Lifetime Duration of Breastfeeding and Cardiovascular Risk in Women With Type 2 Diabetes or a History of Gestational Diabetes: Findings From Two Large Prospective Cohorts

Cardiovascular disease (CVD) is the leading cause of morbidity and mortality among people with type 2 diabetes and accounts for >75% of hospitalizations and >50% of all deaths (1). The American Heart Association embraces primordial prevention to avoid the development of CVD risk factors, as treating manifest clinical risk factors is not equivalent to avoiding them because the risk of CVD remains elevated even with optimal treatment (2). Type 2 diabetes is associated with up to 50% higher relative CVD risk in women compared with men (1,3). In addition to type 2 diabetes, the onset or recognition of diabetes for the first time in pregnancy (gestational diabetes mellitus [GDM]) confers higher risk for CVD (4). GDM affects 5–10% of pregnancies in the U.S. alone (5) and is a cardiovascular risk factor even among women who do not develop type 2 diabetes (4). Thus, it is important to identify early modifiable determinants of CVD risk specific for these high-risk women.

Breastfeeding provides short-term and long-term health benefits to women and children. In mothers, breastfeeding is associated with decreased risks of breast and ovarian cancers (6,7), type 2 diabetes (8), CVD (9), and all-cause and cause-specific mortality (10). Exclusive breastfeeding is recommended by the World Health Organization (WHO) for the first 6 months of life and continued breastfeeding for at least first 2 years of life, with complementary foods being introduced at 6 months postpartum (11). However, <50% of postpartum women worldwide breastfeed according to WHO recommendations.

Whether breastfeeding as a modifiable risk factor is associated with lower CVD risk in the high-risk population of women with type 2 diabetes or GDM history has not been investigated so far. To address these knowledge gaps, we assessed the association of lifetime duration of breastfeeding and CVD risk among women with diabetes in two large U.S. cohorts, the Nurses’ Health Study (NHS) and the Nurses’ Health Study II (NHS II). Additionally, we examined the relationship of total and exclusive breastfeeding duration with CVD risk among women with prior GDM in NHS II.

The analysis was conducted in two ongoing prospective cohort studies. The NHS was initiated in 1976 and included 121,701 female nurses aged 30–55 years. The NHS II was established in 1989 recruiting 116,429 female U.S. nurses aged 25–42 years. In both cohorts, mailed questionnaires were administered biennially to assess reproductive and lifestyle factors and health status, with follow-up rates >90%. Study procedures have been approved by the Institutional Review Boards of the Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health (Protocol number: 2009-P-002375), and those of participating registries as required.

For analyses among women with type 2 diabetes, we included parous women who reported physician-diagnosed type 2 diabetes on any questionnaire between 1986 (retrospective breastfeeding report across all pregnancies) and 2016 in the NHS and between 1989 and 2017 in the NHS II. Type 2 diabetes could be diagnosed before or after pregnancies but had to be diagnosed before the CVD event. GDM was defined as glucose intolerance with onset or first recognition during pregnancy. The population for analysis among women with prior GDM comprised NHS II participants who reported a history of GDM in 1989 or were diagnosed with GDM during the follow-up pregnancies. After exclusion of nulliparous participants, participants with prevalent CVD or cancer at baseline, without breastfeeding data, missing date of birth, who never returned follow-up questionnaires, the analytical samples comprised 9,606 NHS and 5,540 NHS II nurses with type 2 diabetes, and 4,537 NHS II nurses with prior GDM. Type 2 diabetes subsequently developed in 505 women (11%) with prior GDM and available breastfeeding data, and stratified analyses according to interim diabetes status were performed among women with prior GDM.

