Prevalence of diabetes in pregnancy among Indigenous women in Australia, Canada, New Zealand, and the USA: a systematic review and meta-analysis

Background Indigenous peoples in countries with similar colonial histories have disproportionate burdens of disease compared with non-Indigenous peoples. We aimed to systematically identify and collate studies describing the prevalence of pre-existing diabetes and gestational diabetes, and compare the prevalence of these conditions between Indigenous and non-Indigenous pregnant women in Australia, Canada, New Zealand, and the USA.


Introduction
Diabetes during pregnancy, including both pre-existing diabetes (type 1 or type 2 diabetes) and gestational diabetes, are associated with adverse pregnancy outcomes, 1-3 poor maternal outcomes (caesarean section, preeclampsia, postpartum haemorrhage, and gestational hypertension), and poor perinatal outcomes (macrosomia, preterm birth, a large-for-gestational age birth, congenital anomalies, stillbirth, injury during birth, neonatal death, and neonatal admission to an intensive care unit). [3][4][5] Gestational diabetes is associated with an increased risk of type 2 diabetes in the mother and an increased risk of early-onset type 2 diabetes in the child. [6][7][8][9] The risk of having diabetes during pregnancy varies across different populations, with a prevalence of 2-5% in low-risk ethnic groups. 10 One population of individuals who are at a high risk for developing diabetes during pregnancy are Indigenous women. 11,12 A previous systematic review 13 of the prevalence and effect of diabetes in pregnancy among Indigenous women from multiple countries found that the prevalence of diabetes during pregnancy was not the same for all Indigenous peoples. 65% of studies included in the review 13 estimated that the prevalence of diabetes was higher in Indigenous women than in non-Indigenous women. However, this review 13 did not consider pre-existing diabetes and gestational diabetes separately, nor did it focus on Indigenous women from countries with a similar history of colonialism.
Colonialism has been identified as a crucial determinant of the health of Indigenous peoples. 14 The effects of colonialism include systematic discrimination, poverty, racism, and a rapid transition to a so-called western lifestyle. Indigenous women in Australia, Canada, New Zealand, and the USA have similar experiences of colonialism, and health outcomes have been compared in

Research in context
Evidence before this study Before this study, we searched MEDLINE and Google Scholar on Feb 20, 2018 to identify systematic reviews on the prevalence of pre-existing diabetes and gestational diabetes that had been done previously. Three previous systematic reviews have provided important data on this topic. The first study assessed the prevalence of diabetes in pregnancy and the maternal and infant outcomes of diabetes in pregnancy in Indigenous women from multiple countries. In the same review, the authors combined the prevalence of diabetes in pregnancy into one estimate, with a subcategory for gestational diabetes, and comparison groups consisted of either non-Indigenous women or national estimates for non-Indigenous women. To our knowledge, no meta-analysis comparing the prevalence of diabetes in pregnancy in Indigenous women with that in non-Indigenous women has been done. The second study examined the prevalence of gestational diabetes among Indigenous women in Australia. Although a meta-analysis was done, the prevalence of pre-existing diabetes was not analysed, and an analysis of the prevalence of diabetes during pregnancy in Indigenous women from countries other than Australia was not done. The third study examined the evidence for early screening for diabetes in Indigenous pregnant women in Australia, Canada, New Zealand, and the USA. This study summarised a large amount of data but it did not systematically compare the prevalence of pre-existing diabetes and gestational diabetes between Indigenous and non-Indigenous women. Our systematic review and meta-analysis focused on the comparing the prevalence of pre-existing diabetes and gestational diabetes in Indigenous women with that in non-Indigenous women in Australia, Canada, New Zealand, and the USA. These four countries are all industrialised and share a similar history of colonialism.

