Rheumatoid arthritis (RA) and juvenile idiopathic arthritis (JIA) are chronic autoimmune inflammatory diseases supposed to develop as untoward consequences of interaction(s) between genetic susceptibility and environmental risk factors, in a manner hitherto incompletely understood. The decreasing incidence of RA during the last decades suggests the importance of environmental exposures that have changed over time.1 2 In addition, the shift towards higher age at diagnosis suggests a birth cohort effect, indicating that risk factors acting early in life might be of importance.3 4 Recent data on the existence of antibodies to citrullinated peptides (CCP) many years before the onset of adult RA further strengthen the hypothesis that critical risk factors for RA may act long before the clinical onset of disease.5 6

There are numerous studies relating early life events to the risk of adult disease, including epidemiological as well as experimental data. The concept of “the developmental origin of adult disease” is based on the fetal response to its environment, inducing changes in the functional development that alter later disease risk. The period during which these early life events influence lifelong consequences can extend from conception to infancy.7

In the context of allergy, factors acting in utero or during infancy may profoundly modify adult functioning of the immune system.8^–10 More recently, it has also become clear that the timing and nature of environmental exposures in susceptible individuals may also modify the risk of autoimmune inflammatory disease, such as coeliac disease, type 1 diabetes mellitus and inflammatory bowel disease.11^–16 Little is known about early life exposures, infections in particular and subsequent risks of RA or JIA. The scarce data so far available do, however, strengthen the hypothesis that factors early in life are important in the aetiologies of RA as well as JIA. In a US study and in a small previous study from our group, high birth weight was identified as a risk factor for RA.17 18 In a UK study, rheumatoid factor positivity in adult women was dependent on their childhood living conditions, and in a Finnish study increased risks for RA and JIA were observed in girls exposed to maternal smoking during pregnancy.19 20

To assess the role of birth characteristics further and to explore the role of early life infections in the aetiology of RA and JIA, respectively, we performed a large population-based case–control study using national registers on pregnancy and delivery characteristics, infections, RA and JIA.

SUBJECTS AND METHODS

Design and setting

We performed a nationwide case–control study using prospectively recorded register data in Sweden 1973–2002. Using the national registration number (a unique 10-digit number assigned to all Swedish residents alive in 1947 and born thereafter),21 we linked data from the following registers: (1) The Swedish inpatient register, which is a population-based register, nationwide since 1987, which contains individual information on dates of hospital admission and discharge and discharge diagnoses, classified according to the Swedish version of the International Classification of Diseases (ICD), versions 8 to 10.22 Validation of the RA diagnosis against the American College of Rheumatology (ACR) criteria suggests that 90% of the registered RA diagnoses are correct.23 (2) The early arthritis register (EAR), which is a seminational register established in 1997 including incident (less than one year of symptom duration) cases of rheumatologist-diagnosed RA fulfilling the ACR criteria.24 (3) The Swedish medical birth register (MBR), which is a population-based register covering more than 99% of all births in Sweden since 1973, including prospectively recorded information on maternal characteristics, complications during pregnancy and delivery, mode of delivery and information on the newborn.25 (4) The Swedish population register, which is a weekly updated register of the vital status of all residents in Sweden.

Cases and controls

We defined two groups of cases, RA and JIA, respectively. Using the Swedish inpatient register, we identified all individuals born in 1973 or later who were first discharged with RA at age 16 years or above (n  =  208). Through linkage with the EAR we identified 143 partly overlapping cases with RA born in 1973 or later, such that the total number of cases with RA was 333. The study population of RA patients only includes cases between 16 and 29 years of age, which limits the analysis to “early adult onset RA”. Rheumatoid factor (RF) status was determined as specifically recorded in the EAR and/or by the ICD code for RA used in the Swedish inpatient register. Of the 333 cases with RA, 153 were RF positive, 107 RF negative and 73 were not specified (table 1). In the Swedish inpatient register, we further identified 3334 cases discharged with JIA. Of these, 230 had at some later stage also been discharged with an RA diagnosis, but were analysed according to the first diagnosis made (JIA).

