Infant Growth and Risk of Childhood-Onset Type 1 Diabetes in Children From 2 Scandinavian Birth Cohorts

Importance  Type 1 diabetes mellitus is one of the most common chronic diseases with onset in childhood, but environmental risk factors have not been convincingly established.

Objective  To test whether increased growth during the first year of life is associated with higher risk of childhood-onset type 1 diabetes.

Design, Setting, and Participants  This is a cohort study using information from 2 population-based cohort studies in Norway and Denmark, the Norwegian Mother and Child Cohort Study (MoBa) and the Danish National Birth Cohort (DNBC), of children born between February 1998 and July 2009. The current study was conducted between November 2014 and June 2015.

Exposures  Change in weight and length from birth to age 12 months.

Main Outcomes and Measures  Unadjusted and adjusted hazard ratios (HRs) of type 1 diabetes, classified based on nationwide childhood diabetes registers, obtained using Cox proportional hazards regression.

Results  A total of 99 832 children were included in the study, with 59 221 in MoBa (51.2% boys and 48.8% girls; mean age at end of follow-up, 8.6 years [range, 4.6-14.2 years]) and 40 611 in DNBC (50.6% boys and 49.4% girls; mean age at end of follow-up, 13.0 years [range, 10.4-15.7 years]). The incidence rate of type 1 diabetes from age 12 months to the end of follow-up was 25 cases per 100 000 person-years in DNBC and 31 cases per 100 000 person-years in MoBa. The change in weight from birth to 12 months was positively associated with type 1 diabetes (pooled unadjusted HR = 1.24 per 1-SD increase; 95% CI, 1.11-1.39; pooled adjusted HR = 1.24 per 1-SD increase; 95% CI, 1.09-1.41). There was no significant association between length increase from birth to 12 months and type 1 diabetes (pooled unadjusted HR = 1.06 per 1-SD increase; 95% CI, 0.93-1.22; pooled adjusted HR = 1.06 per 1-SD increase; 95% CI, 0.86-1.32). The associations were similar in both sexes.

Conclusions and Relevance  This is the first prospective population-based study, to our knowledge, providing evidence that weight increase during the first year of life is positively associated with type 1 diabetes. This supports the early environmental origins of type 1 diabetes.

Type 1 diabetes mellitus is among the most common chronic diseases with onset in childhood, and the Nordic countries have the highest disease burden.^1-3 Despite the well-known role of genetic susceptibility, a changing incidence during the past decades suggests a role for environmental factors in development of type 1 diabetes.¹ While no single environmental factor has been established as a risk factor for type 1 diabetes, it is proposed that environmental factors might operate early in life.⁴

In 1975, Baum et al⁵ reported that children with type 1 diabetes had a higher weight at age 12 months compared with a group of children without type 1 diabetes. Since then, a few other studies have reported similar findings, but these studies varied in the age at anthropometric measurements and were retrospective case-control studies.^6-10

Prospective cohort studies are less susceptible to selection and information bias compared with case-control studies. A few cohort studies of genetically susceptible children have reported associations between growth and islet autoimmunity.^11-13 However, associations identified in high-risk populations are not necessarily transferable to the general population. The previous studies are briefly summarized in eTable 1 in the Supplement.

We aimed to test whether increased growth during the first year of life predicts higher risk of childhood-onset type 1 diabetes in 2 Scandinavian birth cohorts, the Norwegian Mother and Child Cohort Study (MoBa) and the Danish National Birth Cohort (DNBC), which are 2 of the largest birth cohorts in the world.

Box Section Ref ID

The MoBa is a population-based birth cohort administered by the Norwegian Institute of Public Health.^14,15 This cohort study recruited pregnant women across Norway between June 1999 and December 2008, at approximately 18 gestational weeks, where 40% of eligible women participated. We used data available in March 2014, comprising 95 267 mothers and 114 761 children, and information from questionnaires administered at 18 gestational weeks and when the child was aged 6, 18, and 36 months. The Norwegian Data Inspectorate and the Regional Ethics Committee for Medical Research of South East Norway approved this study. All participants provided written informed consent.

