Two major hypotheses pertain to environmental causes of the increasing incidence of Type 1 DM in many populations around the world. The “Hygiene” hypothesis suggests that improved hygiene and living conditions have decreased the frequency of childhood infections, leading to a modulation of the developing immune system and increasing risk for autoimmune and allergic diseases such as Type 1 DM and asthma.(12,13) The “Overload” or “Accelerator” hypothesis suggests that overload of the pancreatic beta cells early in life make them more prone to autoimmunity and cell death. (14) “Overload” may be caused by a high growth rate in fetal and early life (15), or by early-life stress, such as complicated pregnancy (16), neonatal hospitalization, or even childhood psychological stress. (17,18) Factors previously examined for an association with Type 1 DM, and which are potentially related to these hypotheses, include maternal age (19), caesarian section (20), birth order (21), birth weight (22), maternal gestational diabetes and pre-existing diabetes (23), parental education (23), smoking (23,24), and socioeconomic characteristics affecting care access and health status (25, 26).

Exposure information for the study was obtained from each subject’s birth record. Preliminary evaluation of risk estimates was conducted by stratified analyses. Subsequently, we used multivariable logistic regression to estimate the odds ratios (OR) and 95% confidence intervals (CI) for the associations of factors related to the hygiene hypothesis, including mothers’ age (<18, 18–24, 25–34, >35 years), race (White, Black, Asian, Hispanic, and “Other non-white”), education (<12, 12, and 13+ years), marital status, caesarian-section delivery, prenatal smoking, number of prior births and pregnancies (0, 1, 2, 3+), number of older siblings (estimated by number of prior births now living), use of prenatal care based on the Kotelchuck index of adequacy of prenatal care (inadequate, intermediate, adequate, adequate plus) (27), and trimester prenatal care began (1^st, 2^nd, 3^rd or none). Additional information about the subject’s insurance status (Medicaid or charity insurance at the birth hospitalization) was obtained from the linked CHARS record for the mother’s delivery hospitalization (this linkage is routinely performed annually in Washington since 1987). Maternal factors potentially related to the overload hypothesis included age, diabetes status (established or gestational), BMI calculated from prepregnancy weight and height, and based on WHO classification (<18.5 undernourished, 18.5–24.99 normal, 25–29.99 overweight, ≥30 obese) (28) and pre-pregnancy weight (<100, 100–149, 150–199, 200+ lb.), caesarian-section, and prenatal smoking. Potentially related infant factors included gestational length (<37, 37–42, 42+ weeks), birth weight (<2500, 2500–3999, >= 4000 grams) and size for gestational age (small for gestational age (SGA), appropriate for gestational age (AGA), large for gestational age (LGA), with upper and lower 10^th percentiles calculated (29) using Washington State data 1989–2002 as a standard). Factors evaluated for their potential effects on the relationships of interest included maternal age, race, educational level, marital status, medical insurance at the birth hospitalization, BMI, prenatal smoking, number of prior live births, and infant gender, and birth weight. Subanalyses were also conducted to determine if results varied by birth year categories; they did not. Levels of missing data were generally similar for cases and controls; missing data for all infant variables, maternal age, race, and marital status were less than 5%. For 3 variables (maternal education, pre-pregnancy weight, and BMI), information was available only for birth records from 1992 or later. The greatest level of missing data was for BMI (33% of cases, 38% of controls). Among subjects with missing data for BMI, pre-pregnancy weight, and maternal education, maternal and infant characteristics were distributed similarly to these distributions in the overall study population. Analyses were restricted to subjects with known relevant information for each risk estimate.

The mean age at the hospitalization which identified cases was 7.5 years (SE 4.2, range 0–18 years, similar to the mean age of 7.6 years for all children <19 years hospitalized with Type 1 DM in Washington), with 75% of cases hospitalized at 10 years of age or younger; 49% of both cases and controls were females (data not shown).

Mothers of cases were slightly less likely than mothers of controls to be non-white, unmarried, to have smoked prenatally, to have used private insurance at the birth hospitalization or to have had prior live births (Table 1).

The OR for Type 1 DM increased with increasing maternal age (Table 1). Infants of mothers with BMI > 30 (OR=1.29, 95% CI 1.01–1.64), or with pre-pregnancy weight > 200 lbs also had increased ORs (1.62, 95% CI 1.22–2.14).

Infants of non-white mothers in all race categories had decreased ORs for Type 1 DM, ranging from 0.26 (95% CI 0.18–0.38) for Asians to 0.73 (95% CI 0.53–1.00) for African Americans.

Infants of a mother with less than a high school education had a decreased OR for Type 1 DM, as did those with an unmarried mother, or whose mother smoked prenatally, used Medicaid insurance, or had inadequate or late prenatal care. Relative to having no prior live births, infants of women with 1 or more prior births (total and now living) had ORs for the association with Type 1 DM that were all less than one, and there is evidence of a decreasing OR with increasing number of prior live births (for 3 or more live births, OR=0.56, 95% CI 0.45–0.70, test for trend p< 0.05). Among all subjects as well as among first-born children, being born by caesarian-section delivery was associated with a modestly, but not statistically significant, increased OR for Type 1 DM. When analyses were restricted to the 1305 cases aged less than 10 years at hospitalization, our results did not change, except for 3 instances where the previous ORs that were of borderline statistical significance became statistically significant. This included the ORs for the associations of Type 1 DM with prenatal smoking (OR = 0.83, 95% CI 0.70–0.99), prepregnancy weight of 150–199 lbs (OR = 1.21, 95% CI 1.01–1.46), and C-section (OR = 1.17, 95% CI 1.01–1.35, data not shown). Results were unchanged when children with birth weights of 4000 grams or greater were excluded.

