Bioavailability of phosphorus and kidney function in the Jackson Heart Study

ABSTRACT

Background

High phosphorus (P) exposure may have negative effects on kidney function. Nutrient databases provide total P, but bioavailability varies by source.

Objectives

We aimed to assess natural, added, and bioavailable P intake, and to relate these to estimated glomerular filtration rate (eGFR) in the Jackson Heart Study (JHS).

Methods

A total of 3962 African-American participants of the JHS, aged 21–84 y, with urine albumin:creatinine ratio < 30 mg/g, and eGFR ≥ 60 mL · min⁻¹ · 1.73 m⁻², and without self-reported kidney disease, were included. Diet was assessed by FFQ. We assigned P in foods as naturally occurring or added, and weighted intake by P bioavailability, based on published literature. Relations between P variables and eGFR were assessed using multivariable regression.

Results

Mean ± SE intakes were 1178 ± 6.7 mg and 1168 ± 5.0 mg for total P, 296 ± 2.8 mg and 291 ± 2.1 mg for bioavailable added P, and 444 ± 2.9 mg and 443 ± 2.2 mg for bioavailable natural P, in participants with eGFR = 60–89 and ≥90 mL · min⁻¹ · 1.73 m⁻², respectively. Major sources of total P included fish, milk, beef, eggs, cheese, and poultry; and of added P, fish, beef, processed meat, soft drinks, and poultry. After adjustment for confounders, P intakes, including total (β ± SE: −0.32 ± 0.15; P = 0.03), added (β ± SE: −0.73 ± 0.27; P = 0.01), bioavailable total (β ± SE: −0.62 ± 0.23; P = 0.01), and bioavailable added (β ± SE: −0.77 ± 0.29; P = 0.01), were significantly associated with lower eGFR. However, neither total nor bioavailable P from natural sources were associated with eGFR.

Conclusions

Added, but not natural, P was negatively associated with kidney function, raising concern about P additives in the food supply. Further studies are needed to improve estimation of dietary P exposure and to clarify the role of added P as a risk factor for kidney disease.

Introduction

Phosphorus (P) is found naturally in foods, particularly those containing protein, such as dairy products, red meat, poultry, fish, and legumes (1, 2). P is also contributed from food additives (3). Phosphate additives are used to enhance the taste, texture, and/or color of food, as well as other properties of processed foods (4, 5). Phosphoric acid in soft drinks is an acidifying agent, and P additives act as melting agents in the production of soft cheeses, as anticaking agents in pudding and instant coffee, as leavening agents in baked goods, and as emulsifiers in many food products (6–8). P compounds are added to fish and poultry to extend plumpness and fresh appearance (2). The US FDA classifies P additives as “generally recognized as safe” (GRAS), meaning that they do not require premarket approval for use in food (9). Although P additives are often (but not always) listed in the small-font ingredient list, reporting of the P content on nutrition facts labels is not compulsory, and differences in P content across brands are often not known (10).

In the United States, the RDA for P is 700 mg/d in healthy adults (11). However, national surveys show that some adults consume 2–3 times this amount (12, 13). One study in Ohio estimated that 44% of foods analyzed contained food additive phosphates (14). This high P content is particularly concerning for patients with chronic kidney disease (CKD), where P is not easily excreted (15). High P from the diet may cause hyperparathyroidism, loss of calcium from bone (16, 17), and calcium-phosphate deposition in the vascular wall, contributing to cardiovascular disease (CVD) (18). Studies have raised concerns about effects in the general population, particularly in relation to the development of kidney disease (18, 19), but more studies of dietary P and clinical risks are needed.