The DNBC is a population-based birth cohort administered by the Statens Serum Institut in Denmark.^16,17 Pregnant women were recruited at their first antenatal visit across Denmark between January 1996 and October 2002, and 60% of eligible women participated. The cohort consists of 91 745 mothers and 103 118 children. We used information gathered through telephone interviews at 18 and 30 gestational weeks and when the child was aged 6 and 18 months. Data collection in the DNBC is approved by the Danish National Ethics Board. All participants provided written informed consent.

In both cohorts, additional information was obtained from national birth registers and disease registers. The current study was conducted between November 2014 and June 2015.

In MoBa and DNBC, birth weight and length were collected from the medical birth registers. Through the MoBa questionnaires, the mothers reported anthropometric measurements from their child’s health records. These were recorded at 6 weeks and 3, 6, 8, 12, 15 through 18, 18, 24, and 36 months. In DNBC, the mothers reported the child’s weight and length from their child’s health records at 5 and 12 months.

The a priori–specified primary exposures were absolute change in weight and length from birth to 12 months. Secondary exposures included change in weight and length from birth to age 5 or 6 months, change in weight and length from ages 5 or 6 months to 12 months, and absolute body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) at age 5 or 6 months and age 12 months. Because additional anthropometric measurements were available in MoBa, we explored absolute change in weight and length from 12 to 24 months, absolute change in weight and length from 24 to 36 months, peak weight and length velocity by 36 months, and age at peak BMI (eAppendix in the Supplement).

The outcome was clinical diagnosis of type 1 diabetes, defined as the first day of insulin treatment in accordance with the EURODIAB criteria.³ This information was obtained from the Norwegian Childhood Diabetes Registry² and the Danish Childhood Diabetes Registry.¹⁸ For the MoBa participants, additional linkage to the Norwegian Patient Register was done. Details of the registers are described in the eAppendix in the Supplement.

Characteristics that might influence both postnatal growth and risk of type 1 diabetes were included as potential confounders. These included the child’s sex, birth weight and length, breastfeeding duration, maternal age, maternal parity, maternal education, maternal smoking during pregnancy, maternal height, maternal prepregnancy BMI, paternal height, paternal BMI, and maternal diabetes (maternal type 1 diabetes in MoBa and all types of diabetes in DNBC). Data on paternal type 1 diabetes were available in MoBa.

We used Cox proportional hazards regression to examine the associations of absolute change in weight and length from birth to age 12 months with diagnosis of type 1 diabetes, reporting hazard ratios (HRs) and 95% confidence intervals. Participants were followed up from the day the child turned 1 year of age until diagnosed as having type 1 diabetes or the end of follow-up (April 2, 2014, for MoBa and October 31, 2013, for DNBC). We tested for nonlinearity of the associations by evaluating second-order terms. Because we found no indication of nonlinear associations, the exposures were standardized to examine the association per 1-SD increase. The proportional hazards assumption was evaluated by examination of Schoenfeld residuals, which indicated no deviations from this model assumption. Because some participants were siblings, we used robust cluster variance estimation. We adjusted the multivariable models for all of the covariates selected a priori based on a plausible association with both infant growth and development of type 1 diabetes. Results from the 2 cohorts were combined using a random-effects model by the inverse variance method implemented in the metan function in Stata (StataCorp LP). Heterogeneity of associations between the 2 cohorts was examined using the I² statistic.

We explored a number of sensitivity analyses. The main results presented are from complete case analyses. Approximately 10% to 15% of the observations had missing covariate information in the multivariable analyses. Missing covariate information was therefore imputed using the multivariate normal distribution in DNBC and chained equations in MoBa, generating a total of 20 imputed data sets. Second, we conducted sensitivity analyses excluding children of mothers with diabetes, children of mothers who reported smoking during pregnancy, children with celiac disease, and children who were preterm (born at <37 gestational weeks).

In the secondary analysis of growth between 12 and 24 months, follow-up started when the child was aged 24 months. For growth between ages 24 and 36 months, peak weight and length velocity by age 36 months, and age at peak BMI by age 36 months, follow-up started when the child was aged 36 months.

