This literature review suggests that PFAAs exposure can influence maternal TSH levels. Moreover, several PFAAs were inversely associated with maternal T4 and T3 levels. The studies reported some associations between PFAAs exposure and infant TH levels although the data was less conclusive.
PFHxS, PFOS, PFNA, PFDA, PFUnA and PFDoA exposure seem to affect maternal TSH levels; however, only PFHxS, PFOS and PFNA were found significantly associated with TSH in more than one of the reviewed studies. Thus, five studies found significant increased maternal TSH levels upon exposure to PFOS, PFNA and PFHxS [41,42,43,44,45], which might indicate a (subclinical) hypothyroid maternal state. Two studies found significantly decreased maternal TSH levels with increasing PFOS, PFNA, PFDA, PFUnA and PFDoA concentrations [38, 46], which might indicate a hyperthyroid maternal state. None of the ten studies investigating the association between PFOA concentration and maternal TSH found any significant results [38,39,40,41,42,43,44,45,46,47]. However, it should be noted that all these PFAAs are often highly intercorrelated.
Three studies found increased TSH levels in infant upon PFOS exposure [38, 47, 51], while one study also found increased infant TSH levels with PFOA, PFNA and PFDA concentrations . A single study found decreasing infant TSH with PFOS, PFNA, PFDA, PFUnA and PFDoA concentrations . None of the studies investigating the association between PFHxS (n = 5) and infant TSH found any significant associations [44, 46, 47, 49, 52] and seven of the eight studies investigating the association with PFOA found non-significant results [38, 44, 46, 48, 49, 51, 52].
T4 and T3
Five studies found decreasing maternal TT3, TT4, FT3, FT4 and FT4I levels associated with PFHxS, PFOA, PFNA, PFDA, PFUnA and PFDoA concentrations [39, 43,44,45,46]. Low levels of T4 and T3 (and high TSH) could indicate a hypothyroid maternal state. One study showed a positive association of maternal TT3 upon PFDA exposure .
Four studies found decreasing infant TT4, TT3 and FT3 with exposure to any of the investigated PFAAs [39, 44, 51, 52], and four studies found increasing infant FT3, FT4, FT4I, TT3 and TT4 levels with exposure to PFOS, PFOA, PFDA, PFUnA and PFDoA [46, 47, 50, 52].
In overall, most studies showed a positive association of maternal TSH and a possible negative association of maternal T4 and T3 upon PFAAs exposure. This indicates that pregnant women may be at risk of developing a hypothyroid state with increasing PFAAs exposure. Infant TH might also be affected, mainly increasing infant TSH levels with PFOS. More studies are needed to determine the direction of the association with infant T4 and T3 levels.
Evaluation of TH across pregnancy trimesters with PFOS and PFOA exposure
During the 1st trimester, most studies on maternal TH [38, 39] (Table 7) showed that TSH was inversely related with PFOS, although, only one reached statistical significance . Non-significant positive relationships were found with PFOA concentrations. PFOS and PFOA concentrations was mainly non-significantly but positively associated to maternal FT4 levels [38, 40] and TT4 levels  (Table 8). FT4I was inversely associated to PFOA but not significantly to PFOS .
In the 2nd trimester, studies on maternal TH [41, 42] (Table 7) showed that TSH was positively associated with PFOS exposure, being statistically significant in one study . No consistency was found for the relation between PFOA and TSH, with both positive and inverse non-significant associations [41, 42]. Moreover, results on maternal T4 (Table 8) did not show any significant associations .
In the 3rd trimester, studies on maternal TH [44, 46, 47] (Table 7) demonstrated that TSH was significantly and inversely associated with maternal PFOS concentration in one study . No consistent direction was found for PFOA exposure. For the maternal T4 and T3 levels [44, 46, 47] (Table 8), FT4, TT4, FT3 and TT3 were mainly positively, but non-significantly, associated with PFOS and PFOA.
Two studies conducted a longitudinal assessment of TH measurements across the trimesters and postpartum (Tables 2, 7 and 8) [43, 45]. Repeated measurements of maternal TSH were positively associated with PFOS exposure being significant in one study  (Table 7). A non-significantly inverse association was found for FT4, but positive for FT3 upon exposure to PFOS and PFOA (Table 8).
