Association Between PFOA Exposure and Liver Enzymes: a Systematic Review and Meta-analysis of Longitudinal Cohort Studies

Background: Peruorooctanoic acid (PFOA) has been the focus of research regarding its effects on liver function due to its potential for bioaccumulation in human body, widespread exposure, and the ability acting as an endocrine disruptor. Although the hepatotoxicity of PFOA has been identied, conclusions from population-based studies remain inconsistent, and the purpose of this study was to systematically review the epidemiological evidence linking PFOA to the four most common liver enzyme indicators and to quantitatively pool the effects of existing longitudinal cohort studies. Methods: Longitudinal cohort studies on PFOA and the four liver enzyme indicators (alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT) and alkaline phosphatase (ALP)) were searched in three electronic databases (PubMed, Web of Science and EMBASE) until July 10, 2021. Random-effects meta-analysis was used to calculate pooled effect estimates for continuous exposure. We also assessed the risk of study bias and the overall quality and level of evidence for each exposure and outcome combination. Results: The initial search identied 18 studies, 9 of which were eventually included. PFOA showed signicant relationships with all four liver enzyme metrics: ALT (N=7, β=0.012, 0.001-0.023), AST (N=6, β=0.041, 0.002-0.079), GGT (N=6, β=0.010, 0.001-0.020), ALP (N=4, β=0.007, 0.002-0.011) with low to high concomitant heterogeneity. The certainty of the combined evidence for each exposure-outcome combination was considered "moderate". Conclusions: We conducted the rst meta-analysis and systematic review based on a longitudinal cohort study, and the results support the hypothesis that PFOA increases levels of four common liver enzymes, and that this causal relationship was observed within the context of a population-based study area. Further clinical or epidemiological research, especially longitudinal studies, are needed to expand the sample size and further determine the effect of different types and doses of confounders of PFOA exposure on liver enzyme indices, with adequate consideration of sex, age, and other confounders as well as reverse causality. review based on a longitudinal cohort study conrming the effect of PFOA exposure on liver enzyme indicators in humans, which demonstrated that PFOA exposure showed a statistically signicant association with increased levels of four liver enzyme indicators, accompanied by some heterogeneity. Further clinical or epidemiological studies, especially longitudinal studies, are needed to expand the sample size and further identify the effects of different types and dosages of confounding factors of PFOA exposure on liver enzyme indicators, which are essential to verify the evidence of causality and clear dose-response curves. In addition, future studies should use mediation analysis with full consideration of gender, age and other confounders and reverse causality.