Exposure to high-P processed foods is increasing (14, 20–22). About 60% of organic P in meat and dairy is absorbed (23, 24), whereas absorption from phytate-rich foods, including whole grains, legumes, and nuts, is lower, owing to lack of the enzyme phytase which hydrolyzes P compounds (25). In vegetables and fruit, bioavailability is 20%–50%, based on phytate content (26). However, absorption of P from additives is 80%–100% (5, 27). Existing food composition databases do not distinguish between naturally occurring and additive P, yielding only total P, with unknown bioavailability. Values for recipes in databases are often based on raw materials (28) and may underestimate true P exposure. A recent study distinguished total, added, and natural P from food using general assumptions; however, this approach does not capture added P from mixed dishes and frozen meat products (29). Further, between 1988–1994 and 2015–2016, it was reported that total P intake increased, whereas added P declined (29), which is unlikely, owing to the growing preference for fast foods (30). The objective of the present study was to develop a new tool to estimate bioavailability of P, using data from the Jackson Heart Study (JHS). We created an algorithm to calculate new P metrics based on the estimated contributions of natural and added P, using weights reflecting the estimated P bioavailability of different food sources. Preliminary evidence for the validity of the new dietary P metrics was assessed by associations with estimated glomerular filtration rate (eGFR) and, when available, single-day urinary P.

Methods

The JHS is a community-based prospective cohort study that began in 1998, to address the burden of and risk factors for CVD in the African-American community. During enrollment, from September 2000 through March 2004, ∼5306 African-American adults, aged 21–84 y, were recruited from the Jackson, MS Metropolitan area (Hinds, Madison, and Rankin counties) (31). At baseline, information on sociodemographic and clinical characteristics was collected through in-person interviews and examinations. We used data from the first examination, from 2000–2004, when dietary data were collected. Additional information was gathered on CVD risk factors, socioeconomic and sociocultural factors, and biochemical measures (31). Details of the study design and data collection have been described elsewhere (31–33). The study was approved by the Institutional Review Boards of Jackson State University, Tougaloo College, and the University of Mississippi Medical Center (34). Written informed consent was obtained from all participants and standard protocols were followed as described by the Declaration of Helsinki.

Nutrient intake was assessed at baseline, using an FFQ designed for this region of the country and validated against dietary recalls and blood nutrient indicators (35, 36). Only participants who completed the FFQ were included in these analyses (n = 4797) (Figure 1). Participants were excluded if they left >5 responses blank on the FFQ, had an estimated energy intake outside of 600–4800 kcal/d (n = 268), had evidence of kidney disease (urine albumin:creatinine ratio ≥ 30 mg/g in 24-h urine samples, eGFR < 60 mL · min⁻¹ · 1.73 m⁻², or self-reported kidney disease or dialysis) (n = 594), were missing covariate data (n = 141), or reported P intakes >3 SDs from the mean (n = 100). Individual daily nutrient intakes were extracted from Nutrition Data System for Research (NDSR) software, version 2017 (Nutrition Coordinating Center, University of Minnesota). Kidney function was assessed as eGFR, using the CKD-EPI equation for serum creatinine (37). A total of 3962 participants were included in the analysis. Twenty-four-hour urinary P was measured, also at baseline, using a colorimetric assay on a Roche c311 analyzer at the JHS Core Laboratory among a subset of participants (n = 1026). The interassay CV for this assay was <3.6%. Urine samples were stored at −70°C before analysis. Among the 3962 participants included, those with missing 24-h urinary P (n = 3216) or 25-hydroxyvitamin D [25(OH)D] concentration (n = 2), or who were influential outliers based on Cook's D (n = 7), were excluded from 24-h urine P validation.

FIGURE 1.

Flowchart of participants. eGFR, estimated glomerular filtration rate; JHS, Jackson Heart Study; 25(OH)D, 25-hydroxyvitamin D.

Estimation of original (natural and added) and bioavailable (natural and added) P

P bioavailability of individual food items captured on the FFQ was calculated using an algorithm (Table 1, Box 1), which considered the content of added P in foods, with weights based on the P bioavailability of natural foods and additives from the literature. Published sources included organic P from phytate-rich foods, such as whole grains, legumes, and nuts (10, 24, 25); organic P from meat and dairy (24); plant-based P (24); inorganic phosphate salts (5, 24, 38); ready-to-eat cereals (14, 25, 39); packaged meat and poultry products (2, 14, 21, 40, 41); breads and rice (14, 38, 42); beverages (27, 40); milk (40); cheese (40); and cakes (14, 40). Supplemental Methods 1 and Supplemental Table 1 provide further details.

