Intrauterine growth pattern and risk of childhood onset insulin dependent (type I) diabetes: population based case-control studyBMJ 1996; 313 doi: https://doi.org/10.1136/bmj.313.7066.1174 (Published 09 November 1996) Cite this as: BMJ 1996;313:1174
- Gisela Dahlquist, professor of paediatricsa,
- Sara Sandberg Bennich, MD studentb,
- Bengt Kallen, professorc
- a Departments of Paediatrics and Epidemiology and Public Health, University of Umea, S-901 85 Umea, Sweden
- b Department of Internal Medicine, Uppsala University, Uppsala, Sweden
- c Tornblad Institute, University of Lund, Lund, Sweden
- Correspondence to: Professor Dahlquist.
- Accepted 6 September 1996
Objective: To investigate whether prenatal growth affects the risk of development of childhood onset insulin dependent (type I) diabetes mellitus.
Design: Population based case-control study.
Setting: Data from a nationwide childhood diabetes case register were linked with data from the nationwide Swedish Medical Birth Registry.
Subjects: Data from a total of 4584 diabetic children born after 1973 and diagnosed with diabetes from 1978 to 1992 were studied. For each child with insulin dependent diabetes three control children were randomly selected from among all infants born in the same year and at the same hospital as the proband.
Main outcome measures: Birth weight, gestation, maternal age and parity, number of previous spontaneous abortions, and sex specific birth weight by gestational week expressed as multiples of the standard deviation (SD).
Results: There was a clear trend in the odds ratio for childhood onset diabetes according to SD of birth weight. The odds ratio (95% confidence interval) for small for gestational age after stratification for maternal age, parity, smoking habits, and maternal diabetes was 0.81 (0.65 to 0.99) and for large for gestational age after similar stratification was 1.20 (1.02 to 1.42).
Conclusions: Intrauterine conditions that affect prenatal growth seem also to affect the risk of development of childhood diabetes in the way previously described for postnatal growth: a poor growth decreases and an excess growth increases the risk. The mechanism for this association is unclear.
By linking two nationwide population based reg- isters the effect of intrauterine growth on the risk for childhood onset insulin dependent diabetes is estimated
There was a clear trend in the risk for childhood onset diabetes according to differences in birth weight by gestational age expressed as multiples of SD from population means
The adjusted odds ratio for babies who were small for gestational age was significantly decreased and for large for gestational age babies was significantly increased
A poor intrauterine growth decreases and an excess growth increases the risk of development of childhood insulin dependent diabetes
There is evidence that both future immune reactivity1 and carbohydrate metabolism2 can be affected by the intrauterine environment of the fetus. Both types of associations may be relevant for the pathogenesis of insulin dependent (type I) childhood onset diabetes. In a previous study we reported on several perinatal risk determinants for insulin dependent diabetes3 among which we found a significant effect of short gestation but no effect of birth weight or body length. We have now studied the importance of disturbances in intrauterine growth on the risk for childhood onset insulin dependent diabetes in greater detail, using extended material and the recently published normal growth chart for Swedish children.4
Subjects and methods
In Sweden all children younger than 15 years with diabetes are referred to paediatric departments. Since 1 July 1977 all incident cases of insulin dependent diabetes have been reported to the Swedish Childhood Diabetes Registry.5 The level of ascertainment has been shown to vary between 96% and 99%.6 7 From 1978 to 1992 a total of 4702 children born in 1973 or later were recorded in the diabetes registry.
The Swedish Medical Birth Registry began in 1973 and stores data on the pregnancy, delivery, and neonatal period for nearly all infants born in Sweden.8 The two registries were linked with the unique personal identification number given to everyone living in Sweden. Linkage was obtained for 4584 children. Thus for 118 (2.5%) linkage failed; either the child was not born in Sweden or there was an error in the identification numbers used for linkage.
For each child with insulin dependent diabetes for whom linkage was successful, three control children were selected randomly from among all infants born that year at the same hospital as the proband child. Only living children were accepted as controls; routinely, dates of death are linked to the birth registry for all children dying within Sweden.
We analysed the following variables from the birth registry: birth weight, gestation, maternal age and parity, maternal diagnosis of diabetes during pregnancy, and maternal smoking in early pregnancy. By using a published graph of sex specific normal birth weight for each gestational week in the Swedish population and based on data from the birth registry4 for each case and control, we calculated the difference from the expected weight and expressed it in standard deviations (estimated as 12% of expected weight). Among all records birth weight was known in 4538 (99%), gestation in 4492 (98%), and maternal age and parity in all. Maternal smoking was known only after 1983 and was then stated in 93% (1238 (27%) of all). Deviation in intrauterine growth could be determined in 4447 (97%). There was no difference between cases and controls in the completeness of the records.
Analyses of birth weight, gestation, and intrauterine growth were restricted to singleton pregnancies. The diagnosis of maternal diabetes before 1987 (ICD-8, (international classification of diseases, eighth revision) did not differentiate between insulin dependent diabetes and gestational diabetes, but from 1987 onwards the two conditions got separate codes. This means that before 1987 there was a slight dilution of maternal insulin dependent diabetes with gestational diabetes.
