OBJECTIVES. Sweden experienced a unique epidemic of celiac disease in children <2 years of age. The epidemic was partly explained by changes in infant feeding over time and indicated a multifactorial pathogenesis. The main aim of this study was to analyze celiac disease risk in epidemic and postepidemic birth cohorts up to preschool age, to explore further the opportunity for primary prevention.
METHODS. A population-based incidence register of celiac disease in children covering the entire nation from 1998 to 2003 and part of the country back to 1973 was analyzed. European Society for Pediatric Gastroenterology, Hepatology, and Nutrition diagnostic criteria for celiac disease were used. The annual incidence rate for each age group and the cumulative incidence according to age for each birth cohort were calculated.
RESULTS. A considerable difference in cumulative incidences of celiac disease at comparable ages was demonstrated between birth cohorts from the epidemic and postepidemic periods. The difference persisted during the preschool years, although it decreased somewhat with age. During the last years of the follow-up period, there was again a successive increase in incidence rate among children <2 years of age.
CONCLUSIONS. The difference in celiac disease risk between birth cohorts at comparable ages suggests an opportunity for primary prevention. This highlights the importance of further exploring the role of infant feeding and exogenous factors besides dietary gluten that might initiate or prevent disease development. Moreover, on the basis of postepidemic incidence trends, we speculate that the Swedish epidemic might not have been as unique as thought previously, although its magnitude was striking.
Celiac disease (CD) is the most common food-related chronic disease in children.1 The pathogenesis is multifactorial, with both genetic and environmental factors contributing, and the disease has autoimmune features. In subjects with CD, consumption of wheat gluten and related proteins in rye and barley causes a small-intestinal inflammatory condition with increased numbers of intraepithelial lymphocytes (IELs), crypt hyperplasia, and villous atrophy, which results in impaired nutrient absorption. The treatment of CD includes life-long exclusion of all foods containing wheat, rye, and barley. Adherence to a gluten-free diet results in relief of symptoms and restored mucosal morphologic features, but the condition may cause a variety of health problems if it is not treated.2
In recent decades, the view on CD has changed from it being regarded as a rare childhood disease to it being considered a worldwide public health problem encompassing all age groups. Screening studies suggested that CD affects ∼1% of most populations,3–8 but the majority of affected individuals are unaware of their disease.
Unique to Sweden, the incidence of CD in children <2 years of age showed an epidemic pattern during the period of 1984–1996.9 The incidence rates reached levels higher than ever reported for any population. At that time, CD was considered unavoidable when dietary gluten was given to genetically susceptible individuals. Therefore, both the abrupt fourfold increase and the subsequent rapid decrease in the relatively genetically stable Swedish population were unexpected. Most cases involved in the rapid increase in incidence demonstrated obvious symptoms suggesting CD; therefore, improved case ascertainment was not a sufficient explanation.10 The subsequent rapid decrease in incidence occurred although awareness of CD and possible vague symptoms had increased considerably during the epidemic, among both health care personnel and the public.
The abrupt increase in the incidence rate of CD in the middle 1980s resulted in a decision by the Swedish Pediatric Association to support initiation of a CD epidemiological surveillance system. The CD register initiated was later used as the basis for an incident case reference study to explore factors underlying the epidemic.11 The main finding in that study was the decreased CD risk for infants who were introduced to gluten in small to medium amounts while they were still being breastfed. By calculating the attributable population fraction for these exposures, we concluded that 45% of the cases could have been prevented with favorable infant feeding practices.
