Trends and burden of diabetes in pregnancy among Aboriginal and non-Aboriginal mothers in Western Australia, 1998–2015

This retrospective cohort study used population health datasets linked by the Data Linkage Branch (DLB) of the Western Australia Government Department of Health. DLB uses probabilistic matching methods to link datasets and provides researchers with de-identified data via secure data transfer methods [15] . Datasets linked include the Midwives Notification System (MNS), Hospital Morbidity Database Collection (HMDC), Births Registrations and Death Registrations. MNS, our primary data source contains details on the circumstances of all births occurring at ≥20 weeks’ gestation or birth weight of 400 g or more in Western Australia. HMDC includes information about inpatient episodes in all Western Australian private and public hospitals.

The study included all births at ≥20 weeks’ gestation occurring in WA between 1st January 1998 and 31st December 2015. A multiple pregnancy was counted once when calculating DIP rates. Multiple births (n = 15,558) were excluded from other analyses. Records missing GDM (n = 1912, 0.4%) and PGDM (n = 2056, 0.4%) were excluded from descriptive analyses. Multivariable analyses of LGA trends excluded records with missing gestational age (n = 2177, 0.4%) and birth weight (n = 103, < 0.1%). While remoteness and socioeconomic status were missing in 8232 (1.6%) and 21,443 (4.1%) records, respectively, the percentage of missing values for all other variables was below 0.5%.

The term ‘Aboriginal’ is respectfully used throughout this paper to refer to the First Nations peoples of the Australian continent—Aboriginal and Torres Strait Islander peoples. The term is used for the purpose of brevity and in preference to ‘Indigenous’. While ‘Aboriginal’ is often the preferred term in Western Australian settings (and is a more specific term than ‘Indigenous’), we recognise that it is a generic term that excludes any description of language group or Country and that it is not the preferred term among all Aboriginal and Torres Strait Islander peoples.

Characteristics of mothers with PGDM and GDM, stratified by Aboriginal status and six-year birth periods (1998–2003, 2004–2009 and 2010–2015), were described using counts and proportions. P trend was used to assess the statistical significance of linear trends in neonatal outcomes and maternal characteristics over the three periods. P trend relies on calculating a chi-square statistic for the linear trend (regression) of the proportion of the characteristic (as the outcome variable) on the time periods (as the explanatory variable) [20, 21].

We calculated age-standardised rates for PGDM and GDM to account for the vastly different age profile of Aboriginal and non-Aboriginal mothers. Direct age-standardisation was performed using all mothers who gave birth in WA during the study period as the standard population.

There was both a substantially higher prevalence and more rapid rates of change in PGDM over time among Aboriginal women (Fig. 1 and Additional file 2: Fig. S1). Between 1998 and 2015, the age-standardised prevalence rates of PGDM grew from 4.3% (95% CI: 2.7, 5.9) to 5.4% (95% CI: 4.1, 6.7) (Fig. 1) and the crude rates increased from 2.3% (95% CI: 1.7, 3.1) to 3.8% (95% CI: 3.1, 4.8) (Additional file 2: Fig. S1) among Aboriginal mothers. In non-Aboriginal women, age-standardised and crude annual rates of PGDM were similar, and increased from 0.6% (95% CI: 0.5, 0.7) in 1998 to 0.9% (95% CI: 0.8, 1.0) in 2015.

The age-standardised rates of GDM were consistently higher in Aboriginal women during the study period (Fig. 2), although the trends over time in both standardised and crude rates (Additional file 3: Fig. S2) were comparable. Among Aboriginal mothers, the age-standardised rates of GDM increased from 6.7% (95% CI: 4.8, 8.5) to 11.5% (95% CI: 9.7 to 13.3) and the crude rates increased from 4.6% (95% CI: 3.7, 5.7) to 9.1% (95% CI: 7.9, 10.4) between 1998 and 2015. Among non-Aboriginal mothers, the age-standardised rates of GDM increased from 3.5% (95% CI: 3.3, 3.8) to 10.2% (95% CI: 9.9, 10.6) and the crude rates increased from 3.4% (95% CI: 3.2, 3.7) to 10.6% (95% CI: 10.3, 11.0) over this period.

