The association of parity/live birth number with incident type 2 diabetes among women: over 15 years of follow-up in The Tehran Lipid and Glucose Study

Diabetes mellitus (DM) is a leading cause of mortality and morbidity in the world, with a greater burden in low- and middle-income countries [1]. In 2017, the global prevalence number of DM reached 476.0 million, with a 129.7% increase from 1990 [1]. In addition to prevalence, the trends of incidence rate, death rate, and disability-adjusted life-years (DALYs) were also increasing for DM, especially for type 2 DM (T2DM) [1]. The age-standardized prevalence of T2DM among adult residents of Tehran was 13.5% in 2008–2011 [2], and about 1% of them developed T2DM annually [3].

In addition to known DM risk factors such as obesity, genetic susceptibility, unhealthy lifestyle, and environmental risk factors [4], it was suggested that childbearing could associate with T2DM development among mothers [5]. Pregnancy, as one of the most important events during women's life, involves marked alterations in cardiometabolic parameters, including weight gain, increased plasma glucose, insulin resistance, and dyslipidemia. Although these changes are physiological, which are beneficial to both mother and fetus, they may have an impact on the risk of diabetes and other cardiometabolic diseases in future life [5‐8]. In a meta-analysis of prospective studies on this issue, the risk of T2DM was increased by 9% for each birth [5]. Similarly, in another meta-analysis, each additional birth increased the risk by 6% [9]. Moreover, in both meta-analyses, compared to those with lower parity, higher parity (at least 4 live births) was associated with about 40% higher risk for incident T2DM [5, 9]. It should be noted that in both of these meta-analyses, the authors found significant heterogeneities (all I² > 70%) among previous cohort studies conducted in Europe, East Asia, and the US [5, 9]. The authors recommended further investigations with further adjustment for the potential confounders, especially among uninvestigated populations [5, 9], considering that the impact of parity number on incident T2DM is not same in different ethnicity [10]. To the best of our knowledge, there is a lack of information on this issue in the Middle East and North Africa (MENA) region, with a high burden of DM [11]. The current study aims at determining whether the parity/live birth number is an independent risk factor for T2DM development among Iranian women aged 30–65 years. We used collected data from the oldest cohort of the MENA region, the Tehran Lipid and Glucose Study (TLGS).

The study population consisted of 2552 female participants with a mean age of 44.50 (SD: 9.47) years. Baseline characteristics according to the number of parity are presented in Table 1. Generally, by increasing in the number of parity, cardiometabolic risk profiles became worse among continuous variables except for HDL-C. Hence higher parity was associated with older age, higher BMI and WC, increased BP, and higher levels of FPG, TG, and TG/HDL-C. Among categorical variables, women with one parity had a higher prevalence of current smoking but a lower prevalence of having macrosomia baby and taking antihypertensive/lipid-lowering medications. Moreover, higher educated participants were less likely to have a higher parity number.

From baseline phase until phase VI (2015–2018), during a median follow-up of 15.4 years (IQR: 10.7–16.4), 557 incident T2DM cases have occurred. Multivariable HRs and 95% CIs of the association of number of parity and live birth with incident T2DM are shown in Table 2 and Additional file 2: Table S1, respectively. Each additional parity and live birth was associated with 10% and 11% higher risk of incident T2DM in the age-adjusted model 1, respectively (P values < 0.001); the higher risk remained significant until model 3. However, after further adjustment for BMI and WC in model 4, the HR of each additional parity was attenuated significantly and reached 1.05 (95% CI 1.00–1.11; P value: 0.075). A similar attenuation was also occurred for the HR of each additional live birth, although it remained significant even in model 5 (1.06 [1.00–1.12; P value: 0.049]). Compared to women with parity/live birth number of one, those with parity/live birth number of 3 and ≥ 4 were significantly at higher risk of incident T2DM in models 1–3. After adjustment for general and central obesity indices in model 4, having ≥ 4 parity/live birth was associated with higher risk; even after further adjustment for TG/HDL-C in model 5, the risk remained marginally significant for parity ≥ 4 (1.64 [1.00–2.70; P value: 0.051]). Importantly, trends of the HRs across parity/live birth categories were also significant in model 1–4 and marginally significant in model 5 (P value: 0.055 for parity and 0.060 for live birth).

