Calcium Intake Is Inversely Related to Risk of Obesity among American Young Adults over a 30-Year Follow-Up

ABSTRACT

Background

Calcium (Ca) is an essential nutrient that may play an important role in weight maintenance through its involvement in energy or lipid metabolism. However, little is known about the long-term associations of Ca intake with obesity risk.

Objectives

We aimed to prospectively examine the association between cumulative Ca intake and the incidence of obesity among American young adults over 30 y of follow-up.

Methods

Participants were from the CARDIA (Coronary Artery Risk Development in Young Adults) study. A total of 4097 of 5115 black and white individuals aged 18–30 y at baseline in 1985–1986 were included in the current analysis. Dietary and supplemental Ca intake was assessed by the validated interview-based CARDIA diet history at baseline and exam years 7 and 20. Incident cases of obesity were identified when BMI was ≥30 kg/m² for the first time since baseline. A survival analysis was performed using Cox proportional hazards regression models to estimate the HRs and corresponding 95% CIs for obesity incidence during follow-up.

Results

During a 30-y follow-up (mean ± SD: 20 ± 10 y), 1675 participants developed obesity. Cumulative total Ca intake (dietary plus supplemental Ca) was inversely associated with incidence of obesity in multivariable-adjusted analysis [quintile (Q)5 (highest intake) compared with Q1 (lowest intake): HR: 0.68; 95% CI: 0.56, 0.82; P-trend < 0.01]. This inverse association persisted among Ca supplement users (Q5 compared with Q1: HR: 0.53; 95% CI: 0.40, 0.70; P-trend < 0.01), but was not seen among nonusers.

Conclusions

Following a cohort of Americans from young adulthood to midlife, an inverse association between calcium intake and obesity incidence was observed. Further studies are needed to confirm our findings.

Introduction

Obesity remains a major public health challenge in the United States (1, 2). Excessive weight gain or obesity is an established risk factor for a series of chronic diseases, including cardiovascular disease (3), diabetes (4), and cancers (5). To date, numerous studies have been conducted to identify modifiable risk factors such as diet in relation to obesity. However, little is known about the long-term association of specific dietary components and risk of obesity, especially in the period of the lifespan between young adulthood and midlife when people are more likely to gain weight (6).

Calcium (Ca) is an essential nutrient that may play an important role in weight maintenance through its involvement in energy partitioning in adipose tissue and the modulation of adipocyte lipid metabolism (7, 8). Previous in vivo and in vitro studies support the biological effects of Ca on weight regulation. For example, in a model of mice fed a high-fat diet, Ca supplementation inhibited the expression of lipogenesis genes (e.g., fatty acid synthetase gene and lipoprotein lipase gene) and increased the expression of lipolysis genes (e.g., hormone-sensitive lipase), and consequently reduced lipid accumulation (9, 10). In addition, an in vitro study found that a sustained increase in intracellular Ca²⁺ triggered apoptosis in mature adipocytes (11). Some epidemiological studies have linked a deficit in Ca intake to obesity in adults (12–15), although controversy remains (16, 17). A meta-analysis of 33 randomized controlled trials (RCTs) reported that Ca supplementation significantly reduced weight in young adults (18), but all included RCTs had a relatively short follow-up or intervention duration (ranging from 8 wk to 6 y). It is still unclear whether long-term Ca intake is associated with obesity risk. Therefore, using data from the CARDIA (Coronary Artery Risk Development in Young Adults) study, we prospectively examined Ca intake in relation to the risk of obesity over 30 y of follow-up.

Methods

Study population

The CARDIA study followed 5115 men and women who were aged 18–30 y at recruitment primarily from 4 metropolitan areas in the United States: Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA, with the goal of investigating cardiovascular disease risk factors among young adults in the United States. The study population was roughly balanced within the centers by race (black and white), age (18–24 and 25–30 y), sex, and education levels (high school or less and more than high school). Baseline examination [exam year (Y)0] occurred during 1985–1986 at the recruitment stage. Participants were examined again at Y2, Y5, Y7, Y10, Y15, Y20, Y25, and Y30. The retention rate at each follow-up exam was 91%, 86%, 81%, 79%, 74%, 72%, 72%, and 71% of the surviving cohort during each exam period, respectively. All participants enrolled in the analysis were informed and signed the consent forms. The study procedures were reviewed and approved by boards of the CARDIA participating institutions at each field center. Further details of the research design and protocols have been published (19).

