Cumulative breastfeeding duration and incidence of diabetes in the multiethnic cohort study

Abstract

Objectives

To examine the association between breastfeeding duration and incidence of type 2 diabetes (T2D) across five racial and ethnic groups.

Background

Breastfeeding is a potent protective factor for many adverse perinatal outcomes and subsequent chronic disease conditions. Compared to non-breastfeeding parous women, multinational meta-analysis revealed that women who breastfed, especially for a duration extending beyond 12 months, experienced a significant T2D risk reduction later in life. This study will examine the relationship between breastfeeding and T2D in women from five racial and ethnic groups with a range of risks.

Methods

All multiethnic cohort (MEC) women who completed the 2003-2008 follow-up questionnaire when breastfeeding was assessed. Incident T2D was assessed using self-reported data and administrative data sources. Cox proportional hazards regression of incident T2D with age as the time metric stratified by race and ethnicity estimated its association with breastfeeding adjusted for relevant covariates.

Results

After exclusion of prevalent T2D cases, 5122 incident cases were detected. Of these, 53% of women without incident T2D ever breastfed in comparison to 51.8% of those with incident T2D. In women with any number of children, cumulative breastfeeding duration of ≥24 months resulted in an adjusted hazard ratio (HR) of 0.87 [0.78, 0.96], P-value <.01. In models examining any breastfeeding stratified by race and ethnic group, a lower risk was observed only in White women (adjusted HR of 0.87 [0.77,0.98], P-value < .02) while for cumulative duration of ≥24 months a lower risk was observed only in Native Hawaiian women (adjusted HR of 0.65 [0.45,0.93], P-value < .02).

Conclusions

Risk of T2D was reduced with cumulative duration of breastfeeding ≥24 months, especially in Native Hawaiian women. Extended breastfeeding should be encouraged as a preventative strategy for T2D prevention, especially in Native Hawaiian women who have a high risk for T2D.

Introduction

Breastfeeding refers to the process of feeding breastmilk to an infant, either directly from the breast/chest or via bottle-, cup-, or spoon-feeding. Although terminology may vary to describe how breast milk is consumed,¹ the term breastfeeding will be used following previously published guidance.² The American Academy of Pediatrics recommends exclusive breastfeeding for 6 months after each live birth unless prevented by biological or medical circumstances, followed by continued breastfeeding for 2 years or more with the introduction of complementary foods.³ Exclusive breastfeeding is intake restricted solely to breastmilk.⁴

Breastfeeding is a potent protective factor for many adverse perinatal outcomes and subsequent chronic disease conditions⁵ yielding numerous benefits for both child and mother. Babies with a history of breastfeeding have fewer colics and are healthier, as breastmilk is easy to digest.⁶ Breastfeeding also lowers infants’ and toddlers’ risk for ear infections, colds, bronchitis, asthma, pneumonia, eczema, diarrhea, constipation, colitis, allergies, gastroesophageal reflux, meningitis, necrotizing enterocolitis, sudden infant death syndrome, and dermatitis. They also have fewer hospitalizations and higher intelligence quotient.^7-10 Adults who were breastfed as infants have a lower mean blood pressure and total cholesterol and decreased risks for obesity and diabetes.^11,12 Women who breastfed their babies have a reduced risk for various cancers, elevated cholesterol, high blood pressure, heart attack, and stroke.^13-16 Compared to women who never initiated breastfeeding, mothers who engaged in any breastfeeding experienced a 27% reduction in the risk of type 2 diabetes (T2D) later in life in a multinational meta-analysis of 22 studies.¹⁷ Longer duration of breastfeeding (>12 months) has also been reported to reduce the risk for T2D by at least 30% in comparison to shorter duration of breastfeeding (less than 12 months).^17,18 Women who have a history of breastfeeding also tend to have good bone density, reducing their osteoporosis risk later in life.¹⁹ Breastfeeding facilitates bonding, making the woman more attentive to her baby,^20-22 while hormones released during breastfeeding also reduce postpartum depression.^23,24

Certain race and ethnic groups have substantially higher risk for T2D in comparison to others. Native Hawaiian women, for example, have a higher incidence of T2D and risk of gestational diabetes in comparison to non-Hispanic white women.^25-27 These disparate statistics are also found for breastfeeding: Native Hawaiian women are less likely to initiate breastfeeding than all other women (88% vs 95%). Among Native Hawaiian women who did initiate breastfeeding, only 77% were breastfeeding at 8 weeks compared to 98% of Japanese American women and 85% of White women based on data from the Pregnancy Risk Assessment Monitoring Surveys (PRAMS) completed in Hawaiʻi.²⁸ These rates drop to 55% of Native Hawaiians mothers compared to 66% of non-Hispanic White mothers accounting for exclusivity of breastfeeding at 8 weeks postpartum.²⁸ Most research on breastfeeding among racial and ethnic groups such as Native Hawaiians has been hampered by the frequent and problematic aggregation of Asian and Native Hawaiians and Pacific Islander women in national data,^29,30 which limits the availability of meaningful data and masks disparities. Therefore, there is a need to examine the relationship between breastfeeding and risk for diseases like diabetes in data sources in which race and ethnicity are disaggregated.

