Cognitive Function in Children and Subsequent Type 2 Diabetes

The National Child Development Study (NCDS) follows all individuals born between 3 and 9 March 1958 (∼17,000 births) and living in the U.K. (4). At birth, midwives recorded information on sex, birth weight, gestational age, mother's age, mother's age at leaving school, smoking during pregnancy (after the 4th month), and the Registrar General's social class based on occupation (5). All significant diagnoses, impairments, and disabilities and height and weight were recorded at the medical examination conducted at age 7 years. Diabetes in a first-degree relative was recorded at this time. Cognitive function was assessed at age 11 years with tests administered at school for general ability (comprising verbal and nonverbal components) with a range of 0–79 (6) and reading comprehension with a range of 0–35 (7). A medical examination and record review at age 16 years recorded confirmed or possible diagnoses of diabetes. Interviews at ages 33 and 42 years identified type 2 diabetes through a question about “diabetes that did not require insulin injections” and excluding diabetes only present during pregnancy (4). Ethnic origin was recorded at age 42 years.

Despite some attrition, the cohort remains generally representative, with 11,419 subjects participating at age 42 years (4). Those with a confirmed or suspected diagnosis of diabetes (type 1 or 2) by age 16 years, as well as those with type 1 diabetes or insufficient information on diabetes, were excluded from all analyses. Those with missing data for the main measures were also excluded, therefore information for 9,182 subjects was available for analysis.

A diagnosis of type 2 diabetes after age 16 years was the dependent variable in logistic regression analysis. The general ability and reading comprehension test scores were examined separately and modeled using SD units (Z scores). Adjustment was for sex, birth weight, gestational age (a separate category identified 853 without a valid value), parental social class, maternal smoking during pregnancy, age mother left school, mother's age at delivery, presence of mild or severe mental retardation, disability, and ethnic origin (97% white British). Potential confounding factors were modeled as categorical variables, except birth weight. Separate models were adjusted for BMI (available for a subset of subjects) at age 7 years, modeled as a continuous variable. Further models examined type 2 diabetes diagnosed after age 33 years.

Cohort members with a diagnosis of type 2 diabetes after age 16 years had statistically significant lower assessment scores at age 11 years after adjustment for potential confounding factors (Table 1). The associations reflect a general shift in distribution, rather than because of a small number of outlying values. The odds ratios indicate the reduction in risk of later type 2 diabetes–associated with 1 SD increase in test score. Adjustment for BMI in childhood did not eliminate the significant association of lower test scores with increased type 2 diabetes risk. Exclusion of subjects with diabetes with onset before age 33 years reduced the sample size with valid data, but lower test scores at age 11 years was still significantly associated with increased type 2 diabetes risk. The associations were not notably altered by exclusion of cohort members with a first-degree relative with diabetes or by modeling gestational age in days as a continuous variable or birth weight as a categorical variable (data not shown).

Poorer cognitive function at age 11 years was associated with an increased risk of type 2 diabetes by age 42 years. Although “diabetes that does not require insulin injections” is a somewhat imprecise definition of type 2 diabetes, we minimized the risk of confounding by other forms of diabetes by excluding all cohort members with a diagnosis of any form of diabetes by age 16 years (as well as those with insufficient information). The association of poorer childhood cognitive function with type 2 diabetes with onset after age 33 years provides further evidence that this finding is not confounded by other forms of diabetes and that there may be a long delay between impaired cognition in childhood and symptomatic onset of type 2 diabetes.

There was relatively little impact of multiple simultaneous adjustment for indicators of in utero exposures including those associated with type 2 diabetes risk (5), conventional markers of cultural, and material circumstances in childhood, mental, or physical diagnoses in childhood and ethnic origin. Even though it may be intermediate in the causal pathway and thus an over-adjustment, including BMI at age 7 years (before the pubertal growth spurt) in the models did not eliminate the association of childhood cognitive function with subsequent type 2 diabetes. Residual confounding is always possible, but the relatively small impact of adjustment for well-recognized indicators of early-life exposures and childhood conditions is consistent with a more direct association between childhood cognitive function and subsequent type 2 diabetes. Further adjustment for BMI at age 7 years provided additional control for relevant lifestyle factors and obesity risk.

It is possible that cognitive deficits present in childhood influence lifestyle factors that increase the risk of type 2 diabetes. Alternatively, poorer glycemic control or other shared risk factors may influence both cognitive development and the risk of type 2 diabetes. Insulin resistance is proposed as a possible mediator between dietary intake and cognitive deficits, and there is some evidence of altered β-cell function preceding the development of clinical hyperglycemia (8,9). Chronically elevated blood glucose levels have been linked to decreased cerebral blood flow in type 2 diabetes, and animal models suggest that myelin production is impeded by high blood glucose levels (10). It is possible that such mechanisms, even before overt clinical hypoglycemia, could impair cognitive development in childhood, thus helping to explain the lower levels of cognitive function in type 2 diabetic patients and greater susceptibility to risks for cognitive decline. This suggests that very early detection of subclinical disease and treatment may be of value in protecting against cognitive deficits.

The funding for this project was obtained through the British Economic and Social Research Council (grant L326253061), which forms part of the Capability and Resilience Network. G.M.O. and A.-L.H. were funded by AFA Insurance.