Introduction
Per uorooctanoic acid (PFOA) is one of the most used per uoroalkyl acid (PFAA) class of chemicals, an environmentally widespread uoropolymer [1], characterized by high stability and persistence, which is widely used in food packaging, furniture, textiles, and other commercial manufacturing applications [2] on account of its surfactant property. However, environmental and health issues related to PFOA have attracted increasing public attention in recent years, and an increasing number of experimental studies have revealed the extensive deposition of PFOA in serum and liver after absorption by humans, as well as the long half-life and hyper-biotoxicity, especially hepatotoxicity, as a typical endocrine disruptor [3,4]. Extensive population-based epidemiological studies have identi ed its possible association with adverse birth liver disease, pancreatic cancer, biliary obstruction, and other liver diseases such as hepatocellular carcinoma, and the association of biological markers including liver enzymes, bilirubin, and lipid level markers with PFOA has been routinely reported [5][6][7]. A community study in [8] and a longitudinal cohort study in [9,10] both found a linear positive association between PFOA serum levels and simultaneously measured ALT, AST, and GGT concentrations, and this relationship was validated in another National Health and Nutrition Examination study (NHANES) [11], but [12,13] and [9] failed to detect this effect and [14] even found an inverse association between ALP and PFOA, and it is worth noting that most current epidemiological studies have clustered in cross-sectional studies and small occupational population cohort studies [15]. Despite the fact that many developed countries especially in the United States have banned industrial PFOA emissions and manufactured new short-chain alternatives with similar e cacy, the high bio-chain accumulation of PFOA and its non-degradable character result in varying levels of PFOA still being detectable in the serum of most humans, even in the Arctic [16][17][18][19].
Since the four biological liver enzymes (ALT, AST, ALP, GGT) are the most important biological enzymes re ecting liver function and disease severity, they are also the most commonly used clinical indicators of liver function and are capable of playing an important auxiliary diagnostic role in the areas including hepatocellular injury and hepatitis (ALT, AST, albumin), biliary tract disease and bile duct obstruction (ALP, GGT), bile secretion and lipid metabolism (bilirubin) [20,21]. Besides, these four liver enzyme activities become generally elevated in extrahepatic diseases such as syphilis [22] and obesity [23], and a cohort study revealed that ALT, AST and GGT levels were signi cantly higher in patients with severe COVID-19 than in those without severe disease [24,25]. However, environmental factors such as persistent pollutants (e.g., PFOA, PFOS) with liver enzymes has still lacked su cient attention [11,12,15]. Animal experiments have shown multiple potential mechanisms suggesting the role of PFOA in altering liver enzyme levels, including that exposure to PFOS signi cantly enhances the expression of various genes in the mTOR signaling pathway, disrupting speci c biotransformation pathways [26,27]. Yi Wen et al. [28] found considerable inhibitory effects on both anti-apoptotic and lipid metabolism genes in HepG2 hepatocytes after 48 h exposure to PFOA, and [12] and [29] also demonstrated lipid regulation, multiple hormone regulation and hepatocyte metabolic toxicity of PFOA in rodents. In contrast, population-based studies remain inconsistent, a 10-year U.S. population study found that changes in liver enzyme levels with PFOA exposure corresponded to different stages of renal function [30], Most cross-sectional studies have found that PFOA exposure is not accompanied by statistically signi cant changes in liver enzymes [8, 10,11]. Therefore, it remains controversial whether population-based studies of PFOA exposure cause changes in liver enzyme levels [15].
Considering the current small sample size of individual epidemiological studies and the methodological differences in exposure and liver enzyme measurements, inconsistencies in the selection of study contexts and populations as well as the large number of cross-sectional studies that may have confounded potential causality determination, these may have contributed to the con icting ndings of epidemiological studies of PFOA and liver enzyme levels [10,15,31,32]. Of environmental epidemiology research, systematic reviews and meta-analyses (SRMAs) are increasingly applied to quantitatively combining data across studies and assisting in translating evidence into policies. Sparse data were available on the relevance of PFOA and liver enzyme, not even systematic reviews as well as meta-analyses during the literature search. Although the overall volume of relevant literature is encouraging, the number of studies assigned to speci c combinations of different exposure outcomes is inadequate. Therefore, our study aims to address these inconsistent ndings and summarize the latest evidence from longitudinal cohort-based studies in a more standardized and detailed format to provide policymakers and healthcare providers with a more comprehensive assessment of the association between PFOA and liver enzyme indicators, thereby improving the preventive health of the public.

Search strategy
Our reporting process is in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines [33] ( Table S1). Initially, we formulated a speci c research question based on the PECOS (Population, Exposure, Comparator, Outcome, Study) principles as follows "Do the levels of the four liver enzymes (ALT, AST, GGT and ALP) change (outcome) in people (population) when exposed or presenting with PFOA (exposure) compared with all people not exposed to PFOA or at low risk/level of exposure to PFOA (comparators) based on cohorts (study design)? " We used the Web of Science, PubMed and EMBASE databases for search using terms: (PFOA OR pfoa OR Per uorooctanoic Acid OR Pentadeca uorooctanoic acid OR Per uorocaprylic acid) and (liver enzyme* OR hepatase OR hepatic enzyme* OR ALT OR Alanine transaminase OR AST OR aspartate transaminase OR GGT OR gamma-glutamyl transferase OR ALP OR alkaline phosphatase) applied to ascertain the relevant studies. The search criteria were not restricted to language and were last updated on July 10, 2021. The speci c search methods and results are shown in (Table S2).

Study selection
Following the guidance of the PECOS statement we have established the following eligibility criteria: (P): People exposed to per uorinated compounds (living, working or schooling in the factory, near contaminated water, in a contaminated atmosphere, etc.); (E): Exposed to PFOA OR pfoa OR Per uorooctanoic Acid OR Per uorocaprylic acid in the body (body uids, e.g., serum, plasma, etc.); (C): Individuals assessed as non-exposed or at low risk/level of exposure to PFOA; (O): Levels of four enzymes including ALT, AST, GGT, and ALP; (S): All cohorts, cohort-based reports with well-documented data were regarded as quali ed for review. Firstly, we collected the searched literature and independently ltered the complete titles and abstracts after removing duplicates; after the initial screening, we downloaded the full text of the initial screened literature and secondary screened to decide whether the literature should be retained or removed via a thorough evaluation. All disagreements during this period will be resolved through group discussions.

Data extraction and quality assessment
The author extracted the data independently using a prede ned data template. We focused on the analysis of the association between PFOA and the four liver enzyme indicators. The characteristics of the extracted eligible studies are shown in (Table 1). On this basis, the rst author/corresponding of the selected study will be communicated with to obtain necessary data that are missing or cannot be extracted from their supplementary material or relevant publications.
We assessed the quality of each study using the corrected list used in the [34]. The list was divided into three panels of high, medium, and low quality (quality scores of 1.00 to 0.80, 0.79 to 0.60 and < 0.60, respectively). Each element of the list was scored 0.00 to 1.00, then the total score for all elements of each study was calculated, and nally the unit-weighted quality score was calculated (Table S3).