The trimester data evaluation showed negative associations between PFOS and maternal TSH in the 1st and 3rd trimester, but positive during the 2nd trimester (Table 7), whereas PFOA exposure did not elicite significant association to TSH in any of the trimesters. The trimester data evaluations for maternal T4 and T3 versus PFOS and PFOA concentrations showed inverse association between PFOA and FT4I in first trimester  (Table 8). Longitudinal TH studies found a positive association between TSH and PFOS  while data on PFOA exposure did not imply a strong association with maternal thyroid hormones. It must be noted that only few studies with information on each trimester were found, thus, the direction of the estimates might change when more studies are performed. Interestingly, it has been shown that associations of maternal TSH with exposure to PFHxS and branched PFOS isomers were strongest in early pregnancy (1st trimester) and weakened over subsequent trimesters (data not shown) .
Stratifying by infant sex
Seven studies stratified on infant sex [38, 39, 47,48,49, 51, 52] (Suppl. Tables 1 and 2).
PFOS concentration was positively and significantly associated with TSH levels in boys [38, 47, 51] and girls . PFNA was found to be inversely significantly associated with TSH levels in boys  and girls , while another study found positive associations in both genders  (Suppl. Table 1a-b). One study observed a negative association of TSH with cord blood PFOS exposure in girls and PFDoA in boys (Suppl. Table 1a-b) but they did not perform the statistical analysis on all PFAAs congeners in each sex due to their choice of statistical model (sparse partial least squares) .
In boys, TT4 levels were inversely associated with three PFAAs congeners (PFHxS, PFOS and PFOA) measured in maternal blood and cord blood samples (Suppl. Table 2a) [39, 51] and PNFA maternal concentrations were positively associated with FT4I . In girls, TT3 levels were positively associated with cord blood PFHxS , and FT4I and TT4 levels positively associated with maternal PFOS and PFOA and cord blood PFOA (highest quartile) [47, 50] (Suppl. Table 2b). In boys, a study found significant positive associations for FT4 with several cord blood PFAAs (PFOA, PFNA, PFDA, PFDoA) concentration (Suppl.Table 2a), the analysis was however not performed in girls due to their choice of statistical model (sparse partial least squares) . In the pooled infant analyses (Table 10) this study observed similar significant positive association between FT4 level, but inversely association with FT3 level, and cord blood PFDoA concentration .
The reviewed studies do not allow us to establish a distinct tendency or greater sensitivity regarding infant sex. Further studies must be conducted to state any conclusion on gender specific sensitivity for PFAAs exposure and TH disruption.
Previous review literature
The possible link between PFAA exposure and thyroid outcomes in pregnant women was previously reviewed by Ballesteros et al. in 2017 . They included 10 epidemiological studies on PFOS, PFOA, PFHxS and PFNA exposure during prenatal life, childhood and adolescence (up to 19 years), with sample size ranging from 40 to > 10.000 participants. In accordance to this current review, Ballesteros et al. suggested a positive association between PFHxS and PFOS and maternal TSH levels, as well as, between PFNA and TSH levels measured in boys aged > 11 years. Compared to Ballesteros et al. , the current review includes recent epidemiological studies (enrolment year: 1996–2016 vs. 1987–2013), more studies (n = 15 vs. n = 10), bigger sample size (participants > 100 vs. > = 43) and covers more PFAAs congeners (7 vs. 4), thus it might add a stronger study power and support to the previous observations. We also found stronger evidence of a positive relationship between PFAAs exposure (especially PFOS) with maternal TSH, and a possible inverse relationship with maternal T4 and T3 levels.
PFAAs are endocrine disrupting compounds that impact on endogenous hormone homeostasis . Different pathways have been suggested to explain thyroid disrupting effects: PFAAs may induce increased excretion of T4 by alteration of TBG due to competitive binding , and/or increased conversion of T4 to T3 by type 1 de-ionidase in hepatic cells, as well as increased hepatic metabolism of T4 . PFAAs could furthermore alter the responsiveness of the HPT axis , as well as, alter the HPG homeostasis and their signaling pathways which might affect foetal growth [58, 59].
Several animal studies have shown reduced foetal growth upon PFAAs exposure [60,61,62]. Furthermore, animal studies have reported PFAAs disrupted HPT [60, 63, 64] and HPG-axis [65, 66]. However, comparison among studies is challenged because interspecies diversity (e.g. half-lives, excretion and plasmatic carrier proteins), PFAA congeners and levels of exposure, and extrapolation from animal studies to human risk assessment [64, 67].
