1 BACKGROUND

The prevalence of gestational diabetes mellitus (GDM) in New Zealand, based on regional reports, has increased from 1.3% in 2001¹ to 6.2% in 2019² largely due to the obesity epidemic¹ and changes in population demographics.³ Each year 3000 to 4000 women in New Zealand are diagnosed with GDM.¹ Compared with European/other women, whose prevalence of GDM is 6%,⁴ the prevalence is significantly higher for Indian (28%), Asian (17%), Pacific (19%), and Middle Eastern, Latin American, and African (14%) women.^{5, 6} It is also increased for Māori women (10%), which is an underestimate, largely due to a lack of screening.^{5, 6}

Women with GDM typically develop elevated glucose levels during the second trimester when women are unable to compensate for the increase in insulin resistance in the second half of pregnancy.⁷ In addition to the increased risk for maternal and neonatal complications during pregnancy and delivery,^{1, 8} these women have increased risks of subsequent beta cell dysfunction^{9, 10} and insulin resistance¹¹ and adverse cardiometabolic outcomes in the long term.

Compared with control women, there is a 10–20 fold increase in the incidence rate for developing type 2 diabetes in women with GDM.^{8, 12} Twelve large population studies of mostly European women show an increased risk for women with GDM developing cardiovascular disease (CVD) from Canada,^13-15 the United States,^16-19 Europe,^20-23 the United Kingdom,²⁴ and Israel.²⁵ A small number of studies have reported a positive association between GDM and renal disease,^26-29 and among women who also developed type 2 diabetes.^{13, 30, 31} Although the relationship between GDM and renal disease is not well understood, genetic studies show several loci are associated with down regulation of renal nuclear factor erythroid 2-related factor 2 and reduced glomerular filtration.³¹ Animal modeling also suggests that increasing advanced glycation end products bind to and reduce tubule cell function, as they filtrate through the nephrons.³² A meta-regression pooled analysis reported increased risks for women with GDM of developing hypertension and dyslipidemia and a risk ratio of 1.98 (95% confidence interval [CI] 1.57–2.50) for major cardiovascular events,³³ 1.98 (1.57–2.50) for fatal and nonfatal myocardial infarction, and 1.25 (1.07–1.48) for stroke,³⁴ and at an earlier age, compared with women who were not diagnosed with GDM during pregnancy.^{33, 34}

The aim of this study is to identify all women, in a multiethnic population, diagnosed with GDM in New Zealand between 1 January 2001, and 31 December 2010, and compare cardiometabolic (diabetes, cardiovascular, mortality, hypertension, and dyslipidemia) and renal outcomes in the 10 to 20 years following delivery with women who were not diagnosed with any diabetes before or during pregnancy (matched for age and year of delivery).

2 METHODS

A retrospective cohort study was carried out utilizing the National Maternity Collection (MAT) dataset that included all births in New Zealand between 1 January 2001 and 31 December 2010 (n = 604 398). MAT data are collected from primary maternity services, lead maternity carer claims for payment, and hospital records using the National Minimum Dataset (NMDS) for inpatients and day-patients during pregnancy, birth, and the postnatal period. Adolescent girls <15 years, women ≥50 years, women whose gender was incorrectly recorded and women with type 1 or type 2 diabetes prior to pregnancy were excluded from this study (Figure 1).

2.3 Statistical Analyses

SAS version 9.4 (SAS Institute, Cary, NC, 2013) was used for all statistical calculations. PROC FREQ was used to generate frequency and percentage for categorical variables. PROC MEANS was used to generate mean and SD for continuous variables. Chi square or t tests were used to compare women during their first GDM pregnancy with control women for all categorical and continuous baseline characteristics. PROC PHREG was utilized to compute Cox proportional HR and 95% CI for time-to-first event for the six outcomes for women with GDM and control women between 2001 and 2021. Time-dependent diabetes was included in the fully adjusted Cox model for other outcomes. The proportional hazards assumption of the Cox models was assessed using the graphical approach (Kaplan–Meier curves and Log–Log Survival curves) and indicate that this was not violated by exposed and unexposed women (Figures 2 and 3 and Appendix Figure A1). PROC LIFETEST was used to calculate Kaplan–Meier cumulative incidences for each cardiometabolic outcome.

3 RESULTS

Figure 1 outlines the number of total maternal birth events (N = 604 398) in New Zealand between 1 January 2001, and 31 December 2010. Women excluded from the study (n = 3232) included those under 15 or ≥50 years of age, incorrect gender records and those identified with type 1 or 2 diabetes (utilizing dispensed glycemic medications) prior to pregnancy. Of the total birth events, 18 793 were identified with mothers diagnosed with GDM. After excluding duplicate mothers, 11 459 women were included in the study and 11 447 were able to be matched with control women, who were not diagnosed with GDM during their pregnancy.

