Association Between Dietary Intake of Phosphorus and Measures of Obesity in the Jackson Heart Study

Abstract

Background

The relation between phosphorus (P) intake and obesity is equivocal, with hypotheses in both directions.

Objectives

We investigated the relationship between P intake, assessed from a current database, and calculated bioavailable P intake and obesity among African-American adults.

Methods

We examined associations between original and bioavailable P (total, added, and natural) and BMI and waist circumference (WC) in a cross-sectional study of 5306 African-American adults (21–84 y) from the Jackson Heart Study. A total of 3300 participants had complete interviews, valid dietary data, and normal kidney function. Diet was assessed by food frequency questionnaire. A novel algorithm was used to estimate P bioavailability. BMI or WC was regressed on each P variable, adjusting for total energy intake and potential confounders.

Results

After adjusting for covariates, original P (total and added) and bioavailable P (total and added) intakes (expressed/100 mg) were associated with BMI (β: 0.11, 0.67, 0.31, and 0.71, respectively; all P < 0.0001). Neither original nor bioavailable natural P was significantly associated (β: −0.03 and 0.09, respectively; both P > 0.05). When added and natural P were included in the same model, added P (original and bioavailable) intakes remained strongly associated with BMI (0.70 and 0.73, respectively; both P < 0.0001). Similar results were seen for WC. Intake of original added P tended to be more strongly associated with BMI, in females (β: 0.72; P < 0.0001) than in males (β: 0.56; P = 0.003) (P-interaction = 0.06).

Conclusions

We found that greater intake of added, not natural, which may be a proxy for intake of processed foods was associated with higher BMI and WC. These were somewhat stronger when bioavailability was considered and for women than for men. Further investigation is needed to fully understand the mechanisms driving these associations.

Introduction

The prevalence of obesity has tripled over the last 4 decades [1]. Recent National Health and Nutrition Examination Survey (NHANES 2017–2020) data showed that ∼41.9% of American adults aged 20 y and older had obesity [2]. This is a major public health concern as obesity is a risk factor for many chronic conditions, including certain types of cancer, cardiovascular disease, insulin resistance, metabolic syndrome, type 2 diabetes, asthma, hepatic and renal dysfunction, infertility, and sleep disturbances [[3], [4], [5]]. Industrialization and globalization of food markets, as well as changes in dietary behaviors, such as increasing intake of ultraprocessed foods, are significant contributors to the increasing prevalence of obesity in the United States population [[6], [7], [8]]. Mississippi has among the highest prevalence of obesity among adults in the nation, and this has increased dramatically over the past 2 decades, from 15% in 1990 to 39.1% in 2021 [9]. In 2021, 36.1% of the population of Mississippi was African-American [10], and compared with other race/ethnicity groups, African-American adults had the highest prevalence of obesity (47.1%) [11]. This highlights an urgent need to focus on obesity prevention, particularly among African Americans living in this geographic region.

Beyond energy and macronutrient intakes [12,13] micronutrient intakes may also relate to obesity [[14], [15], [16], [17], [18]]. Phosphorus (P) occurs naturally in many foods, particularly those containing protein, such as dairy products and meat [19,20]. Many ultraprocessed foods contain additive P to preserve color, moisture, and texture [[21], [22], [23]]. These compounds are found commonly in fast food, cheese, soft drinks, baked products, packaged meats, poultry and fish, and other ultraprocessed foods [22,23]. Our previous analysis, also using data from the Jackson Heart Study (JHS), showed that fish (17%), milk (6.2%), beef (5.2%), eggs (4.5%), and cheese (4.3%) were the major contributors to total P intake, while the top 5 major contributors to added P intake were fish (35.2%), beef (7.1%), processed meat (6.7%), cola beverages (5.8%), and poultry (5.3%) [24]. Although the recommended dietary allowance for P is 700 mg/d in healthy adult men and women aged 19 y and older [25], the estimated mean total P intakes in the JHS were 1299 mg/d for men and 1093 mg/d for women [24].