TABLE 1.

Methodology for estimating the bioavailable P content of foods¹

Steps	Food groups
Exclude foods with 0 P	Artificial sweeteners
	Oils
	Vitamin and mineral supplements without P
	Water / certain energy drinks / powdered drink mixes
	Sugar and salt
Assign 0 g added P to foods which are either unprocessed or minimally processed with no P additives	100% fresh fruit and vegetables
	100% fresh (unpackaged) meat, seafood, and tofu
	100% fresh or dried legumes / nuts and seeds
	Unprocessed cereals
	Alcoholic beverages
	Whole / 1% / 2% / skim milk or soy milk
	Regular tea / coffee
	Eggs
	Plain rice, pastas, cereal grains, and flours
	Canned fruit and vegetables
Assign 100% total P as added P for foods which contain minimal or 0 natural P	Regular soft drinks, sport drinks, flavored water, and energy drinks
	Instant tea and coffee
	Supplement powder / supplement drinks
	Vitamin / mineral supplements with P
	Commercial syrup, jelly, gravy
Assign values from published literature for added P using specific values when available, compare P values for specific foods from Nutrition Data System for Research (NDSR) to similar foods from published literature, or percentages based on similar products	15% in commercial bread and baked goods (14)
	38% in packaged meat (beef and pork) (14)
	34% in packaged poultry (2, 69)
	32% in catfish and 18% for other packaged fish and seafood (21, 22)
	44% in processed cereals (14)
	28% in snack chips and snack foods (14)
Calculate added P in processed foods based on comparison with the P:protein ratio from comparable fresh foods	Percentage added P in processed foods = {[(P:protein)processed foods – (P:protein)fresh foods] / (P:protein)processed foods} * 100%
Calculate added P in commercial mixed dishes based on comparison with the P:protein ratio from comparable homemade recipe dishes	Percentage added P in commercial mixed dishes = {[(P:Protein)commercial mixed dishes – (P:Protein)homemade recipe dishes] / (P:Protein)commercial mixed} * 100%

¹Multiply appropriate proportions of each food by the bioavailability for respective components using weights derived from the literature and expert consensus (see Box 1, Supplemental Methods 1, and Supplemental Table 1). P, phosphorus.

BOX 1.

Calculation of P variables. P, phosphorus.

Statistical methods

Bioavailable P expressed in mg/100 g for each food item, amount per medium portion, mean intake (g/d) of each food item, and mean total P intake (mg/d) were calculated from available dietary data. Food contributors were ranked based on percentage contribution to total P intake. Pearson's correlations were used to examine relations between P measures.

To partially validate our dietary P metrics, we evaluated correlations and linear associations of original (total, natural, and added) and bioavailable (total, natural, and added) P with 24-h urine P among those completing the 24-h urine collection (n = 737). Baseline characteristics of participants were described by kidney function [low (eGFR = 60–89 mL · min⁻¹ · 1.73 m⁻²) compared with higher function (eGFR ≥ 90 mL · min⁻¹ · 1.73 m⁻²)] (43). Based on prior studies (44–48), variables likely to confound the association between P intake and kidney function were adjusted. These included age, sex, BMI, education, smoking, alcohol use, diabetes, hypertension, glycated hemoglobin, and total protein and energy intakes. Model 1 was adjusted for age, sex, and total energy intake. Model 2 was adjusted for the variables in model 1 plus smoking status, hypertension, educational status, and glycated hemoglobin. Model 3 was adjusted for the variables in model 2 plus diabetes and total protein intake (g/d). Model 4 was adjusted for the variables in model 3, but bioavailable natural and additive P were included in the same model to adjust for each other. P values < 0.05 were considered statistically significant. BMI, alcohol intake, income level, fasting plasma glucose concentration, and blood pressure medication and calcium channel blocker medication use were also considered but were dropped from final models owing to lack of influence on the results. All analyses were performed using SAS version 9.4 (SAS Institute Inc.).