Risk of bias assessment
We a rmed our conclusions by evaluating the risk of bias (RoB) for each study included in the meta-analysis. The RoB assessment describes the quality of the study and may indicate systematic errors in the results related to the assessment of methodological quality [35]. Individual human and animal studies were classi ed using the NTP/OHAT risk of bias assessment tool [36]. Three signi cant factors (confounding bias, exposure, and detection bias in outcome assessment) and four other factors (selection bias, selective reporting bias, consumption/exclusion bias and con ict of interest) were identi ed in observational studies. Each study assessed the risk of bias through a series of questions using "de nitely low," "probably low," "probably high," or "de nitely high" (see Table S4 for detailed questions and Table S5 for the rationale for each study's assessment). We identi ed the main confounding factors between continuous exposure (PFOA) and outcome according to [15], the domains classi ed as confounders in the study were assigned de nitely high RoB if no or only one signi cant confounder (sex or age or alcohol or smoking or BIM) was considered; classi ed as probably high RoB if at least two of the following confounders (age or alcohol or smoking or BIM or educational attainment or income or sex) were considered; and classi ed as potentially probably low RoB if all of the above confounding factors were considered, the domain was classi ed as likely low RoB; if all of the above confounding factors and at least one of the following factors were considered (race; household income; regular exercise; fasting status; air pollution/noise; urbanization; greenness of residence) the domain was classi ed as de nitely low RoB. Studies with potentially higher RoB ratings were assigned without su cient information to determine RoB, and each study was classi ed as an overall level 1, 2, or 3 of RoB nally.

Meta-analysis
The purpose of the SRMA was to quantitatively assess the risk of PFOA exposure and four liver enzyme levels after combining the results of individual studies with appropriate integration. In this study, the included studies were all cohort studies, and the cross-sectional studies were excluded because the causal interpretation was limited by their design.
Pooled effect estimates were calculated unit change per 1 ng/mL or ln (ng/mL) of PFOA for our meta-analysis. We performed meta-analyses of continuous exposures Since most studies on the relationship between PFOA and liver enzyme levels suggest a linear effect [9,10], and the currently available data are insu cient for categorical exposure assessment (i.e., fourth percentile versus rst percentile exposure, or third percentile versus rst percentile exposure). The effect estimates used for evidence synthesis were extracted from the "fully adjusted model" or "main model" from the included studies. If studies used the same population cohort and did not differ in composition and follow-up time, we therefore analyzed only the most recent date, whereas [14] reported different liver enzyme outcome indicators based on the same US birth cohort ( [29], ALT, AST, GGT; [29], ALP, ALT, AST, GGT; [14], ALP, ALT, AST), and the effect size units were not identical (2007, per ln-unit (ng/ml); 2000, per IQR (ng/ml); 2012, per log-unit (ng/ml)), with large differences in sample size (2007, N = 506; 2000, N = 74 ; 2012, N = 179), therefore, these three studies were retained simultaneously for our further quantitative analysis. Similarly, where possible, maximally adjusted effect estimates (regression coe cients and β) were extracted; [37] reported ΣPFOA, n-PFOA, iso-PFOA, we did not consider dividing the PFOA subgroup and only extracted the total PFOA; [9] reported different population cohorts (N = 1605, N = 1176), we selected the later because it excluded drugtakers and reduced the risk of bias from confounders, similar situation was found in [29].
Based on the heterogeneity between studies, we performed a meta-analysis for each exposure-outcome combination in the form of random effects and xed effects [15]. We used the I 2 (greater than 50% is considered to indicate substantial heterogeneity) [38] and the P-value of the Chi-squared test for heterogeneity (less than 0.1 indicates statistical signi cance) [39] to assess heterogeneity. In addition, to identify sources of heterogeneity in the meta-analysis of included studies [38], pooled estimates were evaluated by subgroup analysis for consistency in whether adjusted for smoking, alcohol, and BMI, number of co-variate adjustments, and sample size, since the number of studies included in each of the four exposure outcome combinations was less than 10, preventing the possibility of meta-regression (Supplementary Material B). We also used the Egger test to assess publication bias [13]. Considering that statistical and graphical methods survive publication bias inaccuracy in some studies [40], Doi plots and Luis Furuya-Kanamori (LFK) indices (absolute value less than 1, "no asymmetry"; absolute value between 1 and 2, "slight asymmetry"; and absolute value greater than 2, "severe asymmetry") were evaluated. We performed sensitivity analyses to assess the stability of the results by sequentially excluding individual studies (the leave-one-out method) and recalculating the pooled estimates. Stata version 14 (StataCorp LP, College Station, TX, USA) and MetaXL v.5.3 software (EpiGear International Pty Ltd, Sunrise Beach, Queensland, Australia, www.epigear.com) were used to perform all data analyses. For articles that did not t the meta-analysis, we investigated and analyzed the relevant associations of the studies by descriptive summaries.