A limited number of large data sets are available, in which Native Hawaiians are disaggregated from Pacific Islanders and Asian Americans.^29,30 Outside of PRAMS, which in 2025 was suspended indefinitely,³¹ the MEC³² is the only other large data set with breastfeeding history and duration as well as disease outcomes like T2D in a representative sample of Native Hawaiians in addition to four other major racial and ethnic groups found in Hawaiʻi and California. Approximately 47% of Native Hawaiians live in Hawaiʻi with California having the highest percent of Native Hawaiians living on the US continent.³³ The objective of this study is to examine the association between breastfeeding duration and the incidence of T2D among the women across the five racial and ethnic groups in the MEC.

Materials and methods

Study population

Data for this study are derived from the MEC³² and comprise over 215 000 men and women aged 45 to 75 years at enrollment (1993-1996) from Hawaiʻi and predominantly Los Angeles County in California. The MEC was designed to examine the association of lifestyle and genetics with cancer and other chronic diseases among Black (16%), Japanese American (26%), Latino (22%), Native Hawaiian (7%), and White (23%) men and women.³⁴ The study population was sampled using driver’s license files for Hawaiʻi and Los Angeles, supplemented with Medicare files in Los Angeles and voter registration lists in Hawaiʻi such that findings may be generalizable to these ethnic group for these areas.

Participants were enrolled in 1993-1996, completing a comprehensive self-administered 26-page questionnaire in which study characteristics including place of birth, height and body weight, medical history, family history of cancer, physical activity, a dietary history, and reproductive history (eg, number of children, age of first live birth, ever been pregnant) in women were self-reported. Follow-up questionnaires were sent every 5 years to update information. This study examined cohort participants who completed the 10-year follow-up questionnaire in 2003-2008; it was completed by approximately 50% of cohort members.³⁵ Out of the 54 924 cohort members with 10-year follow-up questionnaire data who were women, those with racial and ethnic backgrounds other than Latino, White, Black, Native Hawaiian and Japanese American, then those confirmed to be T2D cases prior to 10-year follow-up,³⁵ those with a death date prior to answering the 10-year follow-up questionnaire or prior to their T2D diagnosis, discordant responses to the questions on breastfeeding and the number of children, and those where important factors associated with T2D were unable to be ascertained, were excluded (Figure 1). The final dataset consisted of 30 206 participants.

Figure 1.

Multiethnic cohort (MEC) study population flow chart sample exclusions for examining the association between breastfeeding and incidence of type 2 diabetes (T2D).

Breastfeeding assessment

The 2003-2008 follow-up questionnaire asked three questions related to breastfeeding: (1) How many children did you breastfeed for at least one month? (response options: None, 1, 2, 3, 4, 5-7, 8-10, 11 or more); (2) How many months on average did you breastfeed each child? (response options: 1-2, 3-4, 5-6, 7-8, 9-11 months, 12 or more months); (3) Did you breastfeed your first child? (response options: Yes, No). For analysis, breastfeeding responses were coded in three different ways. To assess a positive history of breastfeeding (Yes or No), any response other than none to the question “How many children did you breastfeed for at least one month?” was coded as Yes. No responses were further divided into nulliparous and parous with no breastfeeding. To determine a cumulative lifetime duration of breastfeeding in months, responses to “How many children did you breastfeed for at least one month?” were multiplied to the response to the “How many months on average did you breastfeed each child?”. For the question “How many children did you breastfeed for at least one month?”, the median values of 6 and 9 months, respectively, were used for the response options that were in ranges (eg, 5-7 and 8-10 months) while for the response option of 11 or more, 11 was used. Nulliparous women were excluded from categorization. Age of first breastfeeding was determined by examining the age of first live birth reported in the baseline questionnaire with those who responded yes to “Did you breastfeed your first child?”.

Diabetes incidence

This study followed the previously reported method of determining cases of T2D in MEC participants.³⁵ For this study, women who completed the 10-year follow-up questionnaire with no T2D at that time were included. Incident diabetes was defined as diagnoses occurring after the 10-year follow-up questionnaire. Specifically, women who self-reported a T2D diagnosis (responding affirmatively to the question “Has your doctor ever told you that you had any of the following conditions?”) on a follow-up questionnaire were considered cases as were women with a diagnosis of T2D in one or more administrative data sources (includes the Medicare Chronic Condition Warehouse,³⁶ California hospital discharges, and Hawaiʻi health plan linkage²⁵). The last linkages for T2D diagnosis was December 31, 2019.

Statistical analysis

The percentage of women in the MEC reporting a history of breastfeeding by race and ethnic group was estimated. For those women who reported a history of breastfeeding, duration of breastfeeding was categorized as nulliparous, 0 (parous), 1-6, 7-17, 18-23, ≥24 months, or missing while age of first breastfeeding was categorized into nulliparous, 12-25, 26-30, 31-35, ≥36 year, and missing. When the sample size allowed, breastfeeding duration category alignment with other research examining the association between breastfeeding duration and incidence of T2D^13,18 was prioritized.

Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated using Cox proportional hazards regression of incident T2D with age as the time metric to estimate association with breastfeeding history and then duration. Observation began at age of the 10-year follow-up questionnaire and ends at the earliest age at the following events: T2D, death, or study closure at December 31, 2019. Models were adjusted for the following covariates: age at 10-year follow-up, total energy intake (kcal/day), BMI (kg/m²), physical activity (METs of moderate and vigorous physical activity per day), years of education, neighborhood-level socioeconomic status index at baseline (block group level), sleep duration (hours/day), smoking status (never, former, current), red meat intake (grams/day), coffee intake (grams/day), alcohol intake (grams/day), diet quality (HEI-2015 score), and number of children. Race and ethnicity (African American, Japanese American, Latina, Native Hawaiian, and White) and number of children were used as strata variables in evaluating baseline hazard functions. Latina women were also stratified by US born vs Mexican born. Covariates were selected for inclusion into the model based on a priori knowledge of risk factors associated with T2D incidence in the MEC.^26,36,38 Heterogeneity in the breastfeeding-T2D association by race and ethnicity was evaluated using a Wald test of the interaction terms between race and ethnicity and the breastfeeding variable. All analyses were performed with SAS 9.4 (SAS Institute, Inc., Cary, NC, USA). P-values <.05 were considered as significant.

Results

The mean age of all participants was almost 69 years of age when the 10-year follow-up questionnaire was completed. Japanese American (33%) and White women (33%) were the largest race and ethnic group while Native Hawaiian women (7%) comprised the smallest group. These proportions reflect the racial and ethnic composition at baseline. The incident T2D group had a higher proportion of parous women with no breastfeeding in comparison to the no T2D group. The proportion of participants that ever breastfeeding was over 50%. Of these participants, approximately 15% breastfeeding for 1-6 months or 7-17 months in both the incident T2D group and no T2D group. In total, 5122 incident cases of T2D were observed (Table 1).

Table 1.

Characteristics of women in the multiethnic cohort (MEC) at the 10-year follow-up.

	All women (n = 30 206)	Incident T2D (n = 5120)	No T2D (n = 25 086)	P-value*	Cramer’s V
History of breastfeeding, %				<.01	0.02
Nulliparous	14.1	13.2	14.3
No, parous	32.9	35.0	32.5
Yes	53.0	51.8	53.2
Cumulative breastfeeding duration, %				.01	0.02
Nulliparous	14.1	13.2	14.3
0 months, parous	32.9	35.0	32.5
1-6 months	15.7	15.2	15.8
7-11 months	14.4	14.4	14.4
12-23 months	12.1	11.9	12.1
≥24 months	9.7	9.2	9.8
Missing	1.1	1.1	1.1
Age of first breastfeeding, %				<.01	0.03
Nulliparous	14.1	13.2	14.3
Never (parous)	32.9	35	32.5
12-25 years	35.1	35.2	35.1
26-30 years	12.0	11.4	12.1
31-35 years	3.8	3.3	3.8
≥36 years	1.2	1.0	1.3
Missing	0.9	0.9	0.9
Age (yrs), mean (SD)	68.9 (8.4)	69.0 (8.0)	68.8 (8.4)	.13
Ethnicity, %				<.01	0.08
White	33.7	25.3	35.5
African American	11.5	13.5	11.0
Native Hawaiian	7.2	8.6	6.9
Japanese American	33.8	37.5	33.0
Latina	13.8	15.1	13.6
BMI, kg/m2, %				<.01	0.12
<25	53.1	41.0	55.6
25 to <30	30.4	34.3	29.5
≥30	16.5	24.7	14.9
Smoking status, %				.69	0.00
Never	58.0	57.9	58.0
Former	36.4	36.2	36.5
Current	5.6	5.9	5.5
Alcoholic drinks, %				<.01	0.06
<1 per month	60.2	61.8	59.1
<1 per day	29.9	26.2	30.3
≥1 per day	9.9	12.0	10.6
Physical activity, metabolic equivalents of moderate-vigorous activity, %				<.01	0.04
Below median	51.6	55.9	50.7
Above median	48.4	44.1	49.3
Education status, years, %
≤12 grade	31.9	34.2	31.4	<.01	0.03
Some college or vocational training	32.5	32.7	32.5
College graduate and above	35.6	33.1	36.1
Total energy intake (kcal/day), mean (SD)	1680.8 (720.3)	1710.7 (746.3)	1674.7 (714.7)	<.01
Healthy Eating Index-2015, mean (SD)	72.5 (10.7)	71.7 (10.7)	72.7 (10.7)	<.01
Neighborhood socioeconomic status (nSES) at baseline				<.01	0.05
1 (lowest)	11.2	13.3	10.8
2	16.5	18.5	16.1
3	19.7	20.1	19.6
4	21.2	21.1	21.3
5 (highest)	31.4	27.0	32.2
Parity, %				<.01	0.02
0	14.2	13.2	14.3
1	11.6	10.9	11.8
2 to 3	49.9	49.8	50
≥4	24.3	26.1	23.9
Post-menopausal at 10-year follow-up, %	99.5	99.5	99.5	.87	0.00

Abbreviation: T2D, type 2 diabetes.

