Use of organophosphate flame retardants (OPFRs) including tris(1,3-dichloro-2-propyl) phosphate, triphenyl phosphate, tris(1-chloro-2-propyl) phosphate, and tris-2-chloroethyl phosphate, in consumer products is on the rise because of the recent phase out of polybrominated diphenyl ether (PBDE) flame retardants. Some of these chemicals are also used as plasticizers or lubricants in many consumer products.

To assess human exposure to these chlorinated and non-chlorinated organophosphates, and non-PBDE brominated chemicals in a representative sample of the U.S. general population 6 years and older from the 2013–2014 National Health and Nutrition Examination Survey (NHANES).

We used solid-phase extraction coupled to isotope dilution high-performance liquid chromatography-tandem mass spectrometry after enzymatic hydrolysis of conjugates to analyze 2,666 NHANES urine samples for nine biomarkers: diphenyl phosphate (DPHP), bis(1,3-dichloro-2-propyl) phosphate (BDCIPP), bis-(1-chloro-2-propyl) phosphate (BCIPP), bis-2-chloroethyl phosphate (BCEP), di-n-butyl phosphate (DNBP), di-p-cresylphosphate (DpCP), di-o-cresylphosphate (DoCP), dibenzyl phosphate (DBzP), and 2,3,4,5-tetrabromobenzoic acid (TBBA). We calculated the geometric mean (GM) and distribution percentiles for the urinary concentrations (both in micrograms per liter [μg/L] and in micrograms per gram of creatinine). We only calculated GMs for analytes with an overall weighted frequency of detection >60%. For those analytes, we also a) determined weighted Pearson correlations among the log10-transformed concentrations, and b) used regression models to evaluate associations of various demographic parameters with urinary concentrations of these biomarkers.

We detected BDCIPP and DPHP in approximately 92% of samples, BCEP in 89%, DNBP in 81%, and BCIPP in 61%. By contrast, we detected the other biomarkers much less frequently: DpCP (13%), DoCP (0.1%), TBBA (5%), and did not detect DBzP in any samples. Concentration ranges were highest for DPHP (<0.16–193 µg/L), BDCIPP (<0.11–169 µg/L), and BCEP (<0.08–110 µg/L). Regardless of race/ethnicity, 6–11 year old children had significantly higher BCEP adjusted GMs than other age groups. Females had significant higher DPHP and BDCIPP adjusted GM than males, and were more likely than males to have DPHP concentrations above the 95th percentile (odds ratio = 3.61; 95% confidence interval, 2.01–6.48).

Our results confirm findings from previous studies suggesting human exposure to OPFRs, and demonstrate, for the first time, widespread exposure to several OPFRs among a representative sample of the U.S. general population 6 years of age and older. The observed differences in concentrations of certain OPFRs biomarkers by race/ethnicity, in children compared to other age groups, and in females compared to males may reflect differences in lifestyle and exposure patterns. These NHANES data can be used to stablish a nationally representative baseline of exposures to OPFRs and when combined with future 2-year survey data, to evaluate exposure trends.

Flame retardants are substances added to plastics, furniture, upholstery, electrical equipment, electronic devices, textiles and other consumer goods to reduce product flammability and to comply with strict government fire safety standards and regulations. Polybrominated diphenyl ethers (PBDEs) were the most popular flame retardants used in the United States (

OPFRs and other contemporary flame retardants have been detected in indoor environments, including dust (

In laboratory studies, OPFRs readily metabolize to their dialkyl or diaryl phosphates and to a variety of hydroxylation products (

Little is known about the potential health effects of these emerging contaminants. OPFRs have been associated with adverse reproductive/developmental and neurological effects in animals (

Furthermore, certain OPFRs may also affect human health (

Because of the increased use of contemporary flame retardants in consumer products and the chemicals potential adverse health effects, interest exist in understanding the extent of human exposure. Biomonitoring of contemporary flame retardants will help us understand the extent of exposure to these chemicals and will provide the information needed to set up reference ranges which may be used to evaluate future exposures. In this manuscript, we report for the first time the urinary concentrations of nine contemporary flame retardant metabolites in a representative sample of the U.S. general population 6 years of age and older, stratified by age group, sex, and race/ethnicity.

