Objectives Early antibiotic exposure influences the gut microbiota which is believed to be involved in the pathogenesis of juvenile idiopathic arthritis (JIA). We aimed to investigate the association between systemic antibiotics in prenatal and early life and risk of JIA.
Methods We conducted a register-based cohort study including all children born in Norway from 2004 through 2012. The children were followed until 31 December 2020. Main exposures were dispensed antibiotics to the mother during pregnancy and to the child during 0–24 months of age. The outcome was defined by diagnostic codes indicating JIA. Multivariate logistic regression analyses were performed to estimate the association between antibiotic exposure and JIA.
Results We included 535 294 children and their mothers in the analyses; 1011 cases were identified. We found an association between exposure to systemic antibiotics during 0–24 months and JIA (adjusted OR (aOR) 1.40, 95% CI 1.24 to 1.59), with a stronger association for >1 course (aOR 1.50, 95% CI 1.29 to 1.74) vs 1 course (aOR 1.31, 95% CI 1.13 to 1.53). Subanalyses showed significant associations in all age periods except 0–6 months, and stronger association with sulfonamides/trimethoprim and broad-spectrum antibiotics. There was no association between prenatal antibiotic exposure and JIA.
Conclusions The novel observation of no association with prenatal antibiotic exposure and JIA suggests that the association between antibiotics in early life and JIA is unlikely to be confounded by shared family factors. This may indicate that exposure to antibiotics in early life is an independent risk factor for JIA.
- arthritis, juvenile
- autoimmune diseases
Data availability statement
Data are available on reasonable request. Data are not accessible due to national GDPR practices, but could be made available from the researchers on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
WHAT IS ALREADY KNOWN ON THIS TOPIC
Antibiotic exposure in early life can have lasting impact on the composition and diversity of the gut microbiota.
While environmental risk factors for juvenile idiopathic arthritis (JIA) are largely unknown, early life antibiotics and gut microbiota have been linked to disease development.
WHAT THIS STUDY ADDS
In our nationwide study, antibiotic exposure during age 0–24 months but not prenatal antibiotics was associated with later JIA development.
The novel finding of no association with prenatal antibiotic exposure suggests that the association between antibiotic exposure in childhood and JIA is unlikely to be confounded by shared family factors.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Our findings suggest that antibiotic exposure in early life is an independent risk factor for JIA.
Future studies should aim to capture different infectious agents and antibiotics to separate these potential risk factors.
Biomarkers of the gut microbiome before development of JIA would strengthen a potential causal link.
Juvenile idiopathic arthritis (JIA) is the most common chronic inflammatory rheumatic condition of childhood and is usually categorised into seven clinically heterogeneous subtypes with chronic arthritis as a common characteristic.1 A pooled annual incidence rate of 7.8/100 000 among Caucasians has been reported with the highest incidence in Europe, especially Nordic countries (15–22.6/100 000).2–4 The aetiology of JIA is multifactorial with both genes and environmental factors contributing.5 6 However, since there is a lack of high-quality studies concerning environmental risk factors, knowledge is limited.5–7
There is an increasing interest on the early life gut microbiota and its role in the development of autoimmune diseases, including JIA.7–9 The gut microbiota develops rapidly during the first years of life and stabilises at around three years of age.10 The main determinants of the gut microbiota are mode of delivery, infant feeding and exposure to antibiotics.11 Exposure to systemic antibiotics before a stable mature microbiome is established may have a long-lasting influence on the composition and diversity of the gut microbiota of the child.12
There is an interplay between the gut microbiota and the development of the immune system. Influences on the immune system induced by the microbiota early in life can be lasting, with consequences for disease development later in life.7 Children with JIA were found to have an altered composition and a lower diversity in their gut microbiota compared with healthy children.13 Cross-sectional studies are, however, insufficient to study causality, and it remains unclear whether this is involved in the aetiology of the disease.13 Interestingly, a recent cohort study demonstrated an altered gut microbiota as early as in infancy in 12 children who later developed JIA.14
Some studies have found a positive association between exposure to antibiotics in early life and JIA.15–18 However, the studies have limitations including small sample sizes, case–control designs with high risk of recall bias, significant loss to follow-up, lack of exposure data from the neonatal period, and relatively low age at JIA diagnoses, factors that may limit the generalisability of the studies. The association of prenatal antibiotic exposure on the risk of developing JIA is mainly unexplored.17 Thus, the objective of our study was to investigate the association between exposure to systemic antibiotics in prenatal and early life, with the risk of developing JIA in a cohort study.