Con dence ratings of evidence body and translation of evidence levels
According to the updated NTP/ OHAT [41] and different exposure results, we assessed the body of evidence on the basis of the combined trust under the GRADE guidelines [42]. The NTP/OHAT framework was used to assess the quality and strength of the evidence because it was consistent with the causality criteria with Hill. In summary, observational studies exhibited low to moderate initial con dence ratings because of the lack of control for speci ed exposures.
The initial con dence in the body of evidence was determined by key study design features. We assessed each exposure-outcome combination with four factors that increased initial con dence or ve factors that decreased initial con dence. The nal con dence for each exposure-outcome combination was categorized into four descriptors: "high" "moderate" "low" or "very low". The nal con dence ratings of different combinations of exposure-outcomes will determine whether the level of evidence would be interpreted as "high", "moderate", "low", "insu cient" or "no evidence of health effects" level in epidemiological evidence.

Study selection and characteristics
As shown in Fig. 1, we ultimately included nine relevant cohort study articles and pooled the available data on PFOA, and the four liver enzyme indicators provided in these studies for meta-analysis as well as systematic reviews [9,15,38,39].
In Table 1, we summarize the characteristics of the included cohort studies. The years of publication ranged from 2006 to 2019. Seven studies were from the United States, the other two from China. The minimum sample size of the included studies was 74 and the maximum sample size was 32,254. Most studies did not provide data by gender subgroup, only two studies reported relevant data for both men and women respectively [10,39]. Of the continuous exposure data, seven studies were associated with ALT, four studies investigated ALP, two each of AST and GGT; for categorical exposure data, only one each of ALT and GGT was available [15,29,38]. All nine cohorts evaluated the relevant liver enzyme indicators by clinical examination reports, and eight studies adjusted for age and six studies adjusted for alcohol as independent confounder.

Risk of bias assessment and study quality
The results of the RoB assessment using heat maps are presented in Table 2. For the confounding bias assessment of the included studies, one was rated as "de nitely low" ROB [15], four as "probably low" ROB, and three as "probably high" ROB [14,29,38,40], because fewer major confounders or total confounders were included in the meta-analysis adjustment. For the assessment of detection bias, one study was classi ed as "de nitely low" ROB in the assessment of exposure characterization [12] and eight studies as "probably low"; one study was "de nitely low" in the ROB assessment of outcome characterization [38] and eight studies with "probably low". In the assessment of selection bias, four studies with "absolutely low" ROB [9,10,12,15], as they were all cohort studies with large population sample sizes, and the remaining ve cohort studies were rated as "probably low" ROB. For the assessment of attrition/exclusion bias, all included studies were rated as "probably low" RoB as there were no explicit exclusion bias. For selective reporting bias, one study was rated as "absolutely low" RoB [9], as the results included in the meta-analysis provided su cient detail, and eight studies were rated as "probably low" RoB for indirect reporting of results. In the assessment of con icts of interest, most authors declared no con icts of interest, but two studies did not state whether there were con icts of interest or nancial support thus rated as "probably high" ROB [13,14]. In short, all eligible studies were categorized as either level 1 (n = 7) or level 2 (n = 2) of the RoB, demonstrating the presence of reasonable bias that might cast some doubt on the results. The studies included in our meta-analysis were cohort studies accompanied by relatively explicit exposure and outcome assessments with quality scores between 0.88 and 0.96, so the quality of the studies we included can be considered high (Table S3).  T1  T1  T1  T1  T1  T1  T1  T2  T1 3.3. Data synthesis 3.3.1. Meta-estimates of the relationship between PFOA and ALT Meta-analysis summarized effect estimates for per 1 ng/ml or 1 ln(ng/ml) increase in PFOA, and there was a statistically signi cant increase in ALT (n = 7, β = 0.012 IU/L, 95% CI: 0.001-0.023, P = 0.000) with an overall I 2 of 86.5% which showed the statistically signi cant heterogeneity; when introducing the exposure as high versus low categories, similar results were found, and there was a statistically signi cant increase (n = 1, β = 0.048 IU/L, 95% CI: 0.031-0.066) ( Table 3). 3.3.3. Meta-estimates of the relationship between PFOA and GGT Meta-analysis summarized the six combinations of study for per 1 ng/ml or 1 ln(ng/ml) increase in PFOA showed there was a statistically signi cant increase in GGT (n = 6, β = 0.010 IU/L, 95% CI = 0.001-0.020, P = 0.001) accompanied with statistically signi cant heterogeneity (I 2 = 75.4%); when introducing the exposure as high versus low categories, similar results were found, and there was a statistically signi cant increase (n = 1, β = 0.013 IU/L, 95% CI:-0.010-0.036) (

Analysis of subgroups
By analyzing the data from the included studies, we established four subgroups of whether adjusted for alcohol, smoking or BMI, number of covariate adjustments, and sample size, to explore and assess the sources and characteristics of heterogeneity between exposure-outcome groups (Table 4) Table 4).