As a sensitivity analysis, to address the effect of changing BMI during follow-up on the association between parity/live birth number and incident T2DM, we rerun the Cox regression by further adjustment for BMI change during follow-up (Additional file 3: Table S2). BMI change was defined as: \(\frac{{{\text{Secondary measurement}} - {\text{Baseline measurement}}}}{{{\text{Baseline measurement }} \times {\text{Time interval }}\left( {{\text{year}}} \right)}} \times 100\); secondary measurement was performed at the first follow-up after recruitment, before the individual developed T2DM or censured. The time interval was calculated as the difference of the second measurement year with the baseline recruitment. Due to the limited sample size and decrease in the number of outcomes (154 participants with incident T2DM were excluded due to the lack of secondary BMI measurement before outcome), generally, the P values became non-significant; however, it should be noted that even after adjustment for BMI change, the effect sizes were similar to the model 5 of the main analysis. Moreover, compared to those with parity number of one, parity ≥ 4 had a HR of 1.69 [0.94–3.01; P value: 0.078], and the trend of the HRs across parity categories was marginally significant (P value: 0.064).

As another sensitivity analysis, we reexamined the impact of parity/live birth on incident T2DM during follow-up of less than 10 years. Accordingly, during a median follow-up of 9.4 years (IQR: 8.6–10.4), 323 incident T2DM cases have occurred until phase IV (2008–2011). In model 1–3, each additional parity and live birth caused a higher risk of incident T2DM, and also those with parity/live birth number of ≥ 4 were at higher risk; however, after adjustment for obesity indices in model 4, the higher risk became non-significant (Table 3 and Additional file 4: Table S3).

This is the first population-based cohort study conducted in the MENA region that examined the effect of the parity/live birth number on incident T2DM. During 15.4 years of follow-up, after adjustment for a wide series of important T2DM risk factors, including age, education level, family history of diabetes, SBP, DBP, using anti-hypertensive medications, and reproductive factors, a higher parity/live birth number was associated with higher risk of T2DM development. After further adjustment for obesity indices and TG/HDL-C, the higher risk was attenuated; however, the trends of the HRs across parity/live birth categories were marginally significant, and those with parity ≥ 4 had more than 60% higher risk for T2DM development. Moreover, each additional live birth was associated with 6% significant higher risk for incident T2DM.

Our findings were in line with two previous meta-analysis studies [5, 9]. Li et al., in a meta-analysis of the cohort, cross-sectional, and case–control studies found a nonlinear association between parity and the risk of T2DM in which the combined multivariable RR for T2DM development was 1.06 (95% CI: 1.02–1.09) per live birth, with high heterogeneity (I² = 87.2%) [5]. Also, in their sensitivity analysis for prospective studies only, the RR reached 1.09 [5]. Similarly, in another mate-analysis of cohort studies by Guo et al., each unit increase in parity number had the combined RR of 1.06 (95% CI 1.02–1.09); however, a significant high heterogeneity was shown between included studies (I² = 84.3%) [9]. Moreover, in both meta-analyses, compared to those with lower parity (reference = 0 or 1), participants of prospective studies who had higher parity (at least 4 live births) showed about 40% higher risk for T2DM development [5, 9]. Based on data of 126,721 middle-aged women from eight cohort studies, a U-shaped association was found between the number of children and the risk of DM, in which women with 2 live births had the lowest risk [16]. A similar U-shaped relation was also observed among Danish women older than 33 years of age [17] and Canadian women aged 18–50 years [10].

In the coronary artery risk development in young adults (CARDIA) Study, it was shown that childbearing did not increase the incidence of diabetes as long as women never developed gestational diabetes mellitus (GDM). In other words, the association of childbearing with incident DM can be explained through GDM [18]. On the other hand, among Canadian women, an increased risk of DM with increasing parity was observed, even after adjustment for GDM [10]. Similarly, among Danish women without a history of GDM, a relationship between parity and incident DM was also found [17]. In the current study, even after adjustment for macrosomia, as a surrogate of GDM [19], the association of parity/live birth number and incident T2DM remained significant.

