Pre-and perinatal exposures associated with developing pediatric-onset immune-mediated inflammatory disease A Danish nation-wide cohort study

A


Introduction
Immune-mediated inflammatory diseases (IMID) are characterized by a dysregulated immune response leading to chronic autoimmune inflammation in various target organs such as the gut, liver, joints, and connective tissues.Although various clinical phenotypes exist, recent studies have indicated a close relationship between the IMIDs with shared inflammatory pathways, of which the interleukin-12/23 pathway is likely the most studied pathway [1].In this pathway, multiple susceptibility loci are shared between conditions such as inflammatory bowel disease (IBD), autoimmune liver disease (AILD), and juvenile idiopathic arthritis (JIA).Moreover, studies have reported familial clustering of IMID [2], and other studies have reported that individuals diagnosed with one IMID carry an increased risk of acquiring a second IMID [3].
The exact etiology of IMID remains unknown, but IMID is thought to develop due to an aberrant immune response in genetically susceptible individuals triggered by various environmental factors.The fetal and early life periods are characterized by extensive organ maturation and immunologic plasticity [4].Some authors have hypothesized that adverse exposures during these developmental periods are likely to alter the immune system, making the child susceptible to disease later in life [5].Following this hypothesis, several epidemiological and case-control studies have found that early-life exposures are associated with a higher risk of adult-onset IMID (aIMID).Among others, low birth weight, passive smoking, and antibiotic exposure have been associated with a higher risk of aIMID, whereas breastfeeding seems protective of aIMID [6][7][8][9].
Because the timeframe of possible exposure is limited in pediatriconset IMID (pIMID) compared with aIMID, early-life exposures and genetic susceptibility are thought to play are larger role in pIMID etiology [10]; however, only circumstantial evidence exists supporting this hypothesis [11].Also, mounting evidence points towards different clinical phenotypes of pIMID and aIMID.For instance, pediatric-onset inflammatory bowel disease (pIBD) presents with more extensive disease and requires more aggressive treatment protocols than adult-onset IBD [12].Thus, exposures triggering aIMID cannot necessarily be extrapolated to pIMID.Therefore, we aimed to assess the impact of pre-and perinatal exposures on the development of pIMID in a nationwide registry-based cohort.

Study overview and population
This register-based cohort study included all individuals born in Denmark from 1 January 1994 to 31 December 2014.Individuals were identified from the Medical Birth Register (MBR), where maternal, obstetric, and perinatal outcomes are recorded.Subsequently, data from the MBR were cross-linked to national socioeconomic registers and inand outpatient diagnoses from the National Patient Register (NPR) to obtain data on potential pre-and perinatal exposures.Cross-linking of registers is possible in Denmark because all children are assigned a unique personal identification number (civil registration number) at birth.We excluded children who died before six months of age because pIMID diagnoses before this age are extremely rare (Fig. 1).

Outcomes
The primary outcome was development of pIMID (diagnosis <18 years) which was defined as any of the following diagnoses: pediatriconset Crohn's disease (pCD), pediatric-onset ulcerative colitis (pUC), pediatric-onset autoimmune hepatitis (pAIH), pediatric-onset primary sclerosing cholangitis (pPSC), JIA and pediatric-onset systemic lupus erythematosus (pSLE).Secondary outcomes included diagnosis with disease-specific pIMID or being diagnosed with more than one pIMID.pIMID cases were defined by a registration of a corresponding International Classification of Diseases and Related Health Problems, 8th revision (ICD-8) or 10th revision (ICD-10) in the NPR with onset during childhood (<18 years) (Supplementary Table 1).Only diagnoses confirmed after six months of age were considered.For pCD and pUC, the diagnoses were defined as > 1 entry in the NPR, as previously proposed [13].Moreover, the latest registered diagnosis was used if an individual was diagnosed with both pCD and pUC.The remaining pIMIDs were considered valid with any entry in the NPR.pPSC and pSLE were not included in the disease-specific analysis due to small sample sizes (25 and 61, respectively).
When assessing exposures associated with being diagnosed with more than one pIMID, we defined the healthy control group as (1) the entire background population (those not diagnosed with pIMID and those diagnosed with only one pIMID) or (2) only those not diagnosed with pIMID.
Individuals were followed until 31 December 2014 (end of followup), date of emigration, death, or pIMID diagnosis, whichever came first.
Maternal preconception IMID was defined according to pIMID (Supplementary Table 1), but diagnoses needed to have been assigned before childbirth and rheumatoid arthritis was added to the definition.Rheumatoid arthritis was defined as the ICD-8 codes 712.19, 712.29, 712.39, 712.49, or 712.59; or ICD-10 codes M05 or M06.In analyses on disease-specific pIMID, maternal preconception IMID included all other IMID diagnoses than the investigated diagnosis.Maternal diseasespecific IMID was defined as disease-specific pIMID, but rheumatoid arthritis was included in the definition of maternal JIA.