Our initial mixed-effects multivariable model estimated a 16% increase per decade in the odds of LGA among births to Aboriginal women (OR = 1.16, 95% CI: 1.04, 1.29) after adjusting for maternal age, parity, remoteness and smoking during pregnancy, as well as for clustering by mother (Table 3). The elevated odds was attenuated to 10% by further adjustment for GDM (OR = 1.10, 95% CI: 0.99, 1.23) and to 14% by accounting separately for PGDM (OR = 1.14, 95% CI: 1.02, 1.26). Simultaneous adjustment for both GDM and PGDM resulted in a 7% increase per decade (OR = 1.07, 95% CI: 0.97 to 1.19). Among non-Aboriginal mothers, the odds of LGA per decade decreased by 6% (OR = 0.94, 95% CI: 0.92, 0.97) after adjusting for the same variables in the initial model above. The magnitude of the decrease was 7% (OR = 0.93, 95% CI: 0.91, 0.95) and 6% (OR = 0.94, 95% CI: 0.91, 0.96) after accounting for GDM and PGDM, respectively. Compared to the initial model, adjusting for both GDM and PGDM reduced LGA odds by two percentage points and resulted in 8% overall decrease in the odds of LGA per decade (OR = 0.92, 95% CI: 0.90, 0.94).

The supplementary figure (Additional file 4: Fig. S3) shows that the crude rates of LGA exhibited an increase in the total Aboriginal population and a slight decrease in the non-Aboriginal population over the study period.

In this study, the annual age-standardised and crude rates of PGDM grew considerably (to 5.4 and 3.8%, respectively) in the last two decades among Aboriginal mothers but remained below 1% in non-Aboriginal mothers. This is consistent with the extant evidence-base—which confirms the relatively heavy burden of PGDM in Indigenous populations at national [5, 22, 23] and global [24‐26] levels—with the highest prevalence rates of PGDM reported among Aboriginal mothers in Central Australia (8.4%) [6], while PGDM complicated 6.3% of pregnancies among Arizona’s Pima population in the United States [27].

The higher prevalence of PGDM in Indigenous populations is a reflection of the elevated rates (and contrasting demographic characteristics) of T2DM [28]. T2DM accounts for more than 95% of PGDM cases among Australian Aboriginal mothers compared to approximately 55% in non-Aboriginal women [6, 29]. In 2015, the prevalence of T2DM (which is preventable) in pregnancy was approximately eight times higher in Western Australian Aboriginal mothers compared with their non-Aboriginal counterparts [29]. Disproportionately higher rates of T2DM in Aboriginal youth [30, 31] and the differentially higher incidence of T2DM among Aboriginal females [32, 33] contribute to the disparate burden of PGDM in this population.

Our observation of steeper increasing trends in Indigenous populations (relative to their respective general populations) is a feature of the global health disparities literature [34]. Hare et al., for example, recently reported a ten-fold increase in PGDM among Aboriginal mothers over 30 years in Australia’s Northern Territory, with a crude prevalence of 5.7% in 2016 [6]. They found that PGDM rates remained ≤0.7% through the study period in non-Aboriginal mothers. The reasons for the sharp growth in the burden of T2DM among Indigenous populations are complex and multifaceted—although at least in part reflect the rapid behavioural and environmental changes towards urban, sedentary lifestyle and energy-dense diets [13, 14] and the concomitant longer-term legacies of colonisation that have given rise to social and economic circumstances that shape the affordability and accessibility of nutritious foods, healthcare and other proximal determinants of chronic diseases. The steeper increase in PGDM in the Northern Territory study compared to ours may reflect demographic circumstances, whereby a larger proportion of Aboriginal people in the Northern Territory live in remote areas [35] where a change in lifestyle may be more impactful on life outcomes.

Our results showed that the annual age-standardised rates of GDM increased consistently in Aboriginal mothers and remained higher than that reported in their non-Aboriginal counterparts over the two decades to 2015. Annual crude rates of GDM were comparable in Aboriginal and non-Aboriginal mothers. The discrepancies in age structure between Aboriginal and non-Aboriginal mothers, and the established link between the risk of GDM and maternal age [36] indicate that age-standardised rates offer a more meaningful comparison. Crude rates underestimate the disparities between the two populations.

There is a worldwide increasing trends in the prevalence of GDM [37‐39] that is believed to be explained by an increase in maternal obesity and older maternal age at birth [40, 41]. Temporal increases in GDM have also been reported in Australian Aboriginal mothers [42]. The crude prevalence increased from 3.4 to 13% between 1987 and 2016 in the Northern Territory of Australia [6]. There has been evidence from population-level studies of both increasing and stable rates of GDM in Canadian First Nations mothers [34, 43].

