Serum levels of perfluoroalkyl substances (PFAS) and body composition – A cross-sectional study in a middle-aged population

Background: It has been suggested that per-and polyfluoroalkyl substances (PFAS) are endocrine disruptors with a potential to influence fat mass. Objective: The primary hypothesis tested was that we would find positive relationships for PFAS vs measures of adiposity. Methods: In 321 subjects all aged 50 years in the POEM study, five PFAS (perfluorooctane sulfonic acid (PFOS), perfluorooctanoic acid (PFOA), perfluorohexane sulfonic acid (PFHxS), perfluorononanoic acid (PFNA), per-fluorodecanoic acid (PFDA)) were measured in serum together with a Dual-energy X-ray absorptiometry (DXA) scan for determination of fat and lean mass. Whole-body magnetic resonance imaging scan was performed and the body was divided into > 1 million voxels. Voxel-wise statistical analysis was carried out by a novel method denoted Imiomics. Results: PFOS and PFHxS, did not show any consistent associations with body composition. However, PFOA, and especially PFNA and PFDA, levels were inversely related to most traditional measures reflecting the amount of fat in women, but not in men. In the Imiomics analysis of tissue volume, PFDA and PFNA levels were inversely related to the volume of subcutaneous fat, mainly in the arm, trunk and hip regions in women, while no such clear relationship was seen in men. Also, the visceral fat content of the liver, the pericardium, and the gluteus muscle were inversely related to PFDA and PFNA in women. Discussion: Contrary to our hypothesis, some PFAS showed inverse relationships vs measurements of adiposity. Conclusion: PFOS and PFHxS levels in plasma did not show any consistent associations with body composition, but PFOA, and especially PFNA and PFDA were inversely related to multiple measures reflecting the amount of fat, but in women only.