From 5115 participants, 1 person dropped out after recruitment. Then, we excluded 50 participants with energy intake outside the plausible range (600–6000 kcal/d for women and 800–8000 kcal/d for men) and 595 participants who had obesity at baseline. We further excluded 139 participants whose obesity status could not be determined owing to missing data, 18 participants who did not have data available on dietary and supplemental Ca intake, and 215 women who were pregnant during the follow-up exams. The final data set for the analysis included 4097 participants (Supplemental Figure 1).

Diet assessment

Diet was assessed at baseline, Y7, and Y20 through an interview-based questionnaire on participants’ dietary history in the month before the interview. The CARDIA dietary history (20) was assessed by specialized interviewers using a comprehensive FFQ including food type, serving size, consumption frequency, and common food additives. Foods not listed in the FFQ were also considered. Cue cards and food models were used to assist the participants to estimate portion sizes. Daily nutrient intakes including Ca intake were computed using the food and nutrient content database developed in the Nutrition Coordinating Center (University of Minnesota, Minneapolis, MN). In addition, the intakes of vitamin or mineral supplements were inquired about in the questionnaire. Ca and other nutrient intakes in this study included both dietary and supplemental intakes. The reliability and validity of the FFQ have been evaluated before (21). For estimated energy-adjusted Ca intake, correlation coefficients between the diet history questionnaire and 24-h dietary recalls ranged from 0.56 to 0.69 across ethnic and gender groups (21). Energy-adjusted cumulative average intakes of Ca (per 1000 kcal/d) and other nutrients were therefore calculated and used in all analyses to reduce measurement errors and to represent long-term dietary habits (22). In addition, the A Priori Diet Quality Score (23) was designed to estimate overall dietary quality in relation to health outcomes. From the 46 food groups CARDIA identified, 20 food groups were classified as beneficial (e.g., fish, fruit, low-fat dairy, dark green vegetables, soy products), 13 as neutral (e.g., chocolate, eggs, diet soft drink, fruit juice, lean red meat), and 13 as adverse (e.g., butter, fried foods, processed meat, sauces, whole-fat dairy). Beneficial and adverse food groups were categorized into quintiles of consumption, and were scored as 0–4 or 4–0 depending on each participant's consumption, respectively. The A Priori Diet Quality Score was calculated as the sum of the category scores for all the beneficial and adverse food groups (max: 132 points). A higher value indicates a better overall dietary quality. According to either the time an incident case of obesity occurred or the end of follow-up, the cumulative averages of the A Priori Diet Quality Scores at baseline, Y7, and Y20 were calculated and used in our analyses.

Laboratory analysis

Participants were instructed to fast for ≥12 h, and avoid smoking or vigorous exercise for 2 h before each clinic visit. Blood samples were collected from the antecubital vein and were processed centrally. Plasma glucose was measured at baseline using the hexokinase ultraviolet method by American Bio Science Laboratories, and at Y7–Y30 using hexokinase coupled to glucose-6-phosphate dehydrogenase (Linco Research) (24). An oral-glucose-tolerance test (OGTT) was performed at Y10, Y20, Y25, and Y30. Glycated hemoglobin (HbA1c) was measured by the HPLC method at Y20, Y25, and Y30 (25). Fasting plasma insulin was measured using the RIA method at Y0 and Y7–Y30 (26). At each exam, total cholesterol, HDL cholesterol, and triglycerides in plasma were measured by enzymatic methods (27), and LDL cholesterol was calculated by the Friedewald equation (28). C-reactive protein (CRP) was measured at Y7, Y15, Y20, and Y25 using a BNI nephelometer (Dade Behring) with a particle-enhanced immunonephelometric assay (29).