^*Chi-squared test for categorical variables; Welch’s t-test for continuous variables.

Breastfeeding, parity and post-menopausal status by race and ethnicity

Native Hawaiian women had the lowest proportion (7.8%) of nulliparous while White women had the highest proportion (18.2%). Significant differences were observed across all race and ethnic groups for any breastfeeding, cumulative duration of breastfeeding, age of first breastfeeding, parity, and postmenopausal status (Table 2). Any breastfeeding was highest in Japanese American women (58%) and Native Hawaiian women (56%) while African American women had the lowest proportion (42%). Cumulative breastfeeding duration greater than 24 months was highest among Latina women (16.8%) and Native Hawaiian women (12.6%). A majority (over 70%) of women from all of the race and ethnic groups had 2 or more children and most were post-menopausal at the time of the 10-year follow-up.

Table 2.

Breastfeeding and other reproductive characteristics of women in the multiethnic cohort (MEC) at the 10-year follow-up by race and ethnicity.

	White (n = 10 190)	African American (n = 3456)	Native Hawaiian (n = 2178)	Japanese American (n = 10 204)	Latina (n = 4178)	P-value*	Cramer’s V
History of breastfeeding, %						<.01	0.11
Nulliparous	18.2	13.6	7.8	13.6	9.1
No, parous	31.2	44.4	35.9	28.0	38.0
Yes	50.5	42.0	56.3	58.4	52.8
Cumulative breastfeeding duration, %						<.01	0.10
Nulliparous	18.2	13.6	7.8	13.6	9.1
0 months, parous	31.2	44.4	35.9	28.0	38.0
1-6 months	13.5	13.2	15.1	20.8	11.1
7-11 months	14.1	10.7	14.7	17	11.5
12-23 months	12.9	7.9	12.9	12.5	12.0
≥24 months	9.1	9.0	12.6	7.0	16.8
Missing	0.9	1.1	1.1	1.1	1.5
Age of first breastfeeding, %						<.01	0.12
Nulliparous	18.2	13.6	7.8	13.6	9.1
Never (parous)	31.2	44.4	35.9	28.0	38.0
12-25 years	32.7	35.2	45.6	32.4	42.0
26-30 years	11.3	4.1	6.0	18.8	6.9
31-35 years	4.3	1.6	2.3	4.9	2.2
≥36 years	1.8	0.5	0.9	1.2	0.7
Missing	0.6	0.7	1.6	1.0	1.0
Parity, %						<.01	0.15
0	18.2	13.6	7.8	13.6	9.1
1	12.3	17.3	8.5	11.5	7.2
2 to 3	49.5	41.6	43	58.6	40.5
≥4	20.0	27.6	40.7	16.2	43.3
Post-menopausal at 10-year follow-up, %	99.7	99.6	97.8	99.6	99.7	<.01	0.07

^* Chi-squared test for categorical variables.

Association of breastfeeding with incident T2D by number of children

After adjustment for relevant covariates and stratified by any, one, or two or more children, only cumulative duration of greater than 24 months breastfeeding yielded a significantly reduced risk of T2D in comparison to parous, non-breastfeeding women, with similar magnitude in the any number of children (HR 0.86 [CI 0.78-0.96]) and the 2 or more children categories (HR 0.86 [CI 0.78-0.96]; see Table 3). The P-trend tests of a decreased risk of T2D as cumulative breastfeeding duration category increased in these models were also significant. Earlier age of first breastfeeding (12-25 years) also showed a significantly lower risk of T2D in comparison to first breastfeeding at 26-30 years in the 2 or more children categories (HR 0.89 [CI 0.68-0.99]). No significant association was observed for any history of breastfeeding.

Table 3.

Association between breastfeeding and type 2 diabetes (T2D) incidence in the multiethnic cohort (MEC) with a maximum sample size of 30 206 and a maximum number of T2D cases of 5120.^a 