Statistical methods—Relative risks were calculated by using Mantel-Haenszel9 estimates of odds ratios after various stratifications. We estimated 95% confidence intervals with a test based method. Tests for homogeneity between strata were made with the Breslow and Day test.10 To study trends in a frequency table we used exact trend statistics (version 2.1, CYTEL Software Corporation, Cambridge, MA, USA).
Among the mothers of children with diabetes 73 had a diagnosis of diabetes during pregnancy; among the controls 56 mothers had such a diagnosis. The rate of recorded diabetes in control children was thus 1:246 or 4.1 per 1000.
BIRTH WEIGHT AND GESTATION
Determinants of birth weight and gestation comprised maternal age and parity, maternal smoking, and maternal diabetes.
Figure 1 shows the crude odds ratio for the development of childhood onset insulin dependent diabetes for each birthweight class with 3000 g to 4000 g as reference (1.0); no obvious relation was seen. A dichotomised analysis with stratification for maternal age, parity, smoking habits, and maternal diabetes of the risk associated with a birth weight <2500 g gave an odds ratio (95% confidence interval) of 1.03 (0.86 to 1.24), which was not significantly different when age of onset of diabetes (0–4, 5–9, or 10–14 years) was considered.
Figure 2 shows the crude odds ratio for the development of childhood onset insulin dependent diabetes according to gestation, with weeks 39–41 as reference (1.0). There was an indicated increased risk with short gestation. A comparison of preterm (<37 completed weeks) and term infants, with stratification for maternal age, parity, smoking habits, and maternal diabetes, gave an odds ratio of 1.25 (0.99 to 1.33). There is thus a suggested but not significant excess risk with preterm births.
To compare intrauterine growth in cases and controls, each child's birth weight was expressed as multiples of SD from the sex specific “normal” weight as defined from the medical birth registry.4 Figure 3 shows crude odds ratios at different scores (in 0.5 SD units) with the interval −0.5 to +0.5 SD as a reference (1.0). There is a clear cut trend in this graph with a reduced risk at low multiples of SD and an increased risk at high multiples. A summary of the exact distribution of trend statistics gave trend = 4.1; P<0.001. Exclusion of data from children whose mothers had a diagnosis of diabetes in pregnancy did not change the association.
The odds ratio for small for gestational age, defined as <2 SD after stratification for maternal age, parity, smoking habits, and maternal diabetes, was 0.81 (0.65 to 0.99). Exclusion of data from children whose mothers had a diagnosis of diabetes in pregnancy did not change the odds ratio, and there was no clear cut difference according to age at onset in the child, even though only the group aged 5–9 years gave a significantly deviating odds ratio (0.62; 0.44 to 0.89).
The odds ratio for large for gestational age, defined as >2 SD after similar stratifications, was 1.20 (1.02 to 1.42), again similar according to age at onset. Exclusion of data from children whose mothers had a diagnosis of diabetes in pregnancy hardly changed the odds ratio (1.19; 1.00 to 1.40).
In the present study we have shown that disturbances of intrauterine growth affect the risk of the child developing childhood onset insulin dependent diabetes. Small for gestational age tends to diminish the risk, large for gestational age increases it. Epidemiological studies from different parts of the world consistently show a peak in incidence occurring during the rapid pubertal growth spurt.11 12 Case-control studies have also shown that an increased growth rate before puberty13 and even in infancy14 is associated with an increased risk for childhood onset insulin dependent diabetes. That the opposite may be true—that is, that starvation leading to a reduced growth rate is protective—is indicated by the findings of a very low prevalence of insulin dependent diabetes mellitus in men born during the war and the immediate postwar period in Germany.15 Our data indicate that even prenatal growth can have an effect on the risk of the development of childhood onset insulin dependent diabetes; low birth weight for gestational age reducing the risk and a high birth weight increasing it. This also explains the finding previously reported by us, that maternal smoking during pregnancy is a protective factor for childhood onset insulin dependent diabetes.3
Routinely recorded register data may be incomplete, but as data are recorded long before the child develops diabetes, no systematic bias can be introduced. One problem is that the information on maternal diabetes may be incomplete, which may to some extent question the conclusions drawn after exclusion of data from children whose mothers had diabetes during pregnancy. The important factor that might confound our results is maternal insulin dependent diabetes. Such diagnosis in the mother is, however, less likely to be underreported. Non-insulin dependent (type II) diabetes probably accounts for only a minority of those with a diagnosis of maternal diabetes, but before 1987 some might have been gestational diabetes. The mean prevalence of insulin dependent diabetes in Swedish women of all ages, 1977–87, was reported to be 3.6 per 1000,16 indicating that our findings of maternal diabetes of 4.1 per 1000 among the control mothers is not an underestimate.
The results may be affected by confounders which were not taken into consideration in the analysis. One such possible confounder is socioeconomic class, which is associated with retardation in intrauterine growth, but previous studies in Sweden have shown that low socioeconomic class is associated with an increased risk of childhood onset insulin dependent diabetes17 and would therefore tend to bias the odds ratio towards 1.0.
In conclusion, this study indicates that disturbances in intrauterine growth may affect the risk of a child developing insulin dependent diabetes. If this association is a direct one further studies on the underlying cause are needed.
Funding This project was supported by grants from the Medical Research Council (project No 07531) and the Swedish Diabetes Association.
Conflict of interest None.