Data on the relationship of P with BMI and waist circumference (WC) is inconsistent. Some studies found that lower serum P was associated with higher BMI and WC among adults with normal kidney function [[26], [27], [28]], likely due to decreasing the stimulation of satiation [28], while others have linked high P intakes from processed foods with the rising prevalence of obesity in the United States [17,18,29,30]. To our knowledge, we know of no study that has examined the association between P intake and obesity in the African-American community. Further, we hypothesized that the relationship between obesity and P will be different for natural P compared with added P. Thus, the aim of this study was to investigate the association between different types of P intakes (including original P and calculated bioavailable P) and measures of obesity (WC and BMI) in African-American adults in Mississippi, aged 21 y and older, who were free of kidney disease. We also examined whether these relationships differed by sex.

Methods

The JHS is a community-based observational longitudinal study aimed at identifying cardiovascular disease risk factors in the African-American community. Between 2000 and 2004, ∼5306 African-American men and women, aged 21–84 y, were enrolled from the Tri-county area around Jackson, Mississippi (Hinds, Madison, and Rankin counties) [31,32]. Baseline information on sociodemographic and clinical characteristics was collected through in-person interviews and examinations [32]. For this analysis, data from the first examination (2000–2004) were used, as dietary data were also collected at this time. More information on the study design and data collection methodology can be found elsewhere [33,34]. The study was funded by the National Heart Lung and Blood Institute (NHLBI) and approved by a collaborative partnership among 3 institutions (Jackson State University, Tougaloo College, and the University of Mississippi Medical Center) [35]. All participants provided written informed consent. Standard protocols were followed, as described by the Declaration of Helsinki.

Study population

Approximately 5306 African-American adults, aged 21–84 y, participated in the JHS. Participants were excluded if they did not complete a food frequency questionnaire (FFQ) (n = 237), left >5 responses blank on the FFQ (n = 18), had an estimated energy intake outside of 600–4800 kcal/d (n = 254), had evidence of kidney disease [urine albumin-to-creatinine ratio of ≥30 mg/g in 24-h urine samples or estimated glomerular filtration rate (eGFR) of <60 mL/min/1.73 m², kidney disease history, or self-reported dialysis, n = 593], reported P intake of >3 SDs from the mean (n = 100), or were missing covariate data [education level (n = 682), smoking (n = 28), alcohol use (n = 8), hypertension (n = 1), diabetes (n = 1), glycosylated hemoglobin (HbA1c), n = 84]. A total of 3300 participants, 1192 men and 2108 women were included in the analysis (Figure 1).

FIGURE 1.

Flow chart of study participants. FFQ, food frequency questionnaire; JHS, Jackson Heart Study.

Measures: outcomes

To determine height and weight, participants were asked to wear an examination gown without shoes and were measured by certified technicians and nurses using calibrated devices [36]. BMI was calculated as weight in kilograms divided by height in meters squared and treated as a continuous variable. WC was measured to the nearest centimeter at the umbilicus, using an anthropometric tape. Two measures of the waist were averaged to determine WC for each participant [37].

Predictor variables of interest

An FFQ designed for this region of the country, and validated against dietary recalls and blood nutrient indicators, was used to collect dietary intake at baseline [38,39]. Data were analyzed using the nutrition data system for research (NDS-R version 2016; University of Minnesota) [24]. Phosphorus intakes, including total P, added P, natural P, bioavailable total P, bioavailable added P, and bioavailable natural P (mg/day), were calculated as described previously [24]. Briefly, we excluded foods containing no P, assigned 0 g added P for unprocessed or minimally processed food without P additives, and 100% added P for foods with zero or minimal natural P. For the 230 remaining line items, we assigned added P content from published literature; with the added P percentage of similar products; calculated added P proportion by subtracting natural from total P, compared the P:protein ratios between processed and unprocessed forms of similar foods; or compared the total P content to the sum of P from the ingredients in processed mixed dishes. The proportion of natural P from each food item was calculated by subtracting added P from total P. We then created weights for all foods containing P based on bioavailability, which were derived from literature or based on expert consensus. Individual exposure to added P, bioavailable added P, natural P, bioavailable natural P, and total bioavailable P were then calculated using the following formulas:

