Incidence of Autoimmune Diseases in Children with Atopy Treated with Dupilumab

Abstract

Background:

The type 2 inflammation-blocking agent dupilumab has gained traction as an effective treatment of multiple atopic diseases. New-onset autoimmune manifestations have been infrequently reported in dupilumab-treated atopic patients, primarily among adults. Autoimmune outcomes in atopic children on dupilumab are less clear.

Objective:

We aimed to determine if dupilumab treatment of atopic children is associated with increased autoimmune or autoinflammatory disease diagnosis risk.

Methods:

Using a multi-institutional electronic health record database comprised of 10 PEDSnet academic health systems, we performed a retrospective cohort study of children aged 6–17 with atopy (i.e., moderate-to-severe atopic dermatitis and/or persistent asthma, as defined by diagnosis codes and prescriptions) with or without (n=4,189 and n=4,195 in inverse-probability-of-treatment–weighted cohorts, respectively) dupilumab prescription between October 1, 2018 and November 29, 2023. Poisson regression estimated autoimmune disease incidence rates and rate differences, and Cox proportional hazards models estimated relative associations between dupilumab exposure and disease diagnosis.

Results:

Among atopic children, the adjusted incidence rate difference (dupilumab-treated minus untreated) for any autoimmune disease was 2.47 (95% CI 0.82, 4.30) per 1,000 person-years, driven primarily by cutaneous autoimmune diseases (rate difference 2.20 [95% CI 0.77, 3.70] per 1,000 person-years). The rates of any or cutaneous autoimmune disease within 4 years of dupilumab treatment were 1.45-fold and 1.57-fold higher, respectively, in treated as compared with untreated atopic children.

Conclusion:

We detected a modest association between dupilumab and cutaneous autoimmune disease diagnosis in atopic children. These findings help address knowledge gaps about clinical outcomes in atopic children treated with this biologic.

CAPSULE SUMMARY

This study examined autoimmune disease outcomes in atopic children treated with dupilumab, detecting a modestly increased risk of cutaneous autoimmune disease diagnosis. Results support longitudinal clinical monitoring in this population, the current standard of care.

INTRODUCTION

Difficult to control atopic diseases, such as atopic dermatitis (AD) and asthma, impair the quality of life of affected children and adults.^1,2 Mechanistically, the immune response in atopic disease is driven by interactions among allergic triggers, the epithelium, and innate and adaptive immunity. Classically referred to as Type 2 inflammation, allergic inflammation is distinguished by the involvement of specific cell types (e.g., T helper 2 cells, mast cells, basophils, eosinophils, type 2 innate lymphoid cells), cytokines (e.g., IL-4, IL-5, IL-13), and immunoglobulin E.^3–6 As such, therapies for atopic diseases have focused largely on blocking Type 2 inflammation. A notable example is the biologic dupilumab, a human monoclonal antibody that binds the IL-4Rα chain on immune cells, inhibiting IL-4 and IL-13 signaling.^7–9 Since its initial FDA approval in 2017, dupilumab has gained recognition as an effective treatment of multiple atopic diseases – including AD, asthma, and eosinophilic esophagitis – and is now approved for use in children as young as 6 months.^8–11

Hypothetically, blockade of Type 2 allergic inflammation with dupilumab might allow for unopposed non-Type 2 inflammatory responses, such as Type 1 and Type 17 inflammation that have been implicated in the immunopathology of autoimmune and autoinflammatory diseases.¹² This prospect is supported by case reports of new-onset autoimmune or autoinflammatory manifestations (e.g., psoriasis, inflammatory arthritis) in patients treated with dupilumab, primarily in the adult populaton.^13,14 It is unclear whether the use of dupilumab for treatment of pediatric atopy may unmask or induce autoimmunity. To address this knowledge gap, we conducted a longitudinal study using electronic health record (EHR) data from 10 institutions participating in PEDSnet to assess whether the use of dupilumab in children is associated with an increased risk of autoimmune disease diagnosis.