The start of the epidemic coincided with new national feeding recommendations for infants, postponing the introduction of gluten-containing foods from 4 to 6 months of age. The new recommendations suggested indirectly that a smaller proportion of children were being breastfed at the introduction of gluten into the diet (eg, 68% were breastfed at 4 months of age, compared with 48% at 6 months of age, among children born in 1986).12 At the same time, there was a twofold increase in the average consumption of gluten from commercially available milk-cereal drinks by children <2 years of age.9 This was mainly attributable to a decrease of milk and an increase of cereals in these products, with the purpose of decreasing the total amount of protein. Milk-cereal drinks provided approximately one half of the total intake of cereal protein for infants during that time period.13
To reduce the risk of CD at the population level, national recommendations on when and how to introduce gluten were changed again in 1996, which coincided with the rapid decrease in CD incidence.9 The introduction of gluten-containing foods in gradually increasing amounts, preferably while the infant was still being breastfed, at 4 to 6 months of age was recommended. Contemporaneously, the gluten content of commercially available milk-cereal drinks was reduced, which implied a decrease in the average consumption of gluten by one third.9
The Swedish epidemic indicates a multifactorial etiology of CD. Environmental factors besides infant feeding also have been suggested as possible risk factors for CD in children. Genetic predisposition and environmental exposures, as well as interactions between them, may together shape the immunologic response to gluten.14 Repeated infectious episodes early in life have been suggested as a possible risk factor, although this has not yet been fully explored.14–16
The present study is the first follow-up study of the Swedish CD epidemic, encompassing the postepidemic period. Trends in the occurrence of CD in childhood were analyzed to shed light on (1) whether birth cohorts of the epidemic period, compared with later cohorts, continue to carry an excess risk for CD after 2 years of age, possibly indicating an opportunity for primary prevention; (2) whether the incidence in children <2 years of age has stabilized on a lower level after the epidemic period; and (3) whether there is an ongoing shift toward older age at diagnosis.
Identification of New Cases
In Sweden, there is a well-established practice of referring children with suspected CD to a pediatric clinic for a diagnostic small-intestinal biopsy.17 In 2002, this practice was formulated in national guidelines for the diagnosis and treatment of childhood CD.
Since 1973, new clinically detected cases in Swedish children <15 years of age have been recorded in a register,9 retrospectively from 1973 to 1990 and prospectively from 1991 onward, covering 15% and 40% of the population, respectively. Since 1998, the register has had nationwide coverage, including 47 pediatric clinics. For case identification, the diagnostic criteria for CD according to the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition (ESPGHAN)18 had to be met, that is, an initial biopsy showing structurally abnormal intestinal mucosa with a gluten-containing diet.
From 1973 to 1997, 2151 children fulfilled the inclusion criteria.9 During the follow-up period from 1998 to 2003, when the register had reached complete nationwide coverage, 2922 children fulfilled the criteria and 295 were excluded (Fig 1). Excluded children were distributed equally among reporting clinics.
Numbers of births per year and mid-year populations for each single-year age group (0–15 years) were obtained from Statistics Sweden (www.scb.SE) for the period from 1973 to 2003. The study base from 1973 to 1997 encompassed 598 262 births and 8 909 501 person-years, and the follow-up period from 1998 to 2003 involved 454 923 births and 9 777 930 person-years.
Children were grouped according to age, that is, 0 to 1.9 years, 2 to 4.9 years, or 5 to 14.9 years, enabling direct comparisons with the epidemic period, when most cases were diagnosed before 2 years of age.9 The annual incidence rate for each age group was calculated by dividing the number of new cases by the number of person-years in the follow-up period, approximated by the mid-year population. The cumulative incidence for each birth cohort at a certain age was calculated by dividing the number of cases diagnosed up to that age by the total number of births in the cohort. Linear regression analysis was used to test for any difference in the increases in cumulative incidence according to age for different birth cohorts. Rate ratios (RRs) with 95% confidence intervals (CIs) were estimated to compare differences in incidence rates. Statistical significance was defined as a P value of <.05 and a RR with the 95% CI excluding 1.0. Excel 2003 (Microsoft, Redmond, WA) and SPSS 15.0 (SPSS, Chicago, IL) were used for basic calculations.
The prospective incidence register has been approved by the research ethics committees of all Swedish medical faculties.
Annual Incidence Rate
In the total childhood population (0–15 years of age), the incidence rate was fairly stable at ∼10 cases per 100 000 person-years between 1973 and 1984. In 1985, at the beginning of the epidemic, the rate started to increase; in 1994, at the peak of the epidemic, it reached 45 cases per 100 000 person-years. At the end of the epidemic, the incidence rate decreased. In 1997, when the epidemic had completely abated, the rate was 16 cases per 100 000 person-years. The rate started to increase again in subsequent years, and the rate reached 44 cases per 100 000 person-years in 2003, the last year of the present follow-up period.