A supplementary questionnaire was mailed to women who reported physician-diagnosed diabetes on a baseline or any biennial questionnaire. A validation study demonstrated a high level of confirmation (98%) of self-reported type 2 diabetes (12). In accordance with the National Diabetes Data Group (13), diagnosed cases were required to meet one of the following criteria: 1) an elevated glucose concentration (fasting plasma glucose ≥7.8 mmol/L, random plasma glucose ≥11.1 mmol/L or plasma glucose ≥11.1 mmol/L after an oral glucose load) and at least one symptom related to diabetes (excessive thirst, polyuria, weight loss, or hunger); 2) in the absence of symptoms, at least two elevated glucose concentrations on different occasions; or 3) treatment with insulin or oral hypoglycemic medication. For cases of type 2 diabetes identified after 1998, the revised American Diabetes Association criteria were applied using the fasting glucose cutoff of 7.0 mmol/L (14).

Women in the NHS II were asked to report the physician diagnosis of GDM on each biennial questionnaire up to 2001 or through a 2009 pregnancy questionnaire. A previous validation study in this cohort suggested a high level of screening testing for GDM (83% reported undergoing a 50-g 1-h glucose challenge test) and high concordance between self-reported and physician-validated GDM diagnoses (94%) (15). GDM history was not assessed in the NHS cohort.

In the NHS, breastfeeding history was assessed once, in 1986, when most of the women had completed their reproductive life span and were asked to report the total lifetime duration of breastfeeding cumulatively for all pregnancies as a categorical variable: cannot remember, did not breastfeed, <1, 1–3, 4–6, 7–11, 12–17, 18–23, 24–35, 36–47, or ≥48 months. Participants in the NHS II reported their breastfeeding duration in three follow-up questionnaires. In 1993, 1997 and 2003, when most women had completed their reproductive life span, women reported their breastfeeding duration as the same categorical variables as NHS women. To allow comparison with the NHS data, in NHS II we calculated the cumulative breastfeeding duration by summing the breastfeeding duration after each birth that the participants reported any breastfeeding for each survey cycle. Using the 1997 and 2003 NHS II questionnaires, we were also able to calculate duration of exclusive breastfeeding based on the reported timing of the introduction of formula or solid food. In the NHS, exclusive breastfeeding could not be ascertained. Because the sample size in some of the categories was small, we redefined the categories of cumulative breastfeeding in both cohorts as follows: 0 months, 1–6 months, 7–18 months, and >18 months prior to analyses. Categories of lifetime total duration of exclusive breastfeeding were defined as 0 months, 1–6 months, 7–12 months, and >12 months. Previous studies have demonstrated that both self-reported breastfeeding initiation and duration are highly reliable (16,17).

The primary outcome was major CVD, defined as a combined end point of nonfatal or fatal myocardial infarction (MI), stroke, coronary artery bypass graft (CABG) surgery, or percutaneous coronary intervention (PCI). The following secondary outcomes were assessed: total coronary heart disease (CHD), comprising fatal, nonfatal MI, and CABG surgery; and total stroke, including fatal and nonfatal stroke cases (ischemic, hemorrhagic, and undetermined subtypes). When a participant (or family members of deceased participants) reported an incident event, permission was obtained to examine their medical records by physicians who were blinded to the participant risk-factor status. For each end point, the month and year of diagnosis were recorded as the diagnosis date. Nonfatal events were confirmed through review of medical records. MI was confirmed according to the WHO criteria on the basis of symptoms and diagnostic electrocardiogram changes or elevated cardiac enzymes. Strokes were confirmed according to the National Survey of Stroke criteria as a neurological deficit with sudden or rapid onset that persisted for >24 h or until death. Deaths were identified by reports of families, the U.S. postal system, or using death certificates obtained from state vital statistics departments and the National Death Index and confirmed through review of medical records or autopsy reports. Follow-up for deaths was >98% complete (18).

Information on potential risk factors, including medical, demographic, and reproductive histories, lifestyle practices, and body weight was collected and updated through NHS and NHS II biennial questionnaires. The validity of these assessments has been documented previously (19,20). Parity was defined as the number of pregnancies lasting >6 months and updated through follow-up questionnaires. As a surrogate prepregnancy BMI measure, BMI at age 18 years was calculated as self-reported weight (kg) at age 18 years divided by the square of height (m²). Socioeconomic status was captured by a score including nine variables: median household income, median home value, percentage with a college degree, percentage non-Hispanic White, percentage non-Hispanic Black, percentage of foreign-born residents, percentage of families receiving interest or dividends, percentage of occupied housing units, and percentage unemployed (21).

