Associations of dairy intake with circulating biomarkers of inflammation, insulin response and dyslipidemia among postmenopausal women

Abstract

Background:

Cardiometabolic diseases are prevalent in aging Americans. Though some studies implicated greater intake of dairy products, it is not clear how dairy intake is related to biomarkers of cardiometabolic health.

Objective:

To test the hypothesis that associations of dairy foods with biomarkers of lipid metabolism, IGF-signaling, and chronic inflammation may provide clues to understanding how dairy may influence cardiometabolic health.

Design:

A cross-sectional study in the Women’s Health Initiative using baseline food frequency questionnaire data to calculate dairy intake.

Participants/setting.

35,352 postmenopausal women 50–79 years-old in 40 clinical centers in the United States.

Main outcome measures:

Baseline (1993–1998) concentrations of 20 circulating biomarkers.

Statistical analyses:

Multivariable-adjusted linear regression was used to estimate percent difference in biomarker concentrations per serving of total dairy and individual foods (milk, cheese, yogurt, butter, and low-fat varieties).

Results:

Lower triglyceride concentrations were associated with greater intake of total dairy (−0.8%(−1.2%, −0.3%), mainly driven by full-fat varieties. Individual dairy foods had specific associations with circulating lipid components. For example, greater total milk intake was associated with lower concentrations of total cholesterol[TC: −0.4%(−0.7%, −0.2%)] and high-density lipoprotein cholesterol[HDL-C: −0.5%(−0.9%, −0.1%)], whereas greater butter intake was associated with higher TC, 0.6%(0.2%, 1.0%) and HDL-C, 1.6%(1.1%, 2.0%) concentrations. In contrast, higher total yogurt intake was associated with lower TC, −1.1%(−2.0%, −0.2%) and higher HDL-C, 1.8%(0.5%, 3.1%). Greater total dairy intake (regardless of fat content), total cheese, full-fat cheese, and yogurt were consistently associated with lower concentrations of glucose, insulin and C-reactive protein. However, milk and butter were not associated with these biomarkers.

Conclusion:

Higher dairy intake, except butter, was associated with a favorable profile of lipids, insulin response and inflammatory biomarkers, regardless of fat content. Yet specific dairy foods may influence these markers uniquely. Findings do not support a putative role of dairy in cardiometabolic diseases observed in some previous studies.

INTRODUCTION

Historically, dairy products were implicated as risk factors for cardiometabolic disease and obesity, in large part due to their content of saturated lipids and cholesterol^1–3. Yet, the emergence of data from large prospective cohort studies has called into question such a conclusion^4–6, suggesting greater complexity and that additional research is necessary to shed light on potential mechanisms. Dietary Guidelines for Americans recommend intake of fat-free or low-fat dairy foods in place of full-fat dairy products which are a significant source of saturated fat^{7, 8}. Most randomized-controlled trials (RCTs) that have examined the effects of dairy consumption on circulating biomarkers of cardiometabolic health have been limited by small sample sizes and short study duration⁹. In addition, the majority of RCTs studied varying dairy foods against different comparator diets or foods and included varying number of biomarkers. This high heterogeneity among RCT study designs makes it hard to meaningfully synthesize findings across several RCTs. Although observational studies have several limitations compared to RCTs, they have the advantage of large sample sizes, studying dietary exposures in the real-world setting, and have greater potential for generalizing findings to large segments of the population¹⁰.

More recent observational studies and meta-analyses have reported an inverse association between dairy, or dairy fat and risk of cardiovascular disease (CVD) and obesity^{11, 12}. Findings from three recent studies that examined associations of dairy consumption and risk of type 2 diabetes (T2D) or mortality in large cohort studies did not support a harmful association of total dairy consumption^13–15. One study in an Italian cohort found that neither full-fat milk nor low-fat milk was associated with overall mortality¹³. Analyses of data from three U.S. cohorts of healthcare professionals found that long-term changes in full-fat dairy consumption and dairy fat were not associated with T2D risk^{14, 15}, however another study reported a positive association between higher total dairy consumption and slightly higher overall mortality risk¹⁶. A recent meta-analysis of 20 prospective cohort studies concluded that fermented milk (buttermilk, quark and cultured sour milk) consumption was associated with reduced CVD and higher yogurt intake was associated with reduced T2D risk¹⁷. The heterogeneous findings of the association of dairy intake and specific disease endpoints warrant a more comprehensive investigation of how total dairy and specific types of dairy may influence chronic disease development in a variety of relevant cohorts. In addition, studies of underlying mechanisms including an examination of known intermediate biomarkers of these diseases may enhance our understanding of these relationships.

Insulin resistance and inflammatory biomarkers have been extensively studied in the development of many chronic diseases including CVD, T2D and cancers, and several have utility in managing patients in clinical practice^{18, 19}. Hyperinsulinemia, chronic inflammation and dyslipidemia cluster in insulin resistance, and higher intake of hyperinsulinemic or proinflammatory dietary patterns^{20, 21} that include a variety of different dairy foods as components, have been associated with higher risk of CVD²², T2D^{23, 24} and cancer²⁵. Specifically, these hyperinsulinemic and proinflammatory dietary patterns are low in full-fat dairy and high in low-fat dairy foods among other food components. It is therefore possible that dairy foods may affect health outcomes through insulin response, inflammation and lipid signaling pathways. However, findings from RCTs are inconsistent, with some meta-analyses showing that higher dairy intake may improve inflammatory biomarker profiles²⁶, whereas others do not support an impact of dairy on inflammation or insulin resistance⁹.

Findings from these recent studies, taken in totality, seem to challenge some prevailing dietary perceptions and guidance based on dairy foods and their constituents, particularly lipid profiles. Therefore, an investigation of the associations of dairy consumption with concentrations of circulating biomarkers is needed. In the current study, we specifically characterize the associations of total dairy, specific dairy foods including fat composition, in relation to a comprehensive list of circulating lipid biomarkers, insulin-related/insulin-like growth factor hormonal axis, and inflammation-related biomarkers.

SUBJECTS AND METHODS

Study Population

The Women’s Health Initiative (WHI) enrolled 161,808 postmenopausal women 50 to 79 years old in 40 clinical sites in the United States between 1993 and 1998²⁷. Participants were enrolled into an observational study (OS) or one or more of four clinical trials (CT), two of which were hormone therapy (HT) trials. The full WHI-OS consisted of 93,676 women not eligible or unwilling to participate in the CT²⁷. At the baseline clinic visit, trained study nurses drew blood samples and performed physical measurements including blood pressure, height and weight.

In the current cross-sectional study, 61,513 women with data on circulating biomarkers of inflammation, insulin response and cardiometabolic health from WHI OS and CT using baseline blood samples were considered for inclusion. Women with implausible total energy intake values (≤600 kcal/d or ≥5000 kcal/d, n=3323), very low or very high body mass index (BMI) values (<15 kg/m² or >50 kg/m², n = 3599) which may be indicative of morbidity, those with potential acute inflammation (CRP >10 mg/l, n=4817), self-reported diabetes (n=3653), cancer (n=3587), or cardiovascular disease (n=7182) were excluded at baseline. The final analytic sample was composed of 35,352 women. Most of the demographic and lifestyle characteristics, intakes of total dairy, other foods and nutrients were comparable between the final analytic sample and excluded sample (data not shown). The WHI protocol was approved by the institutional review boards at the Clinical Coordinating Center at the Fred Hutchinson Cancer Research Center in Seattle, WA, and at each of the 40 Clinical Centers²⁷.

Assessment of dairy intake

At WHI baseline, dairy intake was calculated from a 122 line-item self-administered semi-quantitative food frequency questionnaire (FFQ) that asked about habitual dietary intake in the preceding 3-month period. Foods (servings per day) and nutrient intakes were then estimated using the University of Minnesota’s Nutrition Data System for Research (NDSR version 2005) database released in May 2005^{28, 29}. The WHI FFQ has produced results reasonably comparable to those from four 24-h dietary recall interviews and 4 days of food diaries recorded within the WHI, including intakes of nutrients such as saturated fat, protein and calcium where dairy is an important source³⁰. Low-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza¹⁵. Full-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza. Total dairy was calculated as the sum of daily servings of low-fat and full-fat dairy. Total milk was calculated as the sum of whole milk and low-fat milk. Low-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza. Full-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese). Total cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese. Total yogurt included non-fat yogurt and all other yogurt. Butter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking.