Sensitivity analysis and publication bias
Sensitivity analysis of the relationship between continuous exposure to PFOA and four liver enzyme indicators is shown in Table S6. Since most of the analyses did not substantially change the pooled effect estimates, we consider the results are stable to a certain extent. It is noteworthy that when we performed the meta-analysis under the quality effect model, all the above associations that were originally shown to be signi cant under the random effects model disappeared (Figures S1-S8). By Egger's test, we did not identify evidence of publication bias for the exposure-outcome combinations (Table 3). However, evidence from Doi plots and LFK index indicated that all the exposure-outcome combinations showed publication bias ( Figure S21-S36). Table S7 showed a summary of our con dence ratings for each exposure-outcome combination in the body of evidence. Based on the NTP/OHAT framework, only control and experimental studies were initially rated as "high con dence," but the majority of included observational studies with unmeasured risk of confounding would be rated as "moderate con dence" throughout the rating process, and a few other studies were assessed as "low con dence" due to low con dence in the risk of bias.

Discussion
To our knowledge, our study is the rst meta-analysis and systematic review in the relevant eld to incorporate all relevant cohort studies exploring the association of PFOA with liver enzyme indicators. The results of our cohort-based study support the hypothesis that PFOA increases the levels of four common liver enzymes, and this causal relationship was observed to the extent of the population-based study eld. However, based on subgroup and descriptive analyses, we still detected that our included studies did not provide consistent strength of evidence support, and the heterogeneity of results across liver enzyme indicators and differences in the degree of adjustment for confounders were limiting factors in the interpretation of causality from our results ( Table 4).
The hepatotoxic effects of PFOA in rodents leading to hepatomegaly and hepatocyte histological alterations have been well documented [43,44], and although different mechanisms also play a role in humans, the PPAR-α agonist-mediated pathway remains one of the main mechanisms leading to altered expression of genes involved in peroxisome proliferation, cell cycle control and apoptosis, and this PPAR-α response has similarly predicted in human cellular responses [45][46][47]. Exposure to PFOA resulted in site-speci c DNA methylation of the mTOR pathway inhibitor Pten gene, decreased gene expression and increased expression of Mtor and Kit, promoting apoptosis and liver injury [26]. The experiments with HepG2 cells cultured in vitro revealed that PFOA activated lipid metabolism genes, altered cellular metabolic rate, increased lipid deposition in hepatocytes, and exhibited great cytotoxicity [28]. PFOA disrupted the activity of metabolic detoxi cation enzyme CYP450 by blocking the interaction of nuclear translocation complexes with DNA sequences, affecting the conversion and excretion of toxic substances, which may lead to diminished hepatic detoxi cation [48]. One important mechanism of PFAS interference with fatty acid metabolism and lipid transport is the strong a nity for hepatic fatty acid binding proteins, showing intrinsic hepatotoxicity and bioaccumulation [49,50].
Several studies have shown that PFOA is a speci c risk factor with hepatotoxicity that affects the synthesis and metabolism of liver enzymes, and these studies compared associations between different indicators of liver functional impairment (e.g., liver enzymes, bilirubin, and abnormal lipid metabolism) at different levels of PFOA exposure, but most of this evidence comes from cross-sectional studies as well as cohort studies with small samples, there is a gap in systematic reviews and meta-analyses of longitudinal studies that quantify these associations. Similar to our results, [9,15,29] reported that the relationship between PFOA exposure and ALT elevation (a proxy for hepatocyte injury) in longitudinal studies was consistent across analyses, speci cally for for PFOA continuous consideration (per one ln-unit increase) and quintile (from the rst to the fth quintile) for cumulative and year-speci c serum PFOA, as well as for ALT continuous consideration (ALT level) or as dichotomous results (odds of above-normal ALT), signi cant relationships of increasing ALT were observed. In contrast, however, studies by [12,14,38] suggested no clinical hepatotoxicity associated with PFOA levels, in contrast to previously reported observations. For AST, the existing conclusions of individual cohort studies were not harmonized, with three studies reporting that continuous exposure (per one unit or ln-unit) to PFOA brought a signi cant increase from 0.25% [9] to 2.9% [13], while three other studies failed to reach a signi cance result [14,29,40]. The majority (n = 4) of the six studies that currently examined PFOA and GGT reported the signi cant relationship with continuous exposure to PFOA [9,10,13,29], while two other studies that examined the of the effects of community PFAS exposure similarly gave positive effect sizes (0.003 and 0.000) although without statistical signi cance [15,40]. Among the four longitudinal studies that focused on the association between PFOA and ALP, only one study reported the signi cant association between continuous per one unit exposure to PFOA and an increase in ALP at the 0.73% level [13], and two studies respectively from the United States and China failed to generate a statistically signi cant association [9,40], and [14] even reported a negative association of continuous per one ln-unit exposure to PFOA with a 17.73% (-1.1750, -0.0007) decrease in ALP levels.