METHODS

Data Source

We used EHR data from 10 academic medical institutions that participate in PEDSnet: Children’s Hospital of Philadelphia, Cincinnati Children’s Medical Center, Children’s Hospital Colorado, Lurie Children’s Hospital, Nationwide Children’s Hospital, Children’s National Hospital, Nemours Children’s Health, Seattle Children’s Hospital, Stanford Children’s Health, and Texas Children’s Hospital. Data was extracted in April 2025 and was harmonized to align to the PEDSnet common data model (version 01/2025). EHR data used in this study included information on patient demographics, diagnoses, prescriptions, insurance status, anthropometric measurements, encounters, and death.

Study Sample

Study inclusion and exclusion criteria are detailed in Table E1. The sample included children aged 6 to 17 years with at least one clinical encounter of any type (inpatient, outpatient, emergency department, or other) during the cohort entry period (October 1, 2018 to November 29, 2023). This timeframe includes the approval dates of dupilumab for 12–17-year-olds (October 2018 for asthma and March 2019 for AD) and 6–11-year-olds (May 2020 for AD and October 2021 for asthma).

Children with atopy were defined as meeting criteria for AD and/or persistent asthma, based on diagnosis codes and medication prescriptions (Table E1). These patients were further categorized into dupilumab-treated and dupilumab-untreated groups. Dupilumab-treated patients were those with at least one dupilumab prescription within 4 years following their earliest atopic disease diagnosis (median 6 unique dates of dupilumab prescriptions per patient). Dupilumab-untreated patients had no record of a dupilumab prescription prior to or within 4 years following their earliest atopic disease diagnosis. The cohort entry date (CED) was defined as the date of the earliest dupilumab record for dupilumab-treated patients and as a random visit date during the study period for dupilumab-untreated patients. Cohort entry dates extended from October 1, 2018 to November 29, 2023.

Patients were excluded from the treated and untreated cohorts if they had evidence of autoimmune disease diagnosis before their CED, if they had an emergency department visit or were hospitalized on their CED, or if they lacked an indication of moderate-to-severe atopic disease within 5 years prior to their CED. Cohort attrition steps are illustrated in Figure E1, and code sets are available in a public repository (https://github.com/PEDSnet/dupilumab_t1).

Study Variables

Dupilumab treatment, as defined above, served as the comparator. The outcome was the incidence of a new onset autoimmune disease, measured as a binary indicator. For simplicity, both autoimmune diseases and autoinflammatory diseases are referred to as autoimmune diseases. Autoimmune disease was defined by the presence of a corresponding diagnosis code (code sets available in the public repository: https://github.com/PEDSnet/dupilumab_t1) within 60 days to 4 years post-CED and occurring on at least 2 separate dates. Sub-analyses were also performed considering select autoimmune/autoinflammatory disease groups based on a review of the literature and available diagnosis codes.

Study covariates included age at cohort entry, health system, sex, race/ethnicity, atopic disease type (i.e., AD, asthma, or both), 3-month cohort entry period, number of non-inpatient visit days from 7 days to 1 year prior to the CED, most recent body mass index (BMI) percentile within a year before or on the CED, and medical complexity. Self-identified race and ethnicity were considered as social constructs that may differentially associate with disease outcomes.¹⁵ Medical complexity was determined by applying the Pediatric Medical Complexity Algorithm (PMCA; version 3.0) to data within 3 years prior to the CED. Patients were categorized as non-complex chronic, complex chronic, or neither.¹⁶

Statistical Analyses

To adjust for potential confounding and balance baseline characteristics in the dupilumab-treated and untreated groups, we used inverse probability of treatment weighting (IPTW). Propensity scores, which represent the probability of receiving dupilumab based on baseline characteristics, were estimated using the WeightIt¹⁷ and Cobalt¹⁸ R packages. These scores were estimated using a logistic regression model, with dupilumab treatment as the dependent variable and baseline covariates (as detailed above) as independent variables. Weights were calculated as the inverse of the estimated propensity scores. The target estimand was the average treatment effect in the treated (ATT), in which the treated group was assigned weights of 1 and the untreated group was weighted to reflect the distribution of covariates in the treated group. To reduce the influence of extreme weights, values were capped at the 0.5^th and 99.5^th percentiles. Cohorts were considered balanced if the absolute value of the standardized mean differences (SMDs) was less than 0.1 (Figure E2).