Interpreting prevalence trends of GDM is not straightforward and should be considered in conjunction with the temporal changes in screening policies and diagnostic criteria. In 2013, the Australasian Diabetes in Pregnancy Society (ADIPS) adopted the 2010 recommendations of the International Association of Diabetes and Pregnancy Study Groups (Additional file 5: Table S2) [44]. The implementation of the new guidelines was associated with a substantial increase in the prevalence of GDM globally [45], and we expect a similar contribution to the increasing trends of GDM among both Aboriginal and non-Aboriginal pregnant women in Western Australia. Moreover, between and within jurisdiction discrepancies in the application of diagnostic and screening guidelines [46] may invalidate direct comparisons.

A retrospective audit in rural Western Australia has recently reported inadequate screening for GDM, especially among Aboriginal mothers [47]. Limitations related to the storage of blood samples prior to testing for GDM were also reported in rural and remote Western Australia [48]. Thus, our findings may have underestimated the real prevalence of GDM among the Aboriginal population who are more likely to reside and receive ante-natal care in these areas [29].

The present study reported that DIP may explain a large part of the increasing trend in LGA among the Aboriginal population. Statistically accounting for PGDM and GDM (separately and collectively) attenuated the LGA increase per decade. Similar results were recently reported among the Aboriginal people of Australia’s Northern Territory, but with a stronger impact of PGDM on LGA trends compared to ours, reflecting a sharper rise of PGDM rates among that Aboriginal population over time in the Northern Territory [6].

In line with the impact of DIP on the overall LGA trends among the Aboriginal population, absolute rates of LGA remained high in Aboriginal babies born to mothers with DIP. LGA continued to affect about one-third of Aboriginal births complicated by PGDM, with no signs of improvement over the study period. Despite the improving trends, LGA continued to impact more than 20% of Aboriginal births complicated by GDM.

The association between exposure to intrauterine hyperglycemia and elevated risk for future diabetes in offspring was reported in Indigenous populations in North America [49‐52]. Among Canadian Indigenous women, being born HBW was associated with differentially increased risk of T2DM and GDM, and is believed to play a role in the intergenerational transfer of diabetes risk [9, 53]. Although the present study did not investigate the intergenerational effect of DIP, our findings regarding the over time burden of DIP and its associated LGA are probably in-line with the DIP-driven intergenerational cycle of diabetes in Indigenous populations.

We reported smoking rates of approximately 40–50% among Aboriginal mothers with GDM and PGDM, with no improvement over the study period. Maternal exposure to tobacco smoke during pregnancy significantly increases the risk of T2DM [54, 55], obesity [56] and GDM [57] in offspring. Epigenetic modifications are believed to mediate these effects [58]. The sizeable rates of maternal smoking may thus add to the intergenerational impact of DIP in the Aboriginal population. Likewise, low birth weight may synergistically interact with genetic susceptibility to obesity to heighten the future risk of T2DM [59]. Given the substantially higher prevalence of many health risks in Aboriginal Australians [60], it is plausible that there may be similar interactions between DIP and health risk behaviours with potential intergenerational implications.

The present study also reported some positive findings. The rates of pre-eclampsia decreased in Aboriginal mothers with GDM and PGDM, and LGA decreased in Aboriginal pregnancies complicated by GDM. These positive trends can probably be attributed to improved antenatal care [61], increasing awareness about GDM following the Hyperglycemia and Adverse Pregnancy Outcomes study [62, 63], increased screening rates and the introduction of new diagnostic criteria [44]. These factors might have resulted in better glycemic control and capturing of the less severe cases of GDM.

The use of linked, population-level datasets from multiple sources is a strength of the present study. This eliminates selection bias; allows tracking of babies and mothers upon transfer between different levels of health facilities; and maximises the external validity of the study. Utilizing multiple sources of data also increases the sensitivity of capturing medical conditions by enabling the identification of additional cases. Further, we have used a robust, agreed algorithm for the identification of Aboriginal people in the study data sources [16]. This has been shown to reduce missing data and improve within-individual consistency across time, and is likely to have minimised any potential for under-estimation of increasing DIP trends.

Our study does, however, have some limitations. We used routinely collected data that are not originally designed for research purposes. The capture of DIP over time in the HMDC is impacted by changes in the Australian diabetes coding standards that implies modifications to the documentation of diabetes during a hospital admission [64]. However, we believe that using both MNS and HMDC has improved the capture of DIP. Absence of data on maternal weight is also a limitation of this study. Obesity, which is increasing over time in Australia [65], is linked to both DIP and HBW and can be a source of unmeasured temporal confounding that may explain part of the relationship between DIP and LGA [6]. Moreover, we did not have data on glycemic control biomarkers. Hyperglycemia during pregnancy is an established risk factor for adverse fetal outcomes, and its trends might have explained part of the trend in these outcomes.