figure a

Introduction

A diagnosis of gestational diabetes mellitus (GDM) provides a unique opportunity to identify future risk of cardiometabolic disease in young women long before any clinical manifestation [1]. Indeed, women with GDM have a higher risk than those without GDM of ultimately developing type 2 diabetes and cardiovascular disease (CVD) [1]. In common with non-gestational diabetes, GDM is associated with a propensity for hepatic fat deposition [2,3,4,5]. However, it is not known whether GDM is associated with the risk of serious liver-disease outcomes (such as cirrhosis, liver failure and transplantation) later in life. Thus, our objective was to evaluate the long-term risk of serious liver disease in women with a history of GDM and the potential role of subsequent progression to type 2 diabetes.

Methods

A population-based cohort study was conducted using healthcare administrative databases from the Ministry of Health and Long-Term Care of Ontario, Canada. These databases include discharge abstracts from all hospitalisations in the province, physician service claims for reimbursement for virtually all consultations, procedures and visits, and demographic data for all residents eligible for healthcare in Ontario. The Ontario Diabetes Database (ODD) is a validated registry of physician-diagnosed non-gestational diabetes that is derived from these data [6]. The MOMBABY database is derived from hospitalisation data, and links hospitalisation records of delivering mothers with their neonates. Individuals are linked across data sources through a unique and reproducibly-encrypted health-card number. The use of data in this study was authorised under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a Research Ethics Board.

The study population consisted of all women aged 15–54 years who had a live-birth delivery between April 1994 and March 2002. Individuals who died or emigrated prior to March 2002 were excluded. Women with pre-gestational diabetes (identified through ODD) were excluded, as were those who had liver-disease comorbidity encoded on their delivery hospitalisation records. For women with multiple eligible pregnancies, the most recent was selected. Baseline characteristics that were recorded for each woman were age at index delivery, socioeconomic status (ascertained from the neighbourhood household-income quintile), rurality of residence, hypertension and dyslipidaemia. Individuals with South Asian or Chinese ethnicity were identified, as previously described [7].

The presence or absence of GDM in the index pregnancy was ascertained from diagnostic codes associated with the delivery hospitalisation. Each group was further divided into those who did or did not develop incident diabetes during follow-up, based on postpartum entry into the ODD.

All women were followed for outcome events from April 2002 until March 2016; the available data did not permit outcome ascertainment prior to this window. The primary outcome was serious liver disease, defined as hospitalisation for cirrhosis, liver failure or liver transplantation. The crude incidence of serious liver disease during follow-up was ascertained for GDM and non-GDM groups, and for the four groups that were obtained by further stratification according to the development of diabetes. Baseline characteristics were compared using t tests for continuous variables and χ2 tests for categorical variables. Cox proportional hazards regression with left truncation to April 2002 was used for modelling of the independent association of GDM and/or subsequent diabetes with serious liver-disease outcomes. Data were censored for death, end of healthcare eligibility or end of follow-up. Models were constructed unadjusted or with adjustment for age, income, rurality, hypertension and dyslipidaemia. Because Asian ethnicity is associated with both GDM and liver-disease outcomes, a third model was constructed with further adjustment for ethnicity. Analyses were performed using SAS Enterprise Guide version 6.1 (Cary, NC, USA).

Results

The study population (N = 698,078) consisted of 17,932 women who had GDM and 680,146 who did not have GDM (Table 1). Consistent with known risk factors, women with GDM had a higher mean age than women without GDM, and were more likely to be of Chinese or South Asian ethnicity, more likely to be living in an urban setting, and more likely to have a low income (all p < 0.001). Diagnoses of hypertension and dyslipidaemia were also more common in women with GDM (both p < 0.001). Development of type 2 diabetes in the years after pregnancy was far more common in women with GDM than in those without GDM (47.8% vs 6.5%, p < 0.001).