Poisson regression was used to estimate incidence rates (IRs) of autoimmune disease and disease subcategories per 1,000 person-years. IRs for the treated and untreated atopic groups were computed using the emmeans R package, and the corresponding incidence rate differences were calculated to quantify absolute population impact.¹⁹ 95% confidence intervals were computed for each rate difference using bootstrapping with 2000 iterations.

Using the IPTW-weighted dupilumab-treated and untreated groups, we generated cumulative incidence curves and performed Cox proportional hazards modeling to estimate relative associations between dupilumab exposure and autoimmune disease outcomes by computing disease hazard ratios (HRs). Cumulative incidence curves were smoothed with Loess regression (span = 0.1) to mask small cells. Each disease subcategory was analyzed in a separate model. The Cox proportional hazards assumption was not violated for any model (p>0.05). Models were fully adjusted by all covariates used in the weighting process to estimate HRs for any autoimmune disease and cutaneous autoimmune disease. For autoimmune endocrinologic, gastrointestinal, and rheumatologic disease categories, reduced adjustment (limited to age and sex) was applied due to small (<15) outcome counts. Patients were right-censored at the time of death or at the date of the most recently available data from the patient’s healthcare site, whichever occurred first.

Sensitivity and Secondary Analyses

In addition to fully adjusted Cox models (using all covariates), unadjusted and partially adjusted (for sex and age) models were generated. Sensitivity analyses included models with varying follow-up durations, including follow-up starting 60 days to 1, 2, or 3 years after a subject’s CED; models requiring a single autoimmune disease outcome condition code rather than 2 codes on separate dates; and stratification analyses based on disease severity and comorbidity.

Lastly, we estimated the annual prevalence of autoimmune diseases among children with and without atopy. A single “atopic disease” group was constructed by combining the atopic dupilumab-treated and untreated cohorts, which was then compared against children without atopic disease (Table E2). The non-atopic group was constructed by identifying patients with no evidence of atopic disease conditions or medication. Autoimmune disease was identified by two distinct outcome dates within an autoimmune disease diagnosis on or before December 31 of a given year. Prevalence was calculated by dividing the number of patients with autoimmune disease in a given year by the total number of patients aged 6 to 17 as of December 31 of that year.

Data Availability

Results are based on detailed individual-level patient data compiled by members of PEDSnet. Due to the high risk of reidentification based on unique patterns in the clinical data, even when demographic identifiers have been removed, patient privacy regulations prohibit us from releasing the individual-level data publicly. However, logic and codes used to define the cohort are publicly available at the following link: [https://github.com/PEDSnet/dupilumab_t1/tree/main].

RESULTS

Cohort demographic and clinical characteristics are summarized in Table 1 (pre-weighting) and Table E3 (post-weighting) for the atopic cohorts and Table E2 for the non-atopic cohort. In the dupilumab-treated and untreated (unweighted) atopic cohorts, there were 4,189 and 174,788 children, respectively (Table 1). The median ages were 12.0 and 11.8 years old, respectively. The cohorts were comparably representative of biological sex (48.8% and 47.2% female, respectively). With regard to race and ethnicity, the dupilumab-treated cohort had decreased representation of Hispanic children (18.7% vs. 22.3%) and increased representation of non-Hispanic Asian children (9.1% vs. 5.7%), as compared with the untreated cohort. The majority of atopic children had exclusive AD (57.9% treated, 54.4% untreated), with an increased proportion of children with both asthma and AD in the treated cohort (29.8% vs. 13.2%). The weighted atopic dupilumab-untreated cohort (Table E3) was well-balanced with absolute value standardized mean differences <0.1 for all variables included in the weighting process (Figure E2). Weighted cohorts were used for subsequent analyses to determine incidence of autoimmune disease diagnosis, considered either as a whole category or select disease categories (Table 2).

Table 1.