Descriptive analyses were conducted for baseline characteristics according to breastfeeding duration among parous women by standardizing to participants’ age distribution. We applied time-dependent Cox proportional hazards regression models (22) to estimate the adjusted hazard ratios (aHRs) and 95% CIs for the associations between lifetime total breastfeeding duration and risk of CVD in both cohorts, jointly stratified by age in years at the start of follow-up and calendar years of the current questionnaire cycle that were equivalent due to the way we structured the data and formulated the Cox models. Because of the similarity of study design and population, we also pooled the HRs in both cohorts using an inverse variance-weighted meta-analysis with the fixed-effects model to maximize statistical power (23). Tests for linear trends were evaluated using the Wald test on the continuous breastfeeding duration representing the midpoint values of each category. In addition, we examined the associations between breastfeeding duration as a continuous variable (per 6 months of breastfeeding) and risk of CVD.

Person-time was calculated from the return date of the questionnaire in which the type 2 diabetes or GDM diagnosis was first reported until CVD onset, death, or end of the follow-up, whichever occurred first. End of the follow-up was June 2016 for NHS and June 2017 for NHS II participants. Multivariable Cox models were initially adjusted for age (strata). Model 2 was additionally adjusted for race/ethnicity (non-Hispanic White or other), age at first birth (12–24, 25–29, 30–34, ≥35 years), prepregnancy BMI (≤25, 26–30, >30 kg/m²), parental history of CVD <60 years of age (yes/no), and time-varying variables of smoking status (ever/never smoker), physical activity, and alcohol intake. Time-varying variables with missing values at a given follow-up cycle were replaced with the information from the most recent cycle (mostly <5%); otherwise, we used the missing covariate indicator to deal with missing values.

Stratified analyses were performed to assess the effect modification by family history of CVD, smoking status, parity (one or two or more) and menopausal status (postmenopausal vs. premenopausal) across women’s reproductive life span. Analyses among women with GDM were additionally stratified by interim type 2 diabetes. Stratification by parity was not possible in NHS, since only few women remained primiparous and very few events occurred in that group. Multiplicative interaction was estimated using cross-product terms between dichotomous breastfeeding duration (<12 vs. ≥12 months) and the stratification variables based on the Wald test.

We conducted several sensitivity analyses: 1) including angina in the composite CHD and CVD end points, 2) restricting CHD and CVD outcomes to the hard clinical end points MI and stroke, and 3) taking into account potential confounding by socioeconomic status, menopausal status, and BMI change between 18 years and the return of the questionnaire. In the cohort with prior GDM, we modeled the associations between exclusive breastfeeding (i.e., before introduction of the formula) and CVD in addition to the analyses with total breastfeeding duration.

All data were analyzed using SAS 9.4 for UNIX (SAS Institute), and a two-sided P value of <0.05 was considered statistically significant in all analyses.

The baseline age-standardized characteristics of 15,146 parous NHS and NHS II participants with type 2 diabetes and 4,537 NHS II participants with prior GDM according to breastfeeding duration are shown in Table 1. In total, 6,339 (41.9%), 3,265 (21.6%), 3,197 (21.1%), and 2,345 (15.5%) women with diabetes breastfed for a cumulative total of 0, 1–6, 7–18, and ≥18 months, respectively. Among women with prior GDM, the corresponding distribution for breastfeeding duration was 17.6%, 17.7%, 34.1%, and 30.5%. In both cohorts, the number of live births and physical activity tended to increase, and the prevalence of smoking tended to decrease with a longer duration of breastfeeding. A total of 1,159 CVD events (7.7%) were documented among women with type 2 diabetes during 188,874.4 person-years of follow-up. Among 4,537 NHS II participants with a GDM history, 132 (2.9%) experienced at least one CVD event during 100,217.9 person-years of follow-up.