Studies have suggested that thyroid and oestrogenic/androgenic pathways may be a part conjoined mechanisms. In vitro studies have shown that the HPT axis is affected by PFAAs exposure by inhibiting thyroid receptors, antagonizing thyroid cell proliferation . Furthermore, PFOS and PFOA decreased TPO activity, blocking the ionizing process of thyroglobin , leading to reduced T4 and T3 levels. Cell proliferation of the TH-dependent rat pituitary GH3 cells was shown to be dependent on the involvement of the oestrogen receptor . Moreover, in vitro studies have also shown alterations in the biosynthesis of oestrogens and interference with the oestrogen receptor, after exposure of PFNA, PFDA, PFHxS, PFOS and PFOA, due to their oestrogenic and anti-androgenic potential [2, 69,70,71]. Extracted PFAA mixtures from human serum samples also activated oestrogenic receptors [72, 73] and in an epidemiological study serum PFAAs mixture induced estrogenic activity in pregnant women was associated with significantly lower birth weight and birth length . Most studies examined the effect of single PFAA congeners. More studies are needed for the combined cocktail effect of PFAAs mixtures and/or other POPs, which may have additively and/or synergistic actions [58, 73].
PFAAs exposure and TH outcome assessment
All the reviewed studies determined PFAA concentrations, using small variations of the same methodology (liquid chromatography-tandem mass spectrometry), in mother and/or infant. Maternal biological matrix was either blood plasma or serum and for infant concentrations either cord blood plasma or serum. The LOD and LOQ values were different among studies. The percentages of samples below LOD/LOQ varied among studies, especially for PFHxS, PFNA, PFDA in maternal serum/plasma samples (Table 3) and for most PFAA congeners in cord blood (Table 4). Nevertheless, most selected studies handled the PFAA concentrations below LOD/LOQ with replacement by LOD/LOQ divided by the square root of 2. The sensitivity of the used methods might influence the estimated effect but the exposure levels is expected to be the main factor for the exposure - outcome associations.
Umbilical cord blood samples were also used to determine both infant exposure and thyroid function having the benefit being of representing the fetal environment. However, there are also some limitations of using cord blood for infant thyroid function assessment. Firstly, cord blood thyroid hormone levels likely reflects a combination of maternal and fetal hyroid hormones. Secondly, maternal and umbilical cord blood PFAAs concentrations are highly correlated [74, 75], and the ability of PFAAs to cross the placenta differs between compounds (i.e. PFOA passes more easily than PFOS [20, 75, 76].
Maternal PFAA concentrations might not be constant throughout pregnancy due to physiological hemodynamic changes. PFAAs concentrations decrease during pregnancy due to increased plasma volumes  and higher renal PFAAs excretion . Thus, measurement of PFAAs in late pregnancy might not be directly compared with PFAAs concentration in 1st trimester . Moreover, women are able to transfer some of their PFAAs burden to their offspring by placental [21, 79] and post-partum breast feeding . This complicates comparison between studies with different pregnancy time points of exposure assessment . Even more, one study measured the thyroid hormone outcome before exposure (PFAAs) , which could cause concern of causality, although some retrospective consistency for PFAAs exposure can be assumed due to their persistent nature (long half-lives in human serum) [15, 18]. However, a study showed the PFAA concentrations measured in the 1st and 3rd trimester were highly correlated (r > 0.64)  and argued that a single exposure measurement might be considered as a representative of serum levels during pregnancy.
In most of the studies, the exposure was evaluated as single congeners and not as the real mixture/combined exposure scenario. PFAAs are often highly inter-correlated. It is very diffecult to control for confounding by other PFAAs when assessing the effect of single PFAA congeners and interpret the association in light of the high mutual correlation among PFAA congeners. However, given the structural analogy and similar expected modes of action, PFAAs should be regulated as a chemical family rather than isolated compounds [82, 83]. In this regard, the complementation of exposure biomarkers with biomarkers of combined biological activity to PFAS mixtures in epidemiologic studies may be of great help . Berg et al. (2017)  did not find an association between summed PFAAs exposure (sum of PFHpS, PFHxS, PFOA, PFOS, PFNA, PFDA and PFUnA) and TH outcome. Whereas, high exposure (4th quartile) to the summed POP (sumPFAAs + sumPCBs [sum of PCB 99,118, 138, 153, 163, 170, 180 and 187] + DDE + HCB + cis- and trans-nonachlor) was significantly positively associated with maternal TSH . A study found no effect of the summed PFAAs exposure on maternal risk of hypothyroxinaemia (low free T4 without the compensatory rise in TSH) . Furthermore, one study examined branched isomers of PFOS and found several significant findings, but not with the total PFOS .