	Any number of children						One child				Two or more children
	Unadjusted		Model 1b		Model 2c		Unadjusted		Model 1b		Unadjusted		Model 1b
	HR [95% CI]	P	HR	P	HR	P	HR [95% CI]	P	HR	P	HR [95% CI]	P	HR	P
History of breastfeeding
No, parous	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
Nulliparous	0.87 [0.8, 0.95]	<.01	0.99 [0.9, 1.08]	.78	0.95 [0.86, 1.05]	.33	—	—	—	—	—	—	—	—
Yes	0.91 [0.86, 0.97]	<.01	0.96 [0.9, 1.02]	.18	0.97 [0.91, 1.03]	.26	0.86 [0.73, 1.01]	.07	0.94 [0.79, 1.11]	.45	0.92 [0.86, 0.98]	<.01	0.96 [0.9, 1.03]	.26
Cumulative breastfeeding duration
0 months, parous	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
Nulliparous	0.87 [0.8, 0.95]	<.01	0.99 [0.91, 1.08]	.83	0.98 [0.88, 1.09]	.65	—	—	—	—	—	—	—	—
1-6 months	0.91 [0.83, 0.99]	.02	0.98 [0.9, 1.07]	.65	0.98 [0.9, 1.07]	.61	0.88 [0.74, 1.06]	.17	0.95 [0.79, 1.14]	.6	0.92 [0.84, 1.02]	.1	0.99 [0.9, 1.09]	.9
7-11 months	0.93 [0.86, 1.02]	.11	1 [0.91, 1.09]	.94	1 [0.92, 1.09]	.95	0.84 [0.61, 1.16]	.29	0.95 [0.69, 1.32]	.77	0.93 [0.85, 1.02]	.12	1 [0.91, 1.09]	.96
12-23 monthsd	0.92 [0.84, 1.01]	.07	0.97 [0.89, 1.07]	.58	0.98 [0.89, 1.07]	.63	0.75 [0.5, 1.13]	.17	0.87 [0.58, 1.31]	.51	0.92 [0.83, 1.01]	.08	0.98 [0.89, 1.08]	.73
≥24 monthse	0.89 [0.8, 0.98]	.02	0.86 [0.78, 0.96]	<.01	0.87 [0.78, 0.97]	.01	—	—	—	—	0.88 [0.79, 0.97]	.01	0.86 [0.78, 0.96]	<.01
P-trendf	0.99 [0.99, 1]	.02	0.99 [0.99, 1]	.02	1 [0.99, 1]	.06	0.98 [0.96, 1]	.08	0.99 [0.97, 1.01]	.46	0.99 [0.99, 1]	.01	0.99 [0.99, 1]	.02
Age at first breastfeeding (years)
26-30	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
12-25	1.08 [0.98, 1.18]	.13	0.92 [0.83, 1.01]	.08	0.93 [0.84, 1.03]	.14	1.23 [0.9, 1.68]	.19	1.1 [0.8, 1.51]	.57	1.05 [0.95, 1.15]	.37	0.89 [0.81, 0.99]	.03
31-35	0.93 [0.78, 1.1]	.37	0.94 [0.79, 1.11]	.44	0.93 [0.78, 1.1]	.39	1.37 [0.96, 1.96]	.09	1.37 [0.96, 1.96]	.09	0.84 [0.68, 1.02]	.08	0.84 [0.68, 1.03]	.09
≥36	0.8 [0.6, 1.07]	.13	0.83 [0.62, 1.12]	.22	0.82 [0.61, 1.1]	.18	0.98 [0.63, 1.52]	.92	0.96 [0.61, 1.49]	.85	0.82 [0.53, 1.29]	.4	0.87 [0.56, 1.37]	.55

^aBold = P-value <.05.

^bModel 1 is adjusted for age at 10-year follow-up, total energy intake, body mass index, moderate to vigorous physical activity, education, neighborhood socioeconomic score (nSES) at baseline, sleep duration, smoking status, red meat intake, coffee intake, alcohol intake, and diet quality as measured by Healthy Eating Index 2015.

^cModel 2 is adjusted for everything in model 1 plus parity.

^dFor women with one child, this reflects “≥12 m”.

^eLimited to women with two or more children.

^f P-trend test is limited to parous women and uses the median values of the categorical responses (0, 3.5, 9, 17.5, and 24 months).

Association of breastfeeding with incident T2D by race and ethnicity

Race and ethnic differences were observed when examining the association between history of any breastfeeding and incident T2D in comparison to parous, non-breastfeeding women. A significantly lower risk was observed in White women (HR 0.87 [CI 0.77-0.98]) while a significantly higher risk was observed in Latina women (overall) who either were nulliparous or had any history of breastfeeding. However, after adjustment for birthplace and language spoken, a higher risk was observed only in the nulliparous Latina women (overall) (HR 1.37 [CI 1.03-1.83]; Table 4). A similar finding was observed in the nulliparous Latina women (HR 1.43[CI 1.07-1.92]). In addition, a significantly reduced risk of 35% was observed for Native Hawaiian women who continued breastfeeding for ≥24 months (HR 0.65 [CI 0.45-0.93]). P-trends for cumulative duration were significant in White women and Native Hawaiian women only. No associations were observed with the age of first breastfeeding across the race and ethnic groups.

Table 4.