Through linkage with the Swedish MBR, we randomly selected four same-sexed controls for each unique case, born at the same delivery unit the same year as their case. Controls who died before the time of diagnosis of their case were excluded. The two sets of cases and controls thus consisted of 333 cases with RA and 1332 controls and the 3334 cases with JIA and their 13 336 controls.

Exposures

From the Swedish MBR and the Swedish inpatient register we extracted the following information for cases and controls: maternal age, maternal civil status, parity, feto–maternal blood group incompatibility, mode of delivery (vaginal vs Caesarean section), singleton versus multiple birth, date of birth, birth weight, gestational age, Apgar score at 5 minutes and malformations. Small for gestational age was defined as a birth weight for gestational age below −2 standard deviations (SD) according to the Swedish reference for fetal growth.26 Appropriate for gestational age was defined as birth weight for gestational age from −2 to +2 SD. Large for gestational age was defined as birth weight for gestational age above +2 SD. Information on maternal smoking has been collected in the MBR since 1983. Because of the large proportion of RA cases/controls who therefore had missing information, maternal smoking was only assessed as a risk factor for JIA.

Information on hospitalisation for neonatal and infant infections up to one year of age (below collectively referred to as infections during the first year of life), and maternal infections during pregnancy, was extracted from the Swedish MBR and the Swedish inpatient register using predefined discharge codes.

Statistics

The association between the above exposures and the risk of developing RA and JIA was estimated as univariate odds ratios (OR) with 95% CI using conditional logistic regression in SAS (PROC PHREG). Although considering the tight correlation between factors such as birth weight, gestational age and size for gestational age and the fact that these are rather to be regarded as markers of fetal growth than as actual biological risk factors themselves, we performed multivariate analyses. In these models we included maternal age, civil status, season of birth, number of siblings, mode of delivery, gestational age and birthweight in the model. For JIA we also performed the multivariate analysis with and without maternal smoking.

The study was approved by the Ethics Committee at Karolinska Institutet.

RESULTS

Birth characteristic, RA

Being small for gestational age (OR 0.5, 95% CI 0.3 to 1.0) and having a birth weight less than 3000 g (OR 0.7, 95% CI 0.5 to 1.0) was associated with a decreased risk of RA of borderline statistical significance. Preterm birth (⩽258 days) was associated with a statistically non-significantly decreased risk (OR 0.6, 95% CI 0.4 to 1.1). Non-significantly increased risks were observed for individuals born large for gestational age (OR 1.6, 95% CI 0.7 to 3.3) and for those having three or more older siblings (OR 1.4, 95% CI 0.8 to 2.4). Maternal age, maternal civil status, Apgar score at 5 minutes, multiple birth, malformations, Caesarean section, season of birth and maternal–child blood group incompatibility were all unrelated to the risk of later RA (0.9 < OR < 1.1 and table 2). Alternative categorisation of birth weight (as <2500 g, 2500–3999 g (reference), ⩾4000 g) did not markedly change the point estimates (data not shown) and re-categorising birth weight as less than 3000 g, 3000–4499 g (reference), 4500 g or greater only marginally altered our results (OR ⩾4500 g 1.2, 95% CI 0.6 to 2.4). Multivariate analysis only marginally modified the point estimates, which remained statistically non-significant (not shown, see supplementary table 1, available online only).

