Size for gestational age affects the risk for type 1 diabetes in children and adolescents: a Swedish national case–control study

The incidence of type 1 diabetes, one of the most common chronic diseases among children and adolescents, has rapidly increased over the last decades. Sweden has one of the highest incidences in the world, surpassed only by Finland [1]. Although there is a known genetic association between type 1 diabetes and certain HLA-alleles, such as DQ2 and DQ8 [2], the cause of the disease is mostly unknown. It is generally believed that both genetic and environmental factors contribute to the pathogenesis and that environmental factors are responsible for the steep increase in incidence [3]. Since type 1 diabetes can manifest early in life, and there is a long preclinical phase before clinically overt disease develops [4], perinatal factors concerning the child have been proposed to contribute to the pathogenesis [5].

There is a great variation in the prevalence of children born small for gestational age (SGA) and large for gestational age (LGA) worldwide and it has changed over time. In most countries there has been an increase in the incidence of children born LGA in recent decades, though in the last few years there have been reports of a shifting trend. On the other hand, the incidence of children born SGA has been stable or decreasing over recent decades [6].

The results of previous studies have been conflicting. A cohort study from the UK, including 991 children with type 1 diabetes, showed an increased risk for type 1 diabetes if the child had a birthweight >4000 g vs <3000 g (RR 1.47 [95% CI 1.16, 1.85]) [7], and a meta-analysis of 11 studies (in total 7491 cases) showed an increased risk for type 1 diabetes if the child was born >4000 g vs <4000 g (OR 1.17 [95% CI 1.09, 1.26]) [8]. A Swedish case–control study, including 4584 cases, found that the OR for type 1 diabetes was 1.20 (95% CI 1.02, 1.42) if the child was born LGA [9]. A high birthweight has also been shown to be associated with a higher weight later in life [10]. Other studies have shown that childhood obesity and a rapid linear growth in childhood increases the risk of type 1 diabetes [11]. A low birthweight, or to be born SGA, has been shown to correlate negatively with the risk of type 1 diabetes [9, 12, 13]. However, many smaller case–control studies have failed to show any significant correlation between birthweight/LGA/SGA and subsequent risk of type 1 diabetes [14‐17]. In addition, a low birthweight (<2500 g) has been shown to correlate with earlier onset of type 1 diabetes [18]. Several studies on this subject have used absolute birthweight, but it is important to correct for gestational age when looking at birthweight since there is a strong correlation between the two [19].

When studying size for gestational age, it is important to adjust for maternal factors known to affect birthweight such as maternal age, BMI, parity, gestational weight gain, maternal smoking and maternal diabetes [6, 20].

We hypothesise that being born LGA or SGA affects the risk of the child later being diagnosed with type 1 diabetes. Our large and nationwide body of material gives us an exceptional opportunity to investigate the potential relationship between these factors in a case–control study. Hence, the aim of this case–control study is to investigate whether size for gestational age affects the future risk of type 1 diabetes, and the age at onset of type 1 diabetes, and to determine whether there is a sex difference or not.

Most children were born AGA (93.8%), whereas 3.7% were born LGA and 2.5% were born SGA, with no significant sex differences. Children with type 1 diabetes were significantly more often born LGA and significantly less often born SGA than control children (4.7% vs 3.5% and 2.0% vs 2.6%, respectively, p < 0.001) (Table 1). Limiting the study population to only children born at term, their size for gestational age was a little more evenly distributed, but the statistical significance remained (Table 2). The same was true when using the broader definition of LGA (weight above the 90th percentile for that gestational age) and SGA (weight below the 10th percentile for that gestational age).

The absolute birthweight revealed that more children who later developed type 1 diabetes had a birthweight of ≥4000 g (macrosomia), and fewer were born with a low (1500–2499 g) or very low (<1500 g) birthweight (p < 0.001). After restricting the analysis only to children born at term, the difference remained (p < 0.001, Table 2). The proportion of children, regardless of sex, with a birthweight <2500 g increased with age at diagnosis; in the youngest age group (0–4 years) it was 2.0%, and this increased stepwise (5–9 years; 2.6% and 10–14 years; 2.8%) to 4.0% in the oldest age group (15–19 years). For children born SGA, the results did not reach statistical significance (p = 0.073).

Furthermore, children with type 1 diabetes more often had an older mother (age >29 years, 46.0% vs 44.8%, p = 0.048), whereas the control children more often had a mother who had smoked during pregnancy (20.3% vs 17.7%, p < 0.001). The gestational age also showed a slightly different distribution between children with type 1 diabetes and control children (p = 0.014), with more children who later developed type 1 diabetes having been born moderately preterm (5.4% vs 4.9%, Table 1).

A linear-by-linear association was found between type 1 diabetes and maternal BMI and the child’s size for gestational age and the child’s absolute birthweight.

In this large nationwide case–control study we found that, regardless of sex, being born LGA significantly increased the risk for type 1 diabetes later in life, but being born SGA significantly decreased the risk. These results were the same when looking only at children born at term and regardless of maternal diabetes and BMI. Nevertheless, if the mother was overweight or obese in early pregnancy, this increased the risk of type 1 diabetes, and maternal diabetes was the strongest risk factor.

Our results are in accordance with a large cohort study by Khashan et al [13], which showed that children with type 1 diabetes were more likely to have been born LGA and less likely to have been born SGA, with RR approximately the same as our adjusted OR. However, in their sibling design study, the association between LGA and type 1 diabetes was no longer statistically significant. Another large cohort study from the UK by Goldacre [25] also found an increased risk for type 1 diabetes for children with a high birthweight for gestational age, and a decreased risk for children with a low birthweight. Although it used birthweight z scores instead of size for gestational age, the results of a recent large case–control study from Sweden by Waernbaum et al [26] are in line with our results, showing that a birthweight z score of less than −1 was associated with a decreased risk, and above 1 with an increased risk of type 1 diabetes. However, that study did not include children of mothers with diabetes.