Association between breastfeeding and type 2 diabetes (T2D) incidence stratified by race and ethnicity in the multiethnic cohort (MEC) with a maximum sample size of 10 190 for white women, 3456 for black women, 2178 for native Hawaiian women, 10 204 for Japanese women, and 4178 for latina. (The maximum number of T2D cases for white women was 1 297  689 for black women, 442 for native Hawaiian women, 1919 for Japanese women and 773 for latina.^a)

	White		Black		Native Hawaiian		Japanese		Latina (overall)		Latina (adjusted for US-born and language)
	HR [95% CI]	P	HR [95% CI]	P	HR [95% CI]	P	HR [95% CI]	P	HR [95% CI]	P	HR [95% CI]	P	Race interaction P
History of breastfeeding													<.01
No, parous	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
Nulliparous	0.84 [0.69, 1.03]	.09	1.09 [0.83, 1.42]	.54	0.96 [0.63, 1.47]	.86	0.9 [0.75, 1.08]	.24	1.43 [1.07, 1.92]	.02	1.37 [1.03, 1.83]	.03
Yes	0.87 [0.77, 0.98]	.02	0.91 [0.77, 1.07]	.26	0.87 [0.71, 1.06]	.16	0.92 [0.83, 1.02]	.1	1.21 [1.03, 1.42]	.02	1.12 [0.95, 1.33]	.17
Cumulative breastfeeding duration													.04
None, parous	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
Nulliparous	0.86 [0.7, 1.06]	.16	1.1 [0.83, 1.45]	.51	1.02 [0.67, 1.57]	.92	0.88 [0.73, 1.06]	.19	1.48 [1.09, 2]	.01	1.45 [1.07, 1.96]	.02
1-6 months	0.89 [0.75, 1.06]	.2	0.88 [0.69, 1.13]	.31	0.92 [0.69, 1.24]	.59	0.9 [0.79, 1.02]	.1	1.2 [0.94, 1.54]	.15	1.15 [0.9, 1.47]	.27
7-11 months	0.91 [0.77, 1.09]	.31	0.92 [0.71, 1.18]	.51	1.05 [0.8, 1.39]	.71	0.89 [0.78, 1.03]	.11	1.25 [0.99, 1.59]	.06	1.18 [0.93, 1.5]	.17
12-23 monthsb	0.84 [0.7, 1.01]	.07	0.96 [0.72, 1.29]	.8	0.77 [0.56, 1.05]	.1	1.01 [0.87, 1.18]	.86	1.21 [0.96, 1.54]	.11	1.15 [0.9, 1.46]	.26
≥24 monthsc	0.81 [0.64, 1.02]	.07	0.83 [0.61, 1.13]	.24	0.65 [0.45, 0.93]	.02	0.87 [0.71, 1.07]	.18	1.18 [0.93, 1.49]	.18	1.02 [0.8, 1.3]	.88
P-trendd	0.99 [0.98, 1]	.02	1 [0.98, 1.01]	.36	0.98 [0.97, 1]	.01	1 [0.99, 1.01]	.58	1.01 [1, 1.02]	.11	1 [0.99, 1.01]	.63
Age at first breastfeeding (years)													.11
26-30	1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)		1 (Ref)
12-25	1.17 [0.95, 1.45]	.15	0.89 [0.58, 1.37]	.59	1.43 [0.85, 2.42]	0.18	0.9 [0.78, 1.03]	.12	0.98 [0.73, 1.31]	.9	1.04 [0.78, 1.39]	.79
31-35	1.16 [0.83, 1.61]	.39	1 [0.48, 2.08]	1	—	—	0.96 [0.76, 1.21]	.75	0.8 [0.46, 1.4]	.44	0.76 [0.43, 1.33]	.34
≥36	0.89 [0.53, 1.48]	.64	—	—	—	—	0.72 [0.44, 1.17]	.19	—	—	—	—

^a Bold = P value <.05 while gray shading indicates fewer than 10 events; models are adjusted for age at 10-year follow-up, total energy intake, body mass index, moderate to vigorous physical activity, education, neighborhood socioeconomic stats (nSES) at baseline, sleep duration, smoking status, red meat intake, coffee intake, alcohol intake, diet quality as measured by Healthy Eating Index-2015, and parity; the test for the race interaction does not adjust for place of birth or language.

^b For women with one child, this reflects “≥12 m”.

^c Limited to women with two or more children.

^d P-trend test is limited to parous women and uses the median values of the categorical responses (0, 3.5, 9, 17.5, and 24 months). 

Sensitivity analyses of breastfeeding with incident T2D

Sensitivity analyses were conducted limiting to parous women (Table S1) and women with two or more children (Table S2). The categorical HRs and the P-values for trend remained consistent with what was found in the main analyses.

Discussion

The findings of this prospective analysis contribute to the growing evidence linking breastfeeding to lower T2D incidence. Longer cumulative duration, earlier initiation of breastfeeding and any breastfeeding may be protective factors in lowering T2D risk in certain women. Key findings indicated that cumulative breastfeeding greater than 24 months, especially in Native Hawaiian women, earlier age of first breastfeeding (12-25 years) particularly in women with two or more children, and any breastfeeding in White women were associated with a significantly reduced risk of T2D. A significantly higher risk for T2D was observed in Latina women (overall) who either were nulliparous or had any history of breastfeeding, however, the latter was reduced to being not significant after adjustment for birthplace and language spoken. These variations by race and ethnicity highlight the need to consider sociocultural and biological factors that may influence the risk of T2D and breastfeeding practices. For example, breastfeeding may result in lower visceral fat,^37,38 lower BMI,³⁹ or better obesity-related biomarker levels⁴⁰ that persists for many years after breastfeeding.