Study population and design
This study was a nationwide register-based cohort study from Norway. All children born between 1 January 2004 and 31 December 2012 and registered in the Medical Birth Register of Norway (MBRN) were included. MBRN is a national health registry containing information about all births in Norway, and the register is based on mandatory reporting.19 We linked data from MBRN to relevant population-based registers on an individual level using the unique national identification (ID) number.
From 2008, The Norwegian Patient Registry (NPR) has received data with personal ID numbers from all Norwegian public hospitals and specialists with public funding,20 and thus captures virtually all patients with JIA. JIA cases were defined by a combination of International Classification of Diseases (ICD) codes (ICD-10 codes M08 Juvenile arthritis and M09 Juvenile arthritis in diseases classified elsewhere like juvenile arthritis in psoriasis and inflammatory bowel disease) reported to NPR between 1 January 2008 and 31 December 2020.
Children with either ≥2 M08, ≥2 M09 or 1 M08 combined with 1 M09 ICD-10 codes were classified as cases. In addition, children with only 1 M08 or 1 M09 were classified as cases if the year of diagnosis was 2020 because this was the last year with data from NPR and it is likely that some of these children only had only one visit before the end of the year.
We had no data on exact dates of registrations, but cumulative data for registrations each year from 2008 and onwards. Year of onset was defined as the first M08 or M09. In cases where M08/M09 were preceded by a code of M13 (other arthritis), the first entry of M13 was used as onset. Due to missing ID numbers on data from NPR before 2008,20 exact year of diagnosis was not available for children diagnosed 2004–2007. Since we only had birth year and birth month available, a replacement birthday on the 15th in each month was constructed to calculate age at exposure and outcome.
For data on exposure to antibiotics, the children and their mothers were linked to The Norwegian Prescription Database (NorPD). Prenatal exposure was defined as any dispensed systemic antibiotic to the mother from a pharmacy during pregnancy. Further, child antibiotic exposure was defined as any dispensed systemic antibiotic to the child from a pharmacy during 0–24 months of age. In addition, data on systemic parenteral antibiotics given during the neonatal period in hospital were available from MBRN and included in the main exposure.
In Norway, antibiotics are only available by prescription. All prescribed drugs dispensed from Norwegian pharmacies are registered at the individual level in NorPD and classified after Anatomic Therapeutic Classification codes.21 The different types of antibiotics were categorised into five groups. In approximately 30% of the prescriptions, there was a lack of ID numbers, precluding the individual linkage.
From MBRN we included child’s sex, birth weight, season of birth, year of birth, mode of delivery, pre-eclampsia, maternal age, parity and smoking in pregnancy. From Statistics Norway (SSB), maternal educational level by end of follow-up was included.
Patient and public involvement
Patient and public involvement was not considered relevant in this registry-based study.
We performed χ2 tests to investigate associations between categorical variables. Further, we performed multivariate logistic regression analyses to estimate the OR for JIA by antibiotic exposure prenatally or during 0–24 months of age. The analyses were performed with cluster correction to adjust for correlation between siblings. In the main analyses, we adjusted for sex (adjusted OR (aOR) because JIA is more common in girls than in boys, and boys more often receive antibiotics in early life.7 21 Further, we performed subanalyses on number of dispensed antibiotics, exposure in different time periods and on types of antibiotics dispensed. In secondary models the effects of adjusting for the following potential confounders were assessed: prenatal antibiotic exposure, season of birth, year of birth, mode of delivery, birth weight, pre-eclampsia, maternal age, maternal parity, maternal educational level and maternal smoking during pregnancy.
Since the most common subtypes of JIA have a biphasic peak in age at onset with the first peak already between 2 and 4 years,7 we could not exclude an overlap between the timing of antibiotic exposure and disease onset in some children. We, therefore, performed a sensitivity analysis excluding children with JIA onset before 3 years of age, and all children with JIA born 2004–2006 since the exact year of diagnosis was unavailable in children diagnosed before 2008. To study the effect of missing exposure data (due to lack of ID number for linkage, which was more common from 2004 to 2007), we assessed the association by year of dispense.
A 95% CI not overlapping the null value 1.00 for the OR was regarded as statistically significant. All analyses were performed by using Stata V.16.1 statistical software (StataCorp).
In the final analyses, 535 294 children and 360 769 mothers were included (figure 1). We identified 1011 JIA cases with 616 girls (60.1%) and 395 boys (39.1%) after a median follow-up (from birth through 31 december 2020) of 12.4 years (range 8.0–16.9). Of all children included, 149 534 (27.9%) were exposed to systemic antibiotics prenatally and 236 340 (44.2%) during 0–24 months of age (table 1). The distribution of selected baseline characteristics and covariates by dispensed antibiotics during 0–24 months of age are also presented in table 1.