Introduction
Per-and polyfluoroalkyl substances (PFAS) are chemicals which have been widely used in our societies for decades.Due to their water-, oil-and dirt-repellent properties PFAS are used in many applications such as clothing, food packaging, firefighting foam, etc.They are widely spread in the environment, and, owing to a half-life of several years, many PFAS are bioaccumulating in the biota.Consequently, measurable levels of PFAS are found in almost all humans in the industrialized part of the world (Domingo and Nadal, 2019).An increasing number of studies have suggested that the exposure to PFAS may be associated with a broad range of adverse health effects (Fenton et al., 2021).A recent evaluation by EFSA highlighted that there are evidences that PFAS are linked to reduced antibody response to vaccination, increased cholesterol levels, increased serum levels of the liver enzyme alanine transferase (ALT) and birth weight (https://efsa.onlinelibrary.wiley.com/doi/full/10.2903/j.efsa.2020.6223).Mainly experimental studies have suggested that PFAS are endocrine disruptors with a potential to influence fat mass, resulting in obesity (Grun and Blumberg, 2009;White et al., 2011).
Mother-child cohort studies have documented that high levels of PFAS in the mothers are associated with increased BMI, or other measures of obesity, in the offspring at various ages up to 20 years (Gyllenhammar et al., 2018;Halldorsson et al., 2012;Jensen et al., 2020;Karlsen et al., 2017;Maisonet et al., 2012), although this has not been a universal finding (Manzano-Salgado et al., 2017;Martinsson et al., 2020).Also, inverse associations have been reported (Shoaff et al., 2018).In some of these studies, the PFAS have been measured in the mothers, while in others the PFAS were measured in cord blood.Also, the follow-up periods of the children have been of different lengths and due to differences in sample sizes, the power to detect relationships varies between the cited studies.
In cross-sectional analyses of the National Health and Nutrition Examination Survey (NHANES), one study conducted in children aged 3-11 years old found inverse relationships between circulating levels of perfluorohexane sulfonic acid (PFHxS), but not other PFAS, and weight and BMI in boys only (Scinicariello et al., 2020).In adults, no consistent relationship between PFAS and body build was noted in NHANES (Nelson et al., 2010).Similar results were observed in adults in the longitudinal Fernald Community Cohort (Blake et al., 2018) and in a Korean cohort (Seo et al., 2018), but in an adult Chinese population high levels of PFAS were related to an increased prevalence of overweight, especially in women (Tian et al., 2019).Differences in age-ranges were seen between the studies, but for example, the NHANES study was not smaller than the Chinese study, so the lack of significant relationships in some of the studies is not likely to be due to a lack of power.
If an environmental chemical were to have effects on body weight and fat mass, a number of factors could be mediating this effect, such as thyroid function, resting energy expenditure, inflammation and adipokine levels, such as leptin.From this perspective, the C8 Health Project reported high levels of perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) to be related to increased levels of thyroxine, the principal thyroid hormone (Knox et al., 2011).Moreover, PFAS levels were also related to higher thyroxine concentrations in the Fernald Community cohort (Blake et al., 2018), a Korean cohort (Seo et al., 2018), as well as in the NHANES study, especially in women (Wen et al., 2013).Despite that the EFSA 2020 report on PFAS conclude there is insufficient evidences for a link between PFAS and thyroid function (https://efsa.onlinelibrary.wiley.com/doi/full/10.2903/j.efsa.2020.6223), it could be of interest to evaluate if PFAS could affect body composition by means of changed thyroid function.
At the present time, only one study has investigated associations of PFAS and resting energy expenditure (EE) among obese subjects, and it reported null findings (Liu et al., 2018).In a recent study, PFAS levels were inversely related to a number of markers of inflammation, such as interleukin-6 (IL-6) (Salihovic et al., 2020).Inverse correlations between PFAS and leptin levels have been reported in Danish 9-year-old children (Domazet et al., 2020).Thus, several possible mediators between PFAS exposure and body composition have been reported.And it could be worthwhile to investigate if those physiological pathways play a role in the relationship between PFAS and body composition.
Being aware that a recent risk assessment from EFSA (https://efsa.onlinelibrary.wiley.com/doi/full/10.2903/j.efsa.2020.6223)concluded that no evidence exist for a relationship between PFAS and obesity in adults, we still found it worthwhile to evaluate this further for three reasons: First, the previous studies have used BMI or some other crude measure of obesity, while we today have more sensitive methods to use, like DXA and MRI.Second, most other studies have evaluated PFOS and PFOA, and we have experience from other studies with other outcomes that long-chained PFAS such as perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA) and perfluoroundecanoic acid (PFUnDA) show associations, while PFOS and PFOA do not (Lind et al. 2014(Lind et al. , 2017)).Third, since we previously have found sex-specific associations between some PFAS and atherosclerosis (Lind et al., 2017), and previous studies on PFAS and obesity did not generally stratify their analysis by sex, we found it worthwhile to perform sex-stratified analyses.
Therefore, the aim of the present study was to explore sex-specific relationships between serum levels of five PFAS and body composition in a middle-aged population (The Prospective study on Obesity, Energy, and Metabolism (POEM)) using both traditional measurements of body composition, such as anthropometry and Dual-energy X-ray absorptiometry (DXA), as well as a novel technique based on whole-body magnetic resonance imaging (WB-MRI) that provides a detailed 3D view of size and fat content in different body compartments, by voxelwise statistical analysis so-called "Imiomics."Since PFAS have been suggested to be endocrine disruptors with potential to cause obesity (Grun and Blumberg, 2009;White et al., 2011), the primary hypothesis tested was that we would find positive relationships for PFAS vs primarily fat mass.As a secondary aim, we also evaluated associations between the five PFAS and possible mediators in the association of PFAS and body composition, such as thyroxine levels, resting energy expenditure (EE), as well as IL-6 and leptin levels.