The National Health and Nutrition Examination Survey (NHANES) is the result of the National Health Survey Act of 1956, which granted legislative authorization for a continuing survey to provide current statistical data on the amount, distribution, and effects of illness and disability in the United States (

For this study, we analyzed 2,666 spot urine samples collected from a random one-third subset of 2013–2014 NHANES participants six years of age and older to maintain the representative design of the survey. NCHS Research Ethics Review Board reviewed and approved the study protocol. All respondents gave informed written consent to participate in the survey; parents or guardians provided consent for participants younger than 18 years (

The urine specimens were collected, aliquoted, and shipped on dry ice to the CDC`s National Center for Environmental Health where they were stored at −70 ⁰C until analysis. We quantified nine urinary biomarkers: BDCIPP, bis(1-chloro-2-propyl) phosphate (BCIPP), bis(2-chloroethyl) phosphate (BCEP), DPHP, di-p-cresylphosphate (DpCP), di-o-cresylphosphate (DoCP), dibutyl phosphate (DNBP), dibenzyl phosphate (DBzP), and TBBA. The analytical method, described in detail elsewhere (

We analyzed the data using Statistical Analysis System (SAS) (version 9.3; SAS Institute Inc., Cary, NC) and SUDAAN (version 11, Research Triangle Institute, Research Triangle Park, NC). SUDAAN incorporates sample weights and design variables to account for the complex, clustered design of NHANES. For concentrations below the LOD, we imputed a value equal to the LOD divided by the squared root of 2 (

We stratified age, self-reported in years at the last birthday, in four groups: 6–11, 12–19, 20–59, and ≥60 years. Based also on self-reported data we defined four race/ethnicity groups: non-Hispanic black, non-Hispanic white, all Hispanic, and Other. We calculated the geometric mean (GM) and distribution percentiles for the urinary concentrations (both in micrograms per liter [μg/L] and in micrograms per gram of creatinine [μg/g creatinine]) using the survey sampling weights, which account for unequal selection probabilities caused by the cluster design and the oversampling of certain groups. We only calculated GMs for analytes with an overall weighted frequency of detection >60%. For those same analytes, we also determined weighted Pearson correlations among the log_{10}-transformed concentrations (not corrected for creatinine).

We used analysis of covariance to examine the associations of age group (6–11, 12–19, 20–59, ≥60), sex, race/ethnicity (four above categories), log_{10} creatinine, and all possible two way interaction terms on the log-transformed concentrations of the target analytes detected in >60% of participants. We calculated the adjusted GM concentrations of these five biomarkers (in μg/L). We log-transformed the concentrations of the target analytes and creatinine because their distributions were right-skewed.

For each analyte, to reach the final multivariate logistic regressions model, we used backward elimination including all the two-way interaction terms, with a threshold of P < 0.05 for retaining the variable in the model, using Satterwaite-adjusted F statistics. We evaluated for potential confounding by adding back into the model each of the excluded variables one by one and examining changes in the β coefficients of the statistically significant main effects. If addition of one of these excluded variables changed a β coefficient by ≥ 10%, the variable was re-added to the model.

Last, we conducted weighted multiple logistic regressions to examine the likelihood of having concentrations above the 95^{th} percentile (an arbitrary value we selected as example of higher than average concentrations) for the five biomarkers detected in >60% participants with sex, age group, race/ethnicity, and creatinine, variables selected on the basis of statistical, demographic, and biologic considerations.

We quantified urinary concentrations of nine contemporary flame retardant metabolites in 2,666 NHANES 2013–2014 samples. In Tables

Bivariate Pearson`s correlation analysis (

The final DPHP linear regression model included sex, race, age, sex*age (P=0.0157), sex*race (P=0.0099) and log creatinine (

In the final multivariate logistic analysis, log_{10} creatinine (P <0.001), sex (P=0.0003), and age (P<0.001) were significantly associated with the likelihood of DPHP exceeding the 95^{th} percentile (^{th} percentile (adjusted odds ratio [OR] = 6.6; 95% CI, 2.5–17.2); we observed no other statistically significant differences by age. Females were 3.6 times more likely than males to have concentrations of DPHP above the 95^{th} percentile (adjusted OR = 3.61; 95% CI, 2.01–6.48).

