In this population-based incident case-referent study we found an increased celiac disease risk in infants with three or more parental-reported early infectious episodes (before 6 months of age), regardless of the type of infection. In addition, we found a synergistic effect between early infections and daily amount of gluten intake, which was more pronounced in infants for whom breastfeeding had been discontinued prior to gluten introduction.
In contrast to previous studies, we found no significant association between having siblings and celiac disease [13, 14]. Unfortunately, we had no data on day care attendance, which could have confounded our findings. We found an increased risk for celiac disease in children from lower socioeconomic strata compared to medium-high strata. The risk decreased when adjusting for early infections and infant feeding, but it remained statistically significant, which implies that part, but not all of the risk can be attributed to these factors. Our findings are in agreement with previous observations in Swedish boys and individuals in the UK [15, 28]. On the contrary, however, a recent large Swedish register-based study found a reverse association; a lower celiac disease risk in the lowest strata . Similarly, Kondrashova et al. reported lower celiac disease prevalence in a Russian area than in a bordering area of Finland, suggesting that the lower socioeconomic status and/or inferior hygienic standard in Russia conveyed a protective effect against celiac disease . Socioeconomic status is related to several underlying exposures that most likely differ both between countries and over time and thus further studies are needed to elucidate the role of these exposures and their combined effect on celiac disease risk.
The seasonal difference in celiac disease diagnosis [11, 12] suggested that weaning during winter, when the infectious load is heavier, conveyed an increased celiac disease risk. Welander et al. investigated in a Swedish prospective cohort whether having an infection at the time of gluten introduction affected the risk for later celiac disease, but could not find any significant associations . This could, however, be due to the fact that their study had the power to detect a 1.9-fold increase in celiac disease risk, and our findings suggest that the risk is lower (1.5-fold). As noted above, frequent rotavirus infections have been suggested to increase celiac disease risk  and rotavirus is a major cause of gastroenteritis in Swedish children . However, in Sweden gastroenteritis is still relatively uncommon early in life . In our study we could not find a significantly increased celiac disease risk due to parental-reported gastroenteritis, although this might be due to too few infants with this infection (Table 2). The finding of an increased celiac disease risk in infants with three or more non-gastrointestinal early infections suggests that infections, irrespective of localization, affect celiac disease risk, although the molecular mechanisms behind this finding remain to be clarified.
On the other hand, our finding that the celiac disease cases experienced more infections early in life could be related to an inherited genetic susceptibility for both celiac disease and infections. Previous studies have reported a moderately increased risk for a number of infections in celiac disease cases, for example hospital admission for influenza, although these findings could also be related to nutritional deficiency, increased mucosal permeability or other factors . It is, however, reasonable to assume that an increased risk could be related to even earlier exposures than the ones investigated in the current study. Neonatal infections have been associated with increased celiac disease risk [13, 32], although this observation was not replicated in another study . Mårild et al. found an increased risk for celiac disease following elective, but not emergency, caesarean section . The former affects the gut colonization and subsequently the microbiota in the infant, which could affect the risk for celiac disease, as differences between celiac disease cases and non-cases in microbiota composition have been shown [33–36].
The microbiota interacts with genetic and environmental factors such as infant feeding, through a complex immunological process resulting in establishment of oral tolerance to gluten, or failure to do so resulting in celiac disease . Infections early in life, when the gut microbiota is developing, could possibly result in a different microbiota, affecting both intestinal immune responses and mucosal barrier function and thereby also the risk for celiac disease . Further, antibiotic treatment affects the gut microbiota and might therefore be a risk factor for celiac disease . We speculated that the increase in celiac disease occurrence during the past decades could be related to the increased use of antibiotics during the same time period. However, in our post hoc approach to antibiotic treatment and celiac disease we found no significant association, although this could be due to lack of data on type of antibiotics used. Similarly to antibiotics and early infections, early vaccinations also influence the immune system. We have previously shown that BCG vaccination is associated with a protective effect for celiac disease, although this has to be interpreted with caution . Adjusting the association between early infections and celiac disease for BCG vaccination status did not affect the association (data not shown). Other vaccinations within the Swedish national program were not risk factors .
Concurrent with the beginning and end of the Swedish celiac disease epidemic, infant feeding changed . As infant feeding affects the microbiota [37, 38], these changes could have resulted in the increased frequency of rod-shaped bacteria seen in jejunal biopsies from both untreated and treated celiac disease cases born during the epidemic, as compared to celiac disease cases born afterwards and healthy controls [41, 42]. Alternatively, the rod-shaped bacteria with celiac disease promoting characteristics constitute an independent phenomenon and part of the celiac disease epidemic may be attributed to them. Furthermore, although there are no available ecological data in Sweden to explore whether the infectious panorama changed during the epidemic period, the synergistic effect between early infections and gluten amount, to which approximately 6% of the cases could be attributed, implies that early infections reinforced the changes in gluten amounts consumed . Thereby early infections probably had a minor contribution to the Swedish celiac disease epidemic, irrespective of changes in the infectious panorama.
A limitation of our study is that it was retrospective and based on parental reports in a questionnaire. Consequently, there was an inherent risk for recall bias, even if the interest in celiac disease was not indicated in the questionnaire and incident recruitment of both cases and referents was used. The infant feeding data has previously been evaluated using a “multiple imputation procedure” without effect on the results . Our study was limited to symptomatic infectious episodes. Difficulties in correctly identifying infectious episodes in young infants were the same for parents of both cases and referents, although the lack of data on specific pathogen and severity of the infection still constitutes a limitation.