Statistical analysis
We present descriptive statistics as proportions and medians with interquartile ranges (IQR), as appropriate.The risk of developing pIMID was assessed by Cox regression analyses and presented by hazard ratios (HR) with 95% confidence intervals (95% CI).Results are presented as both crude and adjusted HR (aHR).In the main multivariable model, prenatal exposures were adjusted for the remaining prenatal exposures and perinatal exposures were adjusted for both prenatal and the remaining perinatal exposures except for smoking.Due to missing data in the covariate smoking, we added smoking in a secondary analysis alongside the perinatal exposures.Because this did not significantly change the conclusions of this paper, we present the results from the secondary analysis as supplementary material only (Supplementary Tables 4 and 5).
In the main analyses on the risk of pIMID, 3.9% (n = 52,071) of the children were excluded due to missing values for one or more covariates (Fig. 1).An adjusted p-value <0.05 was considered statistically significant.Analyses were conducted using RStudio, version 3.6.1.

Ethical approval
The Danish Data Protection Agency approved the data retrieval (DT record number: 2012-58-0004, local record number: AHH-2016-033, Isuite number: 04790).Under Danish legislation, register-based studies do not require approval from a scientific ethics committee.

Results
In total, 1,350,353 children were included in the analyses.This yielded a total follow-up time of 14,158,433 person-years and a median individual follow-up time of 10.4 years (IQR 5.2 to 15.8).Due to death within the first six months, 8,554 children were excluded.A total of 2,728 (2.0‰) children were diagnosed with a pIMID during the study period.Among these, 54 were subsequently diagnosed with additional pIMID diagnoses.Patients with pediatric-onset AILD (pAILD) frequently developed more than one pIMID, with 23.7% of pAIH patients and 60.0% of pPSC patients developing additional pIMIDs during follow-up (Supplementary Table 2).

Prenatal exposures
Risk analyses for any pIMID and for disease-specific pIMID are presented in Table 2 and Table 3, respectively.In general, maternal educational level was not associated with risk of pIMID, though we found a lower risk of pCD in individuals with a maternal educational level of 10-12 years (aHR 0.77 [95% CI: 0.61 to 0.97]) and indication of a lower HR was associated with higher maternal education (aHR 0.77 [95% CI: 0.52 to 1.14]).Previous pregnancies were associated with a higher risk of pIMID in the offspring (aHR 1.11 [95% CI: 1.02 to 1.20).When assessing the disease-specific risks, we found that higher risks of pUC and JIA (aHR 1.26 [95% CI: 1.04 to 1.52] and 1.14 [95% CI: 1.02 to 1.28], respectively) were associated with previous pregnancies.Children of mothers diagnosed with IMID were more likely to develop pIMID during follow-up than were children with no maternal IMID (aHR 3.54 [95% CI: 2.74 to 4.58]).When assessing disease-specific risks, we found that a maternal IMID diagnosis was associated with a higher risk of the same pIMID developing in the child across all analyzed pIMIDs (e.g.,

Table 1
Pre-and perinatal exposures for children born in Denmark 1990-2014 and diagnosed with pediatric-onset immune-mediated inflammatory diseases before their 18th birthday.