Ascertainment of obesity and related clinical indicators

Incident obesity was defined as a participant's BMI (in kg/m²) increasing to ≥30 for the first time since baseline (30). To calculate BMI, body weight was measured (to the nearest 0.2 kg) with a balance-beam scale and height was measured (to the nearest 0.5 cm) with a centimeter ruler at each exam. Participants wore light clothes and had no shoes on while these measurements were taken. Waist circumference was also taken at each exam, measured twice at the midway between the umbilicus and xiphoid process, and around the lateral halfway between the lowest portion of the rib cage and iliac crest (31). Skinfold thickness measurements, including triceps, subscapular, and suprailiac skinfolds, were conducted in duplicate using Harpenden calipers (Quinton) from Y0 to Y10 (32). The mean of the corresponding measurements was recorded and used in this study. After a 5-min sitting rest, diastolic blood pressure (DBP) and systolic blood pressure (SBP) were measured 3 times with 1-min intervals using a Hawksley random-zero mercury sphygmomanometer from Y0 to Y15 and an Omron HEM907XL sphygmomanometer from Y20 to Y25 (33). Mean values of the second and third measurements were used in the analysis.

Assessment of demographic and lifestyle factors and medical history

At each exam, self- or interview-based questionnaires were used to collect information on geo-demographics (age, sex, race, and field center), socioeconomics (education), lifestyle characteristics (smoking, alcohol consumption, and physical activity), and medical histories of diabetes and hypertension. History of diabetes was identified when any individual component of the following criteria was met at any exam: a fasting plasma glucose concentration ≥ 7.0 mmol/L, a nonfasting plasma glucose concentration ≥ 11.1 mmol/L, a 2-h post-OGTT of ≥11.1 mmol/L, an HbA1c ≥ 6.5%, or use of antidiabetic medication (34). History of hypertension was identified as an SBP ≥ 140 mm Hg, a DBP ≥ 90 mm Hg, or using antihypertensive medication at any exam (35). Smoking status was classified as never, former, or current smokers. Physical activity history (PAH) was collected using a PAH questionnaire to query the average amount of time spent in physical activities of various intensity (light, moderate, and vigorous) per week and analyzed through a calculation of “exercise units” (36). Mean milliliters of ethanol consumed per day were estimated to represent habitual alcohol consumption.

Statistical analysis

All analyses were completed using SAS version 9.4 (SAS Institute Inc.), with a 2-sided significance level of P value ≤ 0.05. The differences in sociodemographic and clinical characteristics of participants across quintiles of Ca intake were examined using ANOVA for continuous variables and chi-square or Kruskal–Wallis tests for categorical variables.

To estimate the HRs and corresponding 95% CIs for obesity incidence during follow-up, a survival analysis was performed using Cox proportional hazards regression models. Survival time was calculated as the time from baseline to the exam when obesity was identified, the last follow-up completed, or Y30, whichever came first. The proportional hazards assumption was assessed by testing the coefficient of the interaction between quintiles of total Ca intake (dietary plus supplemental Ca) and survival time (37). Three separate models were built, each sequentially adjusted for additional covariates and using the lowest Ca quintile as the referent. The first model was adjusted for demographic and geographic variables (age, sex, race, and field center). In the second model, socioeconomic and lifestyle variables (education, smoking, alcohol consumption, physical activity, baseline BMI, Ca supplementation, and total energy intake) were controlled in addition. Lastly, clinical measures (SBP, HDL:LDL cholesterol ratio, fasting insulin, and medical histories of diabetes and hypertension) were added as potential confounders in the final model. For each Cox model, medians of the Ca quintiles were used as a continuous variable to test for a linear association between Ca intake and obesity incidence. All covariates with repeated measurements were calculated as the cumulative average depending on the time of incident obesity identified or by the end of the study, unless otherwise specified. In addition, the association between total Ca intake and obesity incidence was examined separately among Ca supplement users and nonusers.

Multiple sensitivity analyses (SAs) were conducted to assess the robustness of the results. First, intakes of nutrients suggested to be related to obesity risk were in addition adjusted in model 3, including baseline crude fiber (38), PUFAs (39), SFAs (40), and total carbohydrates (41). In the CARDIA cohort, crude fiber was measured at baseline, whereas dietary fiber was measured at Y7 and Y20. Because the combination of crude fiber and dietary fiber may bias the results, only intake of crude fiber at baseline was included in the analyses. Second, nutrients highly correlated with Ca intake such as magnesium and potassium were adjusted in addition (42). Third, the A Priori Diet Quality Score was in addition included to reduce potential confounding by overall diet quality. Fourth, because dairy products are the main sources of dietary Ca, the association between low-fat dairy consumption and incidence of obesity was examined with additional adjustment for intakes of solid vegetables, whole fruits, and nonfried fish (43).

