Study design and population

This was a population-based retrospective birth cohort study, using the linked birth and infant death database of the Institut de la Statistique du Québec (ISQ) for singleton births in Quebec, Canada, 1996–2010.21 The validity of the Canadian vital records linkage has been well documented.22 We further linked birth records to pregnancy complications through the mother’s medical insurance number.23 The research protocol was approved by the ISQ, the Commission for Access to Information in Quebec, and major Indigenous community health organisations in Quebec—the Cree Board of Health and Social Services of James Bay, the First Nations of Quebec and Labrador Health and Social Service Commission, and the Nunavik Regional Board of Health and Social Services. In adherence to ISQ data confidentiality rules, all numbers are rounded to the nearest 10.

Indigenous (First Nations, Inuit) births were identified based on three sources of information: mother tongue, community of residence (to identify Indigenous reserves according to residential postal code/municipality name) and Indian Registration System membership status of both parents (a positive in any Indigenous identifier was considered Indigenous).21 The original study cohort included 20 190 First Nations, 4260 Inuit and 229 960 non-Indigenous singleton births at gestational age≥20 weeks. The present study included 17 090 First Nations and 217 760 non-Indigenous births at ≥24 weeks of gestation with linked data available on both pregnancy complications and fetus/infant’s survival status from birth to 364 days of life. We restricted the analyses to births at ≥24 weeks of gestation because GDM is routinely screened and diagnosed at 24–28 weeks of gestation. Inclusion of births at <24 weeks of gestation could introduce an immortal bias (all perinatal and postnatal deaths would not happen to the offspring of GDM women due to the absence of screening diagnosis). We focused on the comparisons of First Nations with non-Indigenous births, since First Nations (but not Inuit) women are known to be at substantially higher risk of PGDM and GDM than non-Indigenous women in Quebec.19 Exploratory analyses confirmed that Inuit women were not at higher risk of diabetes in pregnancy than non-Indigenous women. Inuit births were excluded from the present study.

Linkage to pregnancy complications

In Quebec, each resident has a lifetime unique Régie de l’Assurance Maladie du Québec (RAMQ) (Quebec health insurance) number. The mother’s RAMQ number was used to link pregnancy complications (diabetes, etc) to each birth record, based on delivery discharge medical records in the provincial Maintenance et Exploitation des Données pour l'Étude de la Clientèle Hospitalière database.23 The data linkage used the hospitalisation/diagnostic records throughout the course of the index pregnancy (based on the infant’s date of birth and gestational age at delivery). The linkage to hospitalisation records was successful for 97.9% First Nations, 92.3% Inuit and 98.0% non-Indigenous infants.23

Pregnancy complications were coded according to the International Classification of Diseases Clinical Modification (ICD-9CM) codes in Quebec for records in the years 1996–2005, and the ICD-10 codes in Quebec for records in the years 2006–2010. In the ICD-9CM, PGDM was coded as ‘648.0’ or ‘250.*’, while GDM was coded as ‘648.8’. In the ICD-10, PGDM was coded as ‘O24.5’, ‘O24.6’, ‘O24.7’, ‘E10’ and ‘E11’, while GDM was coded as ‘O24.8’. We did not make a distinction between different types of PGDM (type 1, type 2, other), because available information is insufficient for records based on the ICD-9CM codes in the years 1996–2005, while the numbers of records based on the ICD-10 codes in the years 2006–2010 only were too small in First Nations for such subgroup analyses on the risks of perinatal and postneonatal death.

During the study period (1996–2010), universal screening for GDM was in place in Quebec, following a two-step screening procedure. First, all pregnant women underwent a 50 g 1 hour oral glucose challenge test at 24–28 weeks of gestation. If blood glucose was ≥7.8 mmol/L, the 75 g 2 hours oral glucose tolerance test was then administered. GDM was diagnosed if two of three values equaled or exceeded the following cut-offs: fasting 5.3 mmol/L, 1 hour 10.0 mmol/L, 2 hours 8.6 mmol/L.24

Outcomes

The primary outcomes were perinatal death (stillbirth [fetal deaths≥500 g] plus neonatal death [died in the first 28 days after birth])25 and postneonatal death (died between 28 and 364 days after birth). These are objective outcomes not subjecting to misclassifications. We used perinatal death as a primary outcome because it is not affected by variations in registration practices in cases of deaths of newborns at the borderline of viability (a dead newborn may be registered as a stillbirth or neonatal death).25 We used postneonatal death as a primary outcome to reflect the potential ‘long-term’ impact of intrauterine exposure to diabetes on infant health. Other outcomes include birth defects, preterm (<37 completed weeks of gestation), small for gestational age (SGA, <10th percentile in birth weight for sex and gestational age, according to the Canadian fetal growth standards26) and large for gestational age (LGA, >90th percentile) births. Birth defects included any congenital anomalies according to the ICD-9 codes (for records in 1996–2005) or ICD-10 codes (for records in 2006–2010) on birth registration and infant hospitalisation records in the first year of life. We presented the results on the total birth cohort only (rather than further stratification by year period) since the numbers of events (the primary outcomes: perinatal and postneonatal deaths) were relatively small for First Nations with PGDM or GDM.