It is noteworthy that substantial and statistically signi cant heterogeneity was detected in the results of the available cohort analysis for the remaining three of our four liver enzyme indicators, except for ALP (n = 4, 0.007, I-squared = 0.0%) ( Table 3), and that there is substantial and inevitable heterogeneity in studies of liver function biomarkers related to environmental exposures that must be considered, as suggested by [30,51]. Between-group heterogeneity may arise from the dose of PFOA exposure, statistical methods, different sample sizes, inconsistencies in geographic characteristics and the degree of adjustment for confounding factors (e.g., gender, age, alcohol intake, BMI, smoking history, etc.). This limits to some extent the interpretation of our results in terms of implications.
For the effects of combined exposure to different doses of PFOA, while higher PFAS concentrations are always associated with higher ALT in most crosssectional studies [10,11,14,15,29,32], prospective studies have revealed the most null association [32,52], it also indeed leads to differences in results, as described above [9,53]. There is still no consensus regarding the shape of the dose-response relationship curve between PFOA and liver enzyme indicators [10,12,15,32], Kennedy et al. characterized sub-chronic and chronic toxicity studies in rodents, revealing the liver as the most sensitive target organ for PFOA action and the dose-dependent increase in serum ALP, ALT, and AST levels associated with PFOA administration [54], which has been veri ed in many animal experiments [43,44,55], however unlike rats and mice, no similar situation was observed in cynomolgus monkeys after oral dosing PFOA (6 months) [56].
While in population experiments, chronic exposure to community PFOA exposure among 371 residents from households who had resided in the Little Hocking Water Association district ( [40], PFOA:354(184-571)/(ng/ml)), which is substantially higher than the average PFOA currently observed in general population samples in the United States, but they failed unexpectedly to nd any signi cant positive relationship between serum (PFOA) and markers of liver function (including three classes of liver enzymes) and other markers with potential health effects; In another longitudinal evaluation of clinical parameters with PFOA levels (median) greater than 10 ng/ml, similarly no adverse correlation was found between changes in clinical chemistry of PFOA (50.9 ng/ml), non-HDL cholesterol, HDL and liver enzyme indicators [14]; Another cohort-study from a high- and their low RoB scores in all six domains including selection bias, making the quality of their evidence seriously dubious. The other ve included cohort studies all measured PFOA levels below 10ng/ml and showed the positive and signi cant correlation between PFOA exposure and ALT, except for the study by [12] (-0.3000, -0.7333 to 0.1667 ); in ALP, [9] and [13] reported contradictory ndings (-0.0111, -0.0336 to 0.0119 versus 0.0073, 0.0028 to 0.0118), respectively; in AST, except for [29] (0.18, -0.0400 to 0.4000), the other two studies reported signi cant increases; in GGT, the signi cant relationship was reported in four studies except for [15] (0.0030, -0.0020 to 0.0080). In the absence of explicit prior statements about the dose-response relationship between PFOA and any sort of liver enzyme indicator, this variability strongly suggests a possible nonlinear relationship between them with stronger effects at low dose exposures, but further detailed studies are required.
Effect estimates in our meta-analysis are reported in describing comparisons per one unit or ln-unit change, mainly considering that the available data are mainly clustered under this group. More importantly, Kerger et al. [57] demonstrated that classi cation by chemical dose level, such as utilizing the low exposure group as a valid control group for statistical comparisons of common disease states, raises signi cant di culties in toxicological studies, possibly due to inadvertent selection bias that may have affected the lowest exposure quartile (control group), making the dose-response relationship between PFOA/PFOS and risk of outcome tenuous; also, we have attempted dose-response meta-analysis to investigate a possible dose-response relationship between PFOA and liver enzyme indicators, but the existing number of cohort studies was insu cient to extract the necessary data to proceed. Therefore, in aggregate, our approach is the most reasonable allowing consistent comparison of study-speci c estimates and interpretation of ndings at this stage.
In addition to the dosage differences mentioned above, which may account for much of the heterogeneity, regional socioeconomic status (SES) or potential sociodemographic predictors have emerged as potentially important confounding factors in environmental health studies [58], 7 of the 9 studies we ultimately analyzed for data synthesis were from the United States, with the exception of 2 from China, by simply classifying we can observe that the studies from China showed statistically signi cant associations for all three indicators (ALT, AST, GGT) except ALP, which can be moderately explained by the lower foundational health conditions, while the presence of publication bias for positive results must also be considered; the studies we included ranged from the largest sample size of 32254 [15] to the smallest of 74 [13], the effect of sample size on the results was not demonstrated explicitly in our subgroup and main analyses; other factors such as age and gender, given the small amount of evidence available for analysis, subgroup analyses cannot ensure the number of per group greater than 3, which is insu cient to account for speci c associations. Given the inherent limitations of epidemiological data, these estimates need to be interpreted with caution, although our study considers only the results of longitudinal cohort studies, which can show associations with respect to time and exclude the possibility of reverse causality to some extent [9,14,38].