A 95% confidence interval for the hazard ratio excluding 1.00 was considered statistically significant. The analysis was conducted using SAS version 9.4 (SAS Institute, Inc) and Stata version 13 (StataCorp LP) statistical software.

For the primary analysis of growth from birth to age 12 months, data were available for 99 832 children (Figure). Maternal education tended to be higher, smoking less common, and parity lower among mothers of children with sufficient follow-up information to be included in the primary analysis (eTable 2 in the Supplement). However, the risk of type 1 diabetes was similar (eFigure 1 in the Supplement).

The children in the analysis were born between February 1998 and July 2009 (Table 1). The mean age at the end of follow-up in the study sample was 8.6 years (range, 4.6-14.2 years) in MoBa and 13.0 years (range, 10.4-15.7 years) in DNBC. The children diagnosed as having type 1 diabetes during the first year of life were excluded from the primary analyses (5 children in MoBa and 1 child in DNBC). The incidence rate of type 1 diabetes from age 12 months to the end of follow-up was 25 cases per 100 000 person-years in DNBC and 31 cases per 100 000 person-years in MoBa.

The mean (SD) change in weight from birth to age 12 months was just over 6 (1) kg (Table 2). The change in weight from birth to 12 months was positively associated with type 1 diabetes (pooled unadjusted hazard ratio [HR] = 1.24 per 1-SD increase; 95% CI, 1.11-1.39; pooled adjusted HR = 1.24 per 1-SD increase; 95% CI, 1.09-1.41), with no evidence of heterogeneity between the cohorts (Table 3). The multiple imputation analysis yielded results similar to those of the complete case analysis (Table 3). While there was no significant association between birth weight and type 1 diabetes, the absolute weight at age 12 months also significantly predicted type 1 diabetes with a similar adjusted HR of 1.26 per 1-SD increase (95% CI, 1.10-1.44).

The mean (SD) change in length from birth to age 12 months was about 25 (3) cm (Table 2). There was no significant association between length increase the first year of life and type 1 diabetes (pooled unadjusted HR = 1.06 per 1-SD increase; 95% CI, 0.93-1.22; pooled adjusted HR = 1.06 per 1-SD increase; 95% CI, 0.86-1.32) (Table 3).

There were no significant differences in the associations by sex (eFigure 2 in the Supplement). Similar results were found when excluding children who were preterm, children of mothers who smoked during pregnancy, children of mothers with diabetes, and children with celiac disease (data not shown). We explored the potential effect of the slight variation in exact age at anthropometric measurements by analyzing growth measures as age- and sex-specific z scores in the MoBa cohort, and this produced unchanged results (data not shown). Further adjustment for the child’s year of birth or for cesarean delivery also did not change the observed associations (data not shown). Finally, excluding children of parents with a native language other than Norwegian in the MoBa cohort did not change the results (data not shown).

Growth from birth to age 5 months (DNBC) or 6 months (MoBa) showed results similar to those for growth in the first 12 months of life, whereas growth during the second half year of life was not significantly associated with type 1 diabetes after adjustment for growth in the first half year of life (Table 4). The child’s BMI at both age 5 or 6 months and age 12 months was positively associated with type 1 diabetes (eFigure 3 in the Supplement).

Analyses in MoBa showed that after adjustment for previous growth (up to age 12 or 24 months), there was no significant association between increase in weight or length between ages 12 and 24 months with development of type 1 diabetes, while change in weight and length from ages 24 to 36 months showed a tendency for a positive association (eTable 3 in the Supplement). Peak weight velocity was also not significantly associated with type 1 diabetes. Peak length velocity (mean [SD], 62.9 [15.4] cm/y) occurred at an average of 7 days. Each 1-SD increase in peak length velocity was associated with an adjusted HR of 1.31 (95% CI, 1.08-1.60) (eTable 3 in the Supplement). There was a tendency for a decreased risk in development of type 1 diabetes with increasing age at peak BMI, but the results did not indicate a linear decrease in the risk of type 1 diabetes and were not statistically significant (eTable 3 in the Supplement).

In this analysis of 2 large Scandinavian birth cohorts, we found that weight increase in the first year of life was positively associated with subsequent risk of type 1 diabetes during childhood.