Other foods including process meat, red meat, sugar-sweetened beverages, refined grains, wines, tea, coffee, whole fruit, green-leafy vegetables; and nutrients including fiber, carbohydrate, protein, branched-chain amino acids, fat and saturated fat were included in the current study. Food intake serving sizes are listed in the table footnotes. Health-Eating Index (HEI-2015) calculated using the energy density method as previously described,^{7, 8} was used to describe the overall diet quality in the study sample. Higher scores (maximum=100) indicate higher quality of the overall diet.

Biomarker assessment

Biomarkers assessed at baseline included blood lipid biomarkers including triglycerides, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and total cholesterol (TC). Insulin-related and insulin-like growth factors (IGF) and binding proteins markers including glucose, insulin, c-peptide, IGF-1, free IGF-1, IGF binding proteins (1, 3, 4); and inflammatory biomarkers including c-reactive protein (CRP), interleukins (IL-6, IL-10), tumor necrosis factor (TNF)-alpha, TNF-alpha receptor 1 (TNFR1), TNF-alpha receptor 2 (TNFR2), adiponectin and leptin. Biomarker data were pooled from several case-control studies nested within the WHI.

Covariates

Data on potential confounding variables were collected by self-administered questionnaires on demographics, medical history, and lifestyle factors at baseline, as previously described^{27, 31}. Covariates were selected based on the literature including our previous work on these biomarkers. Covariates included in the models were: total energy intake (kcal/day); age at WHI baseline (years); body mass index [BMI=weight (kg)/(height (m) × height (m))]; self-reported racial/ethnic groups (American Indian or Alaska Native, Asian/Pacific Islander, Hispanic/Latino, African American, European American, and other race groups); educational levels categorized into some high school or lower educational level, high school graduate or some college or associate degree, and ≥4 years of college; pack-years of smoking; alcohol intake, drinks/week; regular use of aspirin and other nonsteroidal anti-inflammatory drugs (NSAIDs) (yes/no), and statins (yes/no) (regular use was defined as: ≥2 times in each of the two weeks preceding the interview); number of supplements used; use of unopposed estrogen and/or estrogen plus progesterone (ye/no); total recreational physical activity, calculated by summing the metabolic equivalent tasks for all reported activities for each individual (e.g., walking, aerobics, jogging, tennis, swimming, biking outdoors, exercise machine, calisthenics, popular or folk dancing) and presented as Metabolic equivalents (MET)-hours/week²⁷; fasting status at blood draw (< 8 hours or ≥8 hours); and hormone therapy (HT) study arms (not randomized to HT, estrogen-alone intervention, estrogen-alone control, estrogen plus progestin intervention, estrogen plus progestin control).

Statistical Analysis

Participants’ characteristics were described using means (standard deviations) for continuous variables, and frequencies (%) for categorical variables across quintiles of total dairy intake. Concentrations of all biomarkers were normalized using natural log transformation prior to analyses. Multivariable-adjusted linear regression analyses were conducted to estimate percent differences in circulating biomarker concentrations in quintiles of dairy intake, using the lowest quintiles as reference and adjusting for all the covariates listed in the covariates section. In addition, we examined percent differences in biomarker concentration per serving increment in dairy intake and interpreted the p value of this continuous dairy intake variable as the p value for linear trend across quintiles of dairy intake. Statistical significance was based on a false discovery rate (FDR)-adjusted P<0.05. All analyses were conducted using SAS^® software version 9.4 (SAS Institute Inc., Cary, NC)³².

RESULTS

Participants’ characteristics

The study sample included 35,352 women in the WHI cohort. Table 1 shows the baseline characteristics of participants by quintiles of total dairy consumption, ranging from a mean intake of 0.24 servings/day in quintile 1 to a mean intake of 4.11 servings/day in quintile 5. Those in the highest quintile of total dairy intake included a higher proportion of European Americans, a lower proportion of African Americans and Asian Americans, a higher proportion of current smokers, and a slightly higher proportion of obese women; although BMI and physical activity did not appear to vary across quintiles of total dairy intake compared to women in the lowest total dairy intake quintile. In terms of other foods, intakes of red meat, processed meat, refined grains, tea/coffee, total alcohol including wine, increased across quintiles of dairy intake. However, overall diet quality, based on HEI-2015 scores, slightly decreased with higher total dairy intake. The nutrient profile of women in the highest quintile of total dairy intake was characterized by higher intakes of total carbohydrate, protein, branched-chain amino acids, total and saturated fat, and higher total fiber, compared to those in the lowest quintile.

Table 1.

Baseline characteristics of the study sample in quintiles of total dairy intake among 35,352 postmenopausal women in the Women’s Health Initiative

	Quintiles of total dairy intake
Characteristic	Quintile 1 0 to <0.4 servings	Quintile 2 0.4 to <0.9 servings	Quintile 3 0.9 to <1.4 servings	Quintile 4 1.4 to <2.6 servings	Quintile 5 2.6 to 13.3 servings
Mean total dairy intake, servings/d	0.2 ± 0.1	0.6 ± 0.1	1.1 ± 0.2	1.9 ± 0.3	4.1 ± 1.6
Race/ethnicity, n (%)
Black or African-American	1337 (18.9)	1022 (14.5)	1051 (14.9)	927 (13.1)	690 (9.8)
American Indian or Alaskan Native	60 (0.9)	61 (0.9)	62 (0.9)	80 (1.1)	58 (0.8)
Hispanic/Latino	476 (6.7)	437 (6.2)	447 (6.3)	509 (7.2)	455 (6.4)
Asian or Pacific Islander	461 (6.5)	222 (3.2)	206 (2.9)	155 (2.2)	81 (1.2)
White (not of Hispanic origin)	4690 (66.2)	5261 (74.6)	5243 (74.2)	5330 (75.3)	5728 (81.0)
Others	59 (0.8)	52 (0.7)	60 (0.9)	74 (1.1)	58 (0.8)
Age, years	64.6 ± 7.3	64.6 ± 7.3	64.6 ± 7.1	64.3 ± 7.2	63.8 ± 7.2
Body mass index (BMI)a, kg/m2, mean± sd	26.6 ± 5.6	26.8 ± 5.6	27.0 ± 5.6	27.1 ± 5.7	27.3 ± 5.6
Underweight (15≤ BMI <18.5) n (%)	335 (4.7)	326 (4.6)	310 (4.4)	320 (4.5)	271 (3.8)
Normal weight (18.5 ≤ BMI <25) n (%)	2756 (38.9)	2572 (36.5)	2500 (35.4)	2446 (34.6)	2349 (33.2)
Overweight (25 ≤ BMI <30) n (%)	2362 (33.4)	2388 (33.9)	2427 (34.3)	2428 (34.3)	2507 (35.5)
Obese (BMI ≥30) n (%)	1630 (23.0)	1769 (25.1)	1832 (25.9)	1881 (26.6)	1943 (27.5)
Physical activity, MET-hours/weekb	12.6 ± 14.3	12.8 ± 13.5	13.2 ± 13.6	13.2 ± 13.8	12.5 ± 13.6
Pack-years of smoking, mean ± sd	9.8 ± 18.1	9.1 ± 17.2	9.2 ± 17.5	9.5 ± 17.5	12.0 ± 20.0
Current Smoking, n (%)	659 (9.4)	504 (7.2)	487 (7.0)	515 (7.4)	707 (10.1)
Aspirin/NSAIDs use, n (%)	909 (12.8)	980 (13.9)	953 (13.5)	1005 (14.2)	1024 (14.5)
Statin Use, n (%)	164 (2.3)	t 148 (2.1)	190 (2.7)	159 (2.3)	133 (1.9)
High cholesterol, n (%)	962 (13.8)	952 (13.8)	941 (13.6)	927 (13.4)	834 (12.0)
Educational level
Less than high school, n (%)	560 (7.9)	451 (6.4)	442 (6.3)	435 (6.2)	420 (5.9)
High school/GED 5, n (%)	4093 (57.8)	4045 (57.3)	3822 (54.1)	3694 (52.2)	3855 (54.5)
≥4 years of college, n (%)	2409 (34.0)	2546 (36.1)	2795 (39.5)	2932 (41.5)	2772 (39.2)
Total alcohol intake, servings/week	2.3 ± 5.2	2.4 ± 4.8	2.4 ± 4.8	2.4 ± 4.7	2.6 ± 5.0
Food intakec, servings/week, mean ± sd
Red meatd	2.8 ± 2.7	3.2 ± 2.9	3.3 ± 3.0	3.5 ± 3.1	3.6 ± 3.1
Processed meate	1.5 ± 1.8	1.7 ± 2.0	1.8 ± 2.0	1.9 ± 2.3	2.0 ± 2.1
Sugar-sweetened beveragesf	1.3 ± 3.8	1.2 ± 3.2	1.2 ± 3.3	1.3 ± 3.5	1.2 ± 3.4
Refined grainsg	10.4 ± 6.9	11.7 ± 6.9	12.7 ± 7.3	13.6 ± 7.8	14.4 ± 8.1
Wineh	1.1 ± 3.2	1.3 ± 3.1	1.3 ± 3.3	1.3 ± 3.0	1.5 ± 3.4
Tea/coffeei	12.3 ± 12.4	13.0 ± 12.5	13.0 ± 11.8	13.5 ± 10.6	21.2 ± 11.8
Whole fruit	16.5 ± 12.7	17.3 ± 11.6	18.2 ± 11.5	19.4 ± 12.4	18.5 ± 12.4
Green-leafy vegetablesj	5.3 ± 4.9	5.7 ± 4.7	6.0 ± 4.7	6.5 ± 5.3	6.4 ± 5.3
Healthy eating index-2015k	67.0 ± 10.6	66.5 ± 10.2	65.8 ± 10.3	64.8 ± 10.6	63.5 ± 10.5
Nutrient profilec, mean ± sd
Total fiber, g/d	13.7 ± 6.2	15.1 ± 6.4	15.9 ± 6.5	16.8 ± 6.9	16.8 ± 6.8
Total carbohydrate, g/d	160 ± 61	181 ± 64	199 ± 67	219±751	227 ± 78
Total protein, g/d	50 ± 20	60 ± 22	66 ± 24	73 ± 27	76 ± 28
Branched-chain amino acids, g/d	8.7 ± 3.6	10.6 ± 3.9	11.7 ± 4.3	13.1 ± 4.9	13.6 ± 5.2
Total fat, g/d	44 ± 22	53 ± 25	57 ± 27	63 ± 30	67 ± 31
Total saturated fat, g/d	13.5 ± 7.3	17.2 ± 8.6	19.1 ± 9.5	21.7 ± 11.0	23.5 ± 11.6