• 1.Individual exposure to added P: Total AP = ∑ (AP% ∗ total P in 100 g food ∗g food consumed/100)
• 2.Individual exposure to bioavailable added P: BioAP = ∑all foods (added P bioavailability% × AP% × total P in 100 g food × g food consumed/100)
• 3.Individual exposure to natural P: TotNP = ∑(NP% × total P in 100 g food × g food consumed/100)
• 4.Individual exposure to bioavailable natural P: BioNP = ∑all foods (NP bioavailability × NP% × total P in 100 g food × g food consumed/100)
• 5.Individual total bioavailable P: TotBP = BioAP + BioNP

More detail on this methodology has been previously published [24].

Covariates

Covariates included sex (male/female); age (in years); education level (<high school degree, high school degree/GED, or attended vocational school, trade school, or college); current cigarette smoking status (yes/no); alcohol use in the past 12 mo (yes/no); hypertension status (yes/no); diabetes status (yes/no); and physical activity. The latter was categorized according to the American Heart Association recommendations based on the number of minutes of moderate and/or vigorous physical activity (MVPA): poor, 0 min of MVPA; intermediate, 1–149 min of MVPA; and ideal, ≥150 min of MVPA [40]. All covariates were assessed by questionnaire during the baseline interview. Kidney function was assessed by eGFR (in ml/min/1.73 m²), using the CKD-Epi equation with serum creatinine. HbA1c (%) was measured in plasma. Calcium intake (in milligrams) and total energy intake (in kilocalories) were assessed from the FFQ.

Statistical methods

Baseline characteristics of participants were described by tertile categories of added P intake, adjusted for total energy using the residual method Q1 (<266 mg/d), Q2 (266–335 mg/d), and Q3 (>335 mg/d). Differences across the 3 added P intake groups were assessed with the Kruskal–Wallis Test for continuous variables and the χ² test for categorical variables. General linear regression PROC GLM was used to explore the effects of P intake on BMI and WC. In model 1, these associations were adjusted for age, sex, and total energy intake. Covariates considered in model 2 included sex, age, education level, current smoking status, alcohol use in the past 12 months, AHA physical activity category, hypertension status, diabetes status, eGFR, HbA1c (%), calcium intake, and total energy intake. The PROC GLMSELECT (stepwise) procedure was used to obtain a parsimonious model (model 2) to determine which predictors played the most important roles for all abovementioned potential covariates. Model 3 was adjusted for variables in model 2, but natural and additive P were included in the same model to adjust for each other. Tests for the difference of slopes between males and females were made by regressing predicted BMI and WC on P intake, including the adjustment variables in model 2. We also performed sensitivity analysis excluding participants with low BMI (including only BMI > 18.5, n = 3288). SAS, version 9.4, was used for all analyses. Plots were created using R version 4.2.2 ggplot2 and ggpubr packages.

We then determined the optimal cutoff point for P intake in predicting individuals with obesity, defined as BMI ≥ 30. Receiver operating characteristic analyses were performed to determine cutoffs for each P intake biomarker that best distinguished individuals with obesity from individuals without obesity. Positive percent agreement was defined as the percentage of individuals with obesity who were high in each P intake measure. Negative percent agreement was defined as the percentage of individuals with nonobesity who were not high in each intake measure. Overall percent agreement was defined as the sum of the individuals with obesity who were high in P intake measure and the individuals with nonobesity who were not high in each intake measure, divided by the entire cohort size. Positive and negative likelihood ratios are also presented. The P intake value with the highest Youden index (positive percent agreement + negative percent agreement − 1) was selected as the cutoff value.