The final BDCIPP linear regression model included race, sex (P=0.004), age, age*race (P=0.0037), and log creatinine (P<0.001) (

In the final multivariate logistic analysis, log_{10} creatinine (P <0.0001) and age (P<0.0001) were significantly associated with the likelihood of BDCIPP exceeding the 95^{th} percentile. Children 6–11 years of age were 10.74 times more likely than adults 60 years and older to have BDCIPP concentrations above the 95^{th} percentile (adjusted OR = 10.74; 95% CI, (4.42– 26.1); we observed no statistically significant differences among other age groups.

Race, sex, age, sex*age (P=0.004), and age*race (P=0.04) were significant in the final BEtCP linear regression model (

Among children, non-Hispanic whites (P=0.0056) and Other race persons (P=0.0411) had significantly higher adjusted GMs than non-Hispanic blacks. Among adolescents, Hispanics had significantly higher adjusted GMs than non-Hispanic blacks (P=0.0435). For the 20–59 years olds, non-Hispanic whites had higher adjusted GMs than non-Hispanic blacks (P=0.0367). Regardless of sex, children and adolescents had significantly higher adjusted GMs than adults 20–59 years old (P <0.01). We observed a significant sex difference for the 20–59 years olds (P=0.0111), but not for the other age groups.

In the final multivariate logistic analysis, log_{10} creatinine (P<0.001) and age (P<0.02) were significantly associated with the likelihood of BCEP concentrations exceeding the 95^{th} percentile. Children, but not adolescents or adults 20–59 years of age, were more likely than adults 60 years and older to have BCEP concentrations exceeding the 95^{th} percentile (adjusted OR = 2.05; 95% CI, 1.07 −3.93).

The final DNBP linear regression model included race (P<0.001), sex (P=0.0005), and age (P<0.001) (

In the final multivariate logistic analysis, log_{10} creatinine (P<0.001) and age (P=0.001) were significantly associated with the likelihood of DNBP exceeding the 95th percentile. Children, but not adolescents or adults 20–59 years of age, were more likely than adults 60 years and older to have concentrations of DNBP exceeding the 95th percentile (adjusted OR = 3.6; 95% CI, 1.98 −6.5).

Race and age were significant in the final BCIPP linear regression model (

The final logistic multivariate analysis included log_{10} creatinine (P<0.001) and age (P<0.003). Children (OR = 3.57; 95% CI,1.97–6.45) and adults 20–59 years old (OR = 1.58, 95% CI, 1.03–2.43) were more likely than adults ≥60 years of age to have concentrations of BCIPP above the 95^{th} percentile; we observed no other age-related differences.

For the first time, we present nationally representative data for five OPFRs among Americans 6 years of age and older. We detected BDCIPP, DPHP, BCEP, DNBP in >80% and BCIPP in >60% of the samples analyzed which suggests widespread exposure to the precursors of these OPFRs. Depending on the OPFR metabolite, concentration ranges spanned 3–4 orders of magnitude. These results are in agreement with previous research involving convenience samples of non-occupationally exposed populations (

The lower detection frequency of BCIPP compared to BDCIPP was unexpected considering the presence of relatively similar levels of their corresponding precursors TCIPP and TDCIPP in house dust (

The moderate correlation between concentrations of DPHP and DBCIPP and, of interest, weaker correlations between DPHP and BCEP, BDCIPP and BCEP, and BDCIPP and BCIPP might reflect the ubiquity of the parent compounds for a wide variety of commercial applications or the combined use of the parent compounds in commerce. TDCIPP is the most common flame retardant for polyurethane foam used in upholstered furniture, automotive products, carpet padding and gymnastic equipment (EPA 2015;

Differences in exposure sources may explain the weak correlations we observed between BDCIPP and DNBP. TNBP is predominantly used as plasticizer in the manufacture of plastics and vinyl resins, as an antifoam agent for concrete (

Conclusions about particular exposures are hard to draw from the correlations among urinary biomarkers because there are many possible household sources of flame retardants and limited information is available on OPFRs exposure sources and metabolism. Biotransformation and kinetics studies of several organophosphate triesters are limited to one in vitro study in herring gulls using a hepatic microsomal metabolism assay that suggested the fastest depletion rate for TNBP followed by TCIPP, TPHP, and lastly by TDCIPP (