^aBMI was subdivided into underweight, normal weight, overweight and obesity according to cutpoints defined by the CDC guideline (https://www.cdc.gov/obesity/adult/)

^bTotal recreational physical activity was calculated by summing the metabolic equivalent tasks for all reported activities for each individual (e.g., walking, aerobics, jogging, tennis, swimming, biking outdoors, exercise machine, calisthenics, popular or folk dancing), and presented as Metabolic equivalents (MET)-hours/week.

^cFoods (servings per day) and nutrient intakes were assessed using the WHI food frequency questionnaire and intake values were estimated using the University of Minnesota’s Nutrition Data System for Research (NDSR) database (2005 version)^{28, 29}

^dred meat (servings/day) includes ground meat including hamburgers, beef, pork, and lamb as a main dish or as a sandwich; stew, pot pie, and casseroles with meat; gravies made with meat drippings.

^eprocessed meat (servings/day) includes hot dog, chorizo, other sausage, bacon, breakfast sausage, scrapple; lunch meat such as ham, turkey; other lunch meat such as bologna.

^fsugar-sweetened beverages (servings/day) include all regular (not diet) soft drinks and fruit juice.

^grefined grains (servings/day) include total grain variable minus whole grain variable, both WHI-computed food groups.

^hwine (servings/day) includes red wine and white wine.

ⁱcoffee or tea (servings/day) includes all types of coffee or tea.

^jgreen leafy vegetables (servings/day) include cooked greens such as spinach, mustard greens, turnip greens, collards; lettuce and plain lettuce salad; mixed lettuce or spinach salad with vegetables.

^kHealthy eating index (HEI)-2015 score is a dietary index designed to measure overall dietary quality based on the Dietary Guidelines for Americans (DGA). It was calculated using 13 food components: total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein food, seafood and plant proteins, fatty acids, refined grains, sodium, added sugar, saturated fats. There are criteria for maximum and minimum points for each component and intakes between the minimum and maximum standards are scored proportionately. The HEI-2015 score is the sum of its thirteen components with a maximum of 100 points. Higher scores reflect closer conformance to the 2015 Dietary Guidance for Americans, and higher quality of the overall diet⁸.

Dairy intake and circulating concentrations of lipids

Higher intakes of total dairy, full-fat dairy, total cheese, full-fat cheese, total yogurt and butter were each associated with lower concentrations of triglycerides (Figure 1). Yogurt intake was associated with the largest percent decrease, −5.4% (−7.8%, −2.9%; FDR-P=0.0003) per serving increase of yogurt compared to total dairy and the other dairy foods. Higher intakes of total dairy, full-fat dairy, total cheese, full-fat cheese, total yogurt and butter were associated with higher HDL-C and in contrast, higher total milk was associated with lower HDL-C level, −0.5% (−0.9%, −0.1%; FDR-P=0.03). Higher milk intake was associated with lower TC level, −0.4% (−0.7%, −0.2%; FDR-P=0.01), whereas higher butter intake was associated with higher TC level, 0.6% (0.2%, 1.0%; FDR-P=0.01). In general, yogurt was associated with a more favorable lipid profile whereas butter was associated a less favorable lipid profile. Also, low-fat total dairy and low-fat cheese were generally not associated with lipids. The absolute concentrations of the lipids in the lowest and highest quintiles of each dairy food, and the percent difference between highest and lowest quintiles are presented in Table 2. The findings from these categorical analyses are consistent with the findings from the analyses with the continuous dairy variables.

Figure 1.

Percentage difference (95% confidence intervals) in lipid biomarker per serving increment in total dairy and specific dairy foods in the Women’s Health Initiative. Low-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza. Full-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza. Total dairy was calculated as the sum of daily servings of low-fat and full-fat dairy. Total milk was calculated as the sum of whole milk and low-fat milk. Low-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza. Full-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese). Total cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese. Total yogurt included non-fat yogurt and all other yogurt. Butter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking. Conversation to SI units: for triglycerides, to convert from mg/dL to mmol/L, multiply by 0.0113; for LDL, HDL, and total cholesterol, to convert from mg/dL to mmol/L, multiply by 0.0259.

Table 2.

Absolute concentrations and percent change (95% confidence intervals) in the absolute concentrations of lipids biomarkers in quintiles of total dairy and specific dairy foods among 35,352 postmenopausal women in Women’s Health Initiative^a,b