As mentioned above, many studies have used absolute birthweight instead of size for gestational age, although the results may then be harder to interpret since appropriate birthweight is influenced by the actual gestational week in which a baby is born. Nevertheless, our results for absolute birthweight are in line with those for size for gestational age, and with results from two meta-analyses that found an increased risk for type 1 diabetes for children with a birthweight >4000 g [27] or >3500 g [8]. However, these meta-analyses did not find a decreased risk for children with a low birthweight (<2500 g) [8, 27]. The fact that several other studies have failed to show a correlation between birthweight and type 1 diabetes might be due to their small size.

The finding that being born with a low birthweight affects the age at onset of type 1 diabetes is interesting and has not, to the best of our knowledge, been reported previously. However, since there was no significant result for SGA, this raises the question of whether the gestational age plays a role. Also, we have previously reported that maternal BMI affects the age at onset of type 1 diabetes in the offspring, and in the oldest age group nearly 50% of mothers had been underweight, which might explain at least part of the increased proportion of children with a low birthweight.

There seemed to be an increased risk in being born moderately preterm, while being born very preterm was protective; however, the significance disappeared in the adjusted model. Khashan et al [13] reported similar findings in their cohort study, though they used slightly different birth categories, and their results remained in the adjusted models. Waernbaum et al [26] also showed an increased risk for children born moderately premature (32–36 weeks of gestation), whereas being born very prematurely (<32 weeks) was protective against the later development of type 1 diabetes.

It is still unknown how the intrauterine environment might affect the risk of type 1 diabetes. One possible explanation is the accelerator hypothesis which states that overnutrition, weight gain, obesity or insulin resistance give a chronically increased beta cell secretory demand leading to activation of intrinsic beta cell stress pathways that trigger or accelerate autoimmunity [28]. Our finding that LGA increases the risk of type 1 diabetes is in agreement with this but our finding that SGA decreases the risk is not. Children born LGA have most likely been subjected to higher circulating glucose levels, which is the most important energy substrate for a fetus and which makes their pancreatic beta cells secrete more insulin [29]. Insulin is an important fetal growth hormone [10] and studies have shown that active beta cells are more susceptible to destruction [8, 27]. A study has shown that LGA children have increased insulin resistance and beta cell activity at birth [30]. Also, metabolic imprinting during fetal life can be transmitted across generations [31]. Children born LGA have an increased risk for obesity in childhood and early adulthood [32] and childhood adiposity has been linked to an increased risk of type 1 diabetes [33].

To be born SGA has previously been found to increase the risk of hypertension, obesity and type 2 diabetes in adulthood [34]. The fact that there seems to be an inverse relationship with type 1 diabetes is intriguing, especially since children born SGA often show catch-up growth in early life [34] and rapid linear growth in childhood has been shown to increase the risk of type 1 diabetes [11]. Children born SGA, like those born LGA, have been shown to have an increased insulin resistance in childhood, but studies have also shown increased insulin sensitivity at birth in SGA children [30]. In a study with NOD mice, one of the most commonly used animal models of autoimmune diabetes, in utero undernutrition protected the diabetes-prone NOD mice from autoimmune diabetes by reduction of insulitis and beta cell apoptosis [35]. Immune modulation is a mechanism that might help explain the association since low birthweight neonates have defects in cell-mediated and humoral immunity, and these deficits have been documented both in childhood and adolescence [35]. Yet it cannot be ruled out that SGA is merely a substitute for some other underlying protective factor.

In the present study, we did not adjust for gestational weight gain since we only had information on about 30% of the mothers. Also, though gestational weight gain has been suggested as a more important risk factor for LGA children, our results did not show that it increased the risk of type 1 diabetes and it may therefore not be regarded as a confounder or mediator.

Our finding that maternal smoking seemed to protect against type 1 diabetes is interesting and has been previously reported [36]. The mechanism is still largely unknown and requires further investigation, and there is of course always the possibility that maternal smoking is merely a confounding factor. The fact that the seemingly increased risk with maternal age disappeared in the adjusted model might be due to the fact that older mothers often had a higher parity, and with parity the birthweight of the child increases [20]. This is also probably the cause of the significant interaction between size for gestational age and maternal age.

The major strength of our study is that we used a large amount of prospectively collected data (therefore free from recall bias) stored in national registries, allowing us to adjust for many potential confounders. There are, of course, other confounders such as the socioeconomic status of the parents or paternal diabetes or BMI, on which we unfortunately have no information, however, the birthweight of the offspring has been shown to be more associated with the mother’s adiposity [10]. The present study had some limitations. With registry data there is always a risk of misclassification caused by incorrect registration; however, if this is present it is random and not systematic and is the same among cases and controls. As always, when looking at several risk factors there is a risk of multiple comparison error, and when using a large body of material like that used in the present study, there is also a risk of finding clinically irrelevant but statistically significant results. Also, we have missing data on 30% of mothers for maternal BMI which means that we have a smaller population in our multivariate analysis than in the univariate.

In conclusion, size for gestational age of Swedish children affects the risk of type 1 diabetes, with an increased risk if the child is born LGA and a decreased risk if the child is born SGA. To be born LGA is an independent risk factor for type 1 diabetes, irrespective of maternal BMI and diabetes. Thus, reducing the risk for a child being born LGA might to some extent reduce the risk for type 1 diabetes.