Perinatal exposures
Children delivered by Cesarean section were more likely to develop pIMID than vaginally delivered children (aHR 1.15 [95% CI: 1.04 to 1.28]); however, this finding was only statistically significant in pCD and JIA when stratified by specific disease (aHR 1.29 [95% CI: 1.03 to 1.61] and 1.23 [95% CI: 1.08 to 1.41], respectively).Children from twin or higher-order births had a lower risk of developing pIMID than other children (aHR 0.71 [95% CI: 0.56 to 0.90]).When assessing diseasespecific risks, we found lower risk estimates in multi-fold than singleton pregnancies across all analyzed pIMID; however, this finding was not statistically significant.Overall, females had a higher risk of developing pIMID than males (aHR 1.46 [95% CI: 1.35 to 1.58]).When stratified by the specific pIMID, the higher risk for females persisted for pAIH and JIA (aHR 1.65 [95% CI: 1.09 to 2.49] and 1.88 [95% CI: 1.69 to 2.09], respectively).

Exposures associated with developing more than one pIMID
The frequencies of developing more than one pIMID are presented in Supplementary Table 2, and the risk assessments are presented in Supplementary Table 3.The risk of developing more than one pIMID was higher in children born by mothers diagnosed with IMID than in the total background population (aHR 9.58 [95% CI: 2.98 to 30.76]).No other pre-or perinatal factors were statistically significantly associated with pIMID; however, the study population was small with only 54 patients developing more than one pIMID.