Lipid Biomarkersh	Total dairy quintilesc		% Diffd	FDRe- P-trend	Low-fat dairy quintilesf		% Diffe	FDR P-trend	Full-fat dairy quintilesg		% Diff	FDR P-trend
	Quintile 1 0.0–0.4 servings	Quintile 5 2.6–15.6 servings			Quintile 1 0.0–0.1 servings	Quintile 5 0.8–6.3 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.2–15.4 servings
Triglycerides, mg/dLh (n=18,838)	143 (138, 149)	(135, 146)	−1.9	0.03	140 (135, 146)	144 (139, 150)	2.7	0.94	146 (140, 151)	140 (135, 145)	−4.1	0.002
Low density Lipoprotein (LDL), mg/dLh (n=16,519)	141 (138, 144)	142 (139, 145)	0.8	0.70	141 (138, 144)	141 (137, 144)	−0.1	0.65	140 (137, 144)	142 (139, 146)	1.4	0.52
High density Lipoprotein (HDL), mg/dLh (n=20,513)	51.8 (50.8, 52.8)	52.7 (51.7, 53.7)	1.6	0.02	52.3 (51.4, 53.3)	52.2 (51.2, 53.2)	−0.2	0.90	51.5 (50.5, 52.5)	52.6 (51.7, 53.7)	2.2	0.03
Total cholesterol, mg/dLh (n=21,383)	227 (224, 230)	229 (226, 232)	0.9	0.26	226 (223, 229)	228 (225, 231)	0.6	0.99	227 (224, 230)	229 (226, 232)	1.0	0.25
	Total cheese quintilesi		% diff	FDR P-trend	Low-fat cheese quintilesj		% diff	FDR P-trend	Full-fat cheese quintilesk		% diff	FDR P-trend
	Quintile 1 0.0–0.3 servings	Quintile 5 2.3–12.9 servings			Quintile 1 0.0 serving	Quintile 5 0.3–3.2 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.1–11.4 servings
Triglycerides, mg/dL (n=18,838)	145 (139, 150)	140 (135, 145)	−3.1	0.01	140 (135, 146)	144 (139, 150)	2.8	0.70	146 (140, 151)	140 (135, 145)	−4.0	0.01
Low density Lipoprotein (LDL), mg/dL (n=16,519)	140 (137, 143)	142 (139, 146)	1.7	0.52	140 (138, 144)	141 (138, 145)	0.3	0.94	140 (137, 143)	142 (139, 146)	1.4	0.54
High density Lipoprotein (HDL), mg/dL (n=20,513)	51.5 (50.6, 52.5)	52.7 (51.7, 53.7)	2.2	0.02	52.3 (51.3, 53.3)	52.4 (51.4, 53.5)	0.3	0.47	51.5 (50.5, 52.5)	52.7 (51.7, 53.7)	2.2	0.03
Total cholesterol, mg/dL (n=21,383)	226 (223, 229)	229 (226, 232)	1.4	0.15	227 (224, 230)	229 (226, 232)	0.8	0.37	227 (224, 230)	229 (226, 232)	1.0	0.21
	Total milk quintilesl		% diff	FDR P-trend	Total yogurt quintilesm		% diff	FDR P-trend	Total butter quintilesn,o		% diff	FDR P-trend
	Quintile 1 0.0 serving	Quintile 5 1.0–18.0 servings			Quintile 1 0.0 serving	Quintile 5 0.3–4.0 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.1–11.4 servings
Triglycerides, mg/dL (n=18,838)	142 (137, 148)	144 (138, 149)	1.0	0.94	145 (140, 150)	139 (134, 145)	−3.9	3.0E-04	145 (140, 150)	137 (132, 143)	−5.5	2.6E-05
Low density Lipoprotein (LDL), mg/dL (n=16,519)	142 (139, 145)	140 (136, 143)	−1.6	0.11	142 (139, 145)	140 (136, 143)	−1.7	0.03	141 (138, 144)	144 (140, 147)	1.9	0.02
High density Lipoprotein (HDL), mg/dL (n=20,513)	52.3 (51.4, 53.3)	51.6 (50.7, 52.6)	−1.4	0.03	52.0 (51.1, 53.0)	52.7 (51.7, 53.7)	1.3	0.03	51.8 (50.9, 52.7)	53.5 (52.5, 54.5)	3.2	1.9E-07
Total cholesterol, mg/dL (n=21,383)	229 (226, 232)	225 (222, 228)	−1.4	0.01	228 (225, 231)	226 (223, 229)	−0.9	0.07	227 (224, 230)	230 (226, 233)	1.0	0.01

^aValues are absolute back-transformed biomarker concentrations since values were natural log-transformed prior to analysis and the bolded numbers represent statistically significant findings (i.e., FDR p value <0.05).

^bModels were adjusted for age, total energy intake, physical activity, body mass index, smoking, total alcohol, nutritional supplements, NSAID, statins, fasting status, education level, race/ethnicity, and case-control status.

^cTotal dairy was calculated as the sum of daily servings of low-fat and full-fat dairy.

^dPercent differences are the difference between highest quintiles (Q5) and the lowest quintile (Q1) in biomarker concentrations. For total yogurt and butter, the lowest quintile included participants in both quintiles 1 and 2 because of zero intake in quintile 1.

^eFDR: false discovery rate

^fLow-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza 12.

^gFull-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza.

^hConversation to SI units: for triglycerides, to convert from mg/dL to mmol/L, multiply by 0.0113; for LDL, HDL, and total cholesterol, to convert from mg/dL to mmol/L, multiply by 0.0259.

ⁱTotal cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese.

^jLow-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza.

^kFull-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese).

^lTotal milk was calculated as the sum of whole milk and low-fat milk.

^mTotal yogurt included non-fat yogurt and all other yogurt.

ⁿButter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking.

^oServings/day of dairy products were defined as: 8-oz glass skimmed or low-fat milk, 1/2 cup sherbet or ice milk, 1 cup yogurt; 8-oz glass whole milk or cream, 1 tablespoon sour cream, 1/2 cup ice cream, 2 slices of 16” pizza, 1 oz cream cheese, 1 oz or 1 slice other cheese.

Dairy intake and circulating concentrations of insulin-related/IGF signaling biomarkers

Among the eight insulin-related/IGF signaling biomarkers, intake of total dairy and specific foods, was most consistently associated with glucose and insulin compared to the other markers, though the direction of association varied by type of dairy food (Figure 2). Higher intakes of total dairy, low-fat dairy, total cheese, full-fat cheese, and total yogurt (1 serving increments) were associated with lower glucose concentrations. Yogurt was associated with the largest percent decrease, −2.1% (−2.8%, −1.3%; FDR-P=2.6E-06) in glucose concentrations. Higher intakes of total dairy, low-fat dairy, total cheese, low-fat cheese, and total yogurt were associated with lower insulin concentrations, again with yogurt showing the largest percent decrease, −11.6% (−14.5%, −6.6%; FDR-P=2.4E-12). In contrast, higher intake of butter was associated with higher insulin concentration, 2.3% (1.0%, 3.5%; FDR-F=0.003). Higher intake of milk was associated with higher IGF-1 concentration, 3.2% (1.1%, 5.3%; FDR-P=0.02), whereas for the IGFBPs, only higher intake of low-fat cheese was associated with higher IGFBP-1 concentration. Higher intakes of low-fat dairy and total yogurt were associated with lower free IGF-1 concentration. The absolute concentrations of the insulin response/IGF signaling biomarkers in the lowest and highest quintiles of each dairy food are presented in Table 3, and are in close agreement with the findings in Figure 2.

Figure 2.

Percentage increase (95% confidence intervals) in insulin-related/IGF signaling biomarker per serving increment in total dairy and specific dairy foods among participants from the Women’s Health Initiative. Low-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza. Full-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza. Total dairy was calculated as the sum of daily servings of low-fat and full-fat dairy. Total milk was calculated as the sum of whole milk and low-fat milk. Low-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza. Full-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese). Total cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese. Total yogurt included non-fat yogurt and all other yogurt. Butter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking. Conversions to SI units: for glucose, to convert from mg/dL to mmol/L, multiply by 0.0555; for C-peptide, to convert from ng/mL to nmol/L, multiply by 0.331; for insulin, to convert from uIU/mL to mIU/L, multiply by 1; for IGF-1, IGFBP-1, IGFBP-4 and free IGF-1, to convert from ng/mL to nmol/L, multiply by 0.131.

Table 3.

Absolute concentrations (95% confidence intervals) and percent change in the absolute concentrations of in insulin-related/IGF signaling biomarkers in quintiles of total dairy and specific dairy foods among 35,352 postmenopausal women in Women’s Health Initiative^a,b