During a 5-year follow-up of Japanese women, Nanri et al. found that the association of parity with T2DM development risk was significantly attenuated after adjustment for BMI. They suggested that a higher risk of T2DM development due to increased parity might be partly explained by weight retention after pregnancy [20]. In our short-term follow-up of less than 10 years, we also found similar results that all P values became non-significant after adjustment for obesity indices in model 4. However, in our long-term follow-up (main analysis), similar to a cohort study among American women [21], although the effect sizes for higher parity/live birth numbers were attenuated prominently in model 4, but remained significant for parity/live birth number of ≥ 4. Furthermore, when we considered BMI change during follow-up as another covariate, generally, the effect sizes did not change, and the P value for the trend of HRs across categories of parity number remained marginally significant.

In addition to weight retention, another explanation for the association of the parity/live birth number with incident T2DM in women can be the insulin resistance (IR) pathway. During a normal pregnancy, the target tissues (liver, muscle, or adipose tissue) of the mother become increasingly insensitive to insulin, mainly due to hormone production by the placenta that antagonizes insulin [22]. This IR may progress to GDM or might be overcome by a sufficient increase in insulin production by pancreatic beta cells [22]; however, this process can lead to an extra burden on β-cell function and affect insulin secretion, even after delivery [9, 23]. In addition to GDM, a well-established risk factor for developing T2DM in the future [24], it is also known that mothers with milder degrees of dysglycemia during pregnancy (i.e., less severe than GDM) were at higher risk for developing prediabetes and T2DM in the future [23, 25]. Moreover, in full-multivariable analysis, Kramer et al., found that insulin sensitivity (as assessed by Matsuda index) decreased during the first 3 years postpartum, even among women who had had a normal glucose-tolerant pregnancy [23]. Moreover, the expected reduction in the physical activity and some cardiometabolic changes during pregnancy (weight gain and dyslipidemia) can exacerbate this IR pathway [22]. By repeating pregnancy, exposure time to IR condition increases and can lead to a higher chance of T2DM development. In the current study, despite the lack of data on insulin measurement, after further adjustment for TG/HDL-C (surrogate of IR [26, 27]), each additional live birth and parity ≥ 4 were generally associated with higher risk. It means that other factors might also have roles in the association of childbearing and T2DM development, especially psycho-socioeconomic factors.

The present study has several strengths, including standardized measurements of traditional risk factors and using oral glucose tolerance test for outcome definition, rather than relying on self-reported data. Moreover, besides the well-known risk factors of T2DM, we also considered BMI change as a strong covariate in our analysis, which was not addressed in previous studies of this field. The current study also has several weaknesses. First, during follow-up, possible changes in risk factors, as well as the parity/live birth number, were not considered. Second, we were not able to show the impact of nulliparity due to the limited number of participants with no live birth. Third, we did not have access to data on participant diet, job status, income, and psychiatric history, which are the potential representative factors of the socioeconomic and lifestyle status. They might affect the association between the parity/live birth number and incident T2DM. Fourth, at the baseline of the current study, only 18 participants reported a positive history of GDM that was too low to be considered as a covariate. Moreover, at the time of recruitment of this study (1999–2001) and before, there was no routine screening program for GDM like oral glucose tolerance test (OGTT) during prenatal care in Iran. Therefore, most of the GDM cases were missed at that time. However, in the current study, we adjusted our models for macrosomia, the main surrogate of GDM. Finally, this study was conducted among residents of Tehran city, and its findings may not be extendable to rural populations.

In this population-based observational cohort study, during a long-term follow-up, after adjustment for potential T2DM risk factors, reproductive factors, obesity indices, and TG/HDL-C (IR surrogate), we demonstrated that higher parity/live birth numbers could be associated with increased risk of T2DM development. Moreover, even after further adjustment for BMI change, the suggestive higher risk was still found. We suggested that weight retention can have a key role between parity/live birth number and incident T2DM, and weight management should be more considered after delivery. Moreover, we recommended that the potential role of psycho-socioeconomic factors should be investigated in future studies.