Atopic Cohort Demographic and Clinical Characteristics^a

Characteristic	Overall N=178,977 (%)	Atopic Dupilumab-treated N=4,189 (%)	Atopic Dupilumab-untreated N=174,788 (%)	Standardized Mean Difference
Median Age in Years [IQR]	11.8 [8.8, 15.2]	12.0 [8.8, 14.6]	11.8 [8.8, 15.2]	−0.04
Age Group
6 to 9	62,873 (35.1%)	1,433 (34.2%)	61,440 (35.2%)	−0.02
10 to 13	56,067 (31.3%)	1,465 (35.0%)	54,602 (31.2%)	0.08
14 to 17	60,037 (33.5%)	1,291 (30.8%)	58,746 (33.6%)	−0.06
Sex
Female	84,551 (47.2%)	2,042 (48.8%)	82,509 (47.2%)	0.03
Male or Other/Unknown	94,426 (52.8%)	2,147 (51.3%)	92,279 (52.8%)	0.00
Race/Ethnicity
Hispanic	39,821 (22.3%)	782 (18.7%)	39,039 (22.3%)	−0.09
Non-Hispanic White	57,878 (32.3%)	1,355 (32.4%)	56,523 (32.3%)	0.01
Non-Hispanic Black	51,232 (28.6%)	1,204 (28.7%)	50,028 (28.6%)	0.01
Non-Hispanic Asian	10,354 (5.8%)	382 (9.1%)	9,972 (5.7%)	0.12
Non-Hispanic Other or Multiple Race	4,993 (2.8%)	189 (4.5%)	4,804 (2.8%)	0.08
Other/Unknown	14,699 (8.2%)	277 (6.6%)	14,422 (8.3%)	−0.07
Site
1	9,244 (5.2%)	278 (6.6%)	8,966 (5.1%)	0.06
2	6,777 (3.8%)	375 (9.0%)	6,402 (3.7%)	0.19
3	7,451 (4.2%)	420 (10.0%)	7,031 (4.0%)	0.20
4	25,432 (14.2%)	396 (9.5%)	25,036 (14.3%)	−0.17
5	44,999 (25.1%)	618 (14.8%)	44,381 (25.4%)	−0.30
6	21,806 (12.2%)	404 (9.6%)	21,402 (12.2%)	−0.09
7	10,588 (5.9%)	274 (6.5%)	10,314 (5.9%)	0.03
8	8,055 (4.5%)	241 (5.8%)	7,814 (4.5%)	0.06
9	2,617 (1.5%)	194 (4.6%)	2,423 (1.4%)	0.15
10	42,008 (23.5%)	989 (23.6%)	41,019 (23.5%)	0.00
Atopic Condition
Asthma only	56,999 (31.9%)	514 (12.3%)	56,485 (32.3%)	−0.61
Atopic dermatitis only	97,582 (54.5%)	2,425 (57.9%)	95,157 (54.4%)	0.07
Asthma and atopic dermatitis	24,396 (13.6%)	1,250 (29.8%)	23,146 (13.2%)	0.36
Chronic Condition Comorbidity
Complex Chronic	20,427 (11.4%)	296 (7.1%)	20,131 (11.5%)	−0.17
Non-Complex Chronic	10,356 (5.8%)	191 (4.6%)	10,165 (5.8%)	−0.06
Not Chronic or Complex Chronic	148,194 (82.8%)	3,702 (88.4%)	144,492 (82.7%)	0.18
Body Mass Index Percentile
Below 5th	4,897 (2.7%)	132 (3.2%)	4,765 (2.7%)	0.02