Strengths of this study include the large size, the prospective data collection, inclusion of 2 relatively homogeneous populations, adjustment for a number of potential confounding factors not included in previous studies, and linkage to nationwide registers with a high level of case ascertainment.^2,3 However, as in any observational study, confounding from unmeasured factors may be present. We also cannot exclude the possibility of a selection bias due to the initial participation rate in the cohorts or due to loss to follow-up. A number of associations were found to be similar among eligible pregnant women and participating women in both cohorts by using nationwide registers.^15,16 Furthermore, the observed incidence of type 1 diabetes among these cohorts was similar to that reported for the general population.^2,18 Using maternal report of the child’s anthropometric measurements might have resulted in misclassification, but the prospective design implies that such misclassification is unlikely to be differential and therefore most likely would attenuate the observed association. By asking the mother to refer to the child’s health records, we attempted to minimize misclassification.

The main result from our cohort study is in line with a few previous case-control studies.^5-10 A Danish study reported a nonsignificant 9% increase in the odds of type 1 diabetes per 1-kg increase in weight at age 12 months, which translates to an odds ratio of 1.22 per 1-SD increase.¹⁰ The mean difference in weight at age 11 months was reported to be 204 g between cases with type 1 diabetes and controls in a Finnish study,⁷ while a mean difference of 160 g at age 12 months was reported in a Swedish study.⁹ This is comparable to a mean difference of 240 g in MoBa and 270 g in DNBC at age 12 months. Notably, a larger mean difference of approximately 700 g at age 12 months was reported in the initial study by Baum et al.⁵

The lack of association with length increase in the first year of life in our study is in line with 1 previous case-control study.¹⁹ In contrast, several studies of length increase and type 1 diabetes indicated a positive association.^6,9,10,20-23 Two previous studies explored a sibling comparison, both reporting a positive association between length increase in the first 3 to 4 years of life and type 1 diabetes.^21,22 Our secondary results on length increase in the third year of life, albeit not statistically significant, suggested a positive association between length increase in the third year of life and risk of type 1 diabetes. However, the analysis of growth after the first year of life in MoBa likely had lower statistical power due to the smaller number of study participants with the necessary follow-up information for this analysis.

Although few studies have reported weight at age 12 months in relation to type 1 diabetes, other studies of anthropometric measurements and islet autoimmunity or type 1 diabetes have been published. The previous cohort studies of genetically susceptible children generally included relatively few cases, with islet autoimmunity as the end point, and anthropometric measurements from age 2 years or older.^11-13 Our findings indicated a tendency for a positive association between peak length velocity and type 1 diabetes, while there was no association with peak weight velocity. Results on BMI from other studies have been mixed, but overall they indicate a positive association in line with our findings. Furthermore, our results indicate no significant association between age at peak BMI and diagnosis of type 1 diabetes.

It has been hypothesized that childhood overweight is a risk factor and explains the temporal changes in incidence of type 1 diabetes. Our results may be taken as support for the hypothesis that weight gain explains at least some of the long-term temporal changes in incidence of type 1 diabetes. However, there is limited evidence for changes in infant weight in the short term, in line with a plateau in the incidence of type 1 diabetes.² We observed no time trends in anthropometric measurements during the first year of life in our 2 cohorts (data not shown), and Norwegian growth data for children at birth to age 4 years indicated little or no change in weight (or length) between 1982-1984 and 2003-2006.²⁴

It should be noted that our observed association was linear across the range of weight increase, with no indication of a threshold for overweight infants. While we decided a priori to focus on growth to age 12 months, infant growth is shown in other studies to be associated with obesity later in life.²⁵ Furthermore, our results suggested that weight change early in infancy (age <6 months) was more important for type 1 diabetes risk than was subsequent weight change.