These findings of this prospective analysis align with the existing studies examining breastfeeding duration and risk of T2D. Various studies have reported an inverse relationship between breastfeeding duration with the risk of T2D later in life among young and middle-aged women.^41-46 In the Reproductive Risk factors for Incontinence Study at Kaiser (RRISK), a population-representative cohort of women from within the Kaiser Permanente Medical Care Program of Northern California, women who never exclusively breastfed were more likely to develop T2D than women who exclusively breastfed for an average of 1 or more months (OR 1.38; 95% CI, 1.06-1.81).⁴³ Among parous women reporting a birth within 15 years of the prospective Nurse’s Health Study I and II assessment (NHS, NHS II), a cohort of mostly White nurses, each additional year of lactation after controlling for relevant T2D risk factors, resulted in a decreased risk of T2D among NHS (adjusted HR of 0.85, 95% CI, 0.73-0.99) in the NHS and NHS II participants (adjusted HR of 0.86, 95% CI, 0.79-0.93). ⁴⁴ In a prospective observational pregnancy cohort study of mostly White women conducted by Mount Sinai Hospital, women who engaged in breastfeeding for 12 months or longer had lower likelihood of T2D at 3 years postpartum compared to women who breastfed for less than 12 months (OR 0.37, 95% CI, 0.18 0.78, P = .009).⁴⁵ A prospective German GDM study followed women for 2 and 9 months postpartum; 2, 5, 8, 11, 15, and 19 years after pregnancy; or until the diagnosis of T2D. Women in this study who breastfed for longer than 3 months had the lowest postpartum T2D risk (15-year risk: 42% reduced risk [95% CI, 28.9-55.1] compared to women with no or less than 3 months of breastfeeding, 72% reduced risk [60.5-84.7]; P = .0002).⁴⁶ The current study, which includes individuals from five major racial and ethnic groups living in California and Hawaiʻi, a population not well represented, adds to the literature demonstrating that longer breastfeeding durations are associated with lower incidence of T2D among women.

In this current study, parity was important to account for when examining the association between breastfeeding and the risk for T2D. This was also true in the Women’s Health Initiative, a large prospective study among postmenopausal women who were mostly non-Hispanic White, which found that in comparison to nulliparous women, parous women with less than five births who did not breastfed had higher risk of T2D (HR = 1.06, 95% CI, 1.01-1.12) while women with five or more births had increased T2D risk regardless of breastfeeding.⁴⁷ The Australian 45 and Up cohort study of 52 731 women found that in comparison to nulliparous women, parous women who did not breastfed had a significantly greater risk of T2D (OR 1.48, [95% 1.26-1.73]). This pattern of association remained consistent regardless of the number of children a woman had given birth to.⁴¹

Furthermore, the woman’s age at the time of their first birth may impact their risk for T2D. Findings from the Korean National Health and Nutrition Examination Survey from 2008 to 2011 of 37 753 participants found that the prevalence of T2D was significantly higher in women who gave birth before 19 years of age (OR 1.492 [95% CI, 1.005-2.215]). Their adjusted models did not account for breastfeeding.⁴⁸ Similarly, another cohort study of 512 891 people from 10 diverse areas of China reported that women who had their first child at age <20 years (OR 1.33, 95% CI, 1.22-1.44) and between the ages of 20 and 24 years (OR 1.16, 95% CI, 1.10-1.23) were more likely to be diagnosed with T2D compared with those who had their first child between the ages of 25 and 29 years.⁴⁹ Breastfeeding was also not accounted for in that study. Conversely, in this study, most of the MEC participants (58.1%) age of first breastfeeding occurred between 12 and 25 years of age and the parous women who started breastfeeding between 12 and 25 years of age had a reduced risk for incident T2D in comparison to parous women who started breastfeeding between 26 and 30 years of age.

Although many studies have shown breastfeeding as a protective factor against T2D, other studies have not found breastfeeding to be protective against T2D.^50-52 Breastfeeding, for durations from one month up to 2 years postpartum, did not protect a group (n = 529) of mostly Latina women from a declining glucose tolerance at two years postpartum in comparison to women who did not breastfeed (P = .08).⁵¹ In addition, breastfeeding was not an independent predictor for glucose intolerance at 12 weeks postpartum in women who were diagnosed with gestational diabetes during pregnancy.⁵³ While in a prospective study (n = 381) conducted in South Korea, breastfeeding was not significantly related to glucose homeostasis at 6 to 12 weeks postpartum (r = -0.016, P = .25).⁵² This study is the first, to the authors knowledge, to find an increased incidence of T2D in Latina women (overall) who experienced any breastfeeding in comparison to parous women with no breastfeeding experience. However, upon adjustment for location of birth and language spoken the association became not significant. This finding may speak to the role that acculturation plays in risk for T2D as well as other unique risk profiles for T2D that were not accounted for in these analyses. For example, one study in Latina women found that women who breastfed have an increased risk of T2D when using a progestin-only oral contraception compared to using a low-dose combination oral contraceptive (adjusted relative risk, 2.87; 95% CI, 1.57-5.27).⁵⁴ A history of gestational diabetes is also associated with a higher risk for developing T2D 5 to 7 years postpartum⁵⁵ with significant differences in risk for women with Cuban or Puerto Rican heritage compared to women with Mexican heritage.⁵⁶ More research is needed to explore these findings further to explain the relationship between breastfeeding and the risk of T2D in Latina women.