Results

Participant characteristics

Compared with those having less energy–adjusted added P intake (Q1 and Q2), those with higher added P intake (Q3) were younger, more likely to consume alcohol, to present with obesity (BMI ≥ 30) and diabetes (all P ≤ 0.02) (Table 1).

TABLE 1.

Baseline characteristics of Jackson Heart Study adults, by added P intake tertile (N = 3300)

Added P intake range (mg/d)1	Q1 (n = 1100)	Q2 (n = 1100)	Q3 (n = 1100)	P2
	<266	266–335	>335
Characteristic	Mean ± SE or n (%)	Mean ± SE or n (%)	Mean ± SE or n (%)
Age (y)	53.6 ± 0.38	52.8 ± 0.36	51.9 ± 0.32	0.002
eGFR (mL/min/1.73 m2)	98.1 ± 0.53	98.8 ± 0.53	98.7 ± 0.54	0.58
HbA1c (%)	5.82 ± 0.03	5.85 ± 0.03	5.97 ± 0.04	0.06
BMI (kg/m2)				<0.001
<25	200 (18)	152 (14)	100 (9.1)
25 to <30	395 (36)	365 (33)	336 (31)
≥30	505 (46)	583 (53)	664 (60)
WC (cm)	97.7 ± 0.47	99.3 ± 0.47	102.3 ± 0.52	<0.001
Sex				0.03
Female	692 (63)	736 (67)	680 (62)
Male	408 (37)	364 (33)	420 (38)
Educational status				0.70
Less than high school	4 (0.4)	1 (<0.1)	3 (0.3)
High school graduate/GED	270 (25)	261 (24)	258 (23)
Attended college	826 (75)	838 (76)	839 (76)
Alcohol use				0.004
Yes	503 (46)	564 (51)	575 (52)
Smoking				0.91
Yes	131 (12)	125 (11)	130 (12)
Physical activity
Poor health	486 (44)	481 (44)	468 (43)	0.46
Intermediate health	389 (35)	391 (36)	374 (34)
Ideal health	225 (20)	228 (21)	258 (23)
Hypertension status				0.89
Yes	563 (51)	553 (50)	563 (51)
Diabetes status				0.02
Yes	171 (16)	195 (18)	222 (20)

¹Adjusted for total energy.

²Kruskal–Wallis rank sum test; Pearson χ² test.

Relationship of P intake with BMI

After adjustment, total P, added P, total bioavailable P, and bioavailable added P intakes were each positively associated with BMI; β ± SE were 0.11 ± 0.05, 0.67 ± 0.11, 0.31 ± 0.08, and 0.71 ± 0.12, respectively, as kg/m²/100 mg P intake (all P < 0.0001). Original natural and bioavailable natural P intakes were not associated with BMI (−0.03 ± 0.06 and 0.09 ± 0.12; both P > 0.40) (Table 2 and Figure 2). When added and natural P were included in the same model, added P and bioavailable added P intake remained strongly associated with BMI (0.70 ± 0.12 and 0.73 ± 0.12, both P < 0.0001) after adjusting for covariates. However, originally measured natural P intake became inversely associated with BMI (−0.10 ± 0.06, P = 0.11) (Table 2). In sensitivity analysis excluding those with low BMI, results did not differ meaningfully than with the full sample (Supplemental Table 1).

TABLE 2.