Sample
The study was approved by the ethics committee at Uppsala University (No. 2009/057 andNo. 2012/143), and the participants gave their informed consent.
The Prospective study on Obesity, Energy, and Metabolism (POEM) recruited 50-year-old men and women from the general population by a random invitation by mail using public population registers for the municipality of Uppsala, Sweden (Lind, 2013).The participants received their invitation one month after their 50th birthday.A total of 502 individuals took part in the study, a participation rate of 25%.The response rate was similar in men and women.The initial aim of the POEM study was to investigate links between metabolism and cardiovascular disease.However, since environmental contaminants have been shown to be associated with both metabolic diseases, such as diabetes (Lind and Lind, 2018), and cardiovascular diseases (Lind and Lind, 2012), also environmental contaminants, such as PFAS, are now investigated in the POEM study.
The participants were examined after fasting overnight.Blood was drawn between 8 and 10 A.M., and plasma was separated after spinning and stored in − 80C freezers for later analyses.Waist and hip circumference were recorded at the umbilical and trochanter levels, respectively, and waist and hip ratio (WHR) was calculated.Fat and lean mass was established using DXA.Whole-body MRI and dedicated imaging of liver and pancreas was performed on those who volunteered for this part of the study.MRI was performed on a separate day, within one month following the main study visit.This study includes only the 326 subjects with a technically appropriate MRI registration.

Biochemistry analyses
Plasma thyroxine was measured as a part of a metabolomics package provided by Metabolon (Morrisville, NC; UAS).The samples were evaluated by Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer.Thyroxine was quantified in the negative mode.Normalized values were used for the P.M. Lind et al. statistical analyses.
IL-6 and leptin were measured in plasma by the dual-antibody-based proximity extension assay (PEA) technique (OLINK, Uppsala, Sweden).The levels are expressed on a relative scale.

Indirect calorimetry
In the fasting state and supine position at rest, metabolic gas exchange was measured during 30 min in a ventilated hood.From the VO2 and VCO2 measurements, basal energy expenditure (EE) was calculated.

Dietary intake
A self-administered semiquantitative food frequency questionnaire (FFQ) consisting of 139 items was completed by the participants to assess usual frequency of consumption over the past year.(Wallin et al., 2014).These data were converted to gram/day and the sum of the intakes of different kinds of fish (cod, tuna, salmon, etc.) was used as a measure of fish intake.

Dual-energy X-ray absorptiometry, DXA
Dual-energy X-ray absorptiometry (DXA) is an X-ray-based technique that gives information on fat and lean mass, as well as bone mineralization.It is frequently use in the clinic for treatment of osteoporosis, but is also widely used in obesity research and is considered the gold standard for fat mass determinations.Total and regional body fat and lean mass were estimated using the same DXA scanner (DXA; Lunar Prodigy, GE Healthcare).To minimize potential operator bias, one experienced nurse performed all scans in the same room.The precision error of the DXA measurements in our laboratory was calculated using triple measurements in 15 subjects with repositioning according to recommendations from the International Society for Clinical Densitometry.Total fat and lean mass evinced precision errors of 1.5% and 1.0%, respectively.In the analysis, automatic edge detection was consistently employed; nevertheless, all scans were carefully checked for errors and manually corrected if necessary.
One potential limitation with the two-dimensional DXA measurement technique may be due to the influence of body weight on the distance between the DXA bed and the body compartment under measure and DXA measurements have been suggested to be elevated with increased body fat which may overestimate size in obese individuals.(Kremer and Gilsanz, 2016).However, the Lunar Prodigy DXA used in the current study corrects the scans to the actual effective object plane to limit the impact of these size differences and also limits the effect of tissue thickness (Bazzocchi et al., 2016;Mazess et al., 2000).Magnification errors seen with ordinary fan beam DXA which directly affect body size measurements (Mazess et al., 2000) are reduced by the Lunar Prodigy by using a narrow fan-beam along the axis.