Insulin response biomarkersh	Total dairy quintilesc		% Diffd	F DRe- P-trend	Low-fat dairy quintilesf		% Diffe	FDR P-trend	Full-fat dairy quintilesg		% Diffe	FDR P-trend
	Quintile 1 0.0–0.4 servings	Quintile 5 2.6–15.6 servings			Quintile 1 0.0–0.1 servings	Quintile 5 0.8–6.3 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.2–15.4 servings
Glucose, mg/dL (n=21,677)	97.5 (96.4, 98.6)	96.1 (95.0, 97.2)	−1.4	3.4E-04	97.5 (96.4, 98.7)	95.6 (94.5, 96.8)	−2.0	1.8E-07	97.0 (95.9, 98.1)	96.3 (95.2, 97.4)	−0.7	0.06
Insulin, ulU/mL (n=23,763)	9.3 (8.9, 9.7)	8.9 (8.5, 9.3)	−4.8	2.6E-05	9.4 (9.0, 9.8)	8.6 (8.2, 9.0)	−8.5	3.8E-14	8.9 (8.6, 9.3)	8.9 (8.6, 9.4)	−0.1	0.07
C-peptide, ng/mL (n=943)	1.5 (1.2, 1.9)	1.5 (1.2, 1.9)	1.6	0.99	1.5 (1.2, 1.9)	1.5 (1.2, 1.9)	−2.7	0.26	1.5 (1.2, 1.8)	1.6 (1.2, 1.9)	4.9	0.83
IGF-1i, ng/mL (n=3,042)	85.1 (73.0, 99.2)	87.7 (75.4, 102.1)	3.1	0.56	85.1 (73.2, 99.0)	90.3 (77.5, 105.3)	6.0	0.15	88.4 (75.8, 102.9)	86.6 (74.5, 100.7)	−2.0	0.94
IGFBP-1j, ng/mL (n=979)	19.7 (13.5, 28.6)	19.8 (13.7, 28.6)	0.7	0.88	19.1 (13.3, 27.5)	20.7 (14.2, 30.2)	8.3	0.06	21.7 (15.0, 31.5)	18.7 (13.0, 27.0)	−15.0	0.36
IGFBP-3, ng/mL (n=2,627)	3893 (3338, 4541)	3998 (3431, 4659)	2.7	0.99	3915 (3360, 4563)	4057 (3481, 4729)	3.6	0.40	3869 (3317, 4513)	4008 (3441, 4669)	3.5	0.94
IGFBP-4, ng/mL (n=355)	467 (372, 585)	490 (390, 617)	4.9	0.97	461 (370, 575)	511 (412, 633)	10.2	0.76	503 (402, 629)	496 (396, 622)	−1.3	0.99
free IGF-1, ng/mL (n=2,203)	0.76 (0.59, 0.97)	0.78 (0.61, 1.00)	3.0	0.54	0.73 (0.57, 0.93)	0.66 (0.52, 0.84)	−10.1	0.03	0.67 (0.52, 0.85)	0.79 (0.62, 1.00)	16.2	0.16
Insulin response biomarkersh	Total cheese quintilesk		% diff	FDR P-trend	Low-fat cheese quintilesl		% diff	FDR P-trend	Full-fat cheese quintilesm		% diff	FDR P-trend
	Quintile 1 0.0–0.3 servings	Quintile 5 2.3–12.9 servings			Quintile 1 0.0 serving	Quintile 5 0.3–3.2 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.1–11.4 servings
Glucose, mg/dL (n=21,677)	96.9 (95.8, 98.0)	96.3 (95.2, 97.4)	−0.6	0.02	97.0 (96.0, 98.1)	96.6 (95.5, 97.8)	−0.4	0.29	96.9 (95.8, 98.0)	96.2 (95.1, 97.3)	−0.8	0.03
Insulin, uIU/mL (n=23,763)	9.1 (8.7, 9.6)	8.9 (8.6, 9.3)	−2.3	0.02	9.2 (8.9, 9.7)	8.8 (8.4, 9.3)	−4.5	3.6E-06	9.0 (8.6, 9.4)	8.9 (8.6, 9.4)	−0.4	0.09
C-peptide, ng/mL (n=943)	1.5 (1.2, 1.9)	1.5 (1.2, 1.9)	−0.2	0.94	1.5 (1.2, 1.9)	1.5 (1.2, 1.9)	−4.0	0.48	1.5 (1.2, 1.8)	1.6 (1.2, 1.9)	4.8	0.89
IGF-1i, ng/mL (n=3,042)	87.0 (74.7, 101.4)	87.0 (74.9, 101.2)	−0.02	0.76	85.8 (73.8, 99.6)	87.9 (75.4, 102.5)	2.5	0.72	88.9 (76.3, 103.6)	86.2 (74.2, 100.2)	−3.1	0.84
IGFBP-1j, ng/mL (n=979)	20.1 (13.9, 29.2)	19.4 (13.4, 28.0)	−3.9	0.65	19.0 (13.1, 27.5)	22.1 (15.2, 32.1)	15.1	0.01	21.5 (14.9, 31.2)	18.6 (12.9, 26.8)	−14.9	0.37
IGFBP-3, ng/mL (n=2,627)	3886 (3333, 4531)	4002 (3435, 4661)	2.9	0.99	3948 (3386, 4602)	4078 (3496, 4758)	3.2	0.51	3867 (3315, 4511)	4007 (3440, 4668)	3.6	0.94
IGFBP-4, ng/mL (n=355)	497 (399, 619)	479 (383, 600)	−3.7	0.99	499 (401, 621)	496 (396, 622)	−0.6	0.99	494 (395, 617)	493 (394, 617)	−0.1	0.99
free IGF-1, ng/mL (n=2,203)	0.67 (0.52, 0.85)	0.78 (0.61, 0.99)	14.8	0.24	0.74 (0.58, 0.94)	0.71 (0.56, 0.91)	−3.8	0.76	0.67 (0.52, 0.85)	0.79 (0.62, 1.00)	16.5	0.21
Insulin response biomarkersh	Total milk quintilesn		% diff	FDR P-trend	Total yogurt quintileso		% diff	FDR P-trend	Total butter quintilesp,q		% diff	FDR P-trend
	Quintile 1 0.0 serving	Quintile 5 1.0–18.0 servings			Quintile 1 0.0 serving	Quintile 5 0.3–4.0 servings			Quintile 1 0.0 serving	Quintile 5 0.5–9.0 servings
Glucose, mg/dL (n=21,677)	96.8 (95.8, 97.9)	96.9 (95.8, 98.0)	0.06	0.94	97.3 (96.2, 98.4)	95.9 (94.8, 97.1)	−1.4	2.6E-06	96.9 (95.9, 98.0)	97.1 (95.9, 98.2)	0.2	0.42
Insulin, uIU/mL (n=23,763)	9.1 (8.7, 9.5)	9.2 (8.8, 9.6) **	0.7	0.94	9.4 (9.0, 9.8)	8.7 (8.3, 9.1)	−7.6	2.4E-12	9.1 (8.8, 9.5)	9.5 (9.1,9.9)	3.9	0.003
C-peptide, ng/mL (n=943)	1.5 (1.2, 1.9)	1.5 (1.2, 1.9)	−0.1	0.99	1.5 (1.2, 1.9)	1.4 (1.1, 1.8)	−5.6	0.27	1.5 (1.2, 1.8)	1.5 (1.2, 1.9)	3.2	0.99
IGF-1i, ng/mL (n=3,042)	86.9 (74.8, 100.9)	92.0 (79.0, 107.1)	5.7	0.02	85.5 (73.6, 99.3)	89.1 (76.5, 103.7)	4.1	0.94	87.4 (75.4, 101.4)	82.4 (70.7, 96.0)	−5.9	0.11
IGFBP-1j, ng/mL (n=979)	18.7 (12.9, 27.0)	18.8 (13.0, 27.2)	0.8	0.95	18.7 (13.0, 26.9)	20.8 (14.4, 30.1)	10.4	0.33	20.6 (14.3, 9.5)	18.7 (12.9, 27.0)	−9.7	0.24
IGFBP-3, ng/mL (n=2,627)	3914 (3362, 4556)	4095 (3512, 4774)	4.5	0.37	3925 (3373, 4568)	4042 (3469, 4711)	2.9	0.94	4020 (3457, 4674)	3845 (3296, 4484)	−4.4	0.35
IGFBP-4, ng/mL (n=355)	507 (410, 627)	470 (375, 589)	−7.6	0.64	501 (406, 619)	528 (423, 660)	5.2	0.84	506 (413, 620)	474 (381, 591)	−6.5	0.90
free IGF-1, ng/mL (n=2,203)	0.73 (0.67, 0.92)	0.72 (0.56, 0.91)	−1.4	0.95	0.74 (0.58, 0.93)	0.69 (0.54, 0.88)	−6.0	0.03	0.71 (0.56, 0.90)	0.80 (0.62, 1.02)	11.4	0.35

^aValues are absolute back-transformed biomarker concentrations since values were natural log-transformed prior to analysis and the bolded numbers represent statistically significant findings (i.e., FDR p value <0.05).

^bModels were adjusted for age, total energy intake, physical activity, body mass index, smoking, total alcohol, nutritional supplements, non-steroidal anti-inflammatory drug (NSAID), statins, fasting status, education level, race/ethnicity, and case-control status.

^cTotal dairy was calculated as the sum of daily servings of low-fat and full-fat dairy.

^dPercent differences are the difference between highest quintiles (Q5) and the lowest quintile (Q1) in biomarker concentration. For total yogurt and butter, the lowest quintile included participants in both quintiles 1 and 2 because of zero intake in quintile 1.

^eFDR: false discovery rate

^fLow-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza 12.

^gFull-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza.

^hConversion to SI units: for blood glucose, to convert to mmol/L, multiply by 0.0555; for C-peptide, to convert to nmol/L, multiply by 0.331; for insulin, to convert to mIU/L, multiply by 1; for IGF-1, IGFBP-1, IGFBP-4 and free IGF-1, to convert to nmol/L, multiply by 0.131.

ⁱInsulin-like growth factor (IGF)-1.

^jInsulin-like growth factor binding protein (IGFBP)-1/3/4.

^kTotal cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese.

^lLow-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza.

^mFull-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese).