Birth characteristics, JIA

For JIA, an increased risk of borderline statistical significance was observed for individuals born after more than 42 gestational weeks (⩾294 days) (OR 1.2, 95% CI 1.0 to 1.34) and for individuals delivered by Caesarean section (OR 1.1, 95% CI 1.0 to 1.3). A borderline statistically significant reduced risk (OR 0.7, 95% CI 0.5 to 1.0) was observed for individuals with an Apgar score at 5 minutes of 6 or less. Neither maternal age nor civil status, birth weight, small/large for gestational age, multiple births, number of older siblings, malformations, maternal smoking, season of birth or maternal–child blood group incompatibility was associated with the risk of JIA (table 2). Multivariate analysis with and without maternal smoking hardly altered the results (data not shown, see supplementary table 2, available online only).

Infections, RA

Overall, hospitalisation for any infection during the first year of life was associated with a non-significantly increased risk of RA (OR 1.4, 95% CI 0.8 to 2.5). Exploratory analysis of the association between infection during the first year of life and the risk of RA at age 16 years or later displayed a stronger association for RF-negative RA (OR any infection 2.6, 95% CI 1.0 to 7.0) than for RF-positive RA (OR any infection 1.2, 95% CI 0.5 to 2.9), although these associations, as well as those restricted to specific types of infections, were based on small numbers (table 3). Including hospitalisation for any infection in the same multivariate model as above only marginally altered the point estimate (multivariate OR 1.51, 95% CI 0.86 to 2.68). Maternal infections during pregnancy were not associated with the risk of RA (data not shown).

Infections, JIA

Being hospitalised for any infection during the first year of life was associated with the risk of later JIA (OR 1.9, 95% CI 1.7 to 2.1). Statistically significantly increased risks were observed both for respiratory, gastrointestinal and for skin/soft tissue infections in the infant, see table 3. Including being hospitalised for any infection in the same multivariate model as above (including maternal smoking) only marginally altered the point estimate (multivariate OR 1.92, 95% CI 1.69 to 2.18).

Maternal infections during pregnancy were unrelated to the risk of JIA (data not shown).

DISCUSSION

The results of our study suggest that infections leading to hospitalisation during the first year of life act as risk factors for later RA as well as for JIA. With respect to maternal, pregnancy and birth characteristics, a trend of approximately 30% reduced risk of “early adult onset RA” was observed for preterm birth (⩽258 days), birth weight below 3000 g and small for gestational age birth, whereas having more than three older siblings was associated with a non-significant 40% increased risk of RA. For JIA, a gestational age of 294 days or greater emerged as a potential risk factor. Other birth characteristics were not related to the risk of RA or JIA.

Numerous epidemiological and animal studies during the past decades have suggested that the environment during fetal and infant life is of importance for the risk of the later development of several different types of diseases, presumably through several different mechanisms, including fetal “programming” or “imprinting” and epigenetic modifications.7 27 28 Studies on autoimmune diseases such as coeliac disease, inflammatory bowel disease, autoimmune thyroid disease and type 1 diabetes mellitus indicate a potential association between neonatal infections and later disease risk.13 16 29 30 In RA, infections have so far been proposed as acute triggers preceding the clinical onset of disease. Infections may, however, also act more subtly by interfering with the normal maturation of the immune system during its early development. As observed in allergies/asthma, infections and “the microflora” to which the infant is exposed may affect the immune system in different ways, eg, reducing the risk of allergy and increasing the risk of asthma following viral and respiratory infections before the age of 2 years.9 In a study of RF positivity in adult life a reduced risk was observed among women sharing a bedroom during childhood, which was used as a proxy for infectious exposure.19 In our current study we measured infectious exposure as being hospitalised for infection, thus measuring exposure to more severe infections. We noted that the overall modest association between being hospitalised for infection during the first year of life and the risk of RA 15 or more years later was more pronounced in the RF-negative subset than in the RF-positive subset of RA. This observation is particularly important in light of recent observations indicating marked differences in the profile of genetic susceptibility and gene–environment interactions between CCP-positive and negative RA (CCP and RF are highly correlated).31^–33 Recent data linking the interferon system with anti-CCP-negative RA may support a particular role for infection and the risk of RA in this subset.34 The exact nature (common genetic susceptibility, causality, confounding) of the association between early life infection and the later risk of (RF-negative) RA is yet unknown and sample size put a limit to our ability to detect risks associated with specific types of infections (table 3). The non-significant association between the number of older siblings and the risk of RA might reflect exposure to infections. By contrast, we noted no variation in the risk of RA with the season of birth. Exposure to infections could, however, also be a marker for some factors associated with poorer socioeconomic background, a factor that is known to be asscoiated with RA.35 For JIA, we observed increased risks for respiratory tract infection, gastrointestinal infection as well as for skin/soft tissue infections. JIA comprise a heterogeneous group of arthritides but the major part is distinct from RA in the adult and approximately 10% have a late onset oligoarthritis resembling reactive arthritis. Viral agents such as parvovirus B19, rubella, influenza A, coxsackie, Epstein–Barr virus, cytomegalovirus, as well as bacterial agents have been linked to JIA,36^–39 but the role of infection in the pathogensis is unclear and probably varies between the different infectious agents and forms of JIA.40 41 Clearly, the results of these first exploratory analyses on infections call for confirmation and more detailed assessments.

