Gestational diabetes associated with incident diabetes in childhood and youth: a retrospective cohort study

Data sources

We used provincial health administrative data in the Canadian province of Quebec, merged with information from the birth and death registries. Health administrative databases are maintained by the Régie de l’assurance maladie du Québec (RAMQ) for the public health insurance plan that covers outpatient and in-patient physician services and all costs of acute hospital care for all residents. The health administrative Physician Services Claims and Hospitalization Discharge Diagnoses databases include procedural and diagnostic codes for physician visits (International Classification of Diseases, 9th Revision [ICD-9]) and hospital discharge diagnoses (ICD-9 until April 2006; International Classification of Diseases, 10th Revision [ICD-10] thereafter). Birth and death registries are held at the Institut de la statistique du Québec. The birth registry includes data from the Declaration of Birth that parents are legally required to complete within 30 days of delivery. Questions include maternal level of education, country of birth (mother and father) and first language (mother and father); the sex, gestational age and birth weight of the child are also included.

The RAMQ identified women aged 20–44 years with a singleton live birth between Apr. 1, 1990, and Dec. 31, 2007, with 2 or more outpatient physician billing diagnoses for gestational diabetes mellitus within 6 months of delivery, a diagnosis of gestational diabetes mellitus recorded at hospital discharge, or both. Corresponding fathers and offspring were identified. Mother–offspring–father groupings are hereafter termed “triads.” Only complete triads were retained.

For each mother with gestational diabetes mellitus, the RAMQ randomly selected a matched control from among the cohort of mothers with singleton live births between Apr. 1, 1990, and Dec. 31, 2007, who met the following criteria: 1) absence of gestational diabetes mellitus, 2) singleton live birth in the same calendar year, 3) residence in same health administrative territory and 4) same age group (20–24 yr, 25–29 yr, 35–39 yr and 40–44 yr). Corresponding fathers and offspring were identified to form mother–offspring–father triads.

The Institut de la statistique du Québec used a probabilistic matching strategy12 to extract additional data on these gestational diabetes mellitus and non–gestational diabetes mellitus triads from the birth registry and, when applicable, the death registry.

Triads were excluded if either parent or corresponding matched parents had cardiovascular disease or heart failure. This was defined as at least 1 hospital admission, discharge date within the 3 years preceding the calendar birth year of their child, with any diagnosis of myocardial infarction or congestive heart failure, or a procedural code for coronary angioplasty or implantation of a cardiac stent. Diabetes is a risk factor for cardiovascular disease;13 exclusions related to cardiovascular disease or heart failure were based on the possibility of undiagnosed diabetes in these parents. Although these conditions are uncommon in new parents who are usually young to middle-aged, we made these exclusions because we were asked to limit the data requested as much as possible, for privacy and confidentiality reasons, despite anonymization. We excluded all triads with 1 or more diagnostic codes for diabetes in the father during the 3 years before delivery or in the mother for the same period, but excluding the weeks of pregnancy. When we excluded a triad, we also excluded its matched triad.

To further ensure confidentiality, as required by the Commission d’accès à l’information du Québec, the RAMQ did not provide specific birth dates; instead, it provided the offspring’s birth month and year and the age group of mothers and fathers (5-year brackets) at the time of delivery. We defined the 15th day of the birth month as the delivery date. We had accurate information on gestational age as this was directly available from the birth registry, reported in weeks. We considered the pregnancy period to comprise the estimated delivery date and the weeks before this estimated date corresponding to the reported gestational age, plus 2 additional weeks prior.

Statistical analysis

We performed the analyses using SAS (version 9.3). We examined baseline characteristics (means and standard deviations, numbers and proportions), plotted Kaplan–Meier curves, and computed incidence rates of diabetes (number/10 000 person-years) by status of maternal gestational diabetes mellitus. We examined associations of diabetes in offspring with maternal gestational diabetes mellitus through unadjusted and adjusted Cox proportional hazards models. Follow-up was from birth to up to 22 years.

In secondary analyses, we separately considered age groups of birth to age 12 years and age 12 to 22 years. Insulin resistance rises during puberty which, coupled with an increase in weight or low activity levels, could render adolescents particularly vulnerable;21 hence our interest in examining childhood and adolescence separately in secondary analyses. We adjusted multivariate models for sex, gestational age (< 32 wk, 32 to 37 wk compared with > 37 wk reference group), birth weight (< 3000 g, > 4000 g compared with 3000 g to 4000 g reference), ethnocultural background (European or non-European), material deprivation (4 least deprived quintiles or most deprived quintile), previous pregnancy (yes or no) and maternal autoimmune disease (yes or no).