Whole-body magnetic resonance imaging, WB-MRI
Subjects were imaged using a 1.5 T clinical MR system (Philips Achieva, Philips Healthcare, Best, Netherlands) in supine position with a continuously moving bed setup and the integrated body coil.Imaging included a whole-body water-fat imaging protocol which used a spoiled 3D multi gradient echo sequence.Scan parameters were: TR/TE1/ΔTE = 5.9/1.36/1.87ms, 3 unipolar echoes, flip angle 3 • .Imaged field of view (FOV) 530 × 377 × 2000 mm 3 , reconstructed voxel size 2.07 × 2.07 × 8.0 mm 3 in left-right × anterior-posterior × foot-head directions.A dedicated scan of the liver, which also included the pancreas, was undertaken for detailed analysis of fat content in the liver and pancreas.However, this liver scan was not included for the first subjects in the POEM study.Scan parameters were: TR/TE1/ΔTE = 8.66/0.92/1.32,6 unipolar echoes, flip angle 5 • .Imaged FOV384 × × 150 mm 3 , reconstructed voxel size 3.0 × 3.0 × 10.0 mm 3 .Water-fat image reconstruction was performed employing an algorithm developed inhouse.The imaging protocol and the reconstruction method have been described in more detail previously (Andersson et al., 2018).Liver and pancreas fat was quantified using manual delineation of the volume of interest using the software ImageJ (version 1.45s).A trained operator segmented as much as possible of the volume of interest but avoided tissue borders to limit partial-volume effects.Median fat contents of the volumes of interest were used as measurements of tissue fat content.
The visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) depots were quantified by deforming manually defined depots in a male and female reference subject to all other subjects by utilizing the image-registration method used for the Imiomics analysis described below.The deformed regions were further processed by thresholding operations, removing voxels with fat content <50%.

Imiomics
The imiomics technique is a way to quantify relationships between a certain characteristic, such as the plasma level of a PFAS, and the size and fat content of each part of the body in a very detailed manner.A regression model is performed for each of the approximately 1-2 million voxels that determines a 3D image of the human body using MRI.The results of all those models are depicted in color coded maps for either the correlation coefficient and the p-value for a visual interpretation.
Image registration is a key component of the Imiomics technique.In image registration, a target image is deformed to match a (fixed) reference image by computing and applying a deformation field.For each point in the reference image, the deformation field defines a corresponding point in the target image.These point-by-point correspondences are used in Imiomics analyses, in which whole-body MRI images are deformed to a reference whole-body MRI volume.The reference WB-MRI volume constitutes a reference coordinate system, where each point has a corresponding point in all volumes in the cohort.This enables the voxel-wise statistical analysis procedure (Strand et al., 2017), in which associations between tissue volume and fat content from WB-MRI can be related to study non-imaging data.
The Imiomics image registration method employs a tissue-specific handling of bone, lean tissue, and adipose tissue.The degree of elasticity of the deformation required to align two images tends to differ across these different tissue types.This prior knowledge is made use of by performing the image registration of the different tissues sequentially, applying registration parameters appropriate to each tissue.

P.M. Lind et al.
The Imiomics image registration method comprises the following three steps: 1) Articulated, piece-wise affine, registration of bone sections 2) Registration of water images with constraints on bone 3) Registration of fat images with constraints on bone and water.This image registration method has been presented and evaluated (Ekstrom et al., 2020) on WB-MRI water-fat image data and attains image registration results appropriate for Imiomics analyses.We have validated the results regarding fat mass and lean mass vs DXA measurements (Lind et al., 2019).

Statistics
Linear regression analysis (OLS) was used to evaluate the relationships between traditional anthropometrics, fat distribution measures and circulating concentrations of the five PFAS.Skewed variables were ln-transformed to achieve a normal distribution (liver fat, pancreatic fat, VAT, SAT and the PFAS).Measurements of anthropology and fat distribution, as well as the PFAS, were transformed to an SD scale in order to improve the comparison between the regression models.The analyses were stratified by sex and adjusted for exercise habits, smoking, and education level (age same in all subjects).Thus, one model was used for each PFAS vs each index of body composition.For example; a model for PFDA and BMI would be the following: BMI=PFDA exercise smoking education.In addition, we evaluated an additional set of models also adjusted for fish intake." The five PFAS were furthermore analyzed in relation to the four potential mediators' thyroxine, resting EE, IL-6 or leptin levels (lntransformed values, except EE) using linear regression analysis stratified by sex and with exercise habits, smoking, and education level as confounders.Thus, one model was used for each PFAS vs each potential mediator.For example; a model for PFDA and leptin would be the following: Leptin = PFDA exercise smoking education.
STATA 16 was employed for these computations (Stata Corp, College Station, TX, USA).
For the Imiomics analyses, Spearman's rank Correlation coefficient was computed to generate ρand p-values for all voxels and all parameters.MATLAB (version 2016b, The MathWorks Inc., Natick, MA) was employed for these calculations.
In the so-called correlation-map visualizations, each voxel's correlation (ρ) value and p-value are visualized in the following way: Voxels with p-value≥0.05are transparent and the water content is shown.Voxels with p-value <0.05 ("significant", uncorrected) show the corresponding correlation coefficient in the colormap jet inverted, where red represents strong positive correlation and blue represents strong negative correlation.
The Imiomics results were analyzed visually.To minimize the number of false positive findings from the large number of tests performed, we report only associations found in many image elements within a certain anatomical structure, corresponding to p-values below 0.01.