5th to 84th	80,399 (44.9%)	2,023 (48.3%)	78,376 (44.8%)	0.07
85th to 94th	24,363 (13.6%)	630 (15.0%)	23,733 (13.6%)	0.04
95th and above	46,241 (25.8%)	1,046 (25.0%)	45,195 (25.9%)	−0.02
No BMI	23,077 (12.9%)	358 (8.6%)	22,719 (13.0%)	−0.16
Cohort Entry Period
October – December 2018	12,719 (7.1%)	32 (0.8%)	12,687 (7.3%)	−0.75
January – March 2019	10,291 (5.8%)	62 (1.5%)	10,229 (5.9%)	−0.36
April – June 2019	9,417 (5.3%)	108 (2.6%)	9,309 (5.3%)	−0.17
July – September 2019	9,449 (5.3%)	119 (2.8%)	9,330 (5.3%)	−0.15
October – December 2019	9,491 (5.3%)	118 (2.8%)	9,373 (5.4%)	−0.15
January – March 2020	8,330 (4.7%)	125 (3.0%)	8,205 (4.7%)	−0.10
April – June 2020	5,967 (3.3%)	122 (2.9%)	5,845 (3.3%)	−0.03
July – September 2020	7,661 (4.3%)	194 (4.6%)	7,467 (4.3%)	0.02
October – December 2020	7,638 (4.3%)	189 (4.5%)	7,449 (4.3%)	0.01
January – March 2021	7,072 (4.0%)	187 (4.5%)	6,885 (3.9%)	0.03
April – June 2021	7,772 (4.3%)	207 (4.9%)	7,565 (4.3%)	0.03
July – September 2021	8,191 (4.6%)	233 (5.6%)	7,958 (4.6%)	0.04
October – December 2021	8,336 (4.7%)	215 (5.1%)	8,121 (4.7%)	0.02
January – March 2022	7,835 (4.4%)	248 (5.9%)	7,587 (4.3%)	0.07
April – June 2022	7,761 (4.3%)	237 (5.7%)	7,524 (4.3%)	0.06
July – September 2022	8,367 (4.7%)	304 (7.3%)	8,063 (4.6%)	0.10
October – December 2022	8,622 (4.8%)	323 (7.7%)	8,299 (4.8%)	0.11
January – March 2023	8,498 (4.8%)	285 (6.8%)	8,213 (4.7%)	0.08
April – June 2023	8,748 (4.9%)	354 (8.4%)	8,394 (4.8%)	0.13
July – September 2023	9,585 (5.4%)	299 (7.1%)	9,286 (5.3%)	0.07
October – November 2023	7,227 (4.0%)	228 (5.4%)	6,999 (4.0%)	0.06
Non-Inpatient Visit Days in Year Prior to CED
0 to 3 visit days	50,113 (28.0%)	862 (20.6%)	49,251 (28.2%)	−0.19
4 to 7 visit days	46,089 (25.8%)	937 (22.4%)	45,152 (25.8%)	−0.08
8 to 13 visit days	35,114 (19.6%)	848 (20.2%)	34,266 (19.6%)	0.02
14 to 24 visit days	24,526 (13.7%)	734 (17.5%)	23,792 (13.6%)	0.10
25 or more visit days	23,135 (12.9%)	808 (19.3%)	22,327 (12.8%)	0.17