Table 1 Characteristics of the study population according to the presence or absence of GDM
Full size table

Outcome ascertainment began in April 2002 (a mean of 3.6 years after delivery) and continued for 14.0 years. Women who had GDM had an elevated risk of serious liver disease compared with those without GDM (HR = 1.40, 95% CI 1.01, 1.94), although the absolute event rates were modest in both groups (15 and 11 events per 100,000 person-years, respectively). The groups were further stratified on the basis of incident diabetes. Among women with GDM, the event rate was higher in those who subsequently developed diabetes than in those who did not (19 and 12 events per 100,000 person-years, respectively). Similarly, among women who did not have GDM, the event rate was higher in those who developed diabetes than in those who did not (29 and 10 events per 100,000 person-years, respectively).

To explore the relationship between GDM and post-gestational diabetes in the development of serious liver disease, we determined the HRs (relative to a reference group of women with neither GDM nor type 2 diabetes) in (i) women with GDM who later developed type 2 diabetes, (ii) women with GDM who did not develop type 2 diabetes and (iii) women without GDM who later developed type 2 diabetes (Table 2). The risk of serious liver disease was elevated in both groups of women who went on to develop diabetes: HR = 1.88, 95% CI 1.23, 2.87, p = 0.004 in women with GDM; and HR = 2.76, 95% CI 2.35, 3.25, p < 0.0001 in women without GDM (model I). The risk was not elevated in women with GDM who did not develop diabetes (HR = 1.26, 95% CI 0.76, 2.09, p = 0.38). This pattern remained after adjustment for age, income, region of residence, hypertension and dyslipidaemia (model II) and after further adjustment for ethnicity (model III), as shown in Table 2.

Table 2 Cox proportional hazards regression modelling of the association of GDM and/or subsequent diabetes with serious liver-disease outcomes
Full size table

Discussion

Results from several studies have indicated that women with previous GDM have a higher prevalence of fatty liver than those without GDM [2,3,4,5, 8]. Liver fat identified by ultrasonography in early pregnancy can predict subsequent gestational dysglycaemia, suggesting that hepatic steatosis precedes the development of GDM [9]. In addition, pre-gravid serum levels of γ-glutamyl transferase can predict the risk of GDM in a subsequent pregnancy [10]. Altogether, these data suggest that women who develop GDM have a chronic predisposition to hepatic fat deposition that is present both before and after pregnancy. However, it is not known whether these women ultimately progress to the serious clinical outcomes of advanced liver disease.

The results of the current study extend our knowledge by demonstrating that GDM is associated with the risk for future development of serious liver outcomes. Although absolute event rates in our population were modest, these hepatic outcomes are very serious, and are associated with considerable morbidity and mortality. We identified a GDM-associated risk of serious outcomes over a 14.0 year median follow-up in a population with a mean age at baseline of 32.0 years, and it is possible that this risk will increase as the population progresses into middle age.

Among women with GDM, we only identified a significant risk of serious liver disease in the subgroup of those who progressed to post-gestational diabetes. Notably, in a similar population, significant HRs for microvascular pathologies requiring vitrectomy/photocoagulation or renal dialysis were only found to occur in the subgroup of women with GDM who developed type 2 diabetes, whereas significant HRs for macrovascular CVD outcomes occurred in women with GDM with or without subsequent development of diabetes [1]. The clinical implications of these relationships are that although the propensity for hepatic fat accumulation in women with a history of GDM suggests that there may be benefits to some degree of clinical monitoring of this group, efforts that focus on the prevention of post-gestational diabetes in this population could contribute to mitigation of the risk of serious liver-disease outcomes.

We cannot exclude the possibility that women with GDM who did not progress to type 2 diabetes during this study could still be at risk of developing serious liver disease in the future, particularly as the population was fairly young. In addition, it is possible that some women in our population may have developed liver disease in the time between delivery and the onset of outcome ascertainment in April 2002 (a mean of 3.6 years postpartum). However, given the infrequency of the outcome, very few women are likely to have been affected by this possibility. Another limitation of our study is that the administrative databases we used did not track some potentially important clinical risk factors, such as alcohol intake and body weight. Notably, obesity is a risk factor for GDM, type 2 diabetes and liver disease, and could therefore be an important factor in the relationships that we have described. Conversely, however, our use of population-based data made it possible to study the effects of both GDM and post-gestational diabetes in all parous women in the population, and to demonstrate the long-term risk of serious liver outcomes following GDM.