Women with GDM compared with control women were significantly more likely to be Asian or Pacific women, economically deprived and reside in urban areas (Table 1) and particularly in central (18%) and South (19%) Auckland compared with 12% of control women—data not shown. Women were also more likely to be diagnosed with GDM if they were nonresidents, not registered with a lead maternal carer, experienced a first or ≥4 previous births, or develop preeclampsia compared with control women (Table 1). Women with GDM were also more likely to have been dispensed at least one cardiovascular-related medication within the 12 months prior to delivery and have a shorter gestation at delivery compared with control women. Infants of mothers who were diagnosed with GDM were more likely to be premature at birth than infants of control mothers (Table 1).

Table 2 confirms the lifetime risk for women diagnosed with GDM developing diabetes after controlling for ethnicity, in addition to matching for age and delivery year, HR (95% CI): 20.06 (18.46–21.79); (Model 1). Using the same model, women with GDM had over double the risk of being admitted to hospital with a first cardiovascular event 2.19 (1.86–2.58), over six times more likely to be admitted for renal disease 6.34 (5.35–7.51), an increased risk of premature death 1.55 (1.31–1.83), and increased risks of being dispensed at least two antihypertensive 2.77 (2.66, 2.88) or lipid modifying 5.61 (5.31–5.94) medications after delivery. The HR (95% CI) were similar for each outcome after controlling for all significant covariates maternal deprivation status, Regional Health Authority region, resident status, unregistered with a lead maternity carer, number of previous births, preeclampsia, smoking, cardiovascular medications dispensed 1 year prior to delivery, and premature birth (Model 2). The HR (95% CI) were: diabetes 18.89 (17.36–20.56); cardiovascular events 1.79 (1.52–2.12); renal disease 5.42 (4.55–6.45), and all-cause mortality 1.44 (1.21–1.70). The effect of high infant birthweight (>5000 g), using (Model 2) had very little effect on HR for all primary outcomes.

However, when time-dependent development of diabetes was added to the model (Model 3—all variables in model 2 and time-dependent diabetes), HRs were greatly reduced compared to Model 2, indicating that about half of the increased risk for these outcomes was related to a diagnosis of diabetes after pregnancy. The HR for all outcomes remained significant (p < .05) except for all-cause mortality (p = .06).

Figure 2 outlines the cumulative proportions of women diagnosed with GDM who develop diabetes, are hospitalized with a first cardiovascular event and/or renal disease and all-cause mortality were all significantly higher compared with control women after matching for age and year of birth. A linear increased risk with time since delivery was found for all outcomes throughout the 20-year follow-up period. However, the risk of a first cardiovascular event plateaued after an average 15-year of follow-up-period and the increased risk for all-cause mortality occurred from 12 years postdelivery. The risk was almost exponential for renal disease from about 14 years postdelivery.

Figure 3 shows the cumulative proportion of women with GDM compared with control women who were dispensed hypertensive and lipid modifying medications over 20 years following delivery. A similarly large increased risk was found for developing hypertension and dyslipidemia for women diagnosed with GDM, and this continued to increase throughout the 20-year follow-up period, compared with control women.

Table 3 reports the baseline characteristics (ethnicity, socio-economic status, number of previous births and preeclampsia) that significantly modify the association between GDM and developing diabetes following delivery, after controlling for all variables in the adjusted model (p value for interaction <.05). In the subgroup analyses by baseline characteristics, European/other women with GDM had the highest hazard of developing diabetes following delivery (HR = 37.45, 95% CI = 31.79–44.12), followed by Māori women 16.44 (14.16–19.09) compared with their respective controls. Less economically deprived women with GDM (HR = 32.65, 95% CI = 23.40–45.54), primipara women with GDM (HR = 31.64, 95% CI = 25.93–38.61) and women with GDM who did not develop preeclampsia were more likely to develop diabetes (HR = 19.92, 95% CI = 18.19–21.81) than women who also developed preeclampsia (HR = 12.48, 95% CI =9.87–15.78) compared with their respective controls. Subgroup analyses for other primary outcomes (renal disease, CVD, and mortality) also showed similar adjusted HR. Although p values for assessing interaction were close to .05, indicting significance is likely to be a chance finding. The latter subgroup analyses have not been included due to small numbers.