A few prespecified factors including age at baseline (˂25 y compared with ≥25 y), sex (female compared with male), race (black compared with white), and BMI at baseline (˂22.9 compared with ≥22.9) were considered as effect modifiers. Because of a skewed distribution, a continuous variable of Ca intake was created using the median value of each quintile for interaction tests.

The relation of Ca intake to relevant anthropometric indicators (BMI, waist circumference, and skinfold measurements) was also assessed, as were its relations with glucose metabolism (fasting insulin and glucose) and chronic inflammation (CRP), both of which are thought to have underlying mechanisms related to obesity. Generalized estimating equations with identity linkage and an exchangeable correlation structure assumption were used for all associations. The models were adjusted for the same covariates used in the final Cox model.

Results

This analysis includes 4097 participants (51% female and 48% black) with a mean ± SD age of 24.9 ± 3.6 y at baseline. The median amounts of Ca intake across quintile (Q)1 to Q5 were 274, 354, 422, 509, and 675 mg/1000 kcal, respectively. Compared with those in the lowest quintile of Ca intake, individuals in the higher quintiles were more likely to be Ca supplement users, females, white, and never smokers, with relatively higher education levels and physical activity and lower alcohol consumption, BMI, and total energy intake. In addition, they had relatively lower mean blood pressure and fasting insulin concentration and were less likely to have a history of diabetes or hypertension (Table 1).

TABLE 1.

Characteristics of the study population by quintiles of calcium intake amounts: the Coronary Artery Risk Development in Young Adults study, 1985–2015¹

Characteristics	Quintiles of Ca intake amounts					P value
	Q1 (lowest)	Q2	Q3	Q4	Q5 (highest)
n	819	820	819	820	819	—
Ca intake, mg/1000 kcal	274 [246–298]	354 [336–370]	422 [405–442]	509 [484–537]	675 [619–796]	—
Ca supplement use, yes	24.4	34.2	45.7	48.8	65.5	<0.01
Age at Y0, y	24.8 ± 3.7	25.0 ± 3.6	24.8 ± 3.7	24.6 ± 3.5	25.1 ± 3.6	0.12
Female	46.4	43.5	45.8	51.2	66.5	<0.01
Black	77.5	60.7	44.8	35.0	23.2	<0.01
Education, y	13.1 [12.0–15.0]	13.7 [12.0–16.0]	14.3 [12.6–16.1]	15.0 [12.7–16.2]	15.4 [13.0–16.6]	<0.01
Smoking status						<0.01
Never	52.0	56.7	59.9	60.8	63.9
Former	14.6	19.7	19.0	22.3	20.5
Current	33.4	23.6	21.1	16.9	15.6
Alcohol intake, mL ethanol/d	6.9 [1.0–20.9]	6.6 [1.2–18.4]	6.8 [1.2–16.5]	6.8 [1.3–16.1]	5.1 [0.8–12.0]	<0.01
Physical activity, EU	278 [164–443]	325 [202–510]	356 [219–523]	375 [252–535]	363 [243–521]	<0.01
BMI at Y0, kg/m2	23.3 ± 3.2	23.2 ± 3.0	23.1 ± 2.9	23.0 ± 2.7	22.9 ± 3.0	0.03
Total energy intake, kcal/d	3076 ± 1368	2979 ± 1231	2809 ± 1180	2741 ± 1184	2492 ± 1009	<0.01
SBP, mm Hg	113 ± 10	111 ± 9	111 ± 9	109 ± 10	108 ± 9	<0.01
DBP, mm Hg	71 ± 8	70 ± 7	70 ± 8	69 ± 7	68 ± 7	<0.01
Plasma total cholesterol, mg/dL	180 ± 32	182 ± 30	182 ± 29	181 ± 28	180 ± 28	0.36
Plasma triglycerides, mg/dL	85.2 ± 48.5	87.6 ± 53.5	86.3 ± 59.8	84.0 ± 47.3	81.8 ± 47.0	0.32
Plasma HDL:LDL cholesterol ratio	0.55 ± 0.27	0.53 ± 0.23	0.52 ± 0.22	0.54 ± 0.26	0.56 ± 0.21	<0.01
Plasma insulin, μU/mL	9.6 ± 4.4	9.0 ± 4.1	8.9 ± 4.6	8.5 ± 6.9	7.9 ± 3.2	<0.01
Plasma glucose, mg/dL	88.4 ± 10.7	89.4 ± 15.3	89.9 ± 19.1	89.4 ± 14.8	88.8 ± 10.4	0.21
History of diabetes,2 yes	7.7	6.8	5.9	4.4	4.8	0.02
History of hypertension,3 yes	55.9	49.9	48.4	41.7	37.1	<0.01