Added value of this study
We collated data on the prevalence of pre-existing diabetes and gestational diabetes and did a meta-analysis to compare the prevalence of these conditions in Indigenous women and non-Indigenous women from four countries with similar histories of colonialism. The pooled prevalence odds ratio (POR) of pre-existing diabetes mellitus indicated that the prevalence of this condition was higher in Indigenous women than in non-Indigenous women in Australia, Canada, and in the USA. Heterogeneity between studies done in Australia and Canada was high, but when examining the magnitude and direction of the PORs from individual studies, there was an association between indigeneity and pre-existing diabetes during pregnancy. We were not able to derive a POR of pre-existing diabetes in Maori women in New Zealand because none of the studies included this information. The pooled POR of gestational diabetes suggested that the prevalence of this condition was higher in Indigenous women than in non-Indigenous women, indicating an association between indigeneity and gestational diabetes. In the meta-analysis of the prevalence of gestational diabetes across all included studies, high heterogeneity was observed between studies in Australia, Canada, and USA. However, the magnitude and direction of the PORs from individual studies indicated that there was an association between indigeneity and gestational diabetes during pregnancy. We estimated the PORs of pre-existing and gestational diabetes in studies grouped by screening criteria, diagnostic criteria, risk of bias assessment, and by the race or ethnicity of the comparison group; however, none of these subgroup analyses were able to meaningfully reduce the degree of heterogeneity between studies. The results of our study indicate that the prevalence of pre-existing diabetes and gestational diabetes is higher in Indigenous women than in non-Indigenous women in four different countries with similar histories of colonialism.

Implications of all the available evidence
Future research should focus on targeting important risk factors to prevent pre-existing diabetes and gestational diabetes in Indigenous women, and this research should be done in partnership with Indigenous peoples and communities. Given the higher prevalence of diabetes in Indigenous women than in non-Indigenous women during pregnancy in Australia, Canada, New Zealand and the USA, policy makers in these countries should consider implementing policies that aim to reduce this disparity between Indigenous and non-Indigenous pregnant women in terms of the prevalence and outcomes of maternal diabetes. Health practitioners need to be aware of the increased likelihood of having pre-existing diabetes and gestational diabetes in Indigenous pregnant women, so that they can better support these women before pregnancy, prenatally and postnatally.
www.thelancet.com/lancetgh Vol 8 May 2020 e683 previous systematic reviews. 12,15 The purpose of this systematic review is to compare the prevalence of preexisting diabetes and gestational diabetes between Indigenous women and non-Indigenous women in countries with similar colonial histories: Australia (Aboriginal and Torres Strait Islander), Canada (First Nations, Inuit, and Métis), New Zealand (Maori), and the USA (Native American and Alaska Natives). To our knowledge, this systematic review is the first to systematically identify and collate studies that have compared the prevalence of pre-existing diabetes and gestational diabetes in Indigenous women with that in non-Indigenous women in these four countries.

Search strategy and selection criteria
We did this systematic review and meta-analysis in accordance with PRISMA guidelines. 16 An information specialist (SC) did a comprehensive literature search in June 25, 2019, for relevant articles published from inception to June 25, 2019, in the following databases: Ovid MEDLINE, Ovid Embase, Ovid Global Health, CINAHL (EBSCO), Scopus, ProQuest Dissertations and Theses Global, PROSPERO, and the Wiley Cochrane Library. Controlled vocabulary terms (ie, Medical Subject Heading terms) and key words for Indigenous peoples, diabetes in pregnancy, and the included countries (Australia, Canada, New Zealand, and the USA) were used in the searches (see appendix 1 pp 1-8 for a list of the search terms used). 17 BV and SC also searched the grey literature for relevant studies using the Bielefeld Academic Search Engine and Google Scholar. Lastly, BV manually searched the reference lists of relevant articles. No limits on publication type, language of publication, and year of publication were applied. The detailed search strategies are available in appendix 1 (pp [1][2][3][4][5][6][7][8]. Studies were included if they were epidemiological observational studies comparing the prevalence estimates of either pre-existing diabetes (type 1 or type 2 diabetes, or both) or gestational diabetes among Indigenous women in Australia, Canada, New Zealand, or the USA with the prevalence of the same condition in a group of non-Indigenous women in the same country. We excluded non-primary research articles, letters to the editor, case reports or case series, studies that only provided combined estimates of diabetes in pregnancy by grouping preexisting diabetes and gestational diabetes together, studies in which participants were not Indigenous women, studies in which the comparison group did not include non-Indigenous women, and studies in which Indigenous women from countries other than Australia, Canada, New Zealand, and the USA were included.
Two independent reviewers (BV and SR or BV and OS) screened all titles and abstracts identified from the database search. The full-texts of studies identified as potentially relevant were reviewed in duplicate by two independent investigators (BV and SR or BV and OS).
Disagreements about study selection were resolved by discussion, and if consensus could not be reached, a third reviewer (MBO or DE) made the final decision.