Disturbances of intrauterine growth have been linked to autoimmune diseases such as type 1 diabetes mellitus, coeliac disease and asthma/atopy.30 42^–45 In a previous small case–control study by some of us, we found an association between high birth weight (>4000 g) and the risk of developing RA.17 A similar association with high (>4500 g) birth weight has also been reported among US nurses.18 By contrast, no significant association between early growth including birth weight and the presence of RF was found in a UK cohort study.19 The results of our current study do not support a link between high birth weight and the risk of early adult onset RA. By contrast, our current results indicate a reduced risk associated with factors related to fetal growth restriction and short gestational age. There are several tentative explanations for the inconsistent results between studies. In our previous study, the sample size was small and the statistical precision was limited.17 There may be a difference in the accuracy of birth weight (self-reported birth weight in the study by Mandl et al,18 prospectively midwife registered in our two studies).17 The three studies also span over different birth cohorts of patients with RA; patients born between 1921 and 1946 in the study by Mandl et al,18 1940–1960 in our previous study and 1973 or later in our current study.17 This might reflect a decreasing importance of birth weight or birth weight-related factors in the pathogenesis of RA. For JIA, a recent Finnish study observed no significant overall association with birth weight, although non-significantly increased OR were observed both for low and for high birth weight.20 In our study, we observed an increased risk of JIA following gestational age of 294 days or greater, but no association with high or low birth weight or birth weight for gestational age.

Our study has several strengths and limitations. It is a large population-based study with prospectively recorded data, which minimises the risk of recall and information bias. As the MBR started in 1973 the study population is comparatively young, which may be of importance for the generalisability of our results. The RA cases are limited to “early adult onset RA” and thus our study does not address whether the trends of associations with fetal growth and early life infections are also true for later onset RA. The validity of the diagnosis of RA in the Swedish inpatient register has been shown to be high,23 and the EAR includes rheumatologist-diagnosed RA classified according to the ACR criteria, although chart reviews shows a 10% misclassification.23 RF status was partly determined by data from the EAR and partly by ICD codes, of which the latter may be less precise and sensitive. For JIA, we only included cases hospitalised with a hospital discharge listing JIA, which is why a selection of more severe cases cannot be excluded. Furthermore, JIA comprises a heterogeneous group of patients of which only a smaller part have an RA-like condition. Unfortunately, the ICD classifications used to identify JIA during the study period did not allow further phenotypic categorisation. As the median age of diagnosis of JIA was low we cannot exclude that one explanation for the observed association with infection is the occurrence of some form of reactive or infectious arthritis in the very youngest cases, although postinfectious arthritis is usually most frequent in older children.46

In conclusion, our results indicate that infections during the first year of life and possibly also factors related to size at and timing of birth may be of aetiological importance in the pathogenesis of RA and JIA.