^aCED, cohort entry date; IQR, interquartile range; data for weighted cohort is shown in Table E3

Table 2.

Autoimmune Disease Categories^a

Category	Diseases
Cutaneous	Alopecia; Dermatomyositis; Psoriasis; Vitiligo
Endocrinologic	Addison’s Disease; Grave’s Disease; Hashimoto’s Thyroiditis; Type 1 Diabetes
Hematologic	Autoimmune Anemia; Immune Thrombocytopenia
Gastrointestinal	Celiac Disease; Crohn’s Disease; Ulcerative Colitis
Neurologic	Acute Disseminated Encephalomyelitis; Myasthenia Gravis; Multiple Sclerosis; Neuromyelitis Optica; Transverse Myelitis
Rheumatologic	Ankylosing Spondylitis; Behçet’s Disease; Inflammatory Arthritis; Lupus; Polymyositis; Psoriatic Arthritis; Rheumatoid Arthritis; Sjögren’s Disease

^aSelect autoimmune disease categories were considered based on literature review and availability of corresponding diagnosis codes. For simplicity, both autoimmune and autoinflammatory conditions are referred to as autoimmune diseases.

Autoimmune diseases incidence rates between the dupilumab-treated and untreated groups are reported in Table 3. For all autoimmune diseases, these were 8.78 and 6.31 per 1,000 person-years in the treated vs. untreated cohorts, respectively, representing a significant rate difference of 2.47 (95% CI 0.82, 4.30) per 1,000 person-years. Among the autoimmune disease subcategories, there was a comparable significant weighted incidence rate difference of 2.18 (95% CI 0.67, 3.66) per 1,000 person-years for autoimmune cutaneous diseases. In contrast, we observed non-significant weighted incidence rates between dupilumab-treated and untreated children for autoimmune endocrinologic, gastrointestinal and rheumatologic diseases. Consistent with these results, there were higher proportions of both overall autoimmune disease and cutaneous autoimmune disease diagnoses in the dupilumab-treated group compared to the untreated group over the four-year follow-up period (Figure E3). Alopecia and psoriasis were the most frequently observed cutaneous autoimmune diseases (Table E4).

Table 3.

Weighted Incidence Rates of Autoimmune Disease and Disease Subcategories^a

	Atopic Dupilumab-Treated	Atopic Dupilumab-Untreated
Autoimmune Disease Outcome	Weighted IR per 1,000 Person-Years		Rate Difference (95% Confidence Interval)
Any	8.78	6.31	2.47 (0.82, 4.30)
Cutaneous only	6.12	3.94	2.18 (0.67, 3.66)
Non-cutaneous only	2.69	2.49	0.20 (−0.72, 1.20)
Endocrinologic only	0.87	0.79	0.08 (−0.44, 0.68)
Gastrointestinal only	1.13	0.80	0.33 (−0.26, 1.01)
Rheumatologic only	0.69	0.79	−0.10 (−0.56, 0.44)

^aThe incidence rate (IR) of autoimmune diseases and disease subcategories per 1,000 person-years was determined using Poisson regression. IRs were not determined for autoimmune hematologic and neurologic disease categories due to small subject outcome counts (n<5).

Cumulative incidence curves for treated and untreated atopic children are shown in Figure 1, and hazard ratios (HRs) for specific autoimmune diseases and disease categories are reported in Figure 2. In fully adjusted Cox proportional hazards models, there was a modest, significant 1.45-fold (95% CI 1.18, 1.80) increased risk of any autoimmune disease diagnosis and a comparable significant 1.57-fold (95% CI 1.22, 2.03) increased risk of cutaneous autoimmune disease diagnosis in children treated with dupilumab. When psoriasis and alopecia were excluded, there was a non-significant 1.86-fold (95% CI 0.79, 4.38) increased risk of cutaneous autoimmune disease in dupilumab-treated atopic children. Results were not significant for autoimmune non-cutaneous diseases, including endocrinologic, rheumatologic, and gastrointestinal diseases (Figure 1).

Figure 1. Autoimmune Disease Weighted Cumulative Incidence.

Cumulative incidence curve demonstrating the proportion of children with atopy treated or untreated with dupilumab who were diagnosed with autoimmune disease up to four years following cohort entry.

Figure 2. Forest Plot of Hazard Ratios from Weighted Cox Regressions with Full or Reduced Adjustments.

Autoimmune disease hazard ratios (HRs) were computed by comparing atopic dupilumab-treated vs. untreated children. Reduced adjustment was applied to determine HRs for autoimmune cutaneous disease without psoriasis or alopecia, and autoimmune endocrinologic, gastrointestinal, and rheumatologic disease categories due to small outcome counts.

In sensitivity analyses, the increased risk of any autoimmune disease diagnosis and cutaneous autoimmune disease diagnosis for dupilumab-treated patients persisted when only a single instance of a corresponding disease code was required (Figure E4) and when varying the follow-up time to 1, 2 or 3 years, within both unadjusted and reduced-adjustment models (Figure E5). In sensitivity analyses stratified by disease severity and comorbidity, children with asthma comorbid with non-severe (Figure E6) or severe (Figure E7) AD had 1.78-fold (95% CI 1.20, 2.63) increased and 0.72-fold (95% CI 0.55, 0.93) decreased risks of autoimmune disease diagnosis, respectively, associated with dupilumab treatment. In children with any asthma severity, there was a 2.01-fold (95% CI 1.40, 2.88) increased risk of any autoimmune disease diagnosis associated with dupilumab treatment in children who also had comorbid AD, and no significant effect in those with exclusive asthma (Figure E8).