Statistical analysis

Crude relative risks (RR) and absolute risk (AR) with 95% CIs were calculated to illustrate the strength of the associations of PGDM and GDM with adverse birth outcomes, perinatal and postneonatal mortality. Log-binomial models were fitted to estimate the adjusted RRs with 95% CIs accounting for cluster effects (in cases of multiple births to the same mother) in generalised estimating equations. We examined the changes in the adjusted RRs to assess whether the associations of diabetes in pregnancy with perinatal and postneonatal mortality could be explained by maternal characteristics (age [<20, 20–34, 35+years], education [<high school, high school, college, university], marital status [married, common-law union, single/separated/divorced], parity [primiparous: yes/no], rural residence [yes/no]) and other pregnancy complications (chronic hypertension, heart disease, renal disease, genitourinary infections, anaemia, gestational hypertension and pre-eclampsia). All data analyses were carried out using SAS, V.9.2. Two-tailed p values <0.006 were considered statistically significant, considering eight primary comparisons of interest (PGDM or GDM versus non-diabetic pregnancies in First Nations and non-Indigenous populations, two primary outcomes, 2*2*2=8 comparisons, Bonferroni corrected p value cut-off=0.05/8=0.006).

Patient and public involvement

Patients were not involved in the study based on anonymised linked birth data.

Main findings

Our study is the first to reveal differential associations of PGDM with perinatal and postneonatal mortality in Indigenous versus non-Indigenous populations. PGDM was associated with a much greater elevated risk of perinatal death in First Nations than in non-Indigenous populations, but was associated with a substantially elevated risk of postneonatal death in non-Indigenous infants only.

Data interpretation and comparisons with findings in previous studies

First Nations are a high-risk population for diabetes.27 As expectedly, our data revealed over three times increased risk of PGDM and over two times increased risk of GDM in First Nations versus non-Indigenous populations in Quebec.

Our study is the first to reveal that the risk elevation in perinatal death for PGDM was to a much greater extent in First Nations (5.1-fold) than non-Indigenous (1.8-fold) populations. Glycaemic control tends to be worse in First Nations than non-Indigenous patients with diabetes.28 Poor glycaemic control has been linked to perinatal death.29 30 This may partly explain the greater risk elevation in perinatal death for PGDM pregnancies in First Nations. Poorer glycaemic control may result in more frequent unhealthy fetuses, and consequently more perinatal deaths. However, there was a large risk increase (3.5-fold) in postneonatal death in PGDM pregnancies in non-Indigenous infants, in contrast to the absence of risk increase in First Nations infants, contradicting the hypothesis of more frequent ‘unhealthy fetuses’ in First Nations diabetic pregnancies. We speculated that there may be genetic difference in the vulnerability to the long-term adverse impact of fetal hyperglycaemia in First Nations versus non-Indigenous populations. Alternatively, we speculated that this may be due to a ‘delayed’ impact of diabetes from perinatal to postneonatal periods in non-Indigenous infants. Better glycaemic control in PGDM pregnancies might have delayed the impact into the postneonatal period in non-Indigenous population. It should be noted that the higher RR of postneonatal death associated with PGDM in non-Indigenous infants was almost entirely attributable to a much lower postneonatal mortality rate in the non-diabetic group; postneonatal mortality rates in the PGDM group were very similar in First Nations (3.5 per 1000) and non-Indigenous (3.4 per 1000) infants. It appeared that infants of mothers with non-diabetic or healthy pregnancies were the ones to benefit from the more advantageous socioeconomic and infant care conditions in non-Indigenous populations.

There were large differences in maternal characteristics and rates of pregnancy complications other than diabetes between First Nations and non-Indigenous groups. However, the adjusted RRs showed that these factors hardly affected the associations of PGDM and GDM with perinatal and postneonatal mortality in either population group.

We observed a marginally lower risk of perinatal death for GDM pregnancies in non-Indigenous infants, although this risk difference would be not statistically significant if accounting for multiple comparisons. The observation is, however, consistent with the reported reduced risk of perinatal death in diet-controlled GDM pregnancies.31 GDM is a transient hyperglycaemia that may occur in the second half of pregnancy. The relatively increased insulin resistance state in GDM may confer a greater capacity for maternal–fetal nutrient transfer to meet the demands for fetal growth and development32—which may be somewhat ‘beneficial’ for fetal survival if hyperglycaemia is relatively well controlled.33

Consistent with previous reports,8–17 the risk increases in adverse birth outcomes (preterm, LGA, birth defects and perinatal death) were observed to be less pronounced in GDM than PGDM pregnancies. The risk increase in LGA associated with GDM was greater in First Nations (RR=2.0) than non-Indigenous (RR=1.7) infants, indicating poorer glycaemia control (a common cause of LGA34 in First Nations mothers).

We confirmed the substantially elevated risk of birth defects in PGDM pregnancies9–13 in both First Nations and non-Indigenous infants. It remains uncertain whether GDM may increase the risk of birth defects. We did not observe any significantly increased risk of birth defects in GDM pregnancies in both First Nations and non-Indigenous groups.

Strengths and limitations

The main strength is the large population-based linked birth cohort. Our study has limitations. We did not have data on unmeasured risk factors including smoking, obesity, folic acid supplementation, preconception care and glycaemic control in diabetic pregnancies. Both smoking and obesity are known to be much more prevalent in First Nations versus non-Indigenous women.35 Glycaemic control in diabetic pregnancies is known to be worse in First Nations versus non-Indigenous populations.28 Differences in these and other unmeasured risk factors could partly account for the substantially higher perinatal and postneonatal mortality rates in First Nations versus non-Indigenous populations overall.21 However, the adjustment analyses indicated that the RRs of perinatal and postneonatal death comparing PGDM or GDM versus non-diabetic pregnancies within the First Nations or non-Indigenous group would be unlikely to have been affected by the differences in risk factors between the two groups. Our study is observational in nature, and could not determine whether diabetes is causal to any elevated risk of perinatal or postneonatal death. The study was based on a birth cohort in Quebec. More studies in other regions are required to understand the generalisability of the findings to First Nations (North American Indians) and other Indigenous populations.