, we developed a nomogram for predicting the risk of infantile FA (Figure 2). Each variable in the nomogram was assigned a point value according to its position on the corresponding axis. By aligning the variable’s point vertically with the top scale, the score for that variable was determined. The individual scores of all variables were then summed to produce a total score, which corresponded to a specific risk level for FA occurrence on the risk axis, allowing for the prediction of FA occurrence risk.
Table 2. Multivariate logistic regression analysis results using backward stepwise regression and variables used in the model.
| Variable | B | SE | Wald | P | OR | 95% CI |
|---|---|---|---|---|---|---|
| Eczema | ||||||
| Absent | – | – | – | – | – | – |
| Mild | 0.23 | 0.26 | 0.87 | 0.38 | 1.26 | 0.75, 2.11 |
| Moderate | 1.02 | 0.33 | 3.09 | 0.002 | 2.77 | 1.46, 5.32 |
| Severe | 1.35 | 0.44 | 3.04 | 0.002 | 3.85 | 1.62, 9.32 |
| Stool form | ||||||
| Formed or soft | – | – | – | – | – | – |
| Loose | 0.47 | 0.29 | 1.64 | 0.10 | 1.61 | 0.91, 2.84 |
| Watery | 1.42 | 0.31 | 4.64 | <0.001 | 4.15 | 2.28, 7.63 |
| Hematochezia | 1.60 | 0.27 | 5.96 | <0.001 | 4.97 | 2.95, 8.49 |
| FTT | 0.78 | 0.28 | 2.83 | 0.005 | 2.19 | 1.28, 3.79 |
| Respiratory symptoms | 1.19 | 0.29 | 4.14 | <0.001 | 3.28 | 1.88, 5.82 |
| Family history of allergic diseases | 1.36 | 0.23 | 5.92 | <0.001 | 3.88 | 2.49, 6.13 |
| Constant term | −2.80 | 0.28 | −10.17 | <0.001 | 0.061 | – |
CI, confidence interval; FTT, failure to thrive; OR, odds ratio; SE, standard error.
Figure 2.
Nomogram model for predicting the risk of food allergy. FTT, failure to thrive.
Evaluation and validation of the diagnostic model for FA
The training set was randomly classified into 10 parts, and the model was internally validated using ten-fold cross-validation. ROC curve analysis showed that the AUC of the training set was 0.86 (0.83–0.90). When the Youden index reached its maximum, the cutoff value was 3.19, resulting in a sensitivity of 0.72 (0.71–0.73) and specificity of 0.88 (0.87–0.89) (Figure 3A). These results indicate that the model has a good discriminatory ability. The Hosmer-Lemeshow goodness-of-fit test showed that the fit was good (χ2=11.30, P=0.19). Ten-fold cross-validation was performed within the training set, and the AUC of the internal validation was 0.86 (0.76–0.95), with a sensitivity of 0.72 (0.66–0.78) and specificity of 0.92 (0.88–0.95) (Figure 3B). Regarding the external validation, we used the testing set and obtained an AUC of 0.87 (0.83–0.92), a cutoff value of 3.09, a sensitivity of 0.72 (0.65–0.80), and a specificity of 0.89 (0.83–0.95) (Figure 3C). The Hosmer-Lemeshow goodness-of-fit test of the external validation set showed that the fit was good (χ2=10.95, P=0.20). The model demonstrated its capability to effectively model the data in this study. The reliability and validity of the model were evaluated using a calibration curve, which revealed that the predicted probability aligned well with the actual probabilities (Figure 4A). The Brier Score of the diagnostic model was 0.143, surpassing the random guessing baseline of 0.25. DCA demonstrated a high net benefit for clinical intervention in infants with suspected FA when the threshold probability for the training set ranged from 21% to 98% (Figure 4B).
Figure 3.
ROC curves for the training set (A), internal 10-fold cross-validation (B), and the testing set (C). For the 10-fold cross-validation, the training set was randomly divided into 10 subsets. In each iteration, 9 subsets were used to train the diagnostic model, and the remaining subset was used for validation to verify the accuracy rate of the model in diagnosing food allergies. The results of these 10 cross-validations are comprehensively shown in (B). AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.
Figure 4.
Calibration curve (A) and decision curve (B) results. (A) The predicted probability curve (blue line) of the FA diagnostic model closely follows the ideal situation (black dotted line), indicating good calibration of the model for diagnosing food allergy. (B) The area enclosed by the black dotted line, red dotted line, and red solid line represents the net clinical benefit obtained from applying the diagnostic model for clinical decision-making. FA, food allergy.