A longer lifetime duration of breastfeeding was associated with lower CVD risk in women with type 2 diabetes (Table 2). Compared with women who never breastfed, women cumulatively breastfeeding >18 months had 32% lower CVD and 38% lower CHD risks in the pooled cohorts. When breastfeeding duration was treated as a continuous variable per 6-month increase, the association remained: pooled adjusted aHRs were 0.94 (95% CI 0.91–0.98) for CVD and 0.93 (95% CI 0.88–0.97) for CHD. No association with stroke was observed in either of the cohorts.

Including angina in the composite CVD and CHD outcomes slightly attenuated the associations; however, they remained statistically significant: pooled aHRs were 0.77 (95% CI 0.64–0.92; P-trend = 0.009) for CVD and 0.76 (0.61–0.94; P-trend = 0.02) for CHD when comparing extreme categories of breastfeeding duration (Supplementary Table 1). When only the hard clinical end points of MI and stroke were considered in the composite CVD outcome, the associations remained robust: pooled aHRs were 0.94 (95% CI 0.91–0.98) for CVD and 0.93 (0.88–0.98) for MI per 6-month increase in breastfeeding duration (Supplementary Table 2).

Finally, adjustment for the time-varying variables of neighborhood socioeconomic status score (21), menopausal status, and BMI change since prepregnancy virtually did not attenuate the associations between breastfeeding and CVD and CHD risks in the pooled cohort (Supplementary Table 3). BMI change since prepregnancy until the return of the last questionnaire did not appreciably differ across the categories of lifetime breastfeeding duration (data not shown).

In NHS II, women additionally provided information on exclusive breastfeeding. Characteristics of NHS II participants according to quartiles of lifetime exclusive breastfeeding are presented in Supplementary Table 4.

Compared with parous women who never breastfed, those with a cumulative breastfeeding duration >18 months had a 51% lower risk of incident CVD (aHR 0.49 [95% CI 0.28–0.86], P-trend = 0.02 across quintiles of cumulative breastfeeding) and a 62% lower risk of CHD (aHR 0.38 [95% CI 0.20–0.71], P-trend = 0.02) (Fig. 1A and Supplementary Table 5). The analysis for stroke (n = 43 cases) showed an aHR of 0.90 (95% CI 0.74–1.08) per additional 6 months in total duration of breastfeeding, and an aHR of 0.78 (95% CI 0.23–2.67) when comparing the extreme categories of breastfeeding duration (P-trend = 0.25; data not shown).

Longer duration of exclusive breastfeeding versus no breastfeeding was associated with an even lower risk of CVD and CHD among women with GDM (aHR 0.79 [95% CI 0.67–0.94] for incident CVD and aHR 0.72 [95% CI 0.57–0.89] for incident CHD per 6-month increase in exclusive breastfeeding) (Fig. 1B and Supplementary Table 5). Similar to analyses with total duration of breastfeeding, no significant association was observed between the duration of exclusive breastfeeding and stroke (data not shown).

Breastfeeding and exclusive breastfeeding were robustly associated with lower CVD and CHD risks after inclusion of less rigorous clinical events (angina) (Supplementary Table 6), or after restriction of composite CVD outcome to hard clinical end points MI (n = 44) and stroke (n = 43) (Supplementary Table 7). History of breastfeeding and exclusive breastfeeding was associated with lower MI risk among women with GDM history (aHR 0.81 [95% CI 0.67–0.98] per 6-month increase in breastfeeding and aHR 0.71 [95% CI 0.51–0.99] per each 6 months of exclusive breastfeeding) (Supplementary Table 7).