Results
Means and SD for the studied traditional measures of body composition are given in Table 1.

PFOS
Regarding the traditional measures of body composition, inverse associations between PFOS and only weight, hip circumference, total fat  mass, as well as leg and trunk fat showed in women only p < 0.05 (Table 2).
In the Imiomics analysis of tissue volume and fat content, PFOS was weakly correlated to the volume of trunk and hip subcutaneous tissue, liver and VAT volume and the fat content of the liver and subcutaneous tissue in an inverse fashion in women only (Fig. 1 and Supplemental Material 1-4).

PFOA
Regarding the traditional measures of body composition, the associations vs PFOA levels showed the same pattern as those seen for PFOS, but in this case the inverse associations between the measures reflecting the amount of fat and PFOA in women were stronger than for PFOS and with weight, waist and hip circumference, BMI, total fat mass, trunk fat and leg fat showing p < 0.05 (Table 3).
In the Imiomics analysis of tissue volume and fat content, the associations vs PFOA levels showed the same pattern as those seen for PFOS, but also in this case the inverse associations vs SAT volume and fat content were stronger and were mainly seen in the arms, the hip region, and around the thighs, in women (Fig. 2 and Supplemental Material 5-8).

PFHxS
PFHxS levels were not consistently related to any of the traditional measures of body composition in any of the sexes, except that total lean mass and leg lean mass were inversely related to PFHxS in males, but not females p < 0.05 (Table 4).
In the Imiomics analysis of tissue volume and fat content, inverse associations vs PFHxS were seen for SAT volume in the arm, trunk, and thigh regions, in men but not in women.Inverse associations vs PFHxS were also seen for skeletal muscle volume in the arms and legs in the men only (Fig. 3 and Supplemental Material 9-12).

PFNA
Regarding the traditional measures of body composition, PFNA levels in plasma were inversely related to most measures reflecting the amount of fat, such as BMI, waist circumference, fat mass, VAT and SAT in women p < 0.05.No such pattern was seen in men (Table 5).
In the Imiomics analysis of tissue volume and fat content, PFNA levels in plasma were inversely related to the volume of SAT, mainly in the arm, trunk and hip regions in women, while no such clear relationship was seen in men.Also, the fat content of VAT, the liver, the pericardium, and the gluteus muscle were inversely related to PFNA in women (Fig. 4 and Supplemental Material 13-16).

PFDA
Regarding the traditional measures of body composition, PFDA levels in plasma were inversely related to almost all measures reflecting the amount of fat, such as BMI, waist circumference, fat mass, liver fat, VAT and SAT in women p < 0.05.No such pattern was seen in men (Table 6).
In the Imiomics analysis of tissue volume and fat content, PFDA levels in plasma were inversely related to fat content of SAT, mainly in the arm, trunk and hip regions in women, while no such clear relationship was seen in men.Also, the fat content of VAT, the liver, the pericardium, perirenal fat, and the gluteus muscle were inversely related to PFDA in women (Fig. 5 and Supplemental Material 17-20).
The results from the relationships between the five PFAS and the traditional body composition variables are summarized as a heat map in Fig. 6.
When the five PFAS were related to the four potential mediators (thyroxine, resting energy expenditure, IL-6 or leptin) of an effect of PFAS exposure on body composition, PFNA was significant related to all four potential mediators except leptin in a negative fashion in women only.A similar trend was seen for PFOA, but in this case, the relationships were generally weaker and only showed p < 0.05 for thyroxine, EE, and leptin in women.PFOS and a negative relationship vs IL-6 in men were found, while PFHxS did not only show any significant relationships with any of the four potential mediators (Table 7).
Additional adjustment for fish intake regarding the relationships between the five analyzed PFAS and traditional measures of body composition did only have marginal effects on the relationships described above.