Data extraction and quality assessment
Two independent reviewers (BV and SR or BV and OS) assessed the risk of bias of all included studies, and any disagreements were resolved by discussion until a consensus was reached. Cohort studies were assessed by use of the Newcastle-Ottawa Quality Assessment Scale for cohort studies. 18 This scale assesses three main sources of bias, including the selection of study cohorts, comparability of cohorts, and outcome assessment. Studies were classified as having a low risk of bias if they had a score of 3-4 for study selection, 1-2 for comparability, and 2-3 for outcome assessment; studies were classified as having an unclear risk of bias if they had a score of 2-3 for study selection, 1-2 for comparability, and 1-2 for outcome assessment; and studies were classified as having a high risk of bias if they had a score of 0-1 for study selection, a score of 0 for comparability, and a score of 0-1 for outcome assessment. 18 We assessed the risk of bias of cross-sectional studies using a nine-item scale, developed by Hoy and colleagues, 19 that included the domains of sample selection, non-response bias, data collection, and measurement reliability and validity. A low risk of bias was defined as a high risk classification in 0-3 categories, an unclear risk of bias was defined as a high risk classification in 4-6 categories, and a high risk of bias was defined as a high risk classification in 7-9 categories. 19 One independent reviewer (BV) extracted data using a standardised data collection form, and data from the published reports were verified by a second reviewer (OS) for accuracy (appendix 2 pp 1-2). From all inc luded studies, we extracted information about study design, setting, Indigenous and comparison groups, data sources for exposure and the definition of outcomes, study sample size, screening guidelines and the diagnostic criteria for pre-existing diabetes and gestational diabetes, and data on the number of Indigenous and non-Indigenous women with pre-existing diabetes and gestational diabetes, and on the total number of Indigenous and non-Indigenous women, to estimate the prevalence of pre-existing diabetes and gestational diabetes.

Data analysis
We used a Mantel-Haenszel random-effects metaanalysis to combine data from individual studies comparing the prevalence of pre-existing diabetes and gestational diabetes between groups of Indigenous and non-Indigenous women and to calculate crude prevalence odds ratios (PORs). 20 Because of the unique identities and singularities of Indigenous peoples around the world, only studies done in the same country were pooled so that a pan-Indigenous analysis approach was avoided during data synthesis. The summary measures were the pooled unadjusted PORs with 95% CIs. Statistical heterogeneity across the studies was assessed by use of the I² statistic, with an I² of less than 25% indicating low heterogeneity, an I² of 26-74% indicating moderate heterogeneity, and an I² of greater than 75% indicating high heterogeneity. 21 The summary measures and results for statistical heterogeneity are reported separately for each group of Indigenous peoples. Subgroup analyses were done by study design, screening guide lines, diagnostic criteria, and risk of bias. 22 A narrative synthesis of results was done to describe study populations and to explore both clinical and methodo logical heterogeneity across the included studies. Funnel plots were used to assess publication bias. 23 We used Review Manager (version 5.3) to do the meta-analysis, assess heterogeneity across the studies, and generate the funnel plots. 24 This study is registered with PROSPERO, number CRD42018095971.

Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results
Electronic database searches identified 1294 titles and abstracts. The grey literature search and the review of relevant reference lists identified a further 54 titles and abstracts. After removal of duplicates, 809 articles were screened by title and abstract, resulting in 175 potential articles to be included. After reviewing the full texts of these articles, 43 unique studies were included in the systematic review and 39 studies were included in the meta-analysis (figure 1).
Of all 43 unique studies identified, 25 Studies involving populations of Indigenous women in Australia identified women as either Indigenous Australian [25][26][27][28]33,[35][36][37][38][39][40][41]65 or as Aboriginal and Torres Straight Islanders. 30,31,34,43 Studies in Canada identified Indigenous women as First Nation women 8,[45][46][47][48][49]66 or as a group comprising First Nation, Métis, and Inuit women. 44 The study 50 in New Zealand identified women as Maori, and studies done in the USA identified Indigenous women as either Native American 29,51,54-58,61-63 or Native American and Alaska Native (including Yupik and Inupiat). 52,53,59,60,62,64 Most studies involving Indigenous women in Australia included a comparison group of non-Indigenous women, but some studies also included non-Indigenous women sepa rated into subgroups of Australasian, 26,34,42 European, 26,43 Caucasian, 27,37,41 African women, 42 Asian (usually grouped by region), 26,27,34,37,41,42 Pacific Islanders, 26,34 and women from the Americas. 34 Studies involving Indigenous women in Canada included comparison groups of non-Indigenous women identified as either non-First Nation women, 8,[45][46][47][48] non-Indigenous women, 44,66 or as a group of Caucasian, south east Asian and Chinese women. 49 The study 50 in New Zealand included a comparison group of non-Indigenous Pacific Islander and European women. Finally, studies involving Indigenous women in the USA included comparison groups of non-Indigenous women    45,49,66 although Métis and non-status Indigenous peoples (ie, those who are not eligible to be registered with the federal government) could not be identi fied from health insurance records. For the assess ment of outcomes, studies done in Australia used national perinatal data sources, 30,32 state or region data, 25,27,28,31,[33][34][35]40,41,65 or clinic or hospital data. 26,[36][37][38][39]42 Studies done in the USA used national data sources, 53 state-level data, 29,51,52,[54][55][56][57][58]61 or hospital data. 63 Studies done in Canada used databases at the provincial, 8,45-47,49,66 regional, 48 or hospital levels. 44 The risk of bias assessment indicated that 75% of studies had an unclear or low risk of bias (figure 2). 16  Cohort studies received the lowest risk of bias scores in the following categories: representativeness of the exposed cohorts, selection of the non-exposed cohorts, assessment of the outcome, length of follow-up, and adequacy of follow-up. 18 Cohort studies received high risk of bias scores in the following categories: ascertainment of exposure, demonstration that the outcome of interest was not present before the study, and comparability of cohorts. 18 All cross-sectional studies 44,51,52 had an unclear risk of bias. Cross-sectional studies received high risk of bias scores in the following categories: whether the study instrument showed validity or reliability, whether the same mode of data collection was used for all study participants, and whether the study population was a close representation of the national population. 19 The pooled POR of pre-existing diabetes in Indigenous women compared with non-Indigenous women in Australia was 3·63 (95% CI 2·35-5·62), and heterogeneity across studies was high (I²=98%, p<0·00001; figure 3). The PORs of pre-existing diabetes in individual studies done in Australia ranged from 1·40 (1·28-1·52) to 8·24 (5·35-12·69). In Canada, the pooled POR All studies apart from those by Moses and colleagues (prospective cohort), 26 Dyck and colleagues (cross-sectional), 44 Hunsberger and colleagues (cross-sectional), 52 and Hegwood and colleagues (cross-sectional) 51 were retrospective cohort studies. The GCT was a non-fasted test and the OGTT was a fasted test. GD=gestational diabetes. GCT=glucose challenge test. NR=not reported. OGTT=oral glucose challenge test. FBG=fasting blood glucose. *Includes mothers and babies reports from 1996-2010 and 2012-2017. values ranging from 1·40 (95% CI 1·28-1·52) to 8·24 (5·35-12·69). The pooled POR estimates from the metaanalysis of studies done in Australia and Canada should be interpreted with caution because of the high heterogeneity between studies. However, the direction and magnitude of individual study PORs indicated that there was an association between indigeneity and preexisting diabetes. No obvious publication bias was evident from the funnel plots (appendix 3 pp 6-7). The pooled POR of gestational diabetes in Indigenous women compared with non-Indigenous women in Australia was 1·42 (95% CI 1·24-1·63), with high heterogeneity observed between studies (I²=97%, p<0·00001; figure 3). The PORs of gestational diabetes in individual studies done in Australia ranged from 0·44 (0·14-1·39) to 6·67 (5·11-8·72). The pooled POR of