¹Values are means ± SDs, medians [IQRs], or percentages, unless otherwise indicated. P values are for any difference across quintiles of Ca intake amounts by using ANOVA, Kruskal–Wallis test, or chi-square test as appropriate. For covariates with repeated measurements, their cumulative averages were presented, unless otherwise specified. DBP, diastolic blood pressure; EU, exercise unit; Q, quintile; SBP, systolic blood pressure; Y, exam year.

²History of diabetes was identified when any individual component of the following criteria was met at any exam: a fasting plasma glucose concentration ≥ 7.0 mmol/L, a nonfasting plasma glucose concentration ≥ 11.1 mmol/L, a 2-h post–oral-glucose-tolerance test of ≥11.1mmol/L, a glycated hemoglobin ≥ 6.5%, or use of antidiabetic medication.

³History of hypertension was identified as an SBP ≥ 140 mm Hg, a DBP ≥ 90 mm Hg, or using antihypertensive medication at any exam.

During a mean ± SD follow-up of 19.9 ± 10.4 y, 1675 participants developed obesity. Total Ca intake was inversely associated with incidence of obesity after adjustment for potential confounders [Q5 (highest quintile) compared with Q1 (lowest quintile), HR: 0.68; 95% CI: 0.56, 0.82; P-trend < 0.01] (Table 2). The association remained in all SAs (SA1–SA3, Supplemental Table 1). In analyses stratified by supplement use (Table 2), the inverse association was more pronounced among Ca supplement users (Q5 compared with Q1: HR: 0.53; 95% CI: 0.40, 0.70; P-trend < 0.01), but was not found among nonusers (results for dietary Ca intake only). In addition, significantly lower incidence of obesity was observed for higher low-fat dairy consumption independently of consumption of whole vegetables and fruits and nonfried fish (Q5 compared with Q1: HR: 0.77; 95% CI: 0.64, 0.93), but this association was not as large as that observed for the total Ca intake nor consistently monotonic across all quintiles (P-trend = 0.07) (SA4, Supplemental Table 1). The proportional hazards assumption was satisfied in all analyses.

TABLE 2.

Multivariable-adjusted HRs and 95% CIs of obesity by quintiles of calcium intake amounts: the Coronary Artery Risk Development in Young Adults study, 1985–2015¹

	Quintiles of Ca intake amounts					P-trend
	Q1 (lowest)	Q2	Q3	Q4	Q5 (highest)
All participants (n = 4097)
Range, mg/1000 kcal	<319.4	319.4–388.2	388.3–463.5	463.6–576.2	≥576.3	—
Participants, n	819	820	819	820	819	—
Cases, n	385	365	343	296	286	—
Model 1	1 (Ref.)	0.93 (0.81, 1.08)	0.89 (0.77, 1.04)	0.76 (0.65, 0.90)	0.75 (0.63, 0.89)	<0.01
Model 2	1 (Ref.)	1.02 (0.88, 1.19)	0.94 (0.81, 1.10)	0.79 (0.67, 0.93)	0.71 (0.59, 0.84)	<0.01
Model 3	1 (Ref.)	0.99 (0.85, 1.15)	0.96 (0.82, 1.13)	0.77 (0.65, 0.92)	0.68 (0.56, 0.82)	<0.01
Ca supplement nonusers (n = 2307)
Range, mg/1000 kcal	<295.7	295.7–356.8	356.9–423.2	423.3–514.2	≥514.3	—
Participants, n	461	462	461	462	461	—
Cases, n	208	196	192	180	192	—
Model 3	1 (Ref.)	1.08 (0.87, 1.34)	1.05 (0.84, 1.31)	1.01 (0.81, 1.28)	1.00 (0.79, 1.26)	0.78
Ca supplement users (n = 1790)
Range, mg/1000 kcal	<359.7	359.7–432.9	433.0–516.0	516.1–643.3	≥643.4	—
Participants, n	358	358	358	358	358	—
Cases, n	197	151	135	110	114	—
Model 3	1 (Ref.)	0.66 (0.53, 0.84)	0.60 (0.47, 0.76)	0.45 (0.34, 0.58)	0.53 (0.40, 0.70)	<0.01