Lastly, in a secondary analysis, we interrogated the baseline prevalence of autoimmune disease diagnosis in children with atopy (dupilumab-treated and untreated) versus children without atopy. Compared with non-atopic children (range: 1.09–1.45%), children with atopy – regardless of dupilumab treatment – exhibited a higher mean annual prevalence of autoimmune disease diagnosis (range: 4.01–4.34%) between 2017 and 2023 (Figure E9).

DISCUSSION

In this retrospective study, we observed a modestly increased incidence of cutaneous autoimmune disease diagnosis in atopic children aged 6–17 treated with dupilumab as compared with untreated atopic children, an excess of about 2 cases per 1,000 children per year. We note that clinical trials performed in this patient population have proven dupilumab to be an effective treatment modality, improving disease severity, rates of exacerbation, and quality of life.^20–23 These studies have also demonstrated the overall safety of dupilumab, with injection site reactions as the most common adverse events. Other side effects that have been reported with dupilumab across children and adults include ocular manifestations (e.g., conjunctivitis, keratitis, blepharitis), oral and nasopharyngeal symptoms, gastrointestinal symptoms, arthralgias, upper respiratory tract and herpes viral infections, and peripheral eosinophilia.²⁴ Case reports of new-onset conditions such as psoriasis and inflammatory arthritis in predominantly adult patients treated with dupilumab have prompted interest in whether dupilumab might modulate risk of developing autoimmune or autoinflammatory diseases in children, a knowledge gap we addressed in this study.^13,14

Our sub-analyses revealed that psoriasis and alopecia areata played a notable role in driving the increased incidence of cutaneous autoimmune disease diagnosis associated with dupilumab. To date, multiple case reports have documented de novo development of psoriasis and/or psoriasiform lesions in association with dupilumab in adults, and to a lesser extent, children.^13,25 New-onset psoriasis has also been reported in the setting of both AD and, to a lesser extent, in non-AD conditions such as asthma and chronic rhinosinusitis with nasal polyps.^26–30 We note that our findings build on those of a recent retrospective study that documented an increased relative risk of developing psoriasis in adults with AD prescribed dupilumab as compared with those not prescribed dupilumab.³¹ Mechanistically, dupilumab might promote psoriasis through antagonism of IL-4, a negative regulator of T helper 17 (Th17) cells, which are implicated in the immunopathology of psoriasis.³² In our study, some of the increase in autoimmune cutaneous disease may reflect misdiagnosis of mixed AD-psoriasis, leading to control of AD and uncovering psoriasis. Apart from psoriasis, there is literature indicating mixed results for dupilumab in the setting of alopecia areata.^33,34 These observations underscore the complexity and need for continued study of immune responses in autoimmune or autoinflammatory cutaneous conditions.

Reassuringly, comparison of autoimmune disease rates among non-cutaneous autoimmune disease categories in our analysis did not reveal statistically significant differences. A relatively more recognized non-cutaneous condition described in select dupilumab-treated adult patients is inflammatory arthritis, which is characterized by seronegative arthritis, enthesitis, and/or tenosynovitis.^14,35–36 However, we did not observe an increased risk of autoimmune rheumatologic disease in our study. It is possible that this entity is unique to adults or, alternatively, not captured by our study design.