Results

Participant characteristics

Participants with low (60–89 mL/min · min⁻¹ · 1.73 m⁻²) compared with higher eGFR (≥90 mL · min⁻¹ · 1.73 m⁻²) were older (mean age: 61 compared with 51 y), more likely to be male (40% compared with 35%), to have less than high school education (26% compared with 15%), and to have diabetes (22% compared with 19%) or hypertension (67% compared with 47%), whereas fewer were heavy alcohol users (5% compared with 9%) or current smokers (9% compared with 14%) (Table 2).

TABLE 2.

Characteristics of study participants¹

	eGFR (60–89 mL/min)			eGFR (≥90 mL/min)
Characteristics	n	%	Mean ± SD	n	%	Mean ± SD	P value
Age, y	1426		61.3 ± 10.8	2536		51.0 ± 11.7	<0.00012
BMI, kg/m2	1426		31.3 ± 6.54	2535		31.8 ± 7.61	0.032
Glycated hemoglobin, %	1426		5.98 ± 1.10	2536		5.91 ± 1.32	0.112
Total P intake, mg/d	1426		1178 ± 6.703	2536		1168 ± 5.043	0.284
Added P intake, mg/d	1426		312 ± 2.943	2536		306 ± 2.213	0.114
Natural P intake, mg/d	1426		866 ± 5.553	2536		862 ± 4.173	0.634
Total bioavailable P intake, mg/d	1426		740 ± 4.503	2536		734 ± 3.393	0.294
Bioavailable added P intake, mg/d	1426		296 ± 2.793	2536		291 ± 2.103	0.114
Bioavailable natural P intake, mg/d	1426		444 ± 2.903	2536		443 ± 2.183	0.904
Total protein, g	1426		67.8 ± 31.3	2536		75.0 ± 34.8	<0.00012
Total energy intake, kcal	1426		1928 ± 796	2536		2111 ± 893	<0.00012
Sex
Female	862	60.5		1646	64.9
Male	564	39.5		890	35.1		0.015
Educational status
Less than high school	374	26.2		375	14.7
High school graduate/GED	256	18.0		462	18.2
Attended vocational school, tradeschool, or college	796	55.8		1703	67.1		<0.00015
Alcohol use in past 12 mo
None	876	61.7		1239	48.8
Moderate	471	33.2		1066	42.1
Heavy	74	5.21		228	9.13		<0.00015
Current smoker
No	1300	91.2		2173	85.7
Yes	126	8.84		363	14.3		<0.00015
Diabetes status
No	1116	78.3		2056	81.1
Yes	310	21.7		480	18.9		0.035
Hypertension status
No	474	33.2		1347	53.1
Yes	952	66.8		1189	46.9		<0.00015

¹eGFR, estimated glomerular filtration rate; P, phosphoru; GED, General Education Development Tests..

² P difference in population means (t test).

³Least-squares means and SEs, adjusted for age, sex, and energy intake.

⁴ P difference in least-squares means (general linear models).

⁵ P relation across categories (chi-square test).

Food sources

Fish was the top contributor to total (17%) and, particularly, to added (∼35%) P intake in this population (Table 3). Other major contributors to total P intake included milk (∼6%), beef and eggs (∼5%), cheese, poultry, corn bread, hot breakfast cereal, and rice (each ∼4%), and beans, pizza, dairy desserts, white bread, and processed meats (each ∼3%). In addition to fish, important contributions to added P included beef and processed meat (∼7%), soft drinks (∼6%), poultry (∼5%), pork and snack chips (∼4%), corn bread products (∼3%), followed by white bread, pizza, pasta, and dairy desserts (each 1%–2%). Resulting bioavailable P, weighted more heavily by added P owing to its higher bioavailability, was also mainly from fish (∼21%), followed by beef and milk (each ∼6%), poultry (∼5%), cheese, eggs, corn bread, processed meat, pizza, snack chips, pork, and rice (each ∼3%–4%), and soft drinks and pasta (each ∼2%). Owing to common sources, natural and additive P were highly correlated (r = 0.69; P < 0.01), and each was highly correlated with total P (r = 0.97 for natural and r = 0.84 for additive P; P < 0.01 for each). Correlations were more modest after re-expressing these P metrics as nutrient densities indexed to energy intake (r = 0.16 for natural and additive P; r = 0.90 for natural and total P; and r = 0.57 for additive and total P; P < 0.01 for each).