ⁿTotal milk was calculated as the sum of whole milk and low-fat milk.

^oTotal yogurt included non-fat yogurt and all other yogurt.

^pButter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking.

^qServings/day of dairy products were defined as: 8-oz glass skimmed or low-fat milk, 1/2 cup sherbet or ice milk, 1 cup yogurt; 8-oz glass whole milk or cream, 1 tablespoon sour cream, 1/2 cup ice cream, 2 slices of 16” pizza, 1 oz cream cheese, 1 oz or 1 slice other cheese.

Dairy intake and circulating concentrations of inflammatory biomarkers

Among the eight biomarkers of systemic inflammation, intake of total dairy and specific foods was most consistently associated with CRP and IL-6 concentrations (Figure 3). Higher intakes of total dairy, low-fat dairy, full-fat dairy, total cheese, full-fat cheese, and total yogurt (1 serving increments) were associated with lower CRP and IL-6 concentrations, with yogurt showing the largest percent decrease, −10.0% (−13.9%, −6.0%; FDR-P=1.6E-05) in CRP concentrations and −10.0 (−15.8%, −4.2%; FDR-P=0.01) in IL-6 concentrations. Higher intake of low-fat dairy was associated with lower TNF-alpha receptor 1 concentration, −2.5% (−4.3%, −0.7%; FDR-P=0.03). IL-10, TNF alpha, TNF-alpha receptor 2, leptin and adiponectin were not associated with dairy intake. The absolute concentrations of inflammation-related biomarkers in the lowest and highest quintiles of the different dairy foods are shown in Table 4 and are similar to those presented in Figure 3.

Figure 3.

Percentage increase (95% confidence intervals) in inflammatory biomarker per serving increment in total dairy and specific dairy foods among participants from the Women’s Health Initiative. Low-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza. Full-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza. Total dairy was calculated as the sum of daily servings of low-fat and full-fat dairy. Total milk was calculated as the sum of whole milk and low-fat milk. Low-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza. Full-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese). Total cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese. Total yogurt included non-fat yogurt and all other yogurt. Butter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking. Conversions to SI units: C-reactive protein, mg/L is SI unit; for interleukin 6, interleukin 10, TNF alpha, TNF alpha receptors and adiponectin SI units are not readily available; Leptin, μg/L is SI unit.

Table 4.

Absolute concentrations (95% confidence intervals) and percent change in the absolute concentrations of inflammation biomarkers in quintiles of total dairy and specific dairy foods among 35,352 postmenopausal women in Women’s Health Initiative^a,b

Inflammation biomarkersh	Total dairy quintilesc			% diffd	FDRe- P-trend	Low-fat dairy quintilesf		% diffe	FDR P-trend	Full-fat dairy quintilesg			% diffe	FDR P-trend
	Quintile 1 0.0–0.4 servings	Quintile 5 2.6–15.6 servings				Quintile 1 0.0–0.1 servings	Quintile 5 0.8–6.3 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.2–15.4 servings
C-reactive protein, mg/L (n=26,093)	1.9 (1.8, 2.0)	1.7 (1.6, 1.9)		−8.5	5.3E-05	1.9 (1.8, 2.0)	1.8 (1.7, 1.9)	−5.8	0.002	1.8 (1.7, 1.9)	1.8 (1.7, 1.9)		−3.0	0.003
Interleukin-6, pg/mL (n=12,391)	3.0 (2.7, 3.3)	2.7 (2.4, 3.0)		−10.0	5.7E-04	3.0 (2.7, 3.3)	2.6 (2.4, 2.9)	−13.0	1.1E-04	2.8 (2.5,3.1)	2.7 (2.5, 3.0)		−2.9	0.03
Interleukin-10, pg/mL (n=1,986)	3.1 (2.4, 3.9)	3.1 (2.4, 4.0)		1.2	0.94	3.0 (2.3, 3.8)	3.1 (2.4, 4.0)	3.9	0.72	3.1 (2.4, 3.9)	3.0 (2.4, 3.9)		−1.5	0.99
TNFi alpha, pg/mL (n=2,842)	16.5 (11.7, 23.4)	16.6 (11.7, 23.5)		0.4	0.97	18.5 (13.1, 26.1)	15.6 (11.0, 22.0)	−17.0	0.37	15.8 (11.1, 22.3)	16.2 (11.4, 22.9)		2.1	0.94
TNF alpha receptor 1, pg/mL (n=3,909)	1371 (1295, 1452)	1365 (1287, 1448)		−0.5	0.95	1390(1312, 1472)	1331 (1255, 1411)	−4.3	0.03	1352 (1277, 1432)	1356(1279, 1438)		0.3	0.74
TNF alpha receptor 2, ng/mL (n=7,749)	4.8 (3.8, 6.0)	4.8 (3.8, 6.0)		0.1	0.94	4.8 (3.8, 5.9)	4.8 (3.8, 6.0)	0.2	0.94	4.7 (3.8, 5.9)	4.7 (3.8, 5.9)		0.2	0.94
Adiponectin, ng/mL (n=8,202)	8605(7777, 9521)	8590 (7762, 9506)		−0.2	0.94	8642 (7809, 9564)	8622 (7786, 9547)	−0.2	0.41	8476 (7662, 9476)	8463 (7650, 9363)		−0.2	0.99
Leptin, ng/mL (n=8,046)	25.0 (18.2, 34.1)	25.1 (18.3, 34.3)		0.6	0.99	24.5 (18.0, 33.6)	25.2 (18.5, 34.5)	2.8	0.79	25.0 (18.3, 34.1)	25.0 (18.3, 34.2)		0.2	0.99
Inflammation biomarkersh	Total cheese quintilesj			% diff	FDR P-trend	Low-fat cheese quintilesk		% diff	FDR P-trend	Full-fat cheese quintilesl			% diff	FDR P-trend
	Quintile 1 0.0–0.3 servings	Quintile 5 2.3–12.9 servings				Quintile 1 0.0 serving	Quintile 5 0.3–3.2 servings			Quintile 1 0.0–0.1 servings	Quintile 5 2.1–11.4 servings
C-reactive protein, mg/L (n=26,093)	1.9 (1.7, 2.0)	1.8 (1.7, 1.9)		−5.0	0.002	1.8 (1.7, 2.0)	1.9 (1.8, 2.0)	1.1	0.89	1.8 (1.7, 1.9)	1.8 (1.6, 1.9)		−3.4	0.003
Interleukin-6, pg/mL (n=12,391)	2.9 (2.6, 3.2)	2.7 (2.5, 3.0)		−5.2	0.01	2.9 (2.6, 3.2)	2.7 (2.4, 3.0)	−7.0	0.09	2.8 (2.5,3.1)	2.7 (2.5, 3.0)		−3.0	0.02
Interleukin-10, pg/mL (n=1,986)	3.1 (2.4, 3.9)	3.0 (2.4, 3.9)		−1.3	0.99	2.9 (2.3, 3.7)	3.1 (2.4, 3.9)	6.6	0.58	3.1 (2.4, 3.9)	3.0 (2.3, 3.8)		−3.8	0.99
TNF alpha, pg/mL (n=2,842)	15.8 (11.2, 22.3)	16.4 (11.6, 23.1)		3.3	0.96	16.2 (11.5, 23.0)	15.5 (11.0, 22.0)	−4.6	0.46	15.7 (11.2, 22.2)	16.5 (11.7, 23.3)		4.6	0.94
TNF alpha receptor 1, pg/mL (n=3,909)	1351 (1278, 1433)	1371 (1293, 1454)		1.3	0.89	1365(1289, 1445)	1339(1263, 1420)	−1.9	0.42	1351 (1276, 1431)	1359(1281, 1441)		0.6	0.75
TNF alpha receptor 2, ng/mL (n=7,749)	4.8 (3.8, 5.9)	4.7 (3.8, 5.9)		−0.4	0.94	4.8 (3.8, 6.0)	4.7 (3.8, 5.9)	−1.0	0.94	4.7 (3.8, 5.9)	4.7 (3.8, 5.9)		−0.1	0.94
Adiponectin, ng/mL (n=8,202)	8602(7777, 9513)	8536 (7715, 9443)		−0.8	0.99	8711 (7884, 9624)	8576 (7745, 9496)	−1.6	0.94	8527 (7706, 9435)	8434(7623, 9331)		−1.1	0.98
Leptin, ng/mL (n=8,046)	24.7 (18.1, 3.8)	24.9 (18.2, 34.0)		0.6	0.99	25.0 (18.2, 34.1)	24.6 (18.0, 33.6)	−1.5	0.25	25.0 (18.3, 34.2)	25.1 (18.3, 34.3)		0.4	0.95
Inflammation biomarkersh	Total milk quintilesm			% diff	FDR P-trend	Total yogurt quintilesn		% diff	FDR P-trend	Total butter quintileso,p			% diff	FDR P-trend
	Quintile 1 0.0 serving		Quintile 5 1.0–18.0 servings			Quintile 1 0.0 serving	Quintile 5 0.3–4.0 servings			Quintile 1 0.0 serving		Quintile 5 0.5–9.0 servings
C-reactive protein, mg/L (n=26,093)	1.8 (1.7, 1.9)		1.9 (1.8, 2.0)	4.6	0.67	1.9 (1.8, 2.0)	1.8 (1.7, 1.9)	−7.2	1.6E-05	1.8 (1.7, 2.0)		1.9 (1.8, 2.0)	1.7	0.38
Interleukin-6, pg/mL (n=12,391)	2.8 (2.5, 3.1)		2.8 (2.6, 3.2)	2.5	0.99	2.9 (2.6, 3.2)	2.7 (2.4, 3.0)	−6.9	0.01	2.8 (2.6,3.1)		2.9 (2.6, 3.2)	3.6	0.38
Interleukin-10, pg/mL (n=1,986)	3.1 (2.5, 3.9)		3.1 (2.4, 3.9)	−1.9	0.94	3.0 (2.4, 3.8)	3.0 (2.4, 3.9)	0.4	0.94	3.0 (2.4, 3.8)		3.2 (2.5, 4.1)	6.5	0.70
TNF alpha, pg/mL (n=2,842)	16.6 (11.8, 23.4)		15.5 (11.0, 21.9)	−7.1	0.56	17.0 (12.1, 23.9)	15.1 (10.7, 21.4)	−11.8	0.72	16.2 (11.6, 22.8)		17.3 (12.2, 24.6)	6.7	0.38
TNF alpha receptor 1, pg/mL (n=3,909)	1358 (1283, 1437)		1354 (1278, 1434)	−0.3	0.70	1377 (1301, 1456)	1358 (1281, 1440)	−1.4	0.53	1357 (1284, 1434)		1377 (1298, 1461)	1.5	0.21
TNF alpha receptor 2, ng/mL (n=7,749)	4.8 (3.8, 6.0)		4.7 (3.8, 5.9)	−1.6	0.11	4.8 (3.8, 6.0)	4.7 (3.8, 5.9)	−0.7	0.99	4.8 (3.8, 6.0)		4.7 (3.8, 5.9)	−1.4	0.54
Adiponectin, ng/mL (n=8,202)	8593 (7787, 9482)		8602 (7776, 9516)	0.1	0.94	8560 (7760, 9443)	8790 (7944, 9726)	2.6	0.11	8564 (7774, 9434)		8734 (7882, 9678)	2.0	0.91
Leptin, ng/mL (n=8,046)	24.4 (17.8, 33.3)		24.9 (18.2, 34.1)	2.3	0.27	25.2 (18.5, 34.4)	24.5 (18.0, 33.5)	−2.8	0.45	24.7 (18.1, 33.8)		25.5 (18.7, 34.9)	3.2	0.36