Among women with type 2 diabetes, we did not observe any statistically significant effect modification by age at baseline, age at first birth, prepregnancy BMI, smoking status, family history of CVD, and physical activity (P-interaction >0.05 for all). Among women with GDM, the only statistically significant effect modification was observed for menopausal status, and the results indicated that breastfeeding ≥12 months was associated with lower CVD risk only in the premenopausal group (aHR 0.36 [95% CI 0.19–0.71] vs. aHR 1.03 [95% CI 0.61–1.74] in the postmenopausal group, P-interaction = 0.04) (Fig. 2 and Supplementary Table 8). We performed a priori defined stratified analyses by smoking status (ever/never), family history of CVD, and menopausal status (postmenopausal/premenopausal) (Fig. 2 and Supplementary Table 8). Analyses among women with GDM were additionally stratified by interim type 2 diabetes status and parity (primipara vs. two or more live births). Although we did not find a statistically significant interaction with family history of CVD, primiparity, or interim diagnosis of type 2 diabetes among women with GDM history, significant associations were observed only in women who did not progress to overt type 2 diabetes (aHR 0.63 [95% CI 0.40–0.99]), without family history of CVD (aHR 0.44 [95% CI 0.25–0.75]), and those with two or more live births (aHR 0.66 [95% CI 0.43–0.99]), but these analyses had low statistical power. Pooled analyses by parity among women with type 2 diabetes were not possible, because only a few women in the NHS cohort remained primiparous and subsequently developed a CVD event.

In two large prospective cohorts of women at high risk for CVD involving 15,146 women with type 2 diabetes and 4,537 women with a GDM history, we found that longer lifetime duration of breastfeeding was inversely associated with CVD risk, independent of other cardiovascular risk factors, including BMI, family history of CVD, physical activity, alcohol intake, and socioeconomic and smoking status. Among women with prior GDM, this inverse association persisted even in women who did not develop interim type 2 diabetes, and an even stronger inverse association was observed between duration of exclusive breastfeeding and CVD risk.

The observed inverse association between breastfeeding duration and CVD risk among women with type 2 diabetes and GDM is biologically plausible and supported by previous experimental and human studies in low-risk settings. In previous studies, lifetime duration of breastfeeding was inversely associated with MI and CVD in the general population of parous women (9,24) and with incident type 2 diabetes in the background population (8,25) and among women with GDM (26). Other studies also showed an inverse relationship between breastfeeding and subclinical atherosclerosis, including aortic and coronary calcification (27,28).

On the one hand, breastfeeding favorably influences carbohydrate and lipid metabolism by mobilizing stored fat, reestablishing glucose homeostasis (29,30), and improving insulin secretion through increased β-cell proliferation and reduced oxidative stress in β-cells (30). In population-based human studies, breastfeeding is associated with improved glucose metabolism and pancreatic β-cell function (26,31), more favorable lipid and inflammatory profiles (32,33), and greater weight loss and lower risk of metabolic disease, although data linking breastfeeding with long-term weight change in human populations are mixed (34). In our study, we did not observe any substantial differences in long-term weight change between the categories of total lifetime breastfeeding duration or any meaningful attenuation of the association between breastfeeding duration and CVD risk after accounting for BMI change during the follow-up. The significant interaction with menopausal status, however, indicates that among women with prior GDM, the association between longer duration of breastfeeding and lower risk of CVD is stronger among premenopausal women. Thus, breastfeeding may be particularly beneficial for the prevention of early-onset (premenopausal) CVD in this population.

Another possible mechanism through which breastfeeding may contribute to lower CVD risk is oxytocin, a crucial hormone for ejecting breast milk. Oxytocin has recently been demonstrated to exert several beneficial effects on the cardiovascular system, including blood pressure–lowering effects, vasodilatation, glucose-lowering actions, antioxidant effects, inhibition of inflammation, and lowering of fat mass (35,36). Breastfeeding may also modify cardiovascular risk through changes in stress response, with breastfeeding being associated with lower autonomic responses to stressors (37).