Figure 3: Mantel-Haenszel random-effects meta-analysis results showing the pooled prevalence of (A) pre-existing diabetes and (B) gestational diabetes in observational studies comparing Aboriginal or Torres Strait Islander women to non-Indigenous women in Australia
Events are defined as the number of participants in the study with pre-existing diabetes or gestational diabetes, and n is the total number of participants in the study. AIHW=Australian Institute of Health and Welfare. df=degrees of freedom. Bower et al (1992) 65 Moses et al (1994) 26 Markey et al (1996) 41 Yue et al (1996) 37 DeCosta and Child (1996) 38 Powell and Dugdale (1999) 39 Stone et al (2002) 35 Ishak et al (2003) 28 Simmons et al (2005) 36 Sharpe et al (2005) 27 AIHW et al (2010) 30 Zhang et al (2010) 40 Teh et al (2011) 42 Porter et al (2011) 25 Chamberlain et al (2014) 32 Thrift and Callaway (2014) 33 Abouzeid et al (2015) 34 New South Wales Government (2017) 31 Krikham et al (2017) 43

Indigenous
Events n

Non-Indigenous W eight
Events n

Indigenous
Events n

Indigenous
Events n

Non-Indigenous W eight
Events n

Odds ratio
Random 95% CI

Discussion
This systematic review and meta-analysis provided a comprehensive assessment of the prevalence of diabetes in pregnancy among Indigenous women in Australia, Canada, New Zealand, and the USA. We found a large variation in the prevalence estimates of pre-existing diabetes and gestational diabetes among Indigenous women in these four countries. Predominantly, the odds of having pre-existing diabetes or gestational diabetes are greater in Indigenous pregnant women than in non-Indigenous pregnant women. The crude POR estimates of pre-existing diabetes and gestational diabetes in individual studies showed that both types of diabetes appear to be strongly associated with Indigenous ancestry in pregnant women, but the magnitude of this association varies across studies and by type of diabetes. The association between indigeneity and the prevalence of pre-existing diabetes was consistently stronger than the association between indigeneity and the prevalence of gestational diabetes. Pre-existing diabetes primarily consists of type 1 and type 2 diabetes, diagnosed before pregnancy, and overt diabetes, identified at the beginning of pregnancy. 67,68 The higher prevalence of pre-existing diabetes among Indigenous women compared with non-Indigenous women is likely to be because there are a higher number of Indigenous women with type 2 diabetes than non-Indigenous women. Although this analysis was not done in our study, it is likely to be the case because the prevalence of type 2 diabetes has been reported to be higher among Indigenous peoples than in non-Indigenous peoples. 69 The results of our Events are defined as the number of participants in the study with pre-existing diabetes or gestational diabetes, and n is the total number of participants in the study. df=degrees of freedom. Moum et al (2004) 54 Fridman et al (2014) 29 Anderson et al (2016) 54 Ralls et al (2007) 55 Chu et al (2009) 62 Caughey et al (2010) 63 Kim et al (2012) 57 Cabacungan et al (2012) 56 Kim et al (2013) 58 Fridman et al (2014) 29 Anderson et al (2016) 53 Pearson et al (2016) 59 Singh et al (2018) 60 Dennis et al (2019)   systematic review and meta-analysis are similar to those of other reviews done in this field. 11,13 One previous systematic review 11 that assessed the prevalence of gestational diabetes among Indigenous women in Australia found substantial heterogeneity in the prevalence estimates across the included studies. Compared with previous studies, 11-13 our review adds a systematic comparison between the prevalence of preexisting diabetes and gestational diabetes in Indigenous women and non-Indigenous women from four countries (Australia, Canada, New Zealand and the USA) with similar histories of colonialism. The Indigenous people's experience of colonialism is an important factor contributing to loss of identity, language, culture, the rapid transition to industrialisation, as well as ongoing oppression and discrimination. [70][71][72] Increased exposure to unfavourable social determinants of health (ie, income, education, and employment) among Indigenous peoples is also a consequence of colonialism. 14 Social inequalities and colonialism have been recognised as important contributors for the development of diabetes in Indigenous peoples. 73 These factors also create barriers to the effective management of diabetes in Indigenous peoples (eg, increasing physical activity, consuming balanced meals, monitoring blood sugar concentrations, and injecting insulin). 