TABLE 3.

Food contributors to total, bioavailable, and added P in the Jackson Heart Study¹

Food groups	Food intake, g/d	Total P intake, mg/d	Contribution to total P, %	Contribution to added P, %	Contribution to bioavailable P, %
Fish	50.2	204	17.0	35.2	21.1
Milk	83.6	73.6	6.15	0.0	5.9
Beef	38.5	62.8	5.24	7.1	5.9
Eggs, egg sandwiches	31.0	53.5	4.46	2.6	4.1
Cheese	12.5	51.3	4.28	2.4	4.4
Chicken / turkey	33.6	47.2	3.94	5.3	4.5
Corn bread, corn muffins	30.2	43.9	3.66	3.0	3.9
Hot breakfast cereal	121	43.8	3.66	0.3	1.3
Rice	62.9	41.3	3.45	1.4	3.1
Beans / legumes	36.2	39.6	3.31	0.3	1.2
Pizza	11.3	33.0	2.75	2.0	3.0
Ice cream, sherbet, yogurt	30.0	32.1	2.68	1.3	2.7
Ready-to-eat cereal	12.4	31.1	2.60	0.5	1.1
White breads and products	25.8	30.7	2.57	1.7	2.6
Pasta	30.9	27.1	2.26	1.2	2.2
Pork	10.2	24.5	2.05	3.6	2.5
Processed meat, sausage, hot dogs	19.9	30.0	2.50	6.7	3.4
Snack chips	14.7	29.7	2.48	3.8	2.6
Potatoes	37.9	19.8	1.66	0.6	1.2
Whole-grain breads and products	9.66	19.5	1.63	0.9	1.4
Soft drinks2	244	18.3	1.53	5.8	2.3

¹ n = 4797 participants with valid dietary data. P, phosphorus.

²Most of the P in soft drinks is from cola beverages.

Relation of P intake with 24-h urine P

In multivariable linear regression models adjusted for age, sex, eGFR, BMI, 25(OH)D concentration, and total energy intake (Table 4), 24-h urinary P was 0.12 mg higher for each milligram of total P intake (P = 0.002), 0.14 mg higher for each milligram of natural P intake (P = 0.001), 0.12 mg higher for each milligram of bioavailable P (P = 0.04) and 0.21 mg higher for each milligram of natural bioavailable P (P = 0.01).

TABLE 4.

β coefficients for multivariable general linear associations of 24-h urine P and bioavailable, added, total, and natural P (mg/d)¹

	Original phosphorus variables		Bioavailable phosphorus variables
	β ± SE2	P	β ± SE2	P
Total	0.12 ± 0.04	0.002	0.12 ± 0.06	0.04
Added	0.06 ± 0.09	0.53	0.06 ± 0.09	0.53
Natural	0.14 ± 0.04	0.001	0.21 ± 0.08	0.01

¹ n = 737. P, phosphorus; 25(OH)D, 25-hydroxyvitamin D.

²Model adjusted for age, sex, estimated glomerular filtration rate, BMI, 25-hydroxyvitamin D, and total energy.

Relation of P intake with eGFR

After adjustment for total energy and potential confounders, total P intake was significantly and inversely associated with eGFR (−0.32 lower eGFR per 100 mg/d, P = 0.03) (Table 5). Importantly, total added P was more strongly associated with eGFR (−0.73 lower eGFR per 100 mg/d, P = 0.01). Bioavailable total P was also more significantly associated with eGFR (−0.62 lower eGFR per 100 mg/d, P = 0.01) than the original total P. Added bioavailable and added original P had similar results (−0.77 and −0.73 lower eGFR per 100 mg/d, respectively; each P = 0.01). It is also noteworthy that in the final models, where natural and added P were adjusted for each other, the association of added P became even stronger (−0.84 lower eGFR per 100 mg/d, P = 0.004). In contrast, neither original nor bioavailable natural P intakes were associated with eGFR (−0.15 and −0.21 lower eGFR for each 100 mg/d, respectively; P = 0.30 and P = 0.47, respectively).