4 DISCUSSION

A total of 604 398 hospital birth events in New Zealand were identified between 2001 and 2010 and closely correlates with total births in New Zealand over that period.³ The 11 459 women diagnosed with GDM in this cohort was less than the estimated number (n = 15 400), based on the expected prevalence of 1.3% in 2001 and an annual increase of 13.9%.¹ This underestimation is likely due to a lack of screening for GDM, particularly for Māori women.^{5, 6}

Women diagnosed with GDM had a 20-fold increase in developing diabetes, over double the risk for having a first cardiovascular event, a 6-fold increased risk for admission to hospital for renal disease, a 1.5-fold increased risk of a premature death, over twice the risk for developing hypertension and a 5-fold risk of developing dyslipidemia compared with control women. The increased risk for all outcomes in women with GDM remained, after controlling for the development of diabetes following delivery but not for all-cause mortality (Table 2). This suggests that the increased risk of adverse outcomes in GDM women is partly due to developing diabetes after their initial GDM pregnancy, but that other factors besides diabetes are also increasing the risk of adverse outcomes. The finding for cardiovascular events was similar to that separately reported for the Ontario study for women with GDM who developed type 2 diabetes HR (95% CI), 2.82 (2.41–3.30) compared with women with GDM who did not develop type 2 diabetes, 1.30 (1.07–1.59).¹³ Current results for renal disease were similar to four previous studies among American²⁷ and Black American²⁹ women; Israeli women, whose risk increased with each additional GDM pregnancy²⁶; and aboriginal women.²⁸ Similar results were reported for women with GDM who developed diabetes following delivery in Demark,³⁰ Sweden,³¹ and Ontario for women requiring dialysis 7.52 (5.24–10.81) but contrasted with their finding for women who did not develop type 2 diabetes 1.25 (0.62–2.52).¹³ Findings from this NZ study also show that women who develop GDM are more likely to develop hypertension and dyslipidemia following delivery (also reported from the large Danish cohort study),²² both of which are associated with increasing insulin resistance, metabolic syndrome,³⁹ type 2 diabetes, and an increased risk of cardiovascular events.³³

Significant interactions were found for the risk of developing diabetes, mainly between ethnic groups (Table 3). The relevance of this is unclear and could be partly due to variation in risk among the comparison group for European/other women and the risk ratio, as the risk difference between GDM and control women is larger among Māori and Pacific women. Further studies are required to validate findings of the effects of interaction for ethnicity, socioeconomic status, number of previous births, and preeclampsia¹⁷ on modifying the risk of developing diabetes. Subgroup analysis showed European/other women with GDM have a higher risk of being diagnosed with developing diabetes followed by Māori and Asian women whereas Pacific women had the lowest risk compared with women without GDM (Table 3). It is known that rates of diabetes in the community are higher for Pacific, Māori, and Asian women,⁴⁰ so this finding may reflect a lower uptake of long-term screening for diabetes and other CV risk factors in these higher risk populations.^{5, 41} European women who develop GDM may also be more insulin resistant and at higher risk of developing type 2 diabetes compared with European women without GDM.

Two recent reviews reported a higher risk for non-European women with GDM developing type 2 diabetes when comparing populations,^{8, 12} and one review reported a lower risk for Asian women developing metabolic syndrome.³⁹ There are few studies that examine the risk of developing cardiovascular and renal disease for non-European women with a history of GDM.

NZ best practice guidelines recommend all women who develop GDM are funded for an annual diabetes screen following delivery.^{42, 43} Findings from this study highlight the increased risk of cardiovascular and renal disease, and the contributing role of developing diabetes following delivery.⁴⁴ Postpartum screening and managing cardiometabolic risk factors for the 6% of women diagnosed with GDM each year¹ could be funded through the annual diabetes and cardiovascular review and provide an opportunity to support effective lifestyle changes and improve pharmaceutical management to reduce adverse outcomes.^{12, 45} Based on numbers of people with diabetes on the NZ virtual register,³⁷ extending funding for this at-risk group of women will only increase the total proportion of eligible people in New Zealand with diabetes by 1%–1.5%.

Despite recommendations, there is a paucity of reports on postpartum screening for women diagnosed with GDM internationally,^{12, 24} and in New Zealand,^{1, 44} with no recommendations for assessment of blood pressure, lipids, smoking status, and physical activity recommendations are minimal.^{1, 42} Of the few reports on postpartum screening, most show low rates and poor records for cardiovascular risk factors.^{12, 24} Several barriers to screening have been identified, including a lack of knowledge in general practice and guidance on how to systematically screen for type 2 diabetes and communicate the increased risk to women.^{6, 46} Despite these limitations, women do prefer to know of their increased risk⁴⁷ and are receptive to prompts for screening, provided they are contextualized for the target population.⁴⁶