Discussion
The problem of delayed diagnosis and misdiagnosis of infantile FA should be addressed urgently in response to the increasing number of cases annually. In recent years, there have been efforts to improve the early diagnosis of FA. The CoMiSS has shown good diagnostic value in some European countries (38). However, its AUC values were below 0.75 in Asia, indicating limited accuracy as a diagnostic model (9,10). These regional differences may be influenced by variations in the disease spectrum (39). In 2021, a Japanese study developed and validated a predictive model for infants at high risk of FA, but its accuracy was questionable because of inconsistent diagnostic criteria (11). In 2023, another study established several models to predict OFC outcomes, but the highest AUC achieved was ≤0.68, indicating poor diagnostic performance (12). In summary, there is currently no simple, noninvasive, validated, and commonly used diagnostic model for FA. In this study, we established and validated a diagnostic model and its formula for FA in infants, providing clinicians with a simple and noninvasive tool for the early identification of the disease.
The diagnostic model was presented as a formula incorporating six variables as follows: eczema, stool form, hematochezia, FTT, respiratory symptoms, and a family history of allergic diseases. For infants suspected of having FA, clinicians can refer to the Definition and Assessment of Symptoms section to assess the presence and severity of these variables. The results can then be applied to the diagnostic model to calculate the score. If the calculated score exceeds 3.19, FA should be considered. All these variables were consistent with those of previous epidemiological surveys. Eczema, especially moderate-to-severe eczema, in the early stage after birth, had a strong association with FA (38), possibly because of its role in FA pathogenesis. The occurrence of eczema is the first step in the atopic process, and damage to the skin barrier caused by eczema may facilitate the absorption of food allergens through the skin, potentially leading to FA (40). Gastrointestinal FA is a major component of FA in infants due to the immaturity of the gastrointestinal barrier, which may be the reason for the inclusion of three main variables associated with the digestive system in the model (41,42). Our study found that “persistent respiratory symptoms” was an important variable in the model. Respiratory infection usually resolves within 1–2 weeks. If infection indicators and etiological test results are negative and respiratory symptoms persist for >1 week, it is important to consider the possibility of allergic diseases (7). Several studies have highlighted the significance of family history of allergic diseases, especially in first-degree relatives, as suggestive of FA (15). The high weightage assigned to a family history of allergic diseases in the model further emphasizes its importance in diagnosing FA.
The diagnostic model in this study is a noninvasive and standardized diagnostic tool for infantile FA beyond medical conditions and can be adapted by primary healthcare facilities. The model has the following advantages. First, clinicians can accomplish a preliminary diagnosis of FA by calculating the scores of the diagnostic model using the abovementioned formula, where the diagnostic cutoff value for FA is set at 3.19. The ROC results of the model in our study had good diagnostic discrimination and high specificity in identifying FA. Internal and external validation results showed good stability, transportability, and generalizability. DCA results showed a high net benefit value in clinical decision-making. Second, all variables in this model are common, and their acquisition is easy and noninvasive. The weights assigned to all variables were objective and derived from the statistical results of large-sample data rather than clinical experience. Third, the evaluation of the model variables is not prone to errors. Eczema severity (according to the SCORAD index) and stool form (according to the BITSS) are considered graded variables based on validated standardized scales. All other variables are binary, indicating either the presence or absence of the variable, making their evaluation in clinical practice easy.
Our research team has extensive expertise in the diagnosis and treatment of FA and the conduction of OFCs. We included patients with suspected FA from all over the country, providing us with a robust research population. Nevertheless, there are some limitations in this study. First, it was conducted at a single tertiary-care center, which may have introduced selection bias and limits the model’s generalizability. Future multi-center studies involving primary and secondary care settings are essential to validate and calibrate its performance across different healthcare contexts. Second, the non-FA group was defined based on assessment for clinically suspected and common allergens only, which may have introduced classification bias. Future studies employing more comprehensive allergen challenges or extended follow-up would help validate the model’s robustness. Third, retrospective data collection from family members may have caused recall bias in the testing set. Therefore, prospective validation in a consecutive cohort of infants with suspected FA is essential to establish robust clinical validity and represents our primary next step. Fourth, except for IgE, which has diagnostic value for IgE-mediated FA, there is no accepted laboratory test for FA diagnosis (14). Thus, due to the poor compliance and difficulty of obtaining blood samples in infants, half of the IgE results were unavailable, resulting in the exclusion of IgE in establishing the model (41,43). Although IgE has limited diagnostic value because infantile FA is mainly non-IgE-mediated or mixed IgE/non-IgE-mediated without elevated IgE levels (43), its absence limits the model’s ability to differentiate FA endotypes. Future version of this model aim to incorporate IgE and other mechanistic biomarkers as optional or tiered components to predict both the severity and the underlying immunologic phenotype of FA, guiding more personalized management (44).