TABLE 5.

β Coefficients for the general linear associations of estimated glomerular filtration rate (kidney function) and bioavailable, added, total, and natural P (per 100 mg/d)¹

	Original P variables						Bioavailable P variables
	Total		Added		Natural		Total		Added		Natural
Model	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P
1	−0.12 ± 0.10	0.24	−0.43 ± 0.23	0.06	−0.09 ± 0.12	0.42	−0.19 ± 0.15	0.21	−0.45 ± 0.24	0.06	−0.03 ± 0.23	0.89
2	−0.09 ± 0.10	0.36	−0.40 ± 0.22	0.07	−0.05 ± 0.12	0.69	−0.17 ± 0.15	0.24	−0.42 ± 0.23	0.07	−0.02 ± 0.23	0.95
3	−0.32 ± 0.15	0.03	−0.73 ± 0.27	0.01	−0.15 ± 0.14	0.30	−0.62 ± 0.23	0.01	−0.77 ± 0.29	0.01	−0.21 ± 0.29	0.47
4			−0.84 ± 0.28	0.003	−0.23 ± 0.15	0.13			−0.84 ± 0.29	0.004	−0.38 ± 0.30	0.21

¹ n = 3962. Model 1: adjusted for age, sex, and total energy. Model 2: adjusted for model 1 plus smoking status, hypertension, educational status, and glycated hemoglobin. Model 3: adjusted for model 2 plus diabetes and total protein. Model 4: adjusted for model 3 plus natural and additive P are in the same model to adjust for each other. P, phosphorus.

Discussion

Mean total P intakes in the JHS were 1299 mg/d for males and 1093 mg/d for females, well above the 700-mg RDA, and similar to NHANES estimates of ∼1373 mg/d for both males and females aged 20 y or older (49). Based on available evidence from the literature, we weighted P intake by bioavailability and estimated that a median of 30% of the total P consumed was from additive, as opposed to natural, sources. Protein foods, particularly fish, and including milk, beef, eggs, cheese, and poultry were the top contributors to total P intake, followed by grain products, particularly corn bread, hot breakfast cereals, and rice. The top contributors to added P overlapped considerably, but disproportionally included fish, beef, processed meat, soft drinks, and poultry. Our results agree with other nationally representative surveys identifying meats, dairy, and grains as top contributors to total P intake (49). We extend these findings by showing that fish and other protein foods are also a major source of additive and bioavailable P among African Americans in the southeastern United States.

Excessive dietary P may be a health concern because of its ability to drive hormonal changes, including elevation in parathyroid hormone (PTH) and fibroblast growth factor 23 (FGF23), which may ultimately affect bone and cardiovascular health (50–52). Many studies have documented prospective associations between higher concentrations of these hormones, higher circulating P, and CVD events and mortality (52, 53). Emerging evidence from animal models also suggests important effects. In a study in felines, added inorganic P resulted in temporary postprandial elevation in plasma P, with a concentration-dependent trend in diets with a calcium:phosphorus ratio < 1, altering plasma PTH. However, dietary P obtained from natural foods had no effect on postprandial plasma P (54). Although dietary P may be the most modifiable determinant of these risk factors, few studies have looked at dietary P specifically (55–58). This is due, in part, to the difficulty of quantifying the most biologically relevant exposure, highly absorbed food additive phosphates. Many healthy foods, including fish, whole grains, nuts, and legumes, are high in dietary P. Correlations between dietary P and other healthful dietary components may confound these associations with outcomes. On the other hand, food additive phosphates increase P intake without additional nutritional value and are more readily absorbed. Our algorithm provides a novel lens to distinguish P from additive as opposed to natural sources and to consider the total load of absorbed P, or bioavailable P. This will enable deeper studies of dietary P to understand potential risks of highly absorbed food additive phosphates.