Breastfeeding,⁴⁵ lifestyle changes, and metformin for women with GDM, who remain hyperglycemic, are recommended in the postpartum period to reduce the risk of developing type 2 diabetes.^{1, 42} A recent meta-analysis combining 13 lifestyle trials between 4 and 36 months duration computed a risk ratio (0.76 [95% CI: 0.63–0.93]).⁴⁸ A pooled analysis of 16 trials also showed a modest mean weight reduction in these women following delivery (1.8 kg [95% CI: 2.9, 0.6]).⁴⁹ A further challenge for managing and supporting younger women with a history of GDM is the lack of large long-term lifestyle and pharmaceutical trials, and the difficulty in recruiting mothers with new babies,⁵⁰ to establish a core risk management guideline to reduce adverse outcomes.⁵¹ Although the NZ guidelines, since 2013, recommend all women diagnosed with GDM are screened for diabetes at 3 months following delivery with an HbA_1c and then annually,¹ postpartum screening is low^{1, 4} and significant differences remain between ethnic groups and regions.⁴⁴ Although some centers do routinely recommend annual screening for CVD, current guidelines¹ require updating to reflect this, and promote evidence-based lifestyle and pharmaceutical interventions (in addition to metformin) suitable for at-risk women to improve outcomes.

Main limitations of this study include missing data for some key variables (including body mass index and tobacco use), a lower than ideal positive predictive value and sensitivity using the VDR algorithm,³⁶ a high threshold for diagnosis of GDM, a low 50 g polycose sensitivity,¹ regional variations in screening,³⁵ and lower screening rates for Māori women and women residing in regional areas. It is estimated that approximately 4000 women with GDM were not identified in this dataset, accounting for the lower-than-expected number of women with GDM. This limitation increases the potential for women with undiagnosed GDM (about 10%) or women not dispensed with antihyperglycemic agents for type 2 diabetes being selected as control women and reduces the effect size. The underestimation of postpartum cardiometabolic outcomes are expected to be exaggerated for Māori and Pacific women. Although there is a paucity of reports for Māori women diagnosed with GDM developing type 2 diabetes, one report showed that >56% of Māori women diagnosed with GDM developed prediabetes or diabetes postdelivery during an average 3.9 year follow-up and at a mean age of 36 years.⁵² Screening for GDM is notably lower for Māori women,^4-6 and they have increased uptake of the polycose test and are more likely to have a false negative result.⁴¹ This is because Māori and Pacific women are more likely to be diagnosed with GDM on the fasting 75 g OGTT result.⁵³ In New Zealand, we currently have a higher fasting threshold for diagnosis than other countries, so this likely further contributed to the lower than expected rates of GDM in these women.⁴ Postpartum screening rates for Māori and Pacific women are also less than for European and Asian women⁴⁴ and Māori people are less likely to collect dispensed hypoglycemic medication.⁵⁴

The low postpartum smoking prevalence of <5% was far less than the 27% of women with GDM who smoked during pregnancy when initially registered at National Women's Hospital in central Auckland.⁴ The under-reporting of smoking, which is associated with the development of GDM, type 2 diabetes, renal disease, and CVD⁵⁵ could not be completely controlled for in our analyses and is likely to have resulted in overestimates of HR for each of these outcomes, although our sensitivity analyses suggest this may not have been major.

This report attempts to quantify major long-term cardiometabolic events in women diagnosed with GDM that are poorly documented and understood.³⁴ This large national study of an ethnically diverse population of women with a history of GDM showed a 20-fold increased risk of developing diabetes, twice the risk of major cardiovascular events, increased mortality, and a five-fold increase of developing renal disease. Results also quantify the likely contribution of developing cardiovascular and renal disease for women who also developed diabetes following delivery. At-risk women require annual cardiometabolic reviews to screen for, and effectively manage major cardiovascular risk factors (HbA_1c, hypertension, dyslipidemia, and smoking cessation) and urinary microalbuminuria. Further research among ethnically diverse groups of women is required to establish effective cardioprotective lifestyle and pharmaceutical interventions for this younger group of at-risk women.

AUTHOR CONTRIBUTIONS

Barbara M. Daly and Robert K. R. Scragg conceptualized and designed the study and interpreted the results. Barbara M. Daly acquisitioned the data, conducted aspects of the data analysis, and wrote the manuscript. Robert K. R. Scragg developed the method for the data analyses and contributed to writing the manuscript. Zhenqiang Wu developed the method for the data analyses, analyzed the data, interpreted the results and contributed to writing the manuscript. Krishnarajah Nirantharakumar, Lynne Chepulis, and Janet A. Rowan critically appraised the study findings and reviewed and contributed to the manuscript. Barbara M. Daly, Robert K. R. Scragg, and Zhenqiang Wu had full access to the data and take responsibility for the integrity of the data and the accuracy of the data analysis.

FUNDING INFORMATION

This study was funded by the University of Auckland 71604/9450.

CONFLICT OF INTEREST STATEMENT

There are no competing interests relevant to this study.