Discussion
In this nationwide study including all infants born in Denmark from 1994 to 2014, we assessed the association between pre-and perinatal exposures and the risk of developing pIMID.The strongest risk factor across all included pIMID was having a mother diagnosed with IMID.Also, parental education, parity, Cesarean section, and child sex were associated with the risk of developing pIMID.
Our findings confirm the hypothesis that the etiology of pIMID is more likely to be caused by genetic susceptibility than by environmental exposures [10].We report that maternal IMID increases the risk of being diagnosed with pIMID considerably.Furthermore, the highest risk of a specific pIMID befalls to children with a mother with the same disease-specific IMID followed by a mother with any IMID.A recent nationwide Swedish study including more than 50,000 IBD patients and first degree relatives also proposed a large genetic burden [11].The authors found a large degree of familial aggregation of IBD, which was even more pronounced in patients with pIBD.In line with our results, a previous population-based study found that offspring to mothers diagnosed with rheumatoid arthritis were three times more likely to develop the same condition [14].The authors did not restrict the analysis to childhood onset, and the reported risk estimates are lower than the ones we report in this paper (aHR 22.69).To our knowledge, no previous studies reporting on pAILD have been published on this topic.
Among the assessed prenatal exposures, we found that children whose mothers had a secondary education carried a lower risk of being diagnosed with pCD than children whose mothers had a primary education only.The associations for higher educational levels pointed in the same direction but were not statistically significant.This finding is consistent with a previous case-control study including 466 Canadian pCD patients, which showed lower risks of pCD in children to mothers with a college or university level education [15].A population-based Australian study contrasted our finding of a close correlation between socioeconomic status and maternal education [16].This study reported a higher risk of pCD in children with mothers in the highest socio-economic quintile than in mothers in the lowest quintile.However, the study population was small (N = 278) and historical, patients being included between 1983 and 1998.Due to the study design, our study cannot elaborate on the underlying mechanisms of this finding.Different behavioral patterns may be involved, which should be the focus of future studies.
We found a higher risk of pIMID in children of mothers with previous pregnancies.When focusing on disease-specific risks, we found that this was also the case for pUC and JIA.Our study cannot determine whether this finding was due to the immunologic impact of several pregnancies or environmental exposure due to siblings in the same household.An American case-control study, including 1,252 cases, reported a protective effect of multiparity in JIA [17].For pUC, we found no studies reporting on this, but two studies reported a protective effect on aUC from having siblings in the household during childhood [18,19].Both studies hypothesized that these findings were owed to a protective antigenic stimulation and maturation of the immune system.Our results do not support this, which may potentially indicate that decreased antigenic stimulation increases the risk of aIMID to a larger degree than does pIMID where a genetic load may play a more prominent role.However, no studies have studied differences in risk factors between pIMID and aIMID.We found a protective effect of plural pregnancies across all analyzed IMIDs.This was only statistically significant in the pIMID group.These findings run contrary to the previous literature for JIA and pIBD alike [8,17,20].No studies have reported on this topic for pAILD.For JIA, two previous studies (both case-control studies with a limited number of plural pregnancies) reported no association between plural pregnancies and JIA [8,17].One UK population-based IBD study reported no association.However, the study population was limited with only 170 pIBD patients being included [20].One of the most debated early-life exposures possibly leading to IMID is Cesarean section.It has been hypothesized that the altered microbiome caused by reduced exposure to the mother's vaginal flora [21] leaves the child more susceptible to IMID.Previous studies have associated Cesarean section with developing pIMID such as pIBD, JIA, and systemic connective tissue disorders [22].In our study, we found a slightly higher risk of developing pIMID in children delivered by Cesarean section; however, this increased risk was statistically significant only for pCD and JIA.Our finding is supported by findings from a recent meta-analysis on early-life risk factors of JIA [23].Whether or not the higher risk estimates found in these studies constitute a causal effect was debated in a Finnish study published in 2021 [24].The authors reported that the effect may be explained by a larger group of preterm infants in the Cesarean section group, who were more frequently prescribed antibiotics in early life.Unfortunately, our study was not designed to further disentangle the causality of these associations.
A previous population-based study from Finland found a higher prevalence of JIA in female than in male individuals, whereas pIBD was equally prevalent among sexes [24].This is in line with our results.For pAILD, we also found a higher risk in females than males.This has previously been reported in the adult AILD population [25] but not in population-based pAILD studies.Lastly, we found a substantially higher Table 3 Cox regression analyses for children born in Denmark 1990-2014 -Risk of developing Crohn's disease, ulcerative colitis, autoimmune hepatitis, or juvenile idiopathic arthritis.Associations are presented as crude and adjusted hazard ratios.Statistically significant results (p < 0.05) are presented in bold.risk of developing additional pIMID after the first diagnosis of pIMID, particularly in pAILD.This phenomenon is well described for pIBD, where the frequency of an additional IMID (termed extraintestinal manifestation) is 18-25% [26,27].In line with our findings, previous studies reported that up to 50% of children with pAILD developed other pIMID as well.However, these data are based on small cohorts, with the largest study including only 48 patients [28].Patients with multiple pIMID diagnoses have been found to experience more aggressive disease courses [26], and it would be helpful to identify and eventually intervene on risk factors for developing multiple pIMID.However, our study suggests that the genetic load plays a key role in the development of several different IMID, with risk estimates of maternal pIMID being three times those observed for patients developing only one pIMID.To identify target points for intervention, future studies should focus on epigenetics leading to a higher risk of developing multiple IMID diagnoses.
This nationwide birth cohort study included prospective life-course data on 1,350,353 children born in the course of a 20-year period in Denmark.Even within this large population sample, the disease-specific groups were small.Despite this limitation, this study remains among the largest population-based studies reporting on these findings in a pediatric population.We included a broad range of pre -and perinatal factors.However, other factors, such as breast feeding, use of medicine during pregnancy, and environmental exposures, may also have been relevant.These exposures were unfortunately not available from the included registries.Moreover, our data did not allow further elaboration on the maternal disease course during pregnancy.It is plausible that the disease activity and prescribed anti-inflammatory or immunomodulating drugs might impact the risk of pIMID.
In conclusion, we found that maternal IMID was, by far, the strongest risk factor for developing pIMID.It is important to emphasize that the absolute risk of developing pIMID in children with maternal IMID remains low.Therefore, our findings do not support genetic counseling for families with IMID.Our findings further indicate a difference in the etiology of pIMID and aIMID supporting a larger genetic burden in pIMID.This should be acknowledged when conducting future research and calls for studies focusing on epigenetic interactions, particularly in pIMID.Clinically, our study confirmed that Cesarean section is an intervenable risk factor for developing pIMID, but it was associated only with a slightly higher risk of pIMID.Physicians should keep this in mind when caring for high-risk populations such as pregnant women previously diagnosed with an IMID.

Funding
The study was financed by grants from the Research Fund of Rigshospitalet, Copenhagen University Hospital, Denmark.

Fig. 1 .
Fig. 1.Overview of the included study population among individuals born in Denmark from 1 January 1994 to 31 December 2014.

Table 2
Cox regression analyses for children born in Denmark 1990-2014 -risk of developing any pediatric-onset immune-mediated inflammatory disease.Associations are presented as crude and adjusted hazard ratios.Statistically significant results (p < 0.05) are shown in bold.