Gender differences in the association between PFAS and liver enzymes have barely been reported in the literature [15], and only one study of the nine cohorts we included attempted to explore differences between men and women by performing subgroup analyses by gender, Mora et al. [12] reported a prospective Boston area prenatal cohort that revealed in girls during childhood, higher levels of PFOS, PFOA concentrations were associated with deleterious changes (higher TC and/or LDL-C, higher HDL-C, and slightly lower ALT), unlike the differences in PFAS-lipid associations between males and females in different age groups [59][60][61] which have been widely reported. In the majority of cross-sectional as well as cohort studies, no clear relationship between PFOA exposure level and gender was found [15], but [38] reported differences in the gender distribution of two PFAS (PFTA and PFHxS), speci cally higher dosages of exposure were observed in the male subgroup, an earlier NHANES studies also demonstrated higher mean concentrations of PFOS, PFOA and PFHxS in males than in females [17], the signi cant effect of PFAS and liver enzymes at low doses described above might be a reason to explain the sex differences.
Considering the inevitable differences in lifestyle and exposure patterns such as product use, chemical plant work environment, etc. between the genders, further research on this topic is urgently needed.
The four liver enzyme indicators we selected can largely re ect the impairment of liver function [9,12,14]. However, each individual index has its own differences in sensitivity and speci city, it is still challenging to re ect the status of liver function with differences in biomarker changes only, and it remains a hot topic whether the possible effects of PFAS are clinically signi cant compared to other risk factors for liver enzyme effects, speci cally: higher ALT is a marker of hepatocyte dysfunction, and as one of the most common one of the liver function tests, with a reference value of less than 40 units, is the main diagnostic item for hepatocellular parenchymal damage, its high level often parallels the severity of the disease and is commonly used to screen children for nonalcoholic fatty liver disease (NAFLD) [62], but 1) there is a lack of consistency between changes in ALT activity and pathological histological changes in the liver, some patients with severe liver damage do not have elevated ALT; 2 ) ALT suffers from the lack of speci city, and there are various reasons that can cause changes in hepatocyte membrane permeability, such as: fatigue, alcohol consumption, colds and even emotional factors; the normal value of AST is 0-37 µ/L, but it is present in hepatocytes and cardiomyocytes at the same time, and even the levels in cardiomyocytes are higher than in hepatocytes, which largely limits its application, and in clinical work we often combine AST with ALT, when ALT is signi cantly elevated with the ratio (also known as De Ritis ratio) of glutathione (AST)/glutathione (ALT) > 1, it indicates damage to the liver parenchyma and can also be used as an auxiliary test for myocardial infarction and myocarditis [63-65] ; the normal participation value of ALP is 30-90u/L, and the increased level represents impaired biliary excretion in the intrahepatic biliary tract, which is mainly used for obstructive jaundice, primary hepatocellular carcinoma, secondary hepatocellular carcinoma, cholestatic hepatitis, etc. However, this enzyme is also active in bone tissue, and serum ALP can also be elevated in pregnant women, healing fractures, osteochondrosis, osteoporosis, leukemia, and hyperthyroidism; GGT is very low in healthy human serum (less than 40 units), mainly from the liver, with slightly produced by the kidney, pancreas, and small intestine, and GGT is not as good as ALT in re ecting necrotic damage to hepatocytes, and 1) GGT is not as good as ALT in re ecting the necrotic damage of liver cells, but it can be used to distinguish jaundice caused by internal and external liver obstruction, 2) acute and chronic viral hepatitis, cirrhosis, 3) acute and chronic alcoholic hepatitis and drug-related hepatitis: GGT can be signi cantly or moderately elevated (300-1000 U/L), while ALT and AST are only mildly elevated or even normal, 4) alcoholics can have their GGT decreased after they stop drinking, 5) GGT can also be elevated in other toxic liver disease, fatty liver, liver tumors. Thus, even though we simultaneously found that PFOA exposure resulted in elevated levels of all four liver enzymes and was able to suggest a state of hepatic impairment as assessed by quality of evidence as well as bias analysis, it is still challenging to point speci cally to a speci c hepatic impairment disease.
Pending exploration of whether such exposure to long-chain compounds (e.g., 9- Notably, the results of the meta-analysis under the random effects model and the quality effects model were inconsistent, cutting all the statistically signi cant combinations. The con icting results suggest that the need for further research to expanded number of studies and determine the relationship between each exposure and outcome.

