Predictors of using biologics within one year after EEN therapy in pediatric Crohn’s disease

Abstract

Background

Biologics are commonly used as the second-line medication for maintenance of remission after exclusive enteral nutrition (EEN) induction in pediatric Crohn’s disease (CD). However, predictors of using biologics after EEN therapy in CD are yet to be elucidated. This study explores the predictive factors for the application of biologics in children with CD who had finished 8 weeks of EEN induction.

Methods

Seventy-five patients were divided into derivation group (n=54) and validation group (n=21). Baseline characteristics were compared between biological group and non-biological group. Univariate and multivariate analyses were performed to correlate 102 laboratory indicators. Finally, binary stepwise logistic regression was used to construct a prediction model.

Results

The predictor identified for biologic use in the prediction model was complement C4 and red blood cell distribution width (RDW). Complement C4 [odds ratio (OR) =1.04, 95% confidence interval (CI): 1.01–1.07] and RDW (OR =1.41, 95% CI: 1.05–1.91) were identified as significant independent predictors (P<0.05). Patients with baseline complement C4 levels >0.415 g/L and RDW >13.5% were more likely to use biologics within 1 year after EEN.

Conclusions

Complement C4 and RDW are predictors of biologic use within 1 year after EEN-induced clinical remission. Elevated complement C4 and RDW are associated with the use of biologics after induction of remission. However, extensive experimental validation is still required to confirm the above findings.

Highlight box.

Key findings

• Complement C4 and red blood cell distribution width (RDW) are predictors of biologic use within 1 year after exclusive enteral nutrition (EEN)-induced clinical remission. Elevated complement C4 and RDW are associated with the use of biologics after induction of remission.

What is known and what is new?

• EEN is the first-line induction therapy for pediatric Crohn’s disease (CD) and has been proven clinically effective. However, the possibility of disease relapse and complications still exists. Studies have confirmed that the classification at diagnosis, Pediatric Crohn’s Disease Activity Index score and intestinal microbiota composition are correlated with the long-term prognosis of children with CD. Nevertheless, relevant studies on laboratory indicators are still lacking.

• This study investigated laboratory indicators at diagnosis in children with CD that can predict the use of biologics one year after EEN treatment.，

What is the implication, and what should change now?

• Based on the corresponding research findings, relevant laboratory indicators can be applied in clinical practice to improve therapeutic strategies.

Introduction

Crohn’s disease (CD) is characterized by chronic inflammation of gastrointestinal tract, with a steadily increasing incidence worldwide (1). Due to complexity of diagnosis and treatment, around 50% of patients with CD show disease progression in the 10 years after first flare-up (2). Therefore, new therapy strategy is urgently needed.

The treatment of CD consists of induction and maintenance of remission. Conventional therapies include 5-aminosalicylic, corticosteroids and immunomodulators. Exclusive enteral nutrition (EEN) has been proven to exert equivalent efficacy to corticosteroids in the induction phase and has been endorsed as a first-line therapy for induction in children with mild-to-moderate CD (3,4).Anti-TNF-α agents are the main biologics, primarily including infliximab (IFX) and adalimumab (ADA). Studies have confirmed their efficacy in both remission induction and maintenance among pediatric CD patients (5,6). Step-up therapy is indicated when induction with EEN fails to achieve satisfactory disease control, which clinically manifests as the initiation of or switch to biologics. Therefore, during the subsequent long-term follow-up, we defined “retention of the original treatment regimen without biologic initiation” as an important surrogate marker for sustained disease remission, to evaluate the long-term prognosis of pediatric patients. In contrast, some patients (especially those with active perianal fistulizing disease) initiate biologics at the time of diagnosis, which is called up-down strategy (7). However, predictors of using biologics after EEN therapy in CD are yet to be elucidated.