Discussion
The main finding in the present study was that serum levels of the two PFAS with highest concentrations, PFOS and PFHxS, in this sample did not show any consistent relationships vs body composition, while PFOA, and to a larger extent PFNA and PFDA were related to multiple indices of fat volume and content in both subcutaneous adipose tissue and in ectopic fat depots, such as liver, VAT, pericardial and perirenal adipose tissue, but in women only.Thus, the present results were different from our initial hypothesis of PFAS as obesogens.
The present study confirms most previous investigations (Blake et al., 2018;Nelson et al., 2010;Seo et al., 2018;Tian et al., 2019) and the EFSA2020 evaluation in that no general association was seen between PFAS and body composition.The inverse relationships reported were  seen for certain PFAS only, and only found in females.
We have in the past reported that different PFAS show very different relationships with a number of health outcomes, such as diabetes, atherosclerosis, cardiac geometry and function and liver enzymes (Lind et al. 2014(Lind et al. , 2017;;Mobacke et al., 2018;Salihovic et al., 2018).It has generally been the PFAS with >C8 that have shown the most significant associations.The reason for this is not clear.One explanation might be a longer half-life resulting in a prolonged period of exposure in the body.
The elimination half-lives have been shown to be dependent on structure, carbon chain length, and carbon chain branching (Xu et al., 2020).In general, PFAS with long perfluoroalkyl chain lengths (>C4 for the perfluoro sulfonic acids, >C6 for the perfluorocarboxylic acids) have been reported to be more persistent in the body as compared to their short-chain analogues (Zhang et al., 2013).It might therefore be that PFDA and PFNA have a longer half-life compared to PFOS, PFOA, and PFHxS, for which the half-life has been estimated to be in the range of 2-8 years (Dzierlenga et al., 2020;Li et al., 2018;Olsen et al., 2007;Xu et al., 2020).Although a longer half-life for PFNA and PFDA compared to the other measured PFAS in the present study might be expected, this has however to be proven.
Circulating levels of PFNA and PFDA have previously been shown not to decline during the last 10 years in Uppsala, as PFOA and PFOS have done (Stubleski et al., 2016), indicating a continuous exposure of PFNA and PFDA in this sample.This will result in a prolonged period of exposure to PFNA and PFDA, but if that is of pathogenetic importance for fat mass remains to be proven.
Why PFDA (and PFNA) vs fat mass relationship was noted only in females is not obvious.However, previous data have shown the elimination of PFAS to be sex-dependent (Kato et al., 2011) and that PFAS could affect sex-hormone levels (Joensen et al., 2013).We have furthermore demonstrated divergent effects of PFAS in men vs women regarding carotid artery atherosclerosis (Lind et al., 2017), exemplifying that sex-specific relationships could be present for PFAS also for other health outcomes.
Although it could be suspected that an action of a PFAS on fat tissue is due to the well-documented binding of PFAS to the nuclear receptors PPAR alpha and gamma (Lau et al., 2007), two receptors that are key in the regulation of lipids and fat metabolism, it is unclear which downstream mechanisms of these receptors mediate the link we found between PFDA and fat mass in women.Four likely candidates to mediate this relationship were investigated in the present study: basal EE, IL-6, thyroxine and leptin levels.Given the inverse relationship between PFDA (and PFNA) and fat mass in women, we could expect that PFDA would be positively related to the levels of proinflammatory cytokine  IL-6, thyroxine and leptin, as well as basal EE, in order to reduce fat mass.In fact, exactly the opposite pattern was found for all the potential mediators being inversely related to PFDA in females.Thus, it is still not clear by which pathways PFDA (and PFNA) might influence fat mass.Another alternative is that the amount of adipose tissue determines the PFAS levels, especially PFDA and PFNA in women.While this is certainly the case for lipid-soluble organic pollutants, such as PCBs, DDT, or dioxins, PFAS do not accumulate in adipose tissue, and therefore it is not likely that the amount of fat mass would influence PFAS levels in the circulation to a major degree.In a study of profound fat mass reduction induced by bariatric surgery (Jansen et al., 2019), levels of PFAS generally declined one-year post-surgery.The authors of that study suggested this effect of surgery to be due to a reduced food intake, since food is the major source of PFAS exposure.PFAS are normally bound to albumin and other plasma proteins (Jones et al., 2003), but the decline in serum albumin seen following bariatric surgery did not explain the drop in PFAS post-surgery.
An alternative explanation for our findings in women is that a high intake in fish would increase PFAS levels (Augustsson et al., 2021) and reduce fat mass.However, including self-reported fish intake as a confounder in the models did not alter the strength of the associations between PFDA (and PFNA) and fat mass in women.