Conclusions
This study provides a validated, noninvasive diagnostic model to facilitate early identification of infantile FA. Further multicenter studies are needed to validate and optimize the model parameters and diagnostic cutoff value. Additionally, we intend to simplify the model into a calculator to assist primary healthcare facilities in the early identification of FA in infants. We further plan to include more risk factors and biomarkers related to the mechanism of FA to enhance the predictive value of the diagnostic model and facilitate the classification of FA. The ultimate goal is to evolve the model into a multidimensional tool capable of predicting FA endotypes and severity, thereby enabling precision management. Concurrently, it should be remembered that over-reliance on the diagnostic model is not suggested, and follow-up and other auxiliary tests are indispensable for complex cases.
Supplementary
The article’s supplementary files as
Acknowledgments
None.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Peking University Third Hospital Medical Science Research Ethics Committee (approval No. M2020405, approval date: April 28, 2021), and the guardians of the enrolled infants signed an informed consent form.
Footnotes
Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-aw-783/rc
Funding: This work was supported by the Peking University Health Science Center (No. JKCJ202301).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-aw-783/coif). All authors report funding from the Peking University Health Science Center (No. JKCJ202301). The authors have no other conflicts of interest to declare.
Data Sharing Statement
Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-aw-783/dss
References
- 1.Ebisawa M, Ito K, Fujisawa T, et al. Japanese guidelines for food allergy 2020. Allergol Int 2020;69:370-86. 10.1016/j.alit.2020.03.004 [DOI] [PubMed] [Google Scholar]
- 2.Peters RL, Koplin JJ, Gurrin LC, et al. The prevalence of food allergy and other allergic diseases in early childhood in a population-based study: HealthNuts age 4-year follow-up. J Allergy Clin Immunol 2017;140:145-153.e8. 10.1016/j.jaci.2017.02.019 [DOI] [PubMed] [Google Scholar]
- 3.Spolidoro GCI, Amera YT, Ali MM, et al. Frequency of food allergy in Europe: An updated systematic review and meta-analysis. Allergy 2023;78:351-68. 10.1111/all.15560 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Yanagida N, Ebisawa M, Katsunuma T, et al. Accidental ingestion of food allergens: A nationwide survey of Japanese nursery schools. Pediatr Allergy Immunol 2019;30:773-6. 10.1111/pai.13095 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Dilley MA, Rettiganti M, Christie L, et al. Impact of food allergy on food insecurity and health literacy in a tertiary care pediatric allergy population. Pediatr Allergy Immunol 2019;30:363-9. 10.1111/pai.13019 [DOI] [PubMed] [Google Scholar]
- 6.Meyer R. Nutritional disorders resulting from food allergy in children. Pediatr Allergy Immunol 2018;29:689-704. 10.1111/pai.12960 [DOI] [PubMed] [Google Scholar]
- 7.Meyer R, Venter C, Bognanni A, et al. World Allergy Organization (WAO) Diagnosis and Rationale for Action against Cow's Milk Allergy (DRACMA) Guideline update - VII - Milk elimination and reintroduction in the diagnostic process of cow's milk allergy. World Allergy Organ J 2023;16:100785. 10.1016/j.waojou.2023.100785 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Turner PJ, Worm M, Ansotegui IJ, et al. Time to revisit the definition and clinical criteria for anaphylaxis? World Allergy Organ J 2019;12:100066. 10.1016/j.waojou.2019.100066 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Vandenplas Y, Zhao ZY, Mukherjee R, et al. Assessment of the Cow's Milk-related Symptom Score (CoMiSS) as a diagnostic tool for cow's milk protein allergy: a prospective, multicentre study in China (MOSAIC study). BMJ Open 2022;12:e056641. 10.1136/bmjopen-2021-056641 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.El-Shafie AM, Omar ZA, El Zefzaf HMS, et al. Evaluation of Cow's Milk Related Symptom Score [CoMiSS] accuracy in cow's milk allergy diagnosis. Pediatr Res 2023;94:987-95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Sugiura S, Hiramitsu Y, Futamura M, et al. Development of a prediction model for infants at high risk of food allergy. Asia Pac Allergy 2021;11:e5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Zhang J, Lee D, Jungles K, et al. Prediction of oral food challenge outcomes via ensemble learning. Inform Med Unlocked 2023;36:101142. [Google Scholar]