Accordingly, we carried out a retrospective single center investigation to explore the predictive factors for the application of biologics in children with CD who had finished 8 weeks of EEN induction. Our study sought to identify baseline laboratory indicators and evaluate their capacity to predict the use of biologics in 1 year after induction. We present this article in accordance with the TRIPOD reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0515/rc).

Methods

Research design and data acquisition

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of Children’s Hospital, Zhejiang University School of Medicine (No. 2019-IRB-109), and informed consent was obtained from all participants’ parents or legal guardians. We conducted a single-center retrospective cohort study which enrolled 75 children with newly diagnosed CD at Children’s Hospital, Zhejiang University School of Medicine from March 2015 to July 2023. These patients were diagnosed with CD based on their clinical, laboratory, endoscopic, radiological, and histological criteria, according to the ESPGHAN Revised Porto Criteria for the Diagnosis of Inflammatory Bowel Disease in Children and Adolescents. We included them for the following reasons: (I) had not received effective therapy directed against CD prior to diagnosis; (II) EEN was implemented as induction therapy in accordance with the guidelines; (III) completion of Eight-week of EEN subsequent to the first disease flare; (IV) were followed up for 12 months after diagnosis.

Those patients were then randomly divided into derivation cohort (54 patients) and validation cohort (21 patients). Patients in each cohort were divided into UI group (use biologics in a year after diagnosis) or NUI group (not use biologics in a year).

We then collected the basic information including age, gender, height for age (HFA), body mass index for age (BMIFA), disease phenotype was determined according to the Paris classification, PCDAI and CDEIS. Before EEN induction therapy, all patients were assessed with laboratory parameters recorded at the time of diagnosis included fecal calprotectin (FC), procalcitonin (PCT), complete blood count (CBC), the erythrocyte sedimentation rate (ESR), biochemical indices, globulin (GLB), cytokines and CDs, prothrombin charts, trace elements, serum folic acid, vitamin b12, iron metabolism indices and immunoglobulins (detailed detected variables are summarized in Table S1).

Statistical analysis

All the statistical tests were carried out with IBM SPSS 17.0 (SPSS Inc., USA). Continuous variables with normal distribution were reported in the form of mean values ± standard deviation (SD), those deviating from normality were described as median and interquartile, and categorical data were expressed as frequencies and percentages. To guarantee the quality of data, lab measurements with missing values greater than 40% were removed from the analysis (Table S1). The Mann-Whitney U test or Student’s t-test was applied to continuous variables depending on normality testing results. The χ² test or Fisher test was employed for the comparison of categorical variables. The relationships among PCDAI, CDEIS and laboratory parameters were evaluated via the nonparametric Spearman’s rank correlation coefficient, receiver operating characteristic (ROC) curves and univariate binary logistic regression. Any variable with an area under the curve (AUC) >0.7 or P<0.05 in any other statistical methods was retained for subsequent screening. Afterwards, forward stepwise (Binary likelihood) logistic regression tests were conducted to identify independent predictors with statistical significance. Odds ratio (OR) and matching 95% confidence interval (CI) were calculated for every variable within the predictive model. Prognostic capacity of baseline laboratory indices was compared via ROC curves, and model performance was examined. DeLong’s test was performed to compare the AUC between different predictive models. To establish a decision-support model, continuous variables were dichotomized based on the optimal cut-off value. The odds associated with using biologic were estimated with the chosen logistic regression model and transformed into probabilities. Variables with P<0.05 were regarded as statistically significant.

Results

Baseline clinical profiles among pediatric CD patients prior to EEN induction therapy

The derivation cohort consisted of 54 patients, 28 of whom had not received biologic therapy within the past year; the validation group comprised 21 patients, 15 of whom had not received biologic therapy (Figure 1). The overall long-term remission rate was 57.3%. In the baseline data analysis, there were no significant differences between UI group and NUI group in terms of age, sex, HFA, BMIFA, disease location, disease behavior, perianal lesion status, disease progression, and CDEIS scores (Table 1). These results indicate that although EEN can significantly improve symptoms and promote mucosal healing in the short term, some patients still experience a recurrence of disease activity during subsequent follow-up. Therefore, further exploration of key predictors influencing long-term remission is required.