Antibiotic exposure and diagnosed JIA
We found no significant association between prenatal exposure to antibiotics and JIA (aOR 1.10, 95% CI 0.96 to 1.26). In contrast, exposure to systemic antibiotics during 0–24 months of age was positively associated with JIA (aOR 1.40, 95% CI 1.24 to 1.59). The association was dose-dependent with a stronger association for increasing the number of dispensed courses. Subanalyses on different age periods showed significant associations in all age periods except for systemic antibiotics during 0–6 months, including the neonatal period (table 2).
The frequencies of different types of antibiotics are presented in table 3.
Subanalyses of different types of antibiotics showed a tendency towards a stronger association with sulfonamides/trimethoprim and the group containing more broad-spectrum antibiotics (‘systemic antibiotics not included in other subgroups’) (table 4).
The association between exposure to systemic antibiotics during 0–24 months of age and JIA was robust and essentially unchanged after adjustment for other potential confounders (table 5).
In sensitivity analyses excluding children with JIA born 2004–2006 and other children with JIA onset before 3 years of age (online supplemental figure 1), the association with childhood antibiotics also remained, though slightly attenuated (online supplemental table S1). Sensitivity analyses performed to study the effect of missing exposure data showed the strongest association during 2010–2013, the period with the most complete exposure data (online supplemental table S2).
In this nationwide register-based cohort study including more than 500 000 children and over 1000 JIA cases, we found that exposure to systemic antibiotics in early life was positively associated with diagnosed JIA. In contrast, we found no association between prenatal exposure to antibiotics and JIA development.
Our first main finding of an association between antibiotic exposure in early life and JIA is consistent with earlier studies,15–18 including a large register-based study from Finland in which JIA was defined by reimbursement codes for antirheumatic drugs.16 The authors did not describe whether non-steroidal anti-inflammatory drugs (NSAIDs) was included in the case-definition,16 which might have introduced a selection bias towards more severe cases since some children with JIA only receive first line treatments with NSAIDs and intra-articular corticosteroid injections.22 A strength of our study was a longer follow-up time, and we were able to adjust for relevant possible confounders.16
We found a dose-dependent association with a stronger association for an increasing number of dispensed courses, which is also in line with earlier findings.15–18 Subanalyses of different types of antibiotics showed a tendency towards a stronger association between sulfonamides/trimethoprim and the more broad-spectrum antibiotics and JIA. A similar observation in the Finnish study was that antibiotics targeting anaerobic organisms were stronger associated with JIA compared to antibiotics limited to aerobes. These findings were not replicated in the studies from the UK15 and Sweden,17 but small sample sizes15 17 and antibiotic exposure based on parental reporting17 might have limited the ability to show such differences.
Our second main finding was the observation of no association between prenatal exposure to antibiotics and JIA. This association has previously only been investigated in the All Babies in Southeast Sweden (ABIS) study, suggesting an association between antibiotic exposure in pregnancy with the development of systemic-onset JIA (n=7), but not with JIA as a whole group or with other subtypes of JIA.17
It is believed that changes in the gut microbiota may play an important role in the pathogenesis of JIA.6 8 13 23 If there is a causal link between antibiotics and JIA via perturbations on the gut microbiota, our results correspond well with studies showing that exposure to systemic antibiotics in early life can have a long-lasting influence on the composition and diversity of the gut microbiota of the child.12 The findings of a dose–response relationship and a tendency towards a stronger association with more broad-spectrum antibiotics may also support a causal connection between the antibiotic’s disruptive effect on the microbiota and JIA, as repeated antibiotic courses can have cumulative effects on the microbiome, and broad-spectrum antibiotics may have a more profound influence on the microbiota.12
Our findings do not prove any causality, and an alternative explanation to the observed association between antibiotic exposure in early life and JIA could be that children who later develop JIA have inherent higher risk of infections/are more vulnerable to infections. A reverse causality cannot be ruled out, as joint symptoms could be misinterpreted as bacterial infections.24 Our sensitivity analyses do, however, not support the notion of reverse causality. To minimise the risk of reverse causality, we defined the exposure window to the first 2 years of life where the risk of developing JIA is low. Since the most common subgroups of JIA have a biphasic peak of onset with the first peak already between 2 and 4 years,7 and we only knew the year of JIA onset, we also performed sensitivity analyses restricted to children with JIA with known age of onset after the age of 3 years. The associations remained, though slightly attenuated.