^aValues are absolute back-transformed biomarker concentrations since values were natural log-transformed prior to analysis and the bolded numbers represent statistically significant findings (i.e., FRD p value <0.05).

^bModels were adjusted for age, total energy intake, physical activity, body mass index, smoking, alcohol, nutritional supplements, non-steroidal anti-inflammatory drug (NSAID), statins, fasting status, education level, race/ethnicity, and case-control status;

^cTotal dairy was calculated as the sum of daily servings of low-fat and full-fat dairy.

^dPercent difference presents the difference between highest quintiles (Q5) and the lowest quintile (Q1) in biomarker concentration. For total yogurt and butter, the lowest quintile included participants in both quintiles 1 and 2 because of zero intake in quintile 1.

^eFDR: false discovery rate

^fLow-fat dairy foods included low-fat milk (non-fat, skimmed, 1–2% fat milk), part-skim or reduced-fat cheese, low-fat cottage cheese, low-fat or non-fat frozen desserts, low-fat or non-fat yogurt, other yogurt, and low-fat pizza 12.

^gFull-fat dairy foods included evaporated or whole milk, condensed milk on cereal, cottage and ricotta cheese, other cheese (cheddar/swiss/cream cheese), ice cream, milk/cream/creamer in coffee or tea, and pizza.

^hConversion to SI units: for C-reactive protein, mg/L is SI unit; for interleukin 6, interleukin 10, TNF alpha, TNF alpha receptors and adiponectin SI units are not readily available; for leptin, μg/L is SI unit;

ⁱTNF, tumor necrosis factor.

^jTotal cheese was calculated as the sum of daily servings of low-fat cheese and full-fat cheese.

^kLow-fat cheese included part-skim or reduced-fat cheese such as Mexican type cheeses or mozzarella, low-fat cottage cheese and low-fat pizza.

^lFull-fat cheese included cottage and ricotta cheese, and other cheese (cheddar/swiss/cream cheese).

^mTotal milk was calculated as the sum of whole milk and low-fat milk.

ⁿTotal yogurt included non-fat yogurt and all other yogurt.

^oButter included butter added to cooked cereal or grits, butter on bread or tortillas, and butter added as fats during cooking.

^pServings/day of dairy products were defined as: 8-oz glass skimmed or low-fat milk, 1/2 cup sherbet or ice milk, 1 cup yogurt; 8-oz glass whole milk or cream, 1 tablespoon sour cream, 1/2 cup ice cream, 2 slices of 16” pizza, 1 oz cream cheese, 1 oz or 1 slice other cheese.

DISCUSSION

This study examined associations between total dairy and individual dairy foods (milk, cheese, yogurt, butter, and low-fat varieties) with circulating concentrations of lipids, insulin-related/IGF-signaling and inflammation-related biomarkers in postmenopausal women. Interestingly, higher dairy intake (except butter) was associated with a favorable profile of lipids, insulin response, and inflammatory biomarkers, regardless of fat content, and specific dairy foods had varying associations with biomarkers. For example, higher intake of yogurt and cheese (fermented dairy) was associated with a more favorable biomarker profile than higher butter intake. This heterogeneity may imply that grouping all dairy foods into one food group may not be an optimal approach for studying the health effects of dairy foods and may partly explain the weaker associations for total dairy compared to individual dairy foods for the same biomarker, observed in the current study. To our knowledge, this is the first study to assess associations of dairy intake by type and fat content, with a comprehensive set of circulating biomarkers that are strongly associated with major chronic diseases.

Secular trends in dairy intake among Americans have fluctuated over previous decades and may provide more context for interpreting current study findings. According to annual per capita data from the US Department of Agriculture³³, after stagnating at an average of 4.5 pounds (lbs) per person between 1975 and 1995 (the mid-point of WHI enrollment), butter consumption increased by 35% between 1995 (4.6 lbs) and 2019 (6.2 lbs). American type cheese and cheese other than American has also been on an upward consumption trend since 1975 with the exception of cottage cheese that has been on the decline. In 2018, mozzarella and cheddar accounted for more than two thirds of cheese consumption. There has been a dramatic decrease in fluid milk consumption, from 247 lbs per capita in 1975 to 205 lbs in 1995 and 1996 to 141 lbs in 2019. Full-fat ice cream has been on a steady decline whereas reduced-fat or non-fat ice cream has been immune to declines. Yogurt consumption has been on a consistent rise since 1975, with a slight decrease observed since 2016. Yogurt consumption increased by 205% from 2.0 lbs per capita in 1975 to 6.1 lbs in 1995, and by 153% from 5.9 lbs in 1996 to the most recent highest consumption level of 14.9 lbs in 2014. Aggregating dairy per capita consumption across all dairy products based on milk-fat milk-equivalent, has fluctuated since 1975 but has been mainly increasing since the mid-1990s³³. It is therefore possible that if dietary intake were updated in the WHI, stronger associations may be observed between dairy intake and the biomarkers examined.