¹All models were constructed by using the Cox proportional hazards regression model. P-trend was examined by using the medians of Ca intake quintiles. The values of covariates, if not specified, were calculated as the cumulative averages of repeated measurements by incident obesity or last follow-up. Model 1 was adjusted for age (continuous), sex (female or male), race (blacks or whites), and study center. Model 2 was in addition adjusted for education (<12.0, 12.0–15.9, or ≥16.0 y), smoking status (never, past, or current), alcohol consumption (never, 0.1–11.9, 12.0–23.9, or ≥24 mL ethanol/d), physical activity (quintiles), Ca supplement use (yes or no), total energy (quintiles), and baseline BMI (continuous). Model 3 was in addition adjusted for systolic blood pressure (quintiles), HDL:LDL cholesterol ratio (quintiles), fasting insulin (quintiles), and medical histories of diabetes and hypertension (yes or no). In stratified analyses, Ca intake quintiles were calculated separately for each subgroup (Ca supplement nonusers and users). Q, quintile; Ref., reference.

In analyses stratified by some prespecified factors, age, sex, race, or BMI at baseline did not materially modify the observed associations (data not shown). In addition, although not statistically significant, there seemed a trend that total Ca intake was modestly associated with lower levels of some anthropometric indicators relating to obesity, especially suprailiac skinfold. Moreover, Ca intake seemed to be modestly and inversely associated with fasting insulin, but not fasting glucose or CRP concentrations (Table 3).

TABLE 3.

Multivariable-adjusted mean differences (95% CIs) in indicators of anthropometry, glucose metabolism, and chronic inflammation according to quintiles of Ca intake amounts: the Coronary Artery Risk Development in Young Adults study, 1985–2015¹

Indicator	Quintiles of Ca intake amounts					P-trend
	Q1 (lowest)	Q2	Q3	Q4	Q5 (highest)
BMI, kg/m2	0 (Ref.)	0.15 (−0.07, 0.37)	0.02 (−0.20, 0.25)	0.05 (−0.19, 0.29)	−0.04 (−0.30, 0.22)	0.08
Waist circumstance, cm	0 (Ref.)	0.18 (−0.29, 0.66)	−0.04 (−0.53, 0.46)	−0.08 (−0.61, 0.45)	−0.24 (−0.81, 0.34)	0.052
Triceps skinfold, mm	0 (Ref.)	0.08 (−0.23, 0.40)	0.04 (−0.28, 0.36)	−0.18 (−0.52, 0.16)	−0.03 (−0.39, 0.33)	0.85
Suprailiac skinfold, mm	0 (Ref.)	−0.15 (−0.60, 0.30)	−0.21 (−0.67, 0.25)	−0.26 (−0.76, 0.23)	−0.59 (−1.11, −0.07)	0.02
Subscapular skinfold, mm	0 (Ref.)	−0.02 (−0.36, 0.32)	−0.12 (−0.47, 0.23)	−0.16 (−0.53, 0.21)	−0.40 (−0.79, −0.01)	0.08
Fasting plasma insulin, μU/mL	0 (Ref.)	−0.27 (−0.63, 0.09)	−0.10 (−0.52, 0.32)	−0.63 (−1.03, −0.24)	−0.42 (−0.84, −0.0001)	0.08
Fasting plasma glucose, mg/dL	0 (Ref.)	0.91 (−0.07, 1.90)	0.68 (−0.33, 1.68)	0.62 (−0.41, 1.66)	0.72 (−0.35, 1.80)	0.93
Plasma C-reactive protein, μg/mL	0 (Ref.)	0.13 (−0.22, 0.48)	−0.16 (−0.55, 0.23)	0.20 (−0.12, 0.52)	0.06 (−0.22, 0.34)	0.53