Figure 1.

Flow diagram of study participant recruitment. (A) Data collected at first visit and after EEN induction therapy. (B) Sample sizes of the derivation and validation cohorts. The y-axis represents the number of patients. CDEIS, Crohn’s disease endoscopic index of severity; EEN, exclusive enteral nutrition; PCDAI, Pediatric Crohn’s Disease Activity Index.

Table 1. Demographics and clinical features of patients stratified by post-EEN biologic use.

Characteristics	Derivation cohort			Validation cohort
	Not use (n=28)	Use biologic (n=26)	P	Not use (n=15)	Use biologic (n=6)	P
Age (years)	11.45±2.49 (6.33, 15.58)	12.51±1.96 (5.83, 15.67)	0.14	10.96±3.04 (4.5, 15.25)	12.71±2.81 (8.75, 15.50)	0.24
Gender (M/F)	8/20	10/16	0.44	8/7	3/3	0.89
HFA z score	−0.03±1.04 (−2.48, 1.94)	0.02±1.26 (−2.81, 2.04)	0.88	−0.58±1.31 (−2.88, 2.58)	−0.60±0.44 (−1.20, −0.19)	0.97
BMIFA z score	−1.26±1.58 (−4.29, 1.57)	−1.03±1.84 (−4.44, 2.19)	0.62	−1.22±1.87 (−3.67, 1.65)	−1.70±1.11 (−2.91, 0.06)	0.67
Disease location			0.63			0.13
L1	4 (7.1)	2 (3.8)		2 (6.9)	0 (0.0)
L2	2 (3.6)	2 (3.8)		2 (6.9)	0 (0.0)
L3	14 (25.0)	17 (32.1)		6 (20.7)	6 (54.5)
L4a	17 (30.4)	10 (18.7)		9 (31.0)	1 (9.1)
L4b	19 (33.9)	22 (41.5)		10 (34.5)	4 (36.4)
Disease behavior			1			>0.99
B1	27 (96.4)	24 (100.0)		15 (100.0)	6 (100.0)
B2	1 (3.6)	0 (0.0)		0 (0.0)	0 (0.0)
Perianal disease	0 (0)	3 (11.5)	0.11	6 (40.0)	1 (16.7)	0.61
Disease growth			1			>0.99
G0	26 (92.9)	24 (92.3)		14 (93.3)	6 (100.0)
G1	2 (7.1)	2 (7.7)		1 (6.7)	0 (0.0)
PCDAI before	19.55±10.16 (2.50, 40.00)	22.50±11.36 (5.00, 45.00)	0.32	14.17±11.75 (0.00, 45.00)	15.00±14.58 (0.00, 37.50)	0.89
CDEIS before	6.68±7.30 (0.00, 24.60)	9.35±6.83 (0.00, 25.00)	0.1	3.70±3.43 (0.00, 11.84)	10.05±7.50 (0.20, 22.40)	0.23

Data are presented as n (%) or mean ± standard deviation (interquartile range). BMIFA, body mass index for age; CDEIS, Crohn’s disease endoscopic index of severity; EEN, exclusive enteral nutrition; HFA, height for age; PCDAI, Pediatric Crohn’s Disease Activity Index.

Predicting model for using biological within one year after diagnosis

Data from the derivation cohort for UI group and NUI group were analysed to construct a model for long-term prognosis. The study included 102 variables collected prior to EEN treatment. Following screening, 12 variables met the following criteria: the P-value of Student’s test or Mann-Whitney U-test between UI group and NUI group was less than 0.05; and logistic regression tests between whether use biologics in a year after diagnosis and laboratory indicators was below the threshold of 0.05; and the AUC value indicating the predictive performance of laboratory indicators, including complement C4, RDW, complement C3, mean corpuscular haemoglobin (MCH), mean corpuscular haemoglobin concentration (MCHC), haemoglobin (Hb), type I collagen cross-linked carboxy-terminal peptide (CTX-1), CD3-CD16+CD56+, IFN-γ, interleukin-6 (IL-6), parathyroid hormone (PTH) and total bilirubin (TBA) (Table S2).