- 13.Sicherer SH, Warren CM, Dant C, et al. Food Allergy from Infancy Through Adulthood. J Allergy Clin Immunol Pract 2020;8:1854-64. 10.1016/j.jaip.2020.02.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Çelebioğlu E, Akarsu A, Şahiner ÜM. IgE-mediated food allergy throughout life. Turk J Med Sci 2021;51:49-60. 10.3906/sag-2006-95 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.de Jong NW, Elbert NJ, Mensink-Bout SM, et al. Parental and child factors associated with inhalant and food allergy in a population-based prospective cohort study: the Generation R Study. Eur J Pediatr 2019;178:1507-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Gao X, Yan Y, Zeng G, et al. Influence of prenatal and early-life exposures on food allergy and eczema in infancy: a birth cohort study. BMC Pediatr 2019;19:239. 10.1186/s12887-019-1623-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Bird JA, Leonard S, Groetch M, et al. Conducting an Oral Food Challenge: An Update to the 2009 Adverse Reactions to Foods Committee Work Group Report. J Allergy Clin Immunol Pract 2020;8:75-90.e17. [DOI] [PubMed] [Google Scholar]
- 18.Sorensen K, Meyer R, Grimshaw KE, et al. The clinical burden of cow's milk allergy in early childhood: A retrospective cohort study. Immun Inflamm Dis 2022;10:e572. 10.1002/iid3.572 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Martin VM, Virkud YV, Seay H, et al. Prospective Assessment of Pediatrician-Diagnosed Food Protein-Induced Allergic Proctocolitis by Gross or Occult Blood. J Allergy Clin Immunol Pract 2020;8:1692-1699.e1. 10.1016/j.jaip.2019.12.029 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sardecka I, Łoś-Rycharska E, Ludwig H, et al. Early risk factors for cow's milk allergy in children in the first year of life. Allergy Asthma Proc 2018;39:e44-54. 10.2500/aap.2018.39.4159 [DOI] [PubMed] [Google Scholar]
- 21.Zheng H, Zeng X, Xie Q, et al. Early life environmental antibiotic exposure and preschool allergic diseases: A biomonitoring-based prospective study in eastern China. Front Public Health 2022;10:1043942. 10.3389/fpubh.2022.1043942 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Peters RL, Krawiec M, Koplin JJ, et al. Update on food allergy. Pediatr Allergy Immunol 2021;32:647-57. 10.1111/pai.13443 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Spiewak R. Diseases from the Spectrum of Dermatitis and Eczema: Can "Omics" Sciences Help with Better Systematics and More Accurate Differential Diagnosis? Int J Mol Sci 2023;24:10468. 10.3390/ijms241310468 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Keet C, Pistiner M, Plesa M, et al. Age and eczema severity, but not family history, are major risk factors for peanut allergy in infancy. J Allergy Clin Immunol 2021;147:984-991.e5. 10.1016/j.jaci.2020.11.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Yao X, Song ZQ, Li W, et al. Guidelines for diagnosis and treatment of atopic dermatitis in China (2020)#. Int J Dermatol Venereol 2021;4:1-9. [Google Scholar]
- 26.Rosen R, Vandenplas Y, Singendonk M, et al. Pediatric Gastroesophageal Reflux Clinical Practice Guidelines: Joint Recommendations of the North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition and the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition. J Pediatr Gastroenterol Nutr 2018;66:516-54. 10.1097/MPG.0000000000001889 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Sultana Z, Hasenstab KA, Moore RK, et al. Symptom Scores and pH-Impedance: Secondary Analysis of a Randomized Controlled Trial in Infants Treated for Gastroesophageal Reflux. Gastro Hep Adv 2022;1:869-81. 10.1016/j.gastha.2022.06.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Lopez RN, Lemberg DA. Gastro-oesophageal reflux disease in infancy: a review based on international guidelines. Med J Aust 2020;212:40-4. 10.5694/mja2.50447 [DOI] [PubMed] [Google Scholar]