¹Because these indicators were measured repeatedly through the follow-up, all models were constructed by using generalized estimating equations with identity linkage under an exchangeable correlation structure assumption with adjustment for the covariates listed in model 3 in Table 2. P-trend was examined by using the median of each Ca intake quintile. Q, quintile; Ref., reference.

Discussion

In this large longitudinal cohort study, we found that total Ca intake from diet and supplements was inversely associated with the incidence of obesity, independently of other major lifestyle and dietary risk factors. The observed inverse association was more pronounced among Ca supplement users, but was not seen among nonusers. In addition, this association was not materially modified by age, sex, race, or BMI at baseline.

To date, 2 meta-analyses including 41 and 33 RCTs, respectively, have reported no significant overall association between Ca supplementation and obesity (18, 44). However, the authors noticed that Ca supplementation significantly reduced weight in children/adolescents, men and premenopausal women, and individuals with normal weight (18). For postmenopausal women (45) or individuals with obesity (46), the antiobesity effect of Ca supplementation was less pronounced. Our study population consisted of young adults with a mean BMI of 23 at baseline; thus, our results are generally consistent with findings from the pooled RCTs. The different associations observed among Ca supplement users and nonusers may be explained by several reasons. First, the bioavailability of Ca varies across different food sources and supplements. Dairy products supply 50%–80% dietary Ca in most industrialized countries with a good bioavailability, whereas plant foods including vegetables, grains, legumes, and fruits supply ∼25% of dietary Ca (47). However, plant foods also contain a considerable amount of substances (i.e., oxalates, phytates, and tannins) which can bind to Ca and render it insoluble, thus making Ca unavailable for absorption (47). On the other hand, calcium carbonate, a main form of Ca in supplements (48), has been demonstrated to have a bioavailability as good as, if not better than, Ca from milk (49). Second, higher intake of dietary Ca is usually accompanied by higher total energy intake, which may partially offset the effects of Ca on obesity. Although we adjusted for energy intake in all models, the possibility of residual confounding cannot be completely ruled out. Third, Ca supplementation is often formulated with additional vitamin D, which is essential for Ca absorption and lipid metabolism. Fourth, the potential benefit of Ca intake on obesity development may be only seen at a certain amount of Ca (50), which may be achieved by supplementation.

Although dietary Ca intake was not significantly associated with obesity incidence in this study, we found an inverse association between low-fat dairy consumption and obesity, which is consistent with findings from a meta-analysis of 29 RCTs indicating that high dairy product consumption reduced weight and body fat in energy-restricted interventions (51). Considering the null association of dietary Ca in this study, it is possible that other nutrients in dairy products, such as whey protein components (e.g., branched-chain amino acids and angiotensin-converting enzyme inhibitors) (7), or total energy reduction may have synergistic effects with Ca on weight regulation.