Genetic variants may in theory explain our findings, if the same variants are associated with increased risk of type 1 diabetes and increased infant weight gain. Furthermore, either such variants must be very common or there must be a very large number of different rare variants with these properties. This is not likely to be the case. If anything, variants known to influence type 1 diabetes risk seem to be associated with less weight gain in the first 2 years of life.^26,27

Similarly, epigenetic mechanisms may in theory explain the link between infant weight gain and type 1 diabetes, but only if the same pathways influencing infant growth also influence type 1 diabetes. Current evidence does not support the latter.²⁸

Maternal type 1 diabetes is known to be associated with infant weight gain and increased risk of type 1 diabetes, but this is too uncommon in the population to explain our finding, as indicated by the robustness of our results to adjustment for or exclusion of children whose mothers had diabetes. Furthermore, infant feeding and a number of other environmental factors that could potentially explain an association between infant weight gain and risk of type 1 diabetes were also adjusted for in our analysis and are therefore unlikely to be major explanatory mechanisms.

A general mechanism that may plausibly explain our findings is that rapid growth increases the demand on β cells to secrete insulin, and it has been shown that β cells actively secreting insulin are more susceptible to cytokine damage in vitro.²⁹ We can also speculate that perhaps gut microbiota^30,31 or inflammation^32,33 may be involved in relevant pathways to explain our findings.

In conclusion, our study is the first prospective population-based study, to our knowledge, providing evidence that weight increase during the first year of life is positively associated with type 1 diabetes. This supports the early environmental origins of type 1 diabetes.

Corresponding Author: Maria C. Magnus, PhD, Department of Chronic Diseases, Norwegian Institute of Public Health, PO Box 4404, Nydalen, N-0403 Oslo, Norway (maria.christine.magnus@fhi.no).

Accepted for Publication: October 7, 2015.

Published Online: December 7, 2015. doi:10.1001/jamapediatrics.2015.3759.

Author Contributions: Dr M. C. Magnus had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: M. C. Magnus, Joner, Skrivarhaug, Johannesen, Njølstad, Størdal, Stene.

Acquisition, analysis, or interpretation of data: M. C. Magnus, Olsen, Granström, Skrivarhaug, Svensson, Johannesen, Njølstad, P. Magnus, Stene.

Drafting of the manuscript: M. C. Magnus, Skrivarhaug, Johannesen.

Critical revision of the manuscript for important intellectual content: Olsen, Granström, Joner, Skrivarhaug, Svensson, Johannesen, Njølstad, P. Magnus, Størdal, Stene.

Statistical analysis: M. C. Magnus, Granström.

Obtained funding: Olsen, Skrivarhaug, Njølstad, P. Magnus, Stene.

Administrative, technical, or material support: Granström, Skrivarhaug, Svensson, Njølstad, P. Magnus, Stene.

Study supervision: Olsen, Johannesen, Størdal, Stene.

Conflict of Interest Disclosures: None reported.

Funding/Support: The Norwegian Mother and Child Cohort Study is supported by grants UO1 NS 047537-01 and UO1 NS 047537-06A1 from the National Institutes of Health and grant 151918/S10 from the Norwegian Research Council/FUGE. The substudy PAGE (Prediction of Autoimmune Diabetes and Celiac Disease in Childhood by Genes and Perinatal Environment), based on MoBa, is supported by grant 2210909/F20 from the Norwegian Research Council (Dr Stene). Dr M. C. Magnus is supported by grant 2011.2.0218 from the Norwegian Extra Foundation for Health and Rehabilitation. Dr Størdal is supported by an unrestricted grant from the Oak Foundation. The Danish National Birth Cohort is supported by the March of Dimes Birth Defects Foundation, the Danish Heart Association, the Danish National Research Foundation, the Danish Pharmaceutical Association, the Ministry of Health, and the National Board of Health. The cooperation between the 2 cohorts was supported by EARNEST EU FP6 Integrated Project contract 007036 with the European Commission (Dr Olsen was leader of a work package comparing the 2 databases), grant 09-067124 of the Innovation Fund Denmark from the Centre for Fetal Programming, and the Danish Cancer Union.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Inger Johanne Bakken, PhD, Norwegian Institute of Public Health, Oslo, Norway, provided guidance in the use of data from the Norwegian Patient Register; she received no compensation. We are grateful to all families participating in the Norwegian Mother and Child Cohort Study and the Danish National Birth Cohort and for the administrative staff surrounding these 2 cohorts.