β coefficients for the general linear associations of BMI and P intake (g/d) (N = 3300)

Model	Original phosphorus variables				Bioavailable phosphorus variables
	Total		Added		Natural		Total added				Natural
	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P
1	0.17 ± 0.05	0.001	0.75 ± 0.12	<0.0001	0.04 ± 0.06	0.54	0.39 ± 0.08	<0.0001	0.79 ± 0.12	<0.0001	0.22 ± 0.12	0.07
2	0.11 ± 0.05	0.03	0.67 ± 0.11	<0.0001	−0.03 ± 0.06	0.58	0.31 ± 0.08	0.0001	0.71 ± 0.12	<0.0001	0.09 ± 0.12	0.44
3			0.70 ± 0.12	<0.0001	−0.10 ± 0.06	0.11			0.73 ± 0.12	<0.0001	−0.09 ± 0.12	0.44

Model 1: adjusted for age, sex, and total energy; model 2: adjusted for model 1 plus smoking status, HbA1c, hypertension, and diabetes; and model 3: adjusted for model 2 plus natural and additive P are in the same model to adjust for each other.

FIGURE 2.

β estimates of the associations of original total, natural, and added P and bioavailable total, natural and added P, with BMI (A) and waist circumference (B).

The area under the ROC curve of P intakes were total P intake (AUC: 0.67; 95%,CI: 0.66, 0.70), total bioavailable P intake (AUC: 0.68; 95% CI: 0.66, 0.70), added P (AUC: 0.69; 95% CI: 0.67, 0.70), and bioavailable added P (AUC: 0.68; 95% CI: 0.67, 0.70). Based on this, we identified the cutoff points of 1388 mg/d for total P intake, with sensitivity and specificity of 63% and 64%, respectively; of 555 mg/d for total bioavailable P, with sensitivity and specificity of 64% and 63%; of 308 mg/d for added P, with sensitivity and specificity of 63% and 65%; and of 293 mg/d for bioavailable added P, with sensitivity and specificity of 63% and 65%, to determine risk of obesity in the JHS population (Supplemental Table 2). Natural and bioavailable natural P intakes were not associated with BMI.

Relationship of P intake with WC

As with BMI, after adjustment for total energy and potential confounders, total and added, but not natural, P intakes were positively associated with WC. Expressed as cm/100 mg P intake, the coefficients were 0.26 ± 0.11 (total), 1.56 ± 0.26 (added), 0.68 ± 0.17 (total bioavailable), and 1.64 ± 0.27 (added bioavailable) (all P < 0.0001). For natural and bioavailable natural P, the coefficients with WC were −0.07 ± 0.14, and 0.13 ± 0.26, respectively (both P > 0.60) (Table 3 and Figure 2). When added P and natural P were included in the same model, adjusting for covariates, original and bioavailable added P intakes remained associated with WC (1.61 ± 0.26 and 1.70 ± 0.28 cm/100 mg P intake; both P < 0.0001), while natural and bioavailable natural P became more negatively associated (−0.23 ± 0.14 and −0.32 ± 0.27 cm/100 mg P intake; P = 0.09 and 0.23, respectively) (Table 3).

TABLE 3.

β coefficients for the general linear associations of WC and P intake (g/d) (N = 3300)

Model	Original phosphorus variables				Bioavailable phosphorus variables
	Total		Added		Natural		Total Added				Natural
	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P
1	0.38 ± 0.12	0.001	1.76 ± 0.27	<0.0001	0.06 ± 0.14	0.70	0.87 ± 0.17	<0.0001	1.86 ± 0.28	<0.0001	0.38 ± 0.27	0.16
2	0.26 ± 0.11	0.02	1.56 ± 0.26	<0.0001	−0.07 ± 0.14	0.62	0.68 ± 0.17	0.0004	1.64 ± 0.27	<0.0001	0.13 ± 0.26	0.63
3			1.61 ± 0.26	<0.0001	−0.23 ± 0.14	0.09			1.70 ± 0.28	<0.0001	−0.32 ± 0.27	0.23

Model 1: adjusted for age, sex, and total energy; model 2: adjusted for model 1 plus smoking status, HbA1c, physical activity, hypertension, and diabetes; and model 3: adjusted for model 2 plus natural and additive P are in the same model to adjust for each other.