We note that concentrations of DPHP, BDCIPP, DNBP, and BCIPP differ by sex. In general, females had higher adjusted GMs of these OPFR metabolites than males, which may reflect differences in exposure patterns and/or metabolism by sex. Of interest, not only did females had higher adjusted GM concentrations of DPHP, but females were also 3.6 times more likely than males to have concentrations of DPHP above the 95^{th} percentile. Together, these findings suggest that females experience higher exposures to TPHP, perhaps from higher use of nail polish by women compared to men because TPHP is a known ingredient in nail polish (

We also observed that concentrations of certain OPFRs differed by race/ethnicity. Racial differences may relate to lifestyle, diet, and use of OPFRs-containing products. Last, the detection of OPFR metabolites in children suggests that exposure occurs at young ages. Additionally, higher concentrations of OPFRs in children compared to other age groups suggest that exposures in children are higher than in adults, assuming there are no differences in pharmacokinetics. This finding is in agreement with other studies in which concentrations in children were higher than in adults, particularly for BDCIPP and DPHP (

We present the first nationally representative assessment of exposure to several contemporary flame retardants among Americans 6 years of age and older. We found that exposure to several OPFRs is widespread with BDCIPP, DPHP, BCEP, DNBP, and BCIPP detected in the majority of the samples analyzed. These data can be used to stablish a nationally representative baseline of exposures to OPFRs and when combined with more 2-year survey data, to evaluate trends in exposure. Adjusted GMs of OPFR metabolites were higher in females than males, which may reflect lifestyle differences affecting exposure patterns. The detection of OPFR metabolites in children suggests that exposure occurs at young ages. Higher concentrations of OPFRs in children compared to other age groups might reflect that exposures in children are higher than in adults.

We thank Mr. Charlie Chambers and Ms. LaTasha Williams for technical assistance. This work was supported by the Centers for Disease Control and Prevention, U.S. Department of Health and Human Services.

Disclaimer

The findings and conclusions of this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Geometric mean and selected percentiles of bis(13-dichloro-2-propyl) phosphate (BDCIPP) concentrations in urine (in μg/L first row and in μg/g creatinine second row) for the U.S. population six years of age and older. Data from the National Health and Nutrition Examination Survey 2013-2014.

Geometric mean | Selected percentile (95% CI) | N | ||||
---|---|---|---|---|---|---|

50th | 75th | 90th | 95th | |||

2666 | ||||||

2666 | ||||||

6-11 years | 421 | |||||

421 | ||||||

12-19 years | 427 | |||||

427 | ||||||

20-59 years | 1266 | |||||

1266 | ||||||

60 years and older | 552 | |||||

552 | ||||||

Males | 1344 | |||||

1344 | ||||||

Females | 1322 | |||||

1322 | ||||||

All Hispanic | 671 | |||||

671 | ||||||

Non-Hispanic Blacks | 587 | |||||

587 | ||||||

Non-Hispanic Whites | 1015 | |||||

1015 | ||||||

Others | 393 | |||||

393 |

CI, confidence interval; N, sample size; %, detection frequency

Geometric mean and selected percentiles of diphenyl phosphate (DPHP) concentrations in urine (in μg/L first row and in μg/g creatinine second row) for the U.S. population six years of age and older. Data from the National Health and Nutrition Examination Survey 2013-2014.

Geometric mean | Selected percentile (95% CI) | N | ||||
---|---|---|---|---|---|---|

50th | 75th | 90th | 95th | |||

2666 | ||||||

2666 | ||||||

6-11 years | 421 | |||||

421 | ||||||

12-19 years | 427 | |||||

427 | ||||||

20-59 years | 1266 | |||||

1266 | ||||||

60 years and older | 552 | |||||

552 | ||||||

Males | 1344 | |||||

1344 | ||||||

Females | 1322 | |||||

1322 | ||||||

All Hispanic | 671 | |||||

671 | ||||||

Non-Hispanic Blacks | 587 | |||||

587 | ||||||

Non-Hispanic Whites | 1015 | |||||

1015 | ||||||

Others | 393 | |||||

393 |

CI, confidence interval; N, sample size; %, detection frequency

Geometric mean and selected percentiles of bis(2-chloroethyl) phosphate (BCEP) concentrations in urine (in μg/L first row and in μg/g creatinine second row) for the U.S. population six years of age and older. Data from the National Health and Nutrition Examination Survey 2013-2014.

Geometric mean | Selected percentile (95% CI) | N | ||||
---|---|---|---|---|---|---|

50th | 75th | 90th | 95th | |||

2666 | ||||||

2666 | ||||||

6-11 years | 421 | |||||

421 | ||||||

12-19 years | 427 | |||||

427 | ||||||

20-59 years | 1266 | |||||

1266 | ||||||

60 years and older | 552 | |||||

552 | ||||||

Males | 1344 | |||||

1344 | ||||||

Females | 1322 | |||||

1322 | ||||||

All Hispanic | 671 | |||||

671 | ||||||

Non-Hispanic Blacks | 587 | |||||

587 | ||||||

Non-Hispanic Whites | 1015 | |||||

1015 | ||||||

Others | 393 | |||||

393 |

CI, confidence interval; N, sample size; %, detection frequency

Geometric mean and selected percentiles of dibutyl phosphate (DNBP) concentrations in urine (in μg/L first row and in μg/g creatinine second row) for the U.S. population six years of age and older. Data from the National Health and Nutrition Examination Survey 2013-2014.

Geometric mean | Selected percentile (95% CI) | N | ||||
---|---|---|---|---|---|---|

50th | 75th | 90th | 95th | |||

2666 | ||||||

2666 | ||||||

6-11 years | 421 | |||||

421 | ||||||

12-19 years | 427 | |||||

427 | ||||||

20-59 years | 1266 | |||||

1266 | ||||||

60 years and older | 552 | |||||

552 | ||||||

Males | 1344 | |||||

1344 | ||||||

Females | 1322 | |||||

1322 | ||||||

All Hispanic | 671 | |||||

671 | ||||||

Non-Hispanic Blacks | 587 | |||||

587 | ||||||

Non-Hispanic Whites | 1015 | |||||

1015 | ||||||

Others | 393 | |||||

393 |

CI, confidence interval; N, sample size; %, detection frequency

Geometric mean and selected percentiles of bis(1-chloro-2-propyl) phosphate (BCIPP) concentrations in urine (in μg/L first row and in μg/g creatinine second row) for the U.S. population six years of age and older. Data from the National Health and Nutrition Examination Survey 2013-2014.

Geometric mean | Selected percentile (95% CI) | N | ||||
---|---|---|---|---|---|---|

50th | 75th | 90th | 95th | |||

2666 | ||||||

2666 | ||||||

6-11 years | 421 | |||||

421 | ||||||

12-19 years | 427 | |||||

427 | ||||||

20-59 years | 1266 | |||||

1266 | ||||||

60 years and older | 552 | |||||

552 | ||||||

Males | 1344 | |||||

1344 | ||||||

Females | 1322 | |||||

1322 | ||||||

All Hispanic | 671 | |||||

671 | ||||||

Non-Hispanic Blacks | 587 | |||||

587 | ||||||

Non-Hispanic Whites | 1015 | |||||

1015 | ||||||

Others | 393 | |||||

393 |

CI, confidence interval; N, sample size; %, detection frequency

LOD means less than the limit of detection (LOD= 0.1 μg/L).

Not calculated. Proportion of results below limit of detection was too high to provide a valid result.

Weighted Pearson’s correlation coefficients describing bivariate associations among OPFR urinary metabolites^{a}.

Variable | log_DPHP | log_BDCIPP | log_DNBP | log_BCEP | log_BCIPP |
---|---|---|---|---|---|

1 (2447) | |||||

0.45 (2,319) | 1 (2,461) | ||||

0.29 (2,045) | 0.29 (2,026) | 1 (2,151) | |||

0.36 (2,257) | 0.41 (2,256) | 0.20 (1,979) | 1 (2,375) | ||

0.19 (1,580) | 0.28 (1,587) | 0.12 (1,433) | 0.29 (1,563) | 1 (1,632) |

All correlations were statistically significant (p< 0.0001). Number of samples with detectable concentrations are in parentheses. Concentrations were not corrected for creatinine.

Adjusted geometric mean urinary concentrations (95% confidence interval, CI) in μg/L of OPFR biomarkers in various demographic groups.^{a}

VARIABLE | BIOMARKER | ||||
---|---|---|---|---|---|

BDCIPP | DPHP | BCEP | DNBP | BCIPP | |

Male | 0.76 (0.67-0.87) | 0.17 (0.15-0.20) | |||

Female | 0.96 (0.84-1.09) | 0.21 (0.17-0.25) | |||

6-11 years | 0.32 (0.24-0.43) | 0.31 (0.27-0.36) | |||

12-19 years | 0.18 (0.15-0.21) | 0.17 (0.14-0.19) | |||

20-59 years | 0.17 (0.14-0.20) | 0.19 (0.17-0.21) | |||

≥60 years | 0.22 (0.19-0.25) | 0.17 (0.15-0.18) | |||

Hispanic | 0.18 (0.15-0.22) | 0.17 (0.15-0.18) | |||

non-Hispanic white | 0.20 (0.16-0.24) | 0.20 (0.18-0.22) | |||

non-Hispanic black | 0.18 (0.16-0.21) | 0.15 (0.13-0.17) | |||

Other race/ethnicity | 0.14 (0.12-0.17) | 0.21 (0.18-0.25) | |||

Male: 6-11 years | 1.91 (1.60-2.28) | 0.76 (0.62-0.92) | |||

Female: 6-11 years | 2.25 (1.90-2.66) | 0.87 (0.67-1.13) | |||

Male: 12-19 years | 0.91 (0.75-1.11) | 0.52 (0.43-0.63) | |||

Female: 12-19 years | 1.49 (1.30-1.70) | 0.45 (0.36-0.56) | |||

Male: 20-59 years | 0.58 (0.52-0.64) | 0.35 (0.29-0.41) | |||

Female: 20-59 years | 0.95 (0.86-1.05) | 0.42 (0.37-0.49) | |||

Male: ≥60 years | 0.52 (0.44-0.62) | 0.37 (0.30-0.45) | |||

Female: ≥60 years | 0.91 (0.76-1.07) | 0.39 (0.33-0.46) | |||

Male, Hispanic | 0.60 (0.53-0.66) | ||||

Female, Hispanic | 1.07 (0.94-1.22) | ||||

Male, non-Hispanic white | 0.71 (0.64-0.78) | ||||

Female, non-Hispanic white | 1.12 (1.02-1.23) | ||||

Male, non-Hispanic black | 0.57 (0.52-0.62) | ||||

Female, non-Hispanic black | 0.98 (0.79-1.23) | ||||

Male, Other race/ethnicity | 0.61 (0.54-0.69) | ||||

Female, Other race/ethnicity | 0.83 (0.74-0.94) | ||||

6-11 years, Hispanic | 2.36 (1.79-3.11) | 0.83 (0.59-1.18) | |||

6-11 years, non-Hispanic white | 3.48 (2.83-4.27) | 0.88 (0.68-1.13) | |||

6-11 years, non-Hispanic black | 1.82 (1.47-2.25) | 0.54 (0.43-0.67) | |||

6-11 years, Other race/ethnicity | 2.30 (1.33-3.99) | 0.78 (0.54-1.12) | |||

12-19 years, Hispanic | 0.90 (0.75-1.09) | 0.54 (0.40-0.73) | |||

12-19 years, non-Hispanic white | 1.21 (0.92-1.59) | 0.48 (0.36-0.63) | |||

12-19 years, non-Hispanic black | 0.79 (0.63-0.99) | 0.38 (0.28-0.50) | |||

12-19 years, Other race/ethnicity | 1.10 (0.75-1.62) | 0.55 (0.40-0.74) | |||

20-59 years, Hispanic | 0.86 (0.72-1.02) | 0.40 (0.32-0.50) | |||

20-59 years, non-Hispanic white | 0.85 (0.74-0.97) | 0.40 (0.34-0.46) | |||

20-59 years, non-Hispanic black | 0.62 (0.55-0.71) | 0.32 (0.27-0.37) | |||

20-59 years, Other race/ethnicity | 0.60 (0.49-0.74) | 0.37 (0.30-0.45) | |||

≥60 years, Hispanic | 0.58 (0.46-0.73) | 0.42 (0.32-0.54) | |||

≥60 years, non-Hispanic white | 0.59 (0.45-0.75) | 0.36 (0.31-0.42) | |||

≥60 years, non-Hispanic black | 0.43 (0.35-0.54) | 0.36 (0.28-0.46) | |||

≥60 years, Other race/ethnicity | 0.50 (0.32-0.78) | 0.53 (0.32-0.88) |

The final multivariate regression models included: sex, age, race, sex*age, and sex*race (DPHP); race, sex, age, age*race (BDCIPP); race, sex, age, sex*age, and age*race (BCEP); race, age, and sex (DNBP); and age and race (BCIPP).