Table 7
Relationships between five per-and polyfluoroalkyl substances (PFAS) and four potential mediators between PFAS exposure and body composition.EE = basal energy expenditure.
Male Female Beta 95%CI lower 95%CI higher p-value Beta 95%CI lower 95%CI higher p-value

Strengths and limitations
The strength and novel part of this study is its comprehensive characterization of body composition both by DXA and MRI, as well as the newly developed Imiomics technique that allows an evaluation of all parts of the body.Another advantage is that we were able to investigate the role of possible mediators, including basal EE.Multiple tests were performed in the present study.We did however choose not to adjust the p-value for multiple testing since the majority of body composition variables evaluated are closely related.It should however be noted that the majority of the reported significant relationships between PFDA and body composition indices show very low p-values that would be significant also following a strict Bonferroni-correction of the p-value.
A limitation is that the sample consists of individuals of European descent, so studies in other ethnic groups are warranted.A limitation of the present study is that the participation rate was low.If this fact affected the results is not known.Another limitation is that we do not have reproducibility data (such as coefficients of variation) for the variables measured in the present study.except for PFAS measurements.

Conclusions
PFOS and PFHxS levels in plasma did not show any consistent association with body composition, but PFOA, and especially PFNA and PFDA were inversely related to multiple measures reflecting the amount of fat, but in women only.

Fig. 1 .
Fig. 1.Relationships between PFOS and local volume or fat content using voxel-wise analyses.The analysis is from the left side: Males-local volume, Females-local volume, Males-fat content, and Females-fat content.

Fig. 2 .
Fig. 2. Relationships between PFOA and local volume or fat content using voxel-wise analyses.The analysis is from the left side: Males-local volume, Females-local volume, Males-fat content, and Females-fat content.

Fig. 3 .
Fig. 3. Relationships between PFHxS and local volume or fat content using voxel-wise analyses.The analysis is from the left side: Males-local volume, Females-local volume, Males-fat content, and Females-fat content.

Fig. 4 .
Fig. 4. Relationships between PFNA and local volume or fat content using voxel-wise analyses.The analysis is from the left side: Males-local volume, Females-local volume, Males-fat content, and Females-fat content.

Fig. 5 .
Fig. 5. Relationships between PFDA and local volume or fat content using voxel-wise analyses.The analysis is from the left side: Males-local volume, Females-local volume, Males-fat content, and Females-fat content.

Fig. 6 .
Fig. 6.Heat map of significant associations between PFAS vs traditional body composition measurements.A light blue color represents an inverse relationship with p < 0.05.A dark blue color represents an inverse relationship with p < 0.01.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Table 1
Traditional measurements of body composition and per-and polyfluoroalkyl substances (PFAS) in the sample.

Table 2
Relationships between plasma levels of perfluorooctane sulfonic acid (PFOS) and indices of body composition.a a The analyses were adjusted for exercise habits, smoking, and education level.P.M.Lind et al.

Table 3
Relationships between plasma levels of perfluorooctanoic acid (PFOA) and indices of body composition.a a The analyses were adjusted for exercise habits, smoking, and education level.

Table 4
Relationships between plasma levels of perfluorohexane sulfonic acid (PFHxS) and indices of body composition.a a The analyses were adjusted for exercise habits, smoking, and education level.

Table 5
Relationships between serum levels of perfluorononanoic acid (PFNA) and indices of body composition.a The analyses were adjusted for exercise habits, smoking, and education level. a

Table 6
Relationships between serum levels of perfluorodecanoic acid (PFDA) and indices of body composition.a a The analyses were adjusted for exercise habits, smoking, and education level.