For children <2 years of age, the incidence rate was quite stable from 1973 to 1984, with an average of 65 cases per 100 000 person-years (Fig 2). 9 During the epidemic, the rate reached levels of ∼200 cases per 100 000 person-years, with a maximum of 241 cases per 100 000 person-years in 1994. From 1997, the incidence rate was fairly stable at ∼50 cases per 100 000 person-years.9 In 2001, however, it started to increase again, and a few years later it had almost doubled to 98 cases per 100 000 person-years, an increase that was statistically significant (RR: 2.00; 95% CI: 1.55–2.57). Still, this incidence rate was far from that of the epidemic period.
For children 2 to 4.9 years of age, the incidence rate increased slowly from the middle 1970s to 1996, that is, from no case reported in 1973 to an incidence rate of 33 cases per 100 000 person-years in 1996 (Fig 2).9 This trend broke temporarily in 1997, when the rate decreased to less than one half, compared with the previous year.9 However, the increase then continued during the following years, and the incidence rate almost tripled from 1998 to 2003, from 19 to 52 cases per 100 000 person-years.
In the oldest age group, 5 to 14.9 years, CD was rarely diagnosed in the 1970s (Fig 2).9 A slow increase in incidence rate was seen from the middle 1980s to the middle 1990s. Coinciding with the end of the epidemic among the youngest children, the increase accelerated. From 1998 to 2003, the incidence rate more than doubled, from 14 to 31 cases per 100 000 person-years.
Cumulative Incidence According to Age
The cumulative incidence at 2 years of age was almost 3 times higher during the epidemic, compared with the years before and after the epidemic (Fig 3). For the birth cohorts from 1973 to 1995, the cumulative incidence according to age increased most rapidly up to 2 years of age. In contrast, the cohorts from 1996 to 2002, representing children born after the epidemic, had a gradual increase in cumulative incidence during the preschool years, with no sharp increase during the first 2 years of life.
In contrast to the cohorts born before the epidemic period, for which the cumulative incidence generally ceased to increase after 2 years of age, the cohorts born during and after the epidemic period had a gradual increase also after 2 years of age (Fig 3). For the 1993 cohort, born during the peak of the epidemic, the incidence reached 4.4 cases per 1000 births at 2 years, 5.4 cases per 1000 births at 6 years, and 6.6 cases per 1000 births at 10 years of age. The incidence for the 1997 cohort, the first cohort born after the epidemic had completely abated, reached 1.3 cases per 1000 births at 2 years, increasing to 2.9 cases per 1000 births at 6 years of age, which at this point represents the longest possible follow-up period. The increase in cumulative incidence between 2 and 6 years of age was steeper in cohorts born after the epidemic period, compared with cohorts born during the epidemic period. In a comparison of the 1993 and 1997 cohorts, this difference was significant (P = .008).
Age at Diagnosis and Gender Distribution
The median age at diagnosis of CD in Swedish children has increased since the early 1970s. Between 1973 and 1994, the median age at diagnosis was 1.2 years (interquartile range: 1.0–1.7 years); in 1997, it had increased to 3.7 years (interquartile range: 1.3–8.9 years).9 The median age at diagnosis during the period from 1998 to 2003 was 5.8 years (interquartile range: 1.9–9.8 years). Girls accounted for 62% of all cases between 1998 and 2003, which is in accordance with the gender distribution for the period from 1973 to 1997.19
The potential effects on incidence rates of different population coverages and diagnostic criteria were evaluated. In a comparison of incidence rates based on 40% population coverage (ie, the catchment area in 1991–1997) and 100% coverage (for each year from 1998 to 2003), there was no significant difference for the age groups of 0 to 1.9 years and 2 to 4.9 years; for the 5- to 14.9-year age group, there was a difference only in 1999 (RR: 1.29; 95% CI: 1.02–1.63). In addition, there was no significant difference in incidence rates when case ascertainment was based on the ESPGHAN criteria for diagnosis18 or adapted criteria sometimes used in clinical practice today (ie, also including the 195 children with increased numbers of IELs but normal mucosa morphologic features with a gluten-containing diet).