Strengths And Limitations
This study provides the most recent and comprehensive cohort-based review of the evidence on the relationship between PFOA and the four most used liver enzyme indicators. The major strengths of this study are the integration of the most up-to-date research evidence and broad coverage of all longitudinal studies with rigorous, transparent, and reproducible assessment of the evidence and standardized data processing methods that preserve the originality of the data as much as possible, giving a comprehensive and systematic elucidation of the potential role of PFOA exposure in increasing liver enzyme levels. This study makes a timely contribution to a rapidly evolving eld of research, illuminates the changes facing meta-analysis, and provides suggestions for improving the potential utility of further studies.
Nevertheless, our study still has limitations that characterize the lack of research in this area and the remaining open questions that need to be addressed, as detailed below: We considered only one type of PFAS, considering that most long-chain PFAS (e.g., PFTrDA, PFTeDA) as well as short-chain congeners and functionalized per uoropolyether as substitutes (e.g., chlorinated poly uoroalkyl ether sulfonic acids (Cl-PFESAs) or F-53B) are of restricted relevance [70], and additionally Nian et al. [9] revealed that PFOA isomers (straight or branched chains) differences in the relationship with liver enzyme indicators, future studies should focus on the effects of different dose responses and different types of PFAS.
For further grouping of different categories of data (continuous versus categorical), there are still many exposure-outcome combinations less than 2 studies, the limited data prevented us from conducting a meta-analysis to examine these associations.
Given the strong correlation found between PFAS, PFOA may co-exist and play a combined role with other PFAS, so the effects we detected may be the result of exposure to multiple PFASs and need to be understood by further studies.
Liver function biomarkers are only captured at one time point and some of them, such as ALT, are less speci c and sensitive and can be in uenced by other factors (e.g., exertion, emotion), which may lead to misclassi cation of ndings in cross-sectional as well as transient marker values.
Results based on biomarkers rather than liver biopsies remain a challenge in re ecting liver function as well as speci c disease, and liver biopsies would be the gold standard validation test for patients with clinically signi cant liver disease, and further health studies would bene t from the use of more comprehensive liver function assessments, including measurement of other liver function markers (e.g., albumin, total bilirubin, and lipid metabolism markers), and/or liver imaging techniques.
The representativeness of the study results could still be improved, the type and prevalence of liver function impairment is closely related to the level of public health prevention and control as well as the economic and sociocultural aspects of the area of residence. However, the majority (7/9) of our nine publications currently included were from the United States, a developed country/region, which limits the generalizability of the ndings. Future studies may have to supplement to assess the impact of PFOA emissions on liver enzymes in these countries with weak infrastructure, lower economic levels, and less developed health care insurance mechanisms.
Although potential confounders were controlled for as much as possible in the included studies together with subgroup analyses, the effects of unknown or unmeasured confounders such as sex, age, and alcohol intake could not be measured in detail in our summary effect estimates.
Given the limited amount of evidence observed, future studies should aim to elucidate the associations in age and gender.
In addition, the association between changes in liver enzyme markers and high serum PFOA may be due to "reverse causality" [71], even in longitudinal studies, subtle pharmacokinetic differences can lead to differences in biomarkers of exposure and liver impairment, it is possible that an undetermined reverse causality exists [12], all of which could affect the feasibility of conclusions, and further studies should be conducted with a standardized longitudinal cohort study design.

Conclusions
We conducted the rst meta-analysis and systematic review based on a longitudinal cohort study con rming the effect of PFOA exposure on liver enzyme indicators in humans, which demonstrated that PFOA exposure showed a statistically signi cant association with increased levels of four liver enzyme indicators, accompanied by some heterogeneity. Further clinical or epidemiological studies, especially longitudinal studies, are needed to expand the sample size and further identify the effects of different types and dosages of confounding factors of PFOA exposure on liver enzyme indicators, which are essential to verify the evidence of causality and clear dose-response curves. In addition, future studies should use mediation analysis with full consideration of gender, age and other confounders and reverse causality.

Declarations
Ethics approval and consent to participate Not Applicable.

Consent for publication
The manuscript have been seen and approved for submitting to "Environmental Health" by all authors. Figure 1 Flow diagram of study selection process.