Ultimately, two independent prognostic factors with significant differences (P<0.05) were screened out via binary forward step-wise logistic regression and included in the final model (Figure 2). The results showed that complement C4 (OR =1.04, 95% CI: 1.01–1.07) and RDW (OR =1.41, 95% CI: 1.05–1.91) were significant predictors (P<0.05) (Figure 2A). The full regression equation is provided in Appendix 1. The predictive model constructed using the complement C4 level and RDW had a high AUC (0.78, 95% CI: 0.66–0.90, P<0.05) in the training cohort and validation cohort (0.69, 95% CI: 0.38–1.00, P=0.39) (Figure 2B,2C). But DeLong’s test indicated that there were no significant differences in the AUC between the combined model and complement C4, as well as between the model and RDW (P>0.05).

Figure 2.

ROC curve of parameters affecting long‑term prognosis one year after EEN treatment. (A) Variables adopted to forecast clinical remission using logistic regression in the training dataset. The x-axis represents the OR. Model predictive capacity observed in the training cohort (B) and in the validation cohort (C) for single markers and combined markers were assessed via ROC curve and AUC. Data in parentheses represent the 95% CI of the AUC. AUC, area under the curve; CI, confidence interval; OR, odds ratio; RDW, red cell distribution width; ROC, receiver operating characteristic.

Decision-making matrix model

Optimal UI cutoff values were defined and summarized below: complement C4 at 0.415 g/L and RDW at 13.5%. Risk stratification analysis revealed that patients with complement C4 <0.415 g/L and RDW <13.5% have a low probability of biologic use within one year after EEN and exhibits favorable prognostic performance (Figure 3).

Figure 3.

Decision-making model indicating the use of biologic one year after EEN therapy. Complement C4 (g/L), RDW (%), higher percentages in the figure correspond to more favorable prognosis. EEN, exclusive enteral nutrition; RDW, red cell distribution width.

Discussion

Although the therapeutic approaches of CD have been continuously advanced, a certain risk of disease recurrence or concomitant complications still remains. Therefore, it is of great significance to identify biomarkers capable of predicting long-term prognosis. In this study, we sought to define predictors of whether patients use biologics in 1 year and develop a decision-making methodology, which might help clinicians choose the better treatment.

In our study, the long-term remission rate was 57.3%, which was in accordance with previous researches (8,9). During data collection, we endeavored to incorporate comprehensive laboratory indicators related to CD, with a total of 102 parameters included. Given that the number of detected indicators exceeded the sample size, univariate analysis was first performed to identify factors associated with long-term remission. Subsequently, the 12 screened laboratory parameters were incorporated into the binary logistic regression model.

The experimental findings reveal that complement C4 and RDW are capable of predicting the administration of biologics at one year following EEN therapy.

Complement C4 is primarily used to reflect autoimmune responses and inflammatory conditions, and its level may be elevated in the presence of autoimmune diseases, infections, inflammation and other conditions (10,11). And it is mainly encoded by the C4A and C4B genes (12). Increased copy number of the C4B gene is shown to be associated with intestinal inflammation, intestinal flora dysbiosis and excessive activation of the complement system (13). Another study has suggested that variations in the copy number of the C4 gene may be related to the onset of CD (14). Therefore, the elevation of complement C4 may be correlated with the aggravation of intestinal inflammation in CD patients.