Strengths in our study were our design with a nationwide register-based study including all births in a defined time period, making it representative at a population level and reducing the risk of selection bias.25 Compared with most other studies, we had a virtually complete sample size. All data were collected prospectively with objective measures from high-quality population-based registries, leading to minimal risk of misclassification due to recall bias.25 We had a long follow-up, covering the whole age period <16 years for the oldest children in the study. The design minimises loss to follow up,25 which would only occur if a child with JIA was not registered in the NPR in case of moving out of Norway or not diagnosed correctly. This small risk of misclassification would however change the estimates minimally, due to the size of the cohort.
Our study is the first large scale, register-based study investigating prenatal antibiotic exposure with the risk of JIA. We were able to adjust for prenatal antibiotic exposure in our main analyses investigating the effect of antibiotic exposure in early life and risk of JIA, and the associations remained almost unchanged. These observations suggest that the association between antibiotics in early life and JIA is unlikely to be confounded by shared family factors. We were also able to adjust for many other possible confounders including socioeconomic status, which has often been lacking in previous studies.5 In addition, we had data on different types of antibiotics and different age periods including exposure in the neonatal period.
A limitation is the lack of data on infections as a potential confounder. However, previous studies have not been able to show any clear connection between infections and JIA,6 and our findings are also consistent with a case-control study from UK15 and a study from the ABIS cohort in Sweden,17 which had the advantage of adjusting for infections. Since JIA consists of different subtypes considered different diseases with different pathogenesis,26 information about subtypes would be desirable. In our study, approximately 30% of the prescriptions in NorPD from the first year of life lacked ID numbers (online supplemental table S2), precluding individual-level linkage and leading to misclassification of some children exposed to antibiotics as unexposed. Subanalyses on year of dispense show the strongest association between antibiotics and JIA during 2010–2013, the period with highest exposure accuracy, suggesting that missing exposures bias the overall results towards the null. Also the lack of association with exposure during 0–6 months could be explained by missing ID-numbers from NorPD, which was most common in this age period. Consequently, children in this age group were more often misclassified as unexposed to antibiotics. The statistical power was lower for early exposures, and the overlap of CIs should also call for cautious interpretation.
Except in the neonatal period, we had no data on systemic antibiotics administered in hospital. Nevertheless, if an infection requires hospitalisation, the treatment is often completed after discharge, and therefore registered in the NorPD. An exception can be antibiotic treatment given in the neonatal ward, but here we had data available from MBRN.
Another limitation is the lack of validation in the outcome definition, which could potentially lead to misclassification of the JIA cases. It is known that errors in medical coding occur,27 and we, therefore, required ≥2 ICD-10-codes in our case definition, since we consider it less likely that a majority of the children are coded incorrectly more than once.
In the search for causal risk factors in JIA, reference is often made to studies that investigated the gut microbiota in children with JIA demonstrating altered microbial composition.23 Since most of these studies are only cross-sectional, the causal link between these observations and JIA pathogenesis remains uncertain. Future research should focus on longitudinal studies including predisease microbiological samples, which would strengthen potential causality.
In this nation-wide register-based cohort study, we found an association with exposure to antibiotics in early life and JIA development which remained after adjustments, suggesting that this is an independent risk factor for JIA. The novel observation of no association with prenatal exposure to antibiotics suggests that the association between antibiotics in early life and JIA is unlikely to be confounded by shared family factors.
Data availability statement
Data are available on reasonable request. Data are not accessible due to national GDPR practices, but could be made available from the researchers on reasonable request.
Patient consent for publication
The study was approved by the Regional Committee for Medical and Health Research Ethics (REK #18622) and the Norwegian Data Protection Authority. The study was exempted from individual consent because it was a registry-based study with a low risk of personal identification.
We would like to thank the registries; MBRN, NPR, NorPD and Statistics Norway for making data available. We would also thank cosupervisor Anne Marit Selvaag for valuable discussions.
Contributors SH made substantial contribution to analyses, interpretation of data and writing the manuscript, and also contributed to study design. SA made substantial contribution to analyses, interpretation of data and study design. HS made substantial contribution to interpretation of data and writing the manuscript. KS designed the study, made substantial contribution to analyses, interpretation of data, writing the manuscript and was the guarantor of the study. All authors read and revised the article critically and approved the final manuscript.
Funding We would thank the Norwegian Women's Public Health Association and the South-Eastern Norway Regional Health Authority for funding.
Competing interests SH: none declared, SA has received travel funding from Ferring pharmaceuticals., HS: none declared, KS: none declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.