We observed an increasing trend in breastfeeding prevalence and decreasing trends in parity across the follow-up period among women with type 2 diabetes (NHS: mean parity 3.21, and ∼50% of women never breastfed; NHS II: mean parity 2.26, and ∼26% of women never breastfed), which is in line with increasing trends in breastfeeding (38) and decreasing parity rates (39) in the background population globally. However, the prevalence of exclusive breastfeeding remains far below international targets of 70% by 2030 (38), which is also true for our study (in the NHS II, ∼50% of women with type 2 diabetes ever exclusively breastfed during the follow-up period). Since the Framingham study, which was initiated nearly five decades ago, it has been clear that protective female-specific cardiovascular effects are lost once women develop insulin-treated diabetes (40). Subsequently, several other prospective studies demonstrated the augmented relative risk for CVD in women with diabetes compared with men (3). An excess CVD risk has also been shown among women with prior GDM, even without progression to interim overt diabetes (4).

Strengths of the study include the large number of diabetes cases, long duration, and high retention rates of follow-up, use of repeated measurements of lifestyle and reproductive variables across women’s reproductive life span, and analyses of several CVD outcomes. Further, primary results were robust across various sensitivity analyses.

Potential weaknesses should be acknowledged. Some women with diabetes may have been undiagnosed in the cohort because we did not screen for glucose intolerance universally. However, these would not alter case status of women reporting a diagnosis of diabetes, which was validated in a separate substudy (12).

Duration of breastfeeding was assessed through self-reports that may have resulted in misclassification. However, validation studies have consistently shown high validity of self-reported breastfeeding duration (16,17). Further, because of the prospective study design, such misclassification would be nondifferential with respect to CVD cases and possibly dilute our findings toward a more conservative underestimation of the risk association. Because of the observational design, a causal relationship cannot be demonstrated, as residual confounding by socioeconomic status and other variables remains a possibility, even though the analyses were extensively adjusted for potential confounders, including adjustment for a socioeconomic score.

Finally, the analyses were conducted in cohorts of predominantly non-Hispanic White nurses, which minimizes potential confounding by cultural and socioeconomic determinants but may limit the generalizability. However, there is no evidence showing that protective associations between breastfeeding and maternal cardiometabolic health differ by race/ethnicity or socioeconomic status.

In conclusion, in the current study comprising 15,146 parous women with type 2 diabetes from two large prospective cohorts and 4,537 women with a GDM history, we found significantly lower CVD rates among women who had ever breastfed, and women who cumulatively breastfed for >18 months experienced the lowest CVD risk. Our findings strengthen and expand the evidence of lifelong benefits from breastfeeding to parous populations of women at high risk for CVD complications, such as women with type 2 diabetes or GDM, where breastfeeding might mitigate the excess risk of CVD associated with diabetes in women. The findings underscore the need for greater efforts to promote breastfeeding as a primary prevention strategy in high-risk women.

Acknowledgments. The authors thank the NHS and NHS II participants and staff for their dedication and contribution to the research.

Funding. This work was supported by research grants from the National Institutes of Health National Cancer Institute (UM1 CA186107 and U01 CA176726) and National Heart, Lung, and Blood Institute (U01 HL145386, R01 HL034594, and R01 HL088521). A.B. was supported by the German Research Foundation (Deutsche Forschungsgemeinschaft) individual fellowship (no. BI 2427/1-1). M.G.-.F was supported by Novo Nordisk Fonden Research grant NNF18CC0034900. C.W. was supported by the German Research Foundation (Deutsche Forschungsgemeinschaft) individual fellowship (no. WI5132/1-1) and the SciLifeLab & Wallenberg Data Driven Life Science Program (grant KAW 2020.0239).

The funding sources played no role in the design, collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.B. performed statistical analyses, interpreted the data, and drafted the manuscript. A.B., F.B.H., and C.Z. designed the study and interpreted the data. M.G.-F., S.H.L., D.K.T., F.W., C.W., J.Y., J.E.M., and J.E.C. contributed to statistical analyses and interpretation of the data. All authors critically revised the manuscript. A.B. and F.H.B. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analyses.

F.B.H. and C.Z. are editors of Diabetes Care but were not involved in any of the decisions regarding review of the manuscript or its acceptance.

Prior Presentation. Parts of this study were presented as a poster at the annual American Heart Association EPI|Lifestyle Scientific Sessions 2023, Boston, MA, 28 February–3 March 2023.