RDW is a hallmark of anemia and can also reflect chronic inflammation, iron metabolism disorders, and other pathological conditions (15,16). Inflammation upregulates hepatic hepcidin expression via IL-6, thereby resulting in disordered iron metabolism (17). Disorders of iron metabolism further lead to shortened half-life of circulating red blood cells, inhibited differentiation of erythroid progenitor cells, and other alterations (17). The premature erythrocytes are subsequently released into the peripheral blood, which ultimately causes an elevation in RDW (18). Studies have demonstrated that iron deficiency and anemia are highly prevalent in patients with CD (19,20). In adults, RDW has been proven to be a reliable indicator for reflecting active CD (21). In children, the RDW level is significantly higher in CD patients than in healthy controls (22). Thus, it can be speculated that RDW acts as an important biomarker for evaluating CD activity.

Other studies have also identified predictive factors associated with long-term remission after EEN induction. For instance, Buczyńska et al. (23). found that CD patients presenting with stricturing or penetrating phenotypes, elevated PCDAI scores and hypoalbuminemia at diagnosis had a higher likelihood of requiring biologics during subsequent treatment. In addition, with regard to intestinal flora, Jones et al. (24). reported that bacteria including Ruminococcaceae UCG-002, Lachnospiraceae NK4A136, Bacteroides and Parabacteroides could predict long-term remission following EEN therapy. Ghiboub et al. (25). analyzed the alterations of 21 fecal tryptophan metabolites during EEN treatment, and confirmed that sustained remission after EEN intervention was correlated with increased levels of melatonin, N-acetylserotonin and 5-hydroxytryptophan.

Nevertheless, the predictive reliability of our model is subject to the limitations outlined below. First, the relatively small sample size may limit the model’s generalizability and statistical power. And the large number of candidate variables screened in the initial analysis create a substantial risk of overfitting and unstable variable selection. Large, multicenter cohort studies are needed to validate these findings in the future. Second, the effective events-per-variable ratio in our analysis was low, which will increase the likelihood of obtaining optimistic performance estimates that may not generalize to external populations. Third, we adopted whether patients received biologics within one year to reflect disease remission. However, this outcome is affected by physicians’ subjective clinical judgment, clinical practice and other factors, and further evaluations including disease activity scores and endoscopic scores are required to optimize this indicator. Forth, the single-center setting of the study may introduce selection bias and restrict the model’s universal generalizability. What’s more, we failed to take potential predictive factors such as genetic characteristics, microbiome composition and serological immune markers into account. Finally, although we established a risk matrix to support clinical decision-making, prospective trials are still needed to validate its clinical efficacy in actual clinical practice.

A total of 102 biomarkers were incorporated into this study. However, owing to constraints regarding guardian informed consent, not all patients completed all tests. To maintain the reliability of the dataset, biomarkers with >40% missing rate were excluded, removing potentially valuable predictors (e.g., iron metabolism markers, fecal calprotectin, osteocalcin) and possibly reducing the model’s predictive accuracy and may introduce selection bias. Future studies should address these limitations and ensure complete datasets for all participants to enhance model robustness and applicability.

The findings of the present study need to be further validated by experimental research, and in-depth exploration is essential to clarify the specific mechanism by which model indicators affect the prognosis of CD patients. In addition, further exploration of genetics, gut microbiota and immune biomarkers is conducive to optimizing the predictive ability and clinical practicability of the model.

Conclusions

In summary, we identified two laboratory indicators (complement C4 and RDW) that can serve as predictive factors for the requirement of biologics administration after the completion of EEN induction therapy. Nevertheless, large-sample and multicenter studies are warranted to further verify the above results.

Supplementary

The article’s supplementary files as

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. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of Children’s Hospital, Zhejiang University School of Medicine (No. 2019-IRB-109), and informed consent was obtained from all participants’ parents or legal guardians.

Data Sharing Statement

Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0515/dss

tp-15-9-381-dss.pdf ^{(85.1KB, pdf)}

DOI: 10.21037/tp-2026-0515