- 29.Rasquin A, Di Lorenzo C, Forbes D, et al. Childhood functional gastrointestinal disorders: child/adolescent. Gastroenterology 2006;130:1527-37. 10.1053/j.gastro.2005.08.063 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Huysentruyt K, Koppen I, Benninga M, et al. The Brussels Infant and Toddler Stool Scale: A Study on Interobserver Reliability. J Pediatr Gastroenterol Nutr 2019;68:207-13. 10.1097/MPG.0000000000002153 [DOI] [PubMed] [Google Scholar]
- 31.Chehade M, Meyer R, Beauregard A. Feeding difficulties in children with non-IgE-mediated food allergic gastrointestinal disorders. Ann Allergy Asthma Immunol 2019;122:603-9. 10.1016/j.anai.2019.03.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Rivera-Nieves D, Conley A, Nagib K, et al. Gastrointestinal Conditions in Children With Severe Feeding Difficulties. Glob Pediatr Health 2019;6:2333794X19838536. [DOI] [PMC free article] [PubMed]
- 33.Jiménez Buñuales MT, González Diego P, Martín Moreno JM. International classification of functioning, disability and health (ICF) 2001. Rev Esp Salud Publica 2002;76:271-9. 10.1590/s1135-57272002000400002 [DOI] [PubMed] [Google Scholar]
- 34.Goday PS, Huh SY, Silverman A, et al. Pediatric Feeding Disorder: Consensus Definition and Conceptual Framework. J Pediatr Gastroenterol Nutr 2019;68:124-9. 10.1097/MPG.0000000000002188 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Diaferio L, Caimmi D, Verga MC, et al. May Failure to Thrive in Infants Be a Clinical Marker for the Early Diagnosis of Cow's Milk Allergy? Nutrients 2020;12:466. 10.3390/nu12020466 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Anuntaseree W, Mo-suwan L, Vasiknanonte P, et al. Night waking in Thai infants at 3 months of age: association between parental practices and infant sleep. Sleep Med 2008;9:564-71. 10.1016/j.sleep.2007.07.009 [DOI] [PubMed] [Google Scholar]
- 37.Fiocchi A, Bognanni A, Brożek J, et al. World Allergy Organization (WAO) Diagnosis and Rationale for Action against Cow's Milk Allergy (DRACMA) Guidelines update - I - Plan and definitions. World Allergy Organ J 2022;15:100609. 10.1016/j.waojou.2021.100609 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Salvatore S, Bertoni E, Bogni F, et al. Testing the Cow's Milk-Related Symptom Score (CoMiSSTM) for the Response to a Cow's Milk-Free Diet in Infants: A Prospective Study. Nutrients 2019;11:2402. 10.3390/nu11102402 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Thompson G, Zhelev Z, Peters J, et al. Symptom scores in the diagnosis of pediatric cow's milk protein allergy: A systematic review. Pediatr Allergy Immunol 2021;32:1497-507. 10.1111/pai.13537 [DOI] [PubMed] [Google Scholar]
- 40.Yang L, Fu J, Zhou Y. Research Progress in Atopic March. Front Immunol 2020;11:1907. 10.3389/fimmu.2020.01907 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Al-Iede M, Sarhan L, Alshrouf MA, et al. Perspectives on Non-IgE-Mediated Gastrointestinal Food Allergy in Pediatrics: A Review of Current Evidence and Guidelines. J Asthma Allergy 2023;16:279-91. 10.2147/JAA.S284825 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Nowak-Węgrzyn A, Chehade M, Groetch ME, et al. International consensus guidelines for the diagnosis and management of food protein-induced enterocolitis syndrome: Executive summary-Workgroup Report of the Adverse Reactions to Foods Committee, American Academy of Allergy, Asthma & Immunology. J Allergy Clin Immunol 2017;139:1111-26.e4. 10.1016/j.jaci.2016.12.966 [DOI] [PubMed] [Google Scholar]
- 43.Lopes JP, Sicherer S. Food allergy: epidemiology, pathogenesis, diagnosis, prevention, and treatment. Curr Opin Immunol 2020;66:57-64. 10.1016/j.coi.2020.03.014 [DOI] [PubMed] [Google Scholar]
- 44.Jutel M, Agache I, Zemelka-Wiacek M, et al. Nomenclature of allergic diseases and hypersensitivity reactions: Adapted to modern needs: An EAACI position paper. Allergy 2023;78:2851-74. 10.1111/all.15889 [DOI] [PubMed] [Google Scholar]