The selected studies in this review investigated FT4I, FT4, FT3, TT4, TT3, T3RU and TSH in mothers and infants using different methods of chemoluminescent-, radio- or enzyme linked immunosorbents assays, with variable sensitivity (e.g. LOD (TSH) = 0.50 μU/mL  vs. LOD (TSH) = 0.01 μU/mL ). The radio immune sorbent assay seems to be the most reliable method for the quantification of thyroid hormones , but all methods showed good sensitivity, specificity, and accuracy. However, the use of immunosassay for hormone determination has been criticized. Animal studies suggested that the effect of PFOS to decrease FT4 level might be due to negative bias in analog techniques resulting from competitive displacement of FT4 and the labeled analog from serum and assay binding proteins in the presence of PFOS [61, 87]. Nevethrthless, such bias was not observed in human study and such different result between human and animal study might be related to the difference of PFAA level, protein binding to T4 (in rats: TTR, in human: TBG) and the interaction of proteins and PFAAs . Based on these reported data, the use of different methods seems not to be of concern in measurements of human thyroid hormones.
Most studies examined TH at a single time point during pregnancy, impeding the capture of an effect on the thyroidal system during pregnancy since levels vary substantially during different pregnancy stages and stabilises approximately 6 weeks postpartum [34, 89]. However, longitudinal studies measure TH outcomes multiple times having a greater chance of witnessing an effect on the thyroid system [42, 43, 45].
TSH was examined in all studies except one , and T4 and T3 parameters were less consistently studied. The free TH reflects the biologically active hormones, which may give a better estimate of the exposure effect on the thyroid regulatory systems . The FT4I is an estimate of circulating free T4 levels, which accounts for changes in T4 levels due to changes in thyroid binding protein levels and saturation, such as increases in TBG during pregnancy , being a better estimate, but only assessed in one single study . Some of the studies only investigated a single TH parameter [41, 48, 50], however to understand the effect of PFAA exposure on the TH equilibrium at least T4 and TSH measurements are required.
The extent of adjustment for potential confounders varied largely between studies. General confounders were pre-pregnancy BMI, maternal age, socio-occupational status, parity, alcohol and smoking (Table 2). Eight studies excluded participants with thyroid diseases or related medication [38, 39, 41,42,43,44, 46, 48] and one study adjusted for these confounders . However, few studies also took into account other thyroid parameters such as iodine status, TPOAb or thyroid binding capacity. Berg et al. demonstrated that several significant effect estimates lost significance after adjusting for thyroid binding capacity in the statistical analysis , suggesting an area that should be further studied.
FT4I was suggested as a more reliable FT4 estimate during pregnancy . Weiss et al. found evidence that PFAAs might decrease T4 due to their affinity to thyroid hormone transport proteins using transthyretin (TTR) as carrier protein . However, Preston et al. did not support this theory, after showing lower saturation of plasma binding proteins with higher PFAAs exposure . The incoherent results may be due to the fact that experimental studies use TTR as carrier protein, although the major carrier protein in humans is TBG, which has lower PFAA affinity than TTR .
Three studies stratified for TPOAb-status and maternal TSH was associated with several PFAAs in the high TPOAb group [39, 42, 45]. However, the high TPOAb groups differed in sample size (e.g. 14 vs. 98 participants), gestational week (GW) of exposure sampling (median GW 9.6, GW 15–18 and multiple time points), TPO-Ab threshold (Webster et al. and Reardon et al.: > 9 IU/mL vs. Preston et al.: > 2 U/mL) and PFAA concentrations (lower in Reardon et al). Even though results went in different directions, it might indicate a more susceptible group among women with thyroid stress, whom may not be able to compensate for the PFAAs-induced thyroid affections . If these women are highly exposed to PFAAs, which are strongly suspected to affect the thyroid gland, it could increase their risk of immune reactivity and affect the TH production necessary for a normal foetal development.