Notable strengths of this study are the ability to examine breastfeeding, a Healthy People 2030 objective⁵⁷ that is a fundamental protective factor for many adverse perinatal outcomes and chronic conditions,⁵ in disaggregated multiethnic populations which have been neglected in the maternal health literature. Inclusion of a large cohort with ∼3 decades of follow-up in a multiethnic population in which three of the racial and ethnic groups, the Black, Latino, and Native Hawaiian racial and ethnic groups, experience significantly higher incidence of T2D than the White racial and ethnic group²⁶ is another notable strength.

Limitations are the potential for recall bias in breastfeeding history and duration that may be biasing (either increasing or decreasing) the magnitude of the effect. This study is also unable to ascertain exclusivity of breastfeeding, which may have provided further interpretation on the results, as infant formula feeding was not assessed in the MEC. In the mid-19th century, partly attributed to an aggressive marketing campaign by formula manufacturers, infant formula’s popularity as a substitute for human milk rose drastically, resulting in significant declines in breastfeeding through the 1970s.⁵⁸ Approximately 50% of the MEC completed the 10-year follow-up, resulting a significant proportion of the cohort in which breastfeeding could not be ascertained. This may have influenced the results observed due to the small sample sizes in some of the models (eg, age of first breastfeeding stratified by race and ethnicity). Another important limitation relates to the survey question used to assess breastfeeding duration, “How many months on average did you breastfeed each child?”, which capped responses at “12 months or more.” As a result, the duration variable created from this survey item was unable to capture variation beyond 12 months for participants with only one child. However, as there were a minority of participants with one child this may not have resulted in a major effect on the findings. Although this study did not conduct direct validation of T2D diagnoses through fasting glucose, HbA1c levels, or medical records, previous population-based studies have shown high validity in T2D self-reports.^59,60 In addition, previous history of gestational diabetes, a strong risk factor for T2D,⁶¹ is unknown in this study. Finally, the large number of tests as part of the stratified analysis by race and ethnicity may have resulted in false positive findings.

Conclusion

This study indicates that risk of T2D is reduced when cumulative duration of breastfeeding is greater than 24 months, especially among Native Hawaiian women. However, the protective benefit of breastfeeding in certain race and ethnic groups may be attenuated by other factors which require further exploration. Breastfeeding should be encouraged as a preventative strategy for T2D prevention.

Author contributions

Marie K. Fialkowski(Conceptualization [Equal], Funding acquisition [Equal], Writing—original draft [Equal], Writing—review & editing [Equal]), Brandon Quon(Formal analysis [Equal], Writing—review & editing [Equal]), Christina Young(Writing—original draft [Equal], Writing—review & editing [Equal]), Samantha Streicher(Writing—review & editing [Equal]), Kim Yonemori(Writing—review & editing [Equal]), Gertraud Maskarinec(Writing—review & editing [Equal]), Iona Cheng(Writing—review & editing [Equal]), Veronica Wendy Setiawan(Writing—review & editing [Equal]), and Christopher A Haiman(Writing—review & editing [Equal]), Lynne R Wilkens(Conceptualization [Equal], Writing—review & editing [Equal])

Supplementary material

Supplementary material is available at Global Perspectives on Nutrition online.

Conflicts of interest

None declared.

Funding

This work was supported by the Colorado Center for Diabetes Translation Research, through a grant from the National Institute of Diabetes and Digestive and Kidney Diseases (P30DK092923); National Cancer Institute (NCI) for the Multiethnic Cohort Study (U01 CA164973); University of Hawaʻi Cancer Center Shared Resources for Biostatistics and Nutrition Support services (NCI P30 CA071789); US Department of Agriculture (USDA), Agriculture and Food Research Initiative grant No. 2021-68012-35899; Commission on Dietetic Registration (CDR) Doctoral Scholarship; CDR Emerging Researcher Grant; Hawaiʻi Pacific Foundation Doctoral Research Award. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability

The data used in this study cannot be publicly shared due to privacy considerations as they contain identifiable patient information. However, researchers who meet the access criteria can request the data from the Multiethnic Cohort Study (http://www.uhcancercenter.org/research/the-multiethnic-cohortstudy-mec/data-sharing-mec).

Ethical approval

The study protocols were approved by the Institutional Review Boards of the University of Hawaiʻi and the University of Southern California. Return of a completed questionnaire was considered as consent. Written informed consent was received any time in person measurements or biospecimens were collected.