Relation of P intake and BMI and WC by sex

Because we expected there to be differences between males and females in the relation of P intake and BMI, we examined these associations stratified by sex. In both females and males, consumption of additive P was positively associated with BMI, but this was marginally stronger in females (0.72 ± 0.16 kg/m²/100mg P intake; P < 0.0001) than that in males (0.56 ± 0.16 kg/m²/100 mg P intake; P = 0.003; P-interaction = 0.06) (Figure 3). There were no associations between natural P intake and BMI in either females or males (P > 0.50). Similarly, consumption of additive P was positively associated with WC in both sexes but was slightly stronger in females (1.65 ± 0.34 kg/m²/100 mg P) than that in males (1.32 ± 0.38 kg/m²/100 mg P intake; both P < 0.0001; P-interaction = 0.11). Again, there was no association between natural P intake and WC in either females or males (P > 0.30).

FIGURE 3.

Relation between added P intake and predicted BMI, by sex. Predicted BMI was calculated from the regression adjusted for age, sex, total energy intake (in kilocalories), smoking status, HbA1c, hypertension, and diabetes status.

Discussion

In the JHS, mean total P intakes were considerably higher than recommended dietary allowance recommendations (700 mg/d) for all 3 BMI groups, but highest in those with obesity. We found that added, but not natural P intake, was associated with greater BMI and WC. Excess dietary P intake may lead to many health problems, including bone loss, cardiovascular disease, kidney disease, even all-cause mortality [24,27,[41], [42], [43]]. To our knowledge, this is the first study to examine the association between obesity and P intake, particularly in African-American adults. The relation between P intake and obesity has been equivocal, with hypotheses presented in both directions. We found that added P, but not natural P, being more bioavailable P from foods, was significantly associated with BMI and WC in this population of African-American adults without kidney disease. This is important, as many foods contain P additives, elevating P exposure, which may place people at greater risk of obesity. In addition to total P, dietary databases should present P in bioavailable and added compared with natural forms. We found that cutoff points of 1388 mg/d for total P intake (sensitivity and specificity of 63% and 64%), 555 mg/d for total bioavailable P (sensitivity and specificity of 64% and 63%), 308 mg/d for added P (sensitivity and specificity of 63% and 65%), and 293 mg/d for bioavailable added P intake (sensitivity and specificity of 63% and 65%) were associated with obesity in the JHS population.

The findings that added P intake was associated with BMI and WC are consistent with several previous studies [18,29,44]. In one of these, a suggested potential mechanism linking high P with obesity was through its relationship with fibroblast growth factor (FGF) 23. The phosphaturic hormone (FGF23) is secreted primarily in osteoblasts/osteocytes in response to increased serum P [45]. After eating, P is rapidly absorbed and retained in serum within narrow limits through homeostatic actions by the kidney. During postprandial periods, parathyroid hormone and FGF23 reduce blood P concentration through actions on the kidney. However, exposure to brief high-serum P concentration, or elevation of serum FGF23 and parathyroid hormone after eating a high amount of added P, may be harmful to specific cell types, such as adipocytes, central to the control of energy balance, and lipid homeostasis, which may contribute to weight gain and hypertension [17]. Another study also noted that elevation of serum FGF23 was associated with significantly higher risk of obesity among older adults with normal renal function [30]. They noted that high-serum FGF23 was associated with increased fat mass [46] and suggested 2 hypotheses: 1) the indirect impact of FGF23 on metabolic effects beyond the kidneys and parathyroid glands and 2) FGF23 shares common structural and biological features with growth factors FGF19 and FGF21, which directly control fat mass and glucose metabolism in rodents. More studies should be done to explore the mechanisms for this association.