The underlying mechanisms by which Ca may regulate weight are not fully understood, although a body of evidence has been provided by both epidemiological and laboratory studies. Our findings are likely explained by multiple mechanisms and not a single pathway. Ca can bind with fatty acids to form insoluble soaps in the colon, thereby inhibiting fat absorption and consequently contributing to weight reduction. In a double-blind controlled study investigating the effect of Ca supplementation (0, 2, or 4 g/d) on quantitative and qualitative fecal fat excretion in 24 participants, it was found that Ca supplementation increased fecal fatty acids in a dose-dependent manner (52). In a randomized double-blind crossover study including 10 men, 900 mg Ca supplementation within chocolate increased fecal fat by 2-fold (53). In addition, it has been suggested that the concentration of intracellular adipocyte Ca, which is positively affected by Ca intake through 1,25-dihydroxyvitamin D₃, regulates lipid metabolism in adipocytes. In mouse models, administration of Ca through supplementation with calcium carbonate or nonfat dry milk for 6 wk inhibited 27%–51% of expression and activity in adipocyte fatty acid synthesis, and stimulated lipolysis by 3.4- to 5.2-fold (54). Moreover, parathyroid hormone (PTH) is secreted for maintaining Ca homeostasis, and a high Ca ingestion can depress the release of PTH. A low concentration of PTH may decrease intracellular Ca concentrations, resulting in stimulation of lipolysis and inhibition of lipogenesis (55). It has been suggested that a high concentration of PTH, in response to a low Ca intake, may cause fat accumulation and thus increase BMI (56), waist circumference (57), and skinfolds (58). Furthermore, Ca is an essential nutrient involved in glucose metabolism by playing a vital role in insulin-mediated intracellular processes. Inadequate Ca intake may alter the balance between extracellular and intracellular Ca flux, and thus influence insulin release in β cells or insulin sensitivity in insulin-target tissues (59, 60). A number of RCTs have indicated that Ca supplementation reduced fasting insulin concentration and improved insulin sensitivity (61).

One major strength of our study is the long-term follow-up from young adulthood to midlife, the lifespan when individuals are more likely to experience weight change. Large study samples and long-enough intervention periods are needed in RCTs to obtain sufficient statistical power (62). Notably, a 4000-people RCT with a 30-y intervention is not practically feasible, which increases the value of our study. Another strength is the usage of repeated dietary measurements in a longitudinal setting. Cumulative average Ca intake represents long-term dietary habits and reduces the possibility of bias induced by variation of diet over time. In addition, both dietary and supplemental Ca intake were measured, which facilitates our analyses stratifying by Ca supplementation. Finally, the observed inverse associations were robust in multiple SAs, which ensures the rigor of our study.

A few limitations also need to be acknowledged. First, although dietary Ca is widely used in epidemiological studies (63, 64) and the interview-administered diet history used in CARDIA has been validated, measurement error of dietary data is unavoidable. Nevertheless, we are not aware of any systematic measurement errors. Given the strong and consistent findings, our conclusion should not be substantially biased. Second, similarly to other observational studies, residual confounding and confounding from unknown or unmeasured variables cannot be completely ruled out. Of note, our results were robust in SAs with further adjustment for a number of dietary and nondietary factors, which reduces the likelihood that our conclusion is biased. Third, some biomarkers that may affect Ca metabolism, such as 25-hydroxyvitamin D, PTH, and bone mineral density, were not available in this study. These data may help to better understand the biological effects of Ca on adipocyte regulation. Fourth, similarly to other longitudinal studies, loss to follow-up may have introduced bias to our findings. However, the overall retention rate at Y30 of CARDIA was 71%, which is excellent for a 30-y longitudinal cohort. In addition, we did not find the baseline BMI was significantly different between the participants who were lost to follow-up at Y30 and those who remained in the study.

In conclusion, this longitudinal cohort study with 30 y of follow-up indicates that long-term Ca intake is inversely associated with incidence of obesity. Further studies are needed to confirm our findings.

Notes

Supported by NIH grants R01HL081572 (to KK) and R01DK116603 (to KK). The Coronary Artery Risk Development in Young Adults Study is supported by grants from the National Heart, Lung, and Blood Institute in collaboration with the University of Alabama at Birmingham (HHSN268201800005I and HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN26820100004I). Representatives of the funding agency have been involved in the review of the manuscript but not directly involved in the collection, management, analysis, or interpretation of the data.

Author disclosures: DRJ has been a consultant to the California Walnut Commission. All other authors report no conflicts of interest.

Supplemental Table 1 and Supplemental Figure 1 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn.

Abbreviations used: CARDIA, Coronary Artery Risk Development in Young Adults; CRP, C-reactive protein; DBP, diastolic blood pressure; EU, exercise unit; HbA1c, glycated hemoglobin; OGTT, oral-glucose-tolerance test; PAH, physical activity history; PTH, parathyroid hormone; Q, quintile; RCT, randomized controlled trial; SA, sensitivity analysis; SBP, systolic blood pressure; Y, exam year.