A meta-analysis of nine RCTs showed that total dairy and milk did not have an effect on LDL-C levels, in line with current study finding for total dairy⁴. Similarly, some previous studies showed that intake of milk, cheese, or yogurt may not be associated with LDL-C⁹. In contrast, an intervention study showed that butter significantly increased LDL-C, TC/HDL-C ratio and non-HDL-C compared with coconut oil³⁴. While this largely aligns with current study findings, a favorable association with blood lipids for yogurt and an unfavorable association with blood lipids for butter were observed. Butter tended to increase all types of cholesterol whereas yogurt was more discriminatory in increasing only HDL-C while decreasing TC and LDL-C. Another study found that higher intake of butter was associated with lower triglycerides and higher TC, in line with current study findings³⁵. Moreover, a parallel cross-over intervention study that included 69 men and women to determine the effect of a Mediterranean diet supplemented with dairy foods (MedDairy intervention) compared with a low-fat control diet, resulted in significantly lower triglycerides, lower TC/HDL-C ratio and higher HDL-C for the intervention compared to the control group³⁶. However, the MedDairy intervention did not have an effect on circulating glucose, insulin or CRP concentrations³¹. In contrast, strong associations with these three markers were observed in the current study. Also in contrast with the MedDairy intervention study, two previous observational studies found inverse associations between total dairy and fermented dairy products including yogurt and cheese, with fasting glucose and glycated hemoglobin³⁷. Moreover, total and especially full-fat dairy food intake were inversely associated with metabolic syndrome³⁸. In both studies, adjusting for dairy-derived saturated fatty acids (SFA) attenuated the associations, suggesting that SFA in dairy foods may be partly mediating the associations³⁹. This may suggest that dairy has a preventive role in cardiometabolic diseases mediated through the insulin response pathway. Should future well designed intervention studies lead to similar findings, then current study results would run counter to the guideline to limit full-fat dairy foods because of its SFA content. Taken in total, findings from these previous studies are consistent with current study findings indicating that higher dairy intake appears to have a neutral or inverse association with circulating biomarkers implicated in the pathways of major chronic diseases. Also, the current study support those studies showing either neutral or inverse associations between dairy intake (regardless of fat content) and risk and prognosis of CVD^{17, 40, 41}, T2D^{14, 15, 17, 42, 43} and cancer^{6, 16}. Concentrations of circulating lipids, insulin-related and IGF signaling and inflammatory biomarkers have been implicated in the pathogenesis of these diseases^{44, 45}.

The current study found that associations of dairy intake and circulating biomarkers did not necessarily differ based on fat content of dairy but rather based on specific dairy food type, with associations for fermented dairy (yogurt, cheese) largely contrasting with butter. The mechanisms underlying the associations of higher fermented dairy intake with favorable biomarker profiles are not entirely clear but may involve changes in the gut microbiome. A previous study showed that fermented dairy food plays an important role in increasing gut microbial diversity. Some probiotics in fermented dairy food, including Bifidobacterium and Lactobacillus, have been associated favorably with BMI and blood glucose⁴⁶ and Bifidobacterium may potentiate the microbial production of short-chain fatty acids⁴⁷. Dairy fat is composed of approximately 60% SFA⁴⁸, and is a major contributor to SFA intakes in Western countries⁴⁹. For this reason, healthy eating guidelines tend to recommend moderation in saturated fat intake^{8, 9}. Recent studies suggest that although SFA from red and processed meat is associated with detrimental health effects¹², SFA intake from dairy foods is either neutral^{4, 9} or has beneficial health effects^{42, 50}. Indeed, one prospective study found that very long-chain SFA including arachidic acid, behenic acid, lignoceric acid, and their sum, were inversely associated with T2D risk, with dairy fat contributing a significant proportion to the intakes of these SFAs³⁹. In addition, a study investigating associations of monounsaturated fatty acid (MUFA) intake and mortality found that MUFA from plant sources was associated with lower mortality whereas MUFA from animal sources was associated with higher mortality risk⁵¹. Findings from these single nutrient analyses strengthen the relevance of accounting for the context in which the nutrient or food is consumed, i.e., the food matrix or overall dietary pattern. For example, an intervention study found that dairy fat, eaten in the form of cheese, appears to differentially affect blood lipids compared with the same constituents eaten in different matrices, with significantly lower TC when all nutrients are consumed within a cheese matrix than within a butter matrix⁵². To further elucidate the context of dietary pattern, the two empirical hypothesis-oriented dietary patterns – empirical dietary index for hyperinsulinemia (EDIH)^{20, 53} and empirical dietary inflammatory pattern (EDIP)^{21, 54}, were derived by empirically weighting food combinations in relation to relevant circulating biomarkers (C-peptide as a marker of insulin secretion for EDIH, and CRP, IL6, TNFα-R2 for EDIP), in an unbiased manner without preconceived notions of healthy and unhealthy foods. Higher scores on these dietary patterns are contributed by higher intakes of butter and low-fat dairy products and lower intakes of pizza and full-fat dairy in complex combinations with the other food components of the dietary patterns, and higher scores indicate hyperinsulinemic or pro-inflammatory dietary patterns. This is partly reflected in the findings of the current study in relation to insulin, glucose and CRP.

The current study has several strengths including the large sample size and a comprehensive set of circulating biomarkers that are potential causal factors in the pathways of major chronic diseases. Also, several potential confounding variables were adjusted for. Limitations of this study include the overall low intake of some dairy food types in WHI making their evaluation in the current study difficult. For example, only total intakes of milk, yogurt and butter were analyzed, but not the fat content because few women reported consuming full-fat options of these foods in 1993–1998 when dietary intake was assessed in the WHI. Given the increasing trend of dairy intake since WHI baseline, additional studies in cohorts consuming greater quantities of dairy foods rich in fat content are warranted. Another potential limitation is measurement error in the FFQ^{55, 56} though the WHI FFQ was evaluated for measurement characteristics prior to using it³⁰. Also, the FFQ was not validated specifically for dairy food intake, though several nutrients with dairy as an important source had strong correlations between the FFQ and food records and food dairy data. A previous WHI study found that yogurt consumers were younger postmenopausal women, of higher socioeconomic status and had an overall healthier lifestyle compared to non-consumers⁵⁷. In the current study, adjusted for these and several other confounding factors including total energy intake. However, the potential for confounding by unmeasured variables or residual confounding by inadequately measured variables may not be completely eliminated. Though the current study had a multiethnic sample, it was composed of postmenopausal women, therefore future studies are warranted to examine these associations in the broader population of men and women over a wider age range. In addition, though the sample sizes differ between biomarkers, the exposure (dairy) distribution did not vary to an appreciable extent by outcome (biomarker) sample size, suggesting that differing biomarker sample sizes may not have induced selection bias.

CONCLUSION

In this large cross-sectional study among postmenopausal women, higher dairy intake, except butter, was associated with a favorable profile of circulating lipids, insulin response and inflammatory biomarkers, regardless of its fat content. There was heterogeneity in the associations of specific dairy foods and biomarkers, suggesting that a single dairy food may not be representative of all dairy foods and some dairy foods like butter may be associated with unfavorable biomarker profiles whereas others such as cheese and yogurt may be associated with favorable biomarker profiles despite similar saturated fat content. Current study findings do not support a putative role of dairy in cardiometabolic diseases observed in some previous studies.

Research Snapshot

Research Question:

What is the profile of circulating lipid, insulin response and inflammatory biomarkers associated with dairy intake; and are there differences by fat content of dairy or by specific dairy foods?

Key Findings:

In this large cross-sectional study among 35,352 postmenopausal women in the Women’s Health Initiative cohort, higher total dairy (except butter) intake was associated with a favorable profile of circulating lipids, insulin response and inflammatory biomarkers, regardless of its fat content. Individual dairy foods had specific associations with circulating biomarker profile, with butter associated with a less favorable profile, whereas cheese and yogurt were associated with more favorable biomarker profiles.

Abbreviations:

BMI

Body mass index

CRP

C-reactive protein

CT

clinical trial

CVD

cardiovascular disease

EDIH

empirical dietary index for hyperinsulinemia

EDIP

empirical dietary inflammatory pattern

FFQ

food-frequency questionnaire

HDL-C

high density lipoprotein cholesterol

HEI

health eating index

HPFS

health professionals follow-up study

HT

hormone therapy

IGF

insulin-like growth factors

IGF-BP

insulin-like growth factors binding proteins

IL

interleukins

LDL-C

low density cholesterol

MCSFA

medium chain saturated fats

MET

Metabolic equivalent of task

NHS

Nurses’ Health Study

NSAID

nonsteroidal anti-inflammatory drugs

OS

observational study

TC

total cholesterol

TNF

tumor necrosis factor

TNFR

tumor necrosis factor receptor

T2D

type 2 diabetes

WHI

Women’s Health Initiative