73,74 Alternatively, the high prevalence of diabetes among Indigenous peoples has been proposed to be because of genetic susceptibility, but research has yet to find a strong causative gene. 69 Clinical care and public health interventions designed for Indigenous women to prevent diabetes in pregnancy and to reduce the negative effects of this condition should be led by Indigenous women at the community level. Incorporating the culture, traditional practices, and the world views of Indigenous peoples in the design of prevention strategies at the community level will ensure local needs are met. Care for Indigenous women also needs to be equitable to ensure that social inequalities are not preventing them from achieving the same health status as non-Indigenous women. Two previous reviews 75,76 suggest that the results are promising when Indigenous self-determination is incorporated into primary care and maternal child health programmes. The prevalence of pre-existing diabetes is greater in Indigenous women than in non-Indigenous women, therefore, early screening of Indigenous women during pregnancy should be prioritised to identify previously undiagnosed type 1 or type 2 diabetes. Clinicians should be aware that Indigenous women are more likely to have pre-existing diabetes and gestational diabetes in pregnancy, and they should adjust their practice accordingly when treating Indigenous women.
Future research should focus on both prevention of diabetes in pregnancy and adapting clinical care so that it is more meaningful and effective for Indigenous women. Research on innovative solutions to improve social inequalities should be done in partnership with Indigenous communities. Evaluations of adapted care for Indigenous women should focus on improving outcomes and decreasing the effect of pre-existing diabetes and gestational diabetes on their health and the health of their child. Continued surveillance of the prevalence of gestational diabetes and pre-existing diabetes among Indigenous women, by either healthcare systems, governments, or Indigenous communities, is necessary; but future studies should clearly report the screening and diagnostic criteria that are used to diagnose both types of diabetes so that the interpretation of study results can be improved.
The strengths of our systematic review and metaanalysis include the use of a rigorous methodological approach that followed a pre-defined protocol, which was registered before the beginning of the study. We used validated search terms to do a comprehensive database search, 17 and we searched the grey literature and relevant reference lists to reduce selection bias by decreasing the chance of missing potentially relevant studies. The risk of bias assessment and data extraction were done in duplicate to minimise assessor bias.
However, there are several limitations of our study. Included studies were from a wide range of time periods, over which the screening guidelines and diagnostic criteria for pre-existing diabetes and gestational diabetes had changed. The screening and diagnostic criteria for preexisting diabetes and gestational diabetes, and the characteristics of comparison groups, were not always reported in the included studies, which prevented meaningful subgroup analyses. Other reasons for the high heterogeneity between studies could be because of underlying differences in study participants. One could argue that the high heterogeneity across the included studies might make the pooled POR estimates less useful; however, high heterogeneity could also indicate that there is a wide variation in the prevalence of pre-existing diabetes and gestational diabetes across a range of different Indigenous groups. Finally, no formal evaluation of the effect of colonialism and the social determinants of health among Indigenous and non-Indigenous women in this study was done, therefore, a causal association cannot be inferred.
To the best of our knowledge, there are no cohort studies done in New Zealand that have compared the prevalence of pre-existing diabetes in Indigenous women with that in non-Indigenous women. In Canada, available data from cross-sectional studies 77 have shown that 4·0% of all cases of diabetes among Inuit people are gestational diabetes, and among Métis people, 2·2% of all cases of diabetes are gestational diabetes. The generalisability of our results to other Indigenous peoples should be interpreted with caution, until such a time when those studies have been produced, so that pan-Indigenous interpretations are avoided.
In conclusion, Indigenous women in Australia, Canada, New Zealand, and the USA have an increased susceptibility to having pre-existing diabetes and gestational diabetes during pregnancy. Diabetes during pregnancy results in poor birth and delivery outcomes, with serious long-term effects on the health of the mother and the child. System-wide and structural interventions to address the increased prevalence of preexisting diabetes and gestational diabetes in Indigenous women in Australia, Canada, New Zealand and the USA should be considered.

Contributors
BV, DE, and MBO contributed to the design and conceptualisation of the study. BV, SR, OS, DE, and MBO contributed to the screening of studies for inclusion and data extraction. SC searched the databases. BV, MBO, DT, and SC contributed to the analysis and interpretation of the data. All authors contributed to the writing or editing and reviewing of the manuscript.

Declaration of interests
We declare no competing interests.