There was a considerable difference in cumulative incidences at comparable ages between birth cohorts of the epidemic and postepidemic periods, which suggests an opportunity for primary prevention of CD. The difference was still striking at 6 years of age, although it decreased somewhat with age. Therefore, not all cases were permanently prevented, and the onset or diagnosis of CD was merely delayed for some individuals. Interestingly, a significant successive increase in incidence rates among children <2 years of age was once again revealed during the last years of follow-up monitoring, which might be a sign of a new epidemic approaching.
A strength of this study is that is it unique in basing an epidemiological study on a prospective register with nationwide coverage. However, the present study shows that nationwide coverage is not required for estimation of the incidence of CD in a homogeneous population such as that in Sweden. A prospective national register allows more-detailed analyses of variations in disease occurrence according to time, place, and person and serves as an excellent basis for in-depth studies.
Another strength of this study is that case ascertainment was based on the ESPGHAN criteria for diagnosis,18 including an initial small-intestinal biopsy showing structurally abnormal mucosa with a gluten-containing diet and not relying only on elevated levels of serologic markers indicating the disease. The CD diagnosis now is sometimes considered verified also in cases with increased numbers of IELs but normal mucosa morphologic features with serologic marker levels indicating the disease, although this is not universally accepted. However, inclusion of the children with increased numbers of IELs but normal mucosa morphologic features did not change the results significantly.
No policy for general CD screening of the population has yet been adopted in Sweden; consequently, the majority of the CD cases were clinically detected. In addition to liberal serologic testing of clinically suspected CD cases, the national guidelines regarding childhood CD emphasize the increased susceptibility in high-risk populations such as siblings of patients and children with type 1 diabetes mellitus or Down syndrome. However, CD screening of these high-risk children is not yet an established clinical practice. Also, such screening, when performed, is likely to be conducted at >2 years of age and does not explain the variation in incidence over time for the youngest children.
The decreased difference in cumulative incidences at comparable ages between the cohorts born during and after the epidemic could indicate that, with time, the cumulative incidence in the cohorts born after the epidemic might reach the same level as that in the epidemic cohorts. If so, then a favorable exposure with regard to infant feeding11 might only delay the diagnosis because of milder symptoms and/or individuals at risk might contract the disease at older ages. This would support the view that, given genetic susceptibility, CD cannot be avoided in individuals exposed to gluten. A recent Italian study showed a pairwise concordance rate of 71% in monozygotic twins,20 which supports the strong genetic influence, although monozygotic twins commonly also share environmental exposures. Moreover, ∼25% to ∼30% of the population carries genes encoding HLA-DQ2 or HLA-DQ8 molecules, the most important known predisposing genetic factor for CD,21,22 and most individuals never develop CD. Furthermore, both innate and adaptive immunity seems to be involved in disease development.23,24 Therefore, genetic susceptibility is necessary but most likely not sufficient for CD development.25 Indeed, the difference in CD risk between birth cohorts at comparable ages persisted during the preschool years. A recent screening study in preschool-aged children in Sweden indicated a lower prevalence of screening-detected CD after the national change in infant feeding recommendations in 1996,26 which supports the potential role of environmental exposure in primary prevention of CD. However, the option of primary prevention and the potential impact of environmental factors need to be explored further.14
The annual incidence rate in the total childhood population reached the same level at the end of the follow-up period as during the epidemic years but with a shift toward older age at diagnosis. This shift was partly explained by the fact that children carrying excess risk for CD, that is, children originating from the cohorts born during the epidemic period, had grown older at the time of follow-up monitoring. Increased awareness of the disease, resulting in the diagnosis of CD in cases that would previously have been undiagnosed because of the milder clinical symptoms typical for older children, also might have contributed. Less-obvious symptoms at diagnosis also have been implicated for younger children,27 which might account for delayed diagnosis among the youngest children and thus might contribute to the shift toward older age at diagnosis.