The association between inorganic P intake and obesity is also linked to excessive consumption of processed foods, particularly ultraprocessed foods. Higher consumption of refined sugar, saturated fat and transfat, sodium, and preservatives containing P are often used in processed foods, while P in nature comes from both plant (nuts, legumes, and whole grains) and animal sources (dairy, meat, fish, and eggs) [47,48]. Plant sources are associated with a salt of phytic acid (myoinositol 1,2,3,4,5,6-hexakis dihydrogen phosphate) and are poorly absorbed through the digestive tract [49]. In contrast, inorganic PO43-, a foodstuff preserver in most factory-processed diets, is 90%–100% absorbed [50]. Therefore, although natural P is an indicator of whole foods, added P is an indicator of processed food, from which it is more bioavailable. As a result, P serum concentration increases more rapidly after eating foods with added P. In a previous study [51], the differential impact of added, compared with natural P, on FGF23 was noted; FGF23 did not change after 1 wk on a low-additive diet but increased by 23% after 1 wk on a highly processed food diet with enhanced added P content. Previous studies have also shown that overweight and obesity are associated with high P, along with greater intake of energy, from high ingestion of processed foods [30,52].

We found that the associations between added P intake and BMI were marginally stronger in women than that in men. Again, the association between increased consumption of P and obesity may be partially explained by increased concentration of the phosphaturic hormone FGF23. One study found that older female mice fed a casein (higher bioavailable P) diet upregulated their intact FGF23 2-fold higher than older male mice. It is possible that older female mice use increasing FGF23 production to maintain their serum P in the normal range, potentially due to direct effects of estradiol on osteocytes [53]. As a result, elevated FGF23 may be associated with increases in BMI [46]. Further studies are required to understand how metabolic effects as well as unknown confounders by sex affect the association of P intake on BMI and WC.

Findings from this study are important for public health and further support recommendations for consuming whole, unprocessed, or minimally processed foods, rather than processed foods. A strength of our study was investigating the association of obesity and dietary P intake by separating P into total, natural, and added. Further, the sample size of African Americans in this study was large and dietary intake was assessed using a regionally validated FFQ. Our study has some limitations. With a cross-sectional study, causality cannot be inferred. Longitudinal studies should be conducted to more fully understand associations between P intake and obesity. We calculated added P in foods using our novel algorithm, which is dependent on literature values of P content of foods and their bioavailability, which needs further updating and validation in future studies. Further, limitations of the FFQ include a lack of detail on the level of processing of some foods. Finally, although several demographic and health behavior risk factors were adjusted in the study, it is possible that unmeasured confounders remain.

In conclusion, we found strong associations between added P, but not natural P with obesity, measured both as BMI and WC, which were somewhat stronger when bioavailability was considered, and for women than for men. Further investigation is needed to fully understand the mechanisms driving these associations.

Author contributions

The authors’ responsibilities were as follows – KLT, CND: designed and guided the research and had primary responsibility for the final content; CND, OJA: worked with the food database and conducted data analysis; CND, KLT: drafted the manuscript; SEN, OJA: critically reviewed and contributed to the manuscript and literature review; and all authors: have read and approved the final manuscript.

Conflict of interest

The authors report no conflicts of interest.

Funding

The Jackson Heart Study (JHS) is supported and conducted in collaboration with Jackson State University (HHSN268201800013I), Tougaloo College (HHSN268201800014I), the Mississippi State Department of Health (HHSN268201800015I/HHSN26800001) and the University of Mississippi Medical Center (HHSN268201800010I, HHSN268201800011I and HHSN268201800012I) contracts from the National Heart, Lung, and Blood Institute (NHLBI) and the National Institute for Minority Health and Health Disparities (NIMHD). The views expressed in this manuscript are those of the authors and do not necessarily represent the views of the National Heart, Lung, and Blood Institute; the National Institutes of Health; or the United States Department of Health and Human Services.

Data availability

Data described in the manuscript, code book, and analytic code will be made available upon request pending approval of a JHS Manuscript Proposal or Ancillary Study Proposal, found at https://www.jacksonheartstudy.org/

Appendix A. Supplementary data

The following is the Supplementary data to this article: