Association of Perfluoroalkyl Substances, Bone Mineral Density, and Osteoporosis in the U.S. Population in NHANES 2009–2010

1Age-associated osteoporosis is a significant public health concern because it is related to bone fractures and associated morbidities. It is estimated that > 9% of Americans ≥ 50 years of age had osteoporosis either at the femoral neck or at the lumbar spine in 2005–2008 [2% of all U.S. men and 10% of all U.S. women]. Recent evidence suggests that exposure to environmental toxicants such as lead, cadmium, and mercury are associated with higher risks for osteoporosis and fractures.

2Perfluoroalkyl substances (PFASs) have been widely used in protective water- and stain-resistant coatings on clothing, furnishings, and non-stick housewares for > 60 years. PFASs are ubiquitous environmental contaminants and are detectable in humans worldwide. Of the 12 PFASs assayed in the U.S. National Health and Nutrition Examination Survey (NHANES) 1999–2008, 4 PFASs were found in 95% of the U.S. population: perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS), perfluorohexane sulfonic acid (PFHxS), and perfluorononanoic acid (PFNA). The toxicity of PFASs, including tumors, liver damage, and adverse skeletal and reproductive outcomes, has been reported in animal studies. PFOA has recently been categorized as possibly carcinogenic in humans (2B classification). Moreover, PFASs have been characterized as endocrine-disrupting chemicals (ED) based on their hormonal modulation and metabolic associations.

3A recent analysis reported negative associations between environmental exposure to PFOS and LSBMD in premenopausal women using data from NHANES 2005–2008. Furthermore, experimental and human autopsy evidence suggests accumulation of PFASs in the skeleton. In the current analysis, we tested the following hypotheses: a) serum PFAS concentrations are negatively associated with BMD, and b) this association differs by sex.

4Study methods and participants. Publicly available data from the NHANES 2009–2010 cycle were used for this study. Detailed descriptions of the survey design and methods are available on the NHANES website. Briefly, NHANES is an ongoing survey of the non-institutionalized U.S. population collected using a stratified, multistage probability sampling design. After providing informed consent, participants visited a mobile examination center (MEC) for physical assessment, examination, and laboratory measurements. Analysis of PFASs in serum was conducted at the National Center for Environmental Health in a random, one-third subsample of participants ≥ 12 years of age. The present study sample consisted of NHANES participants 12–80 years of age who had BMD measurements available for total femur (TFBMD, n = 1,914), its subregion femoral neck (FNBMD, n = 1,914), and lumbar spine (LSBMD, n = 1,605), in addition to serum concentrations of four PFASs (PFOA, PFOS, PFHxS, PFNA).

5Dual X-ray absorptiometry (DXA) and osteoporosis. BMD (grams per centimeter squared) was measured using DXA Hologic QDR 4500A fan-beam densitometers (Hologic Inc., Bedford, MA, USA) in the MEC. Antero-posterior LSBMD was scanned; mean BMD was computed from the first through fourth lumbar vertebra. For TFBMD and FNBMD, the left hip was routinely scanned. If a left-hip replacement or metal objects in the left leg were reported, the right hip was scanned. Participants were excluded from the femur scan if they had bilateral hip fractures, bilateral hip replacements, or pins. Participants weighing > 300 lbs (136 kg) (DXA table limitation) and pregnant females (positive urine pregnancy test and/or self-report) were ineligible for the DXA examination. Each respondent’s scan was reviewed in the Department of Radiology, University of California, San Francisco, using standard radiologic techniques and NHANES protocols.

6Participants answering “yes” to the question “Has a doctor or other health professional ever told you that you have osteoporosis?” were coded as having “self-reported physician-diagnosed” osteoporosis. The validity of self-reported osteoporosis is moderate in mid-age adults and good in older adults. In a study of Australian women, the agreement between self-reported osteoporosis and medication claims was moderate in women 56–71 years of age (kappa statistic 0.51).

7PFAS assay. Briefly, serum PFASs (non-fasting) were measured using automated solid-phase extraction coupled to isotope-dilution high–performance liquid chromatography–tandem mass spectrometry, as published elsewhere. Serum measurements of four PFASs (PFOA, PFOS, PFHxS, and PFNA), which were detectable in > 98% of the 2009–2010 survey participants, were included in this analysis.

8Covariates. Covariates selected a priori included age, race/ethnicity, sex, body mass index (BMI), smoking (serum cotinine), daily milk intake, physical activity (PA), menopause, and blood lead concentration. Sociodemographic information such as age, sex, race/ethnicity, and reproductive history were recorded using interviewer-administered questionnaires. Age was used concurrently as both a continuous variable and categorized in three levels as 12–20 years, 21–50 years, and ≥ 50 years (referent). Race/ethnicity was self-reported as non-Hispanic white, non-Hispanic black, Mexican American, other Hispanic, and other multiracial (referent).

9Body weight was measured to the nearest 0.01 kg using an electronic load–cell scale, and standing height was measured with a fixed stadiometer. BMI was calculated as body weight (kilograms) divided by height (meters squared). Serum cotinine levels < 1.0 ng/mL were categorized as non-smoker (referent), 1.0–9.9 ng/mL as environmental tobacco smoke (ETS) exposure, and ≥ 10.0 ng/mL as current smoker.

10Menopause status was ascertained as self-reported cessation of regular menstruation over the past 12 months. If women responded “no” to the question “have you had regular periods in the past twelve months?” and stated that the reason for not having regular periods was due to “menopause/hysterectomy,” then they were categorized as postmenopausal.

11Separate questions assessed the history of hysterectomy and bilateral oophorectomy as answering yes to a question about having had a hysterectomy [have you had a hysterectomy, including a partial hysterectomy (that is, surgery to remove your uterus or womb)?] and answering yes to a question about having had both ovaries removed (have you had both of your ovaries removed either when you had your uterus removed or at another time?), respectively. Women who answered that their reason for not having regular periods was “menopause/hysterectomy” also included women who had a hysterectomy and/or bilateral oophorectomy.

12Self-reported vigorous or moderate recreational physical activity (PA) was categorized as “inactive” (< 10 consecutive minutes per week: referent), “low activity” (10 to 149 active minutes per week), “sufficient/medium activity” (150 to 299 active minutes per week, and “high activity” (≥ 300 active minutes per week) according to the 2008 Physical Activity Guidelines for Americans. This activity categorization was derived from six PA variables using an algorithm. These six variables assessed “vigorous” and “moderate” recreational PA using NHANES questionnaire data. For example, vigorous recreational PA was evaluated by participants’ answers to the following questions: a) “Do you do any vigorous-intensity sports, fitness, or recreational activities that cause large increases in breathing or heart rate like running or basketball for at least 10 min continuously?” b)“In a typical week, on how many days do you do vigorous-intensity sports, fitness or recreational activities?” c)“How much time do you spend doing vigorous-intensity sports, fitness or recreational activities on a typical day?” The same three variables were incorporated into the algorithm for moderate intensity PA.

13Milk intake was ascertained as answering “yes” (“no”: referent) to the question, “have you been a regular (five times per week) milk drinker for most or all of your life, including childhood?” Blood lead concentration (micrograms per deciliter), and BMI were used as continuous variables [exception: see Supplemental Material, Table S1, presented as quartiles (lead), or by weight category (BMI)].

14Statistical analysis. Population characteristics, outcomes, and exposures were summarized as the mean ± standard error (SE) or numbers of observations (percent), and differences according to sex were tested using Student’s two-tailed t-test or the Rao–Scott chi-square test as recommended by the National Center for Health Statistics (NCHS). As decided a priori, analyses were conducted to examine the relationships between PFASs, BMD, and osteoporosis stratified by sex. Due to significant nonnormal distribution, natural log–transformation was performed on PFAS concentrations (ln-PFAS). Analyses were repeated by categorizing PFASs in quartiles using sex-specific quartile (Q1: referent, lowest; Q4: highest) cut points for men and women. We constructed full multivariable linear regression models with TFBMD, FNBMD, and LSBMD as dependent variables and individual PFAS concentrations as ln-transformed continuous predictors or in quartiles while adjusting for covariates related to BMD that have been described previously in the literature. Each PFAS was modeled separately. The results are reported as regression coefficients and 95% confidence intervals (CIs).

15In multiple logistic regression models, adjusted odds ratios (aORs) and 95% CIs of osteoporosis were calculated for each PFAS separately as ln-transformed predictors or as sex-specific quartiles (using the lowest quartile as referent). To explore whether PFAS, BMD, and osteoporosis associations differed by menopause status, multiple linear regression analysis stratified by menopause status was performed (as described above). In the Supplemental Material, Table S1 shows the mean untransformed PFAS concentration by covariate categories for men and women. In the Supplemental Material, Table S2 shows unadjusted mean serum PFAS concentration (ln-transformed) by osteoporosis diagnosis in men and women.

16To account for the complex NHANES survey design, sampling weights, strata, and primary sampling units were adjusted in all analyses as recommended by the NCHS. SAS survey procedures were used (SAS Institute Inc., version 9.3) by applying the Taylor series linearization method to calculate SEs. Two-tailed p-values were used for all tests at a 5% significance level.

17Characteristics of the study population. The mean age of the study population was 42.6 (SE: 0.6) years, with no significant difference by sex (Table 1). The study population was predominantly composed of non-Hispanic white participants. BMI was comparable by sex. A higher percentage of men were smokers and had ETS exposure (p < 0.001). In men, milk consumption and blood lead levels were significantly higher than in women (p < 0.05). Less than half of the study participants were physically inactive (42%); there was a statistically significant difference in PA between men and women (p < 0.001). The proportion of inactive women was higher than that of men. High PA was almost three times more common in men than in women.

18TFBMD and FNBMD were 12% and 8% higher in men than in women, respectively (all p < 0.001). LSBMD was also slightly higher in men than in women, but the difference was not statistically significant. A diagnosis of osteoporosis was reported for 5% of the total sample; however, this diagnosis was reported for only 2% of men (n = 17) and was reported more frequently for women (n = 77) (p < 0.001). In men, the average serum PFOA, PFOS, and PFHxS concentrations were 20%, 32%, and 36% higher than in women, respectively (p < 0.001 for all three). PFNA concentrations were comparable between the sexes (p = 0.133).

19In the Supplemental Material, Table S1 summarizes covariates and PFAS concentrations by sex. A significant relationship was noted between age categories and all PFASs in both sexes except for PFHxS in men. PFOS and PFNA were significantly different across race/ethnicity in both sexes (however, this difference was only significant for PFHxS in men). PFAS exposure was not significantly related to BMI categories, recreational PA, or daily milk intake. No significant associations between smoking status and PFASs were noted except for smoking status and PFHxS in men. In women, a significant association was observed between mean blood lead levels and all PFASs. In men, only PFOS and PFHxS had a significant relationship with blood lead levels.

20Serum PFASs and BMD. Adjusted associations between continuous ln-transformed and quartiles of PFAS and each BMD measurement are reported separately for men and women and for pre- and postmenopausal women in Tables 2–4. In most cases, ln-PFAS concentrations and BMD measurements were negatively associated in men, but only the association between ln-PFOS and FNBMD was statistically significant (β = –0.013; 95% CI: –0.024, –0.002). In general, categorical PFAS exposures were not clearly associated with any of the BMD measurements in men, except that compared with Q1, Q4 PFOS exposure had a significant negative relationship with FNBMD (β = –0.046; 95% CI: –0.078, –0.015).

21In women, ln-PFOS was associated with significantly lower TFBMD and FNBMD, and ln-PFOA was associated with significantly lower TFBMD. In premenopausal women, no significant association was observed between any PFASs and BMD. In postmenopausal women, serum ln-PFOS was significantly associated with TFBMD and FNBMD; PFNA was inversely related with TFBMD, FNBMD, and LSBMD (all p < 0.05).

22In quartile analyses, women showed significant negative associations between TFBMD and Q2 of PFOA, Q4 of PFOS, and Q4 of PFHxS; and between FNBMD and Q4 of PFOS, relative to the lowest quartile (Q1) of each exposure. In premenopausal women, TFBMD also had significant negative associations with Q2 of PFOA and Q4 of PFNA. In postmenopausal women, TFBMD was significantly lower in Q4 of PFOS and PFHxS. LSBMD was not significantly associated with any of the categorical PFAS exposures in men or in women, regardless of menopausal status.

23Serum PFAS and osteoporosis. There were 17 cases of osteoporosis in men (Table 1), and there were no significant differences in mean PFAS levels according to case status in men (see Supplemental Material, Table S2). In women, the mean values of all four PFASs were significantly higher in the 77 women with osteoporosis than in the 733 women who did not report osteoporosis. The aORs for PFOA and PFNA were significant for both the continuous ln-transformed exposures (aOR = 1.84; 95% CI: 1.17, 2.90 and aOR = 1.45; 95% CI: 1.02, 2.05, respectively) and for the Q4 versus Q1 comparisons (aOR = 2.59; 95% CI: 1.01, 6.67 and aOR = 3.23; 95% CI: 1.44, 7.21, respectively) (Table 5). In addition, ln-PFHxS was significantly associated with osteoporosis in both the ln-PFAS model (aOR 1.64; 95% CI: 1.14, 2.38) and the quartile analysis, with significant associations for all three quartiles (e.g., aOR = 13.20; 95% CI: 2.72, 64.15 for Q4 vs. Q1). However, there were only eight women with osteoporosis in Q2 of PFHxS, and quartile-specific aORs were very imprecise. PFOS was not significantly associated with osteoporosis in the continuous or quartile exposure models.

24In this representative sample of the U.S. population, PFOS and PFHxS were associated with lower TFBMD and a higher prevalence of osteoporosis among women. In addition, PFOS was negatively associated with FNBMD, PFOA was negatively associated with TFBMD, and PFNA was positively associated with osteoporosis only. In general, associations were stronger among postmenopausal women than among premenopausal women. In men, PFOS was associated with lower FNBMD only. LSBMD was not clearly associated with any of the PFASs in men or women. Although the variance in BMD explained by PFAS was very small (R2 < 1%) in this exploratory investigation, given the study sample size and the nationally representative data, these results are important and support the need for further research to evaluate PFAS toxicity with regard to bone health.

25Human exposure to PFASs has gradually increased since the 1950s, although following regulatory efforts, serum concentrations of some PFASs have decreased in recent years. PFASs are present in many food items including meat, poultry, eggs, fish, and fresh produce, and diet contributes significantly to daily human exposure to PFASs. PFASs are poorly metabolized and slowly eliminated from the human body, with half-lives of 4–8 years, and they partition to bone tissue.

26Limited data from animal models suggest PFAS toxicity to bone. Prenatal PFOS exposure in rodents was associated with fetal bone malformation. In mice, environmentally relevant doses of PFOS were rapidly deposited in bone. Recent data from human autopsy studies suggest that PFASs are sequestered in bone, with PFOA being predominant. Taken together, these studies suggest that PFASs are deposited in bone and may induce osteotoxicity.

27PFASs at low doses are categorized as endocrine-disrupting chemicals (EDCs) in animals, including rats, mice, and fish, and have been characterized as such in some human studies. Data from animal models suggest that bone tissue could be an important target for a number of EDC environmental pollutants because EDCs can disturb the complex hormonal control of bone metabolism. For example, sex-dependent associations of another EDC, polychlorinated biphenyl (PCB), with bone length have been reported in female sheep fetuses. In another study, pregnant ewes exposed to multiple EDCs showed reduced BMD.

28Epidemiological research supports the EDC hypothesis for some forms of PFAS toxicity. PFAS exposure was related to delayed onset of puberty. Older age at menarche has been associated with an increased risk of fractures in women of approximately 20 years of age, with reduced BMD in perimenopausal women, and with an increased risk of hip fractures in older women. Moreover, serum PFOA and PFOS (specifically) were associated with earlier age at menopause in the C8 Health Project and in the NHANES. reported that serum PFOS was negatively associated with serum estradiol concentration. Another pathway that may link PFASs and BMD is through thyroid hormone modulation. Thyroid hormones play a crucial role in bone health and remodeling. PFAS exposure was associated with serum thyroxine (T4) and triiodothyronine (T3) levels in two cross-sectional studies and with altered responses to T3 in a T3-dependent cell line in vitro. Based on a cross-sectional study of adult NHANES participants, reported that serum PFHxS was positively associated with subclinical hyperthyroidism (defined as TSH < 0.24 mIU/mL) in women, which is a risk factor for osteoporosis.

29In the present analysis, men had higher serum levels of PFASs, but women experienced a greater BMD deficit and osteoporosis risk than did men. These observations corroborate the recently published findings of, who reported that PFOS exposure was related to a BMD deficit only in women. The difference between the sexes with regard to the association between PFASs and BMD suggests that reproductive hormones may be involved. Females may be more sensitive to PFAS toxicity than males, or, as animal studies suggest, PFASs may be eliminated differently by males and females.

30We observed a higher prevalence of osteoporosis associated with PFOA, PFNA, and PFHxS in women. The odds ratios and 95% CIs of osteoporosis were large but imprecise for PFHxS in the quartile analysis owing to the small number of observations. We are unable to explain this strong association, and because of the limited precision of the estimates, the associations should be interpreted with caution. PFHxS has the longest half-life (8.5 years) of the four PFASs and has recently been associated with impaired thyroid function and early menopause in epidemiological studies.

31One of the study limitations was its cross-sectional design; we cannot confirm that the exposures preceded the outcomes of interest or rule out the possibility of reverse causation. The strengths of this study include a large sample representative of the U.S. population. Four PFASs commonly detected in U.S. residents were assayed, and to our knowledge, this is the first report assessing the relationship of four PFASs with BMD at three bone sites; the only such previous study, by, was limited to two PFASs (PFOA, PFOS) and BMD at the lumbar spine and total hip.

32We observed significant negative associations between PFOS, PFNA, and PFHxS and TFBMD, FNBMD, and osteoporosis in women. reported a significant negative association between PFOS and LSBMD in premenopausal women only; however, we did not observe any significant association between PFOS and LSBMD in our sample of premenopausal women.

33Some potential reasons for the discrepant relationships between PFOS and LSBMD observed in the present study and those observed by could be attributed to the differences in the NHANES survey cycles examined and to differences in sample size, age range, and covariates included. For example, the study by used data from two combined NHANES surveys, 2005–2006 and 2007–2008, whereas our sample consisted of data obtained from one NHANES survey conducted in 2009–2010. Furthermore, the decreasing mean serum PFOS concentration in the U.S. population could have masked an association with LSBMD in our study. As shown in the NHANES surveys, geometric mean (GM) serum PFOS levels in the general U.S. population, including women, decreased over the 2005–2006 (overall: 17.1 ng/mL, women: 14.4 ng/mL), 2007–2008 (overall: 13.2 ng/mL, women: 10.7 ng/mL), and 2009–2010 (overall: 9.3 ng/mL, women: 7.7 ng/mL) surveys.

34In conclusion, our findings indicate that some PFASs are associated with low BMD and an increased prevalence of osteoporosis in U.S. women. However, these findings must be interpreted with caution given the cross-sectional study design, the large number of comparisons made, and the small number of osteoporosis cases in the study population.