Only longer follow-up monitoring will reveal whether the increase in incidence rates in children <2 years of age seen between 2000 and 2003 is an ongoing trend. Only one half of the epidemic was explained by changes in infant feeding.11 Consequently, other, still-unknown, environmental exposures contributed to the change in disease occurrence. Infections have been suggested as a possible risk factor, at least for initiation of CD in young children.14,15 We recently discovered with scanning electron microscopy that rod-shaped bacteria were associated with the small-intestinal mucosa of patients with both active and inactive disease but not control subjects.24 Furthermore, one recent prospective study of genetically susceptible children suggested that frequent rotavirus infections increased the risk of CD autoimmunity.16 Therefore, it is possible that the epidemic was caused in part by changes in the infectious panorama or an interaction of infant feeding and infections.14
In addition to the Swedish epidemic in young children, it is noteworthy that fluctuations in CD frequency over time have been reported for other European countries.28–31 The magnitude of the Swedish epidemic was striking, however, compared with the other fluctuations. Now Sweden might be facing a new epidemic of CD, inasmuch as the incidence rate for the youngest age group increased during the last years studied. If so, then the epidemic or the “Swedish case” was not as unique as we assumed previously.
The difference in CD risk between birth cohorts at comparable ages suggests an opportunity for primary prevention. This highlights the importance of further exploring the role of infant feeding and other exogenous factors, besides dietary gluten, that might initiate or prevent disease development. Moreover, on the basis of postepidemic incidence trends, we speculate that the Swedish epidemic might not have been as unique as previously thought, although its magnitude was striking.
This work was undertaken within the Centre for Global Health at Umeå University, with support from FAS, the Swedish Council for Working Life and Social Research (grant 2006-1512).
We thank Susanne Walther at the Department of Public Health and Clinical Medicine, Division of Epidemiology and Public Health Sciences, Umeå University, for her skillful administrative work with the register; Phil Lyon for his valuable comments; all of the collaborators at pediatric clinics in Sweden for reporting incident cases; the Swedish Pediatric Association Working Group on Celiac Disease for support; and the Country Council of Västerbotten for financially enabling the register.
- Accepted December 19, 2007.
- Address correspondence to Cecilia Olsson, RD, Department of Food and Nutrition, Umeå University, SE-901 87 Umeå, Sweden. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Celiac disease is today a global public health problem. Sweden experienced a unique epidemic of celiac disease in children <2 years of age, which indicated a multifactorial pathogenesis, including the impact of infant feeding.
What This Study Adds
A considerable difference in celiac disease risk at comparable ages was demonstrated for birth cohorts of the epidemic and postepidemic periods, which suggests an opportunity for primary prevention.
- ↵Carlsson AK, Axelsson IE, Borulf SK, et al. Serological screening for celiac disease in healthy 2.5-year-old children in Sweden. Pediatrics.2001;107 (1):42– 45
- West J, Logan RF, Hill PG, et al. Seroprevalence, correlates, and characteristics of undetected coeliac disease in England. Gut.2003;52 (7):960– 965
- ↵Ascher H, Krantz I, Kristiansson B. Increasing incidence of coeliac disease in Sweden. Arch Dis Child.1991;66 (5):608– 611
- ↵Ivarsson A, Hernell O, Stenlund H, et al. Breast-feeding protects against celiac disease. Am J Clin Nutr.2002;75 (5):914– 921
- ↵National Board of Health and Welfare. Breast-feeding, children born 2003 [in Swedish]. Stat Health Dis.2005;5 . Available at: www.socialstyrelsen.se/statistik/statistik_omne/barn_unga/index.htm. Accessed July 7, 2008
- ↵Ascher H, Holm K, Kristiansson B, et al. Different features of coeliac disease in two neighbouring countries. Arch Dis Child.1993;69 (3):375– 380
- ↵Walker-Smith JA, Guandalini S, Schmitz J, et al. Revised criteria for diagnosis of coeliac disease: report of Working Group of European Society of Paediatric Gastroenterology and Nutrition. Arch Dis Child.1990;65 (8):909– 911
- ↵Nisticò L, Fagnani C, Coto I, et al. Concordance, disease progression, and heritability of coeliac disease in Italian twins. Gut.2006;55 (6):803– 808
- Challacombe DN, Mecrow IK, Elliott K, et al. Changing infant feeding practices and declining incidence of coeliac disease in West Somerset. Arch Dis Child.1997;77 (3):206– 209
- Copyright © 2008 by the American Academy of Pediatrics