Our findings suggest potential value for distinguishing additive from natural, and bioavailable from total P. For instance, we found that higher bioavailable and added P were significantly associated with lower eGFR, whereas natural P was not. Disruption in FGF23, a major hormone regulating phosphorus absorption and excretion, has been associated with poorer kidney function over time (59–61). Thus, it is possible that additive P is a risk factor for kidney function decline.

However, the cross-sectional nature of this analysis means that the direction and causal nature of the effect cannot be inferred. Nonetheless, patients with lower kidney function may also be at higher risk with exposure to highly bioavailable and additive P because of their decreased ability to excrete a P load. For this reason, avoiding P-additive-containing foods is necessary in patients with kidney disease (41). Higher additive P intake at lower eGFR may be a health concern in this population and requires further study. Additional studies in healthy populations, and in those with kidney disease, will be made possible with our novel algorithm.

Although total P was significantly associated with urinary P, bioavailable P correlated only modestly with 24-h urine P in our cohort. The utility of a single 24-h urinary P as an accurate biomarker of habitual intake is controversial, because there is considerable day-to-day variability in urinary P, which will lead to attenuation in associations. Studies in small cohorts from Germany and Japan suggest moderate (r ∼ 0.4–0.5) to high correlation of dietary P intake from weighed dietary records with concurrent 24-h urine P measurement (62, 63). Even under controlled feeding conditions, studies have demonstrated only modest correlation (r = 0.30–0.39), although the food P composition was directly tested (64). Other feeding studies showed wide day-to-day variability in 24-h urine P, with multiple collections needed to increase reliability (65). The attenuated correlation in our study may be a reflection of the discordant timescales of these measurements (habitual intake from the FFQ compared with a single-point measure of urinary P). In fact, correlation between FFQ estimates and 24-h urine markers is often modest, even for nutrients with low day-to-day variability such as total energy and protein, for which correlations range from r = 0.10 to 0.30 (66). It is likely that intake of added P has greater day-to-day variability than natural P and, therefore, would be even less likely captured with a single day of urinary output. Additional validation against multiple days of 24-h urine P and other P biomarkers, such as serum P and FGF23, is needed.

Improved dietary collection methods could further improve this algorithm. Available data on the P content of foods suggest wide variation within food categories (29). Without more detail on the specific quality of each food item purchased (for example, fresh chicken or fish compared with enhanced, homemade compared with commercial baked products), we made assumptions based on the most commonly used products within each category. Based on our knowledge of foods consumed in Jackson, MS, we are confident that the most commonly selected foods within each category will be those that are least expensive. In other environments, where organic and natural foods are more commonly used, this assumption may not hold as strongly. In addition to more detail on the quality of purchased items, food preparation may also affect P availability. For example, soaking, boiling, and roasting may alter the phytate content of foods and thus change the P bioavailability (67, 68). Future studies collecting more detail on specific food products and preparation are needed to provide more precise estimates of additive and natural P to further improve the accuracy of the algorithm.

In conclusion, we developed an algorithm to estimate bioavailable, natural, and additive P from a validated FFQ. This is important, because P exposure may affect the risk of CKD, as well as other potential health conditions. The bioavailability of P varies considerably, and added P is widely used in our food supply. For many foods, the actual P content is unknown or may vary by brand and preparation. Future work will continue to improve this algorithm by capturing these additional relevant details. Moving forward, this algorithm and its future iterations will allow deeper study of the potential health impact of highly absorbable dietary P in population-based studies.

Notes

The Jackson Heart Study is supported by 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) under the foregoing listed contracts from the National Heart, Lung, and Blood Institute (NHLBI) and the National Institute for Minority Health and Health Disparities.

The views expressed in this article are those of the authors and do not necessarily represent the views of the NHLBI, the NIH, or the US Department of Health and Human Services.

Supplemental Methods 1 and Supplemental Table 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/ajcn/.

Abbreviations used: CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; FGF23, fibroblast growth factor 23; JHS, Jackson Heart Study; P, phosphorus; PTH, parathyroid hormone; 25(OH)D, 25-hydroxyvitamin D.

Data Availability

Data described in the article, 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/.