Notably, compared with non-atopic children, children with atopy in our cohort had a higher prevalence of autoimmune disease diagnosis at baseline. In a prior retrospective cohort analysis, AD was found to be associated with an increased risk of subsequent development of autoimmunity, in particular for those with severe AD.³⁷ While atopic diseases are classically associated with Type 2 inflammation, evolving knowledge indicates that that atopy reflects a state of immune dysregulation, with heterogenous involvement of immune responses that also include non-Type 2 immune mediators.^38–42 Future studies will need to delineate the immunologic bases of relevant disease endotypes, to inform primary treatment approaches as well as treatment of subsequent complications. Interestingly, there is evidence that Janus kinase inhibitors may be a successful treatment modality of severe cases of psoriasis and/or inflammatory arthritis noted following initiation of dupilumab.^43,44 Additionally dual biologic therapy targeting distinct immunologic axes may be needed in patients with concurrent Type 2 and non-Type 2 inflammatory diseases.⁴⁵

To our knowledge, our study is the first observational analysis to report on autoimmune disease outcomes following dupilumab treatment in children with atopy, addressing knowledge gaps about the potential side effect profile of dupilumab in the pediatric population. Avoiding reporting bias associated with self-reported studies, our study used EHR diagnostic data across multiple institutions to study disease outcomes. Further, we used weighting and adjustment for covariates, which allowed for the creation of a dupilumab-untreated cohort with similar characteristics as the dupilumab-treated cohort. This weighting also allowed for comparison with a much larger control cohort, optimally powering our study to detect rare disease outcomes. Another strength of our study is that it examined a real-life population which may be more generalizable to children that are prescribed dupilumab in clinical settings than those constrained by strict criteria used in randomized studies.

Limitations

This study is not necessarily nationally representative and is retrospective in nature. Thus, it is equipped to provide insights into disease risk associations and effect sizes but not causal relationships. We note that EHR-based diagnosis codes and prescription information were used as proxies for presence and severity of diseases but, in practice, may not capture true disease states. Moreover, their use may vary from clinician to clinician and could result in outcome misclassification. If there is no systematic effect of dupilumab exposure on autoimmune disease misclassification, this would be expected to bias effect estimates toward the null. We also note that it is not possible to reliably match patients by disease severity. As such, residual confounding by disease severity may remain. As an alternative approach, we conducted stratified sensitivity analyses by disease severity. Future studies would benefit from incorporating validated clinical severity metrics and medications trialed. Additionally, prescription of dupilumab is only a proxy for dupilumab exposure. Since prescription records may overestimate actual medication use, exposure misclassification could underestimate observed associations between dupilumab and disease outcomes. We cannot discount a potential role for ascertainment bias, such that autoimmune diseases may be diagnosed more frequently in dupilumab-treated patients due to clinicians being more attuned to potential side effects compared with clinicians of untreated patients. As most dupilumab-treated patients had AD, one may postulate that closer longitudinal monitoring of cutaneous manifestations may contribute to increased detection of cutaneous autoimmune conditions. Lastly, while we focused on children aged 6–17 due to follow up constraints related to dupilumab approval timeframes, dupilumab is now approved for use in children as young as 6 months.¹⁰ Thus, the generalizability of our findings to children under age 6 is not fully clear, and future studies should evaluate longitudinal outcomes in this youngest age group.

Conclusions

In this study, we used a multi-institutional pediatric EHR database to ascertain the rates of autoimmune disease diagnosis in children with atopy aged 6–17 who were prescribed dupilumab for the treatment of AD and/or asthma. While the weighted incidence rates of autoimmune disease diagnoses were low overall for both dupilumab-treated and untreated children with atopy, we detected an approximately 1.6-fold increased risk of cutaneous autoimmune disease in children treated with dupilumab, as compared with untreated children. At present, our findings support longitudinal monitoring of children being treated with dupilumab, which is already the standard of care. Prospective studies will be valuable in confirming the accuracy of EHR-based diagnoses and more comprehensively understanding dupilumab’s side effect profile, which must be weighed with its established safety and efficacy in managing treatment-refractory atopic diseases.

CLINICAL IMPLICATIONS.

This study found that children with atopy undergoing treatment with dupilumab may exhibit cutaneous autoimmune manifestations. This supports longitudinal clinical monitoring, the current standard of care in patients on biologics.

ABBREVIATIONS

AD

atopic dermatitis

CED

cohort entry date

IPTW

inverse probability of treatment weighting

IR

incidence rate

PMCA

pediatric medical complexity algorithm