We found that maternal factors, but not infant characteristics examined, were more strongly associated with Type 1 DM in children. Many results were supportive of the Hygiene hypothesis. Several maternal factors associated with a decreased OR (low educational level, unmarried status, Medicaid insurance, inadequate prenatal care) are associated with lower socioeconomic status (SES). Having a mother of non-white race was also associated with a decreased risk for Type 1 DM; non-white race likely has a genetic basis for altered risk but is also associated with lower SES. Lower SES has been reported in other studies to be consistently associated with decreased risk for Type 1 DM. For example, a correlation between higher gross domestic product and lower infant mortality with increased incidence of Type 1 DM has been observed (25,26), countries with rapid development have increased incidence of Type 1 DM (34), and within a single country, a greater incidence of Type 1 DM was noted in groups with higher SES (35). Finally, migration studies show an increased Type 1 DM incidence in population groups who move from an area of low-incidence to one of high-incidence. (8,9,36)

We found an inverse association between increasing number of siblings and risk of Type 1 DM, as have multiple prior studies (19–21,24,37,38). This is consistent with the Hygiene hypothesis as more siblings could lead to earlier and more antigenic exposure in life. Similarly, there have been reports of decreased risk of Type 1 DM associated with sharing a room with a sibling (23), more crowded living conditions (39,40) and day care exposure (41); all are factors potentially related to antigenic simulation.

We observed an increased OR for Type 1 DM in children of mothers older than 25 years. Other studies have observed an association with older maternal age (19–21,24,37,38), but have also reported a complex interaction between maternal age and number of prior siblings , an interaction we did not observe. The increased OR associated with older maternal age may be related to higher socioeconomic status and improved living conditions (factors for which we had little information) for children of older mothers compared to children of the youngest mothers. The increased OR for Type 1 DM in children of older mothers might also be attributed to more complicated deliveries, causing “stress”-induced pancreatic dysfunction as suggested by the Overload hypothesis, however our data do not strongly support this speculation since there was no striking dose response noted in the odds of Type 1 DM associated with maternal age.

We observed a borderline increased OR for Type 1 DM associated with caesarian-section delivery. Though it is biologically plausible that vaginal delivery may be an important source of exposure to antigen, as per the Hygiene hypothesis, prior studies have given inconsistent results regarding this association. Some researchers have reported an increased risk for Type 1 DM after caesarian-section (20,42,43), others find no association (16,44).

Findings from our study that support the Overload hypothesis are the increased ORs for Type 1 DM associated with having a mother with BMI > 30 or whose pre-pregnancy weight was greater than 200 lb. Ours is the first study to use BMI data to assess the risk associated with Type 1 DM. The results are consistent with the Overload hypothesis that suggests that over-nutrition, whether pre- or post-natally, may cause overload or stress to the developing pancreas which subsequently predisposes to Type 1 DM. Other studies have reported associations between birth weight (15,16,38,45), being born large for gestational age, and rapid post-natal growth with an increased risk for Type 1 DM (15); however, we did not find statistically significant associations between infant characteristics and Type 1 DM. The reasons for this are unclear, but one possibility is that our exclusion of subjects with diabetic mothers may have excluded the relevant pathway for these associations.

Several other studies have reported decreased risk for Type 1 DM in children of prenatal smokers (23,24); our estimate suggested a slightly decreased risk in children of smokers, however this result was not statistically significant. A decreased OR would support the Overload hypothesis as smokers tend to have smaller infants and other studies have found a decreased risk for Type 1 DM among children who are born small (15,16,38,45) thus, a decreased risk for Type 1 DM among children of smokers may act through this pathway.

Strengths of our study include that it is population-based and one of the largest studies in the US to examine prenatal and perinatal factors associated with Type 1 DM, and provides new information related to maternal BMI. Another strength is that the exposure information was recorded prior to disease onset, and is not subject to recall bias that may affect case-control studies based on interview. Limitations include data misclassification inherent in any vital records database. Underestimation of smoking and BMI information is plausible as these are undesirable traits. Some cases of Type 1 DM may have been included among our controls if they were not hospitalized at diagnosis, if they moved out of state before they were diagnosed, or if they were hospitalized at a federal hospital. We believe that these numbers would be quite small; most children are admitted to the hospital when first diagnosed with Type 1 DM for glucose stabilization and intensive education. Those who are diagnosed early, before onset of diabetic ketoacidosis, likely have better access to medical care and, therefore, have higher SES (46); the result of capturing these higher SES cases would likely be to increase the risk estimates associated with factors relating to the Hygiene hypothesis, a potential ascertainment bias. Census data indicate that out-migration by families with children is about 6% (47), and suggests that this would be a minor issue in our dataset. Our data only captures patients at non-federal facilities; however, we believe the number of diabetic children treated at military hospitals in Washington State represents a small proportion of the total cases. Overall, the effect of these various types of misclassification would tend to drive the risk estimates towards the null. We also had no information about genetic predisposition to Type 1 DM, or other possible risk factors such as infant feeding history. We attempted to address genetic predisposition to some extent by excluding subjects whose mothers were diabetic, however the possible impact of residual confounding by this, or other risk factors, is difficult to ascertain. That we observed no association of Type 1 DM with infant birth weight >4000 g (a possible marker of having a mother with undiagnosed diabetes), and that our results also were unchanged when these large infants were excluded, is some indication that our results are unlikely to be biased by unmeasured genetic predisposition. Finally, because the actual diagnosis of Type 1 vs. Type 2 diabetes may not be clear, particularly among adolescents, it is possible that some of our cases actually had Type 2 diabetes. Although we lacked further information that would allow us to confirm diagnoses, when we restricted our analyses to only children hospitalized at younger than 10 years of age (where chance of Type 2 diabetes is rarer), the results did not change, indicating that any bias due to such misclassification is likely to be small.