Exposure to per- and Polyfluoroalkyl Substances and Markers of Liver Injury: A Systematic Review and Meta-Analysis

Background: Experimental evidence indicates that exposure to certain pollutants is associated with liver damage. Per- and polyfluoroalkyl substances (PFAS) are persistent synthetic chemicals widely used in industry and consumer products and bioaccumulate in food webs and human tissues, such as the liver. Objective: The objective of this study was to conduct a systematic review of the literature and meta-analysis evaluating PFAS exposure and evidence of liver injury from rodent and epidemiological studies. Methods: PubMed and Embase were searched for all studies from earliest available indexing year through 1 December 2021 using keywords corresponding to PFAS exposure and liver injury. For data synthesis, results were limited to studies in humans and rodents assessing the following indicators of liver injury: serum alanine aminotransferase (ALT), nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, or steatosis. For human studies, at least three observational studies per PFAS were used to conduct a weighted z-score meta-analysis to determine the direction and significance of associations. For rodent studies, data were synthesized to qualitatively summarize the direction and significance of effect. Results: Our search yielded 85 rodent studies and 24 epidemiological studies, primarily of people from the United States. Studies focused primarily on legacy PFAS: perfluorooctanoic acid (PFOA), perfluorooctanesulfonic acid (PFOS), perfluorononanoic acid (PFNA), and perfluorohexanesulfonic acid. Meta-analyses of human studies revealed that higher ALT levels were associated with exposure to PFOA (z-score= 6.20, p<0.001), PFOS (z-score= 3.55, p<0.001), and PFNA (z-score= 2.27, p=0.023). PFOA exposure was also associated with higher aspartate aminotransferase and gamma-glutamyl transferase levels in humans. In rodents, PFAS exposures consistently resulted in higher ALT levels and steatosis. Conclusion: There is consistent evidence for PFAS hepatotoxicity from rodent studies, supported by associations of PFAS and markers of liver function in observational human studies. This review identifies a need for additional research evaluating next-generation PFAS, mixtures, and early life exposures. https://doi.org/10.1289/EHP10092


Description of domains in Office of Health Assessment and Translation (OHAT) Risk of Bias tool
Review articles screened for additional eligible articles Table S1. Assessment of study quality by the OHAT approach (human studies). Table S2. Assessment of study quality by the OHAT approach (animal studies). Table S3. Weighted Z-scores for the cross-sectional associations of PFAS with ALT and GGT in humans with selected exclusions. Table S4. Weighted Z-scores for the cross-sectional associations of PFAS with GGT and AST in humans ≥ 12 years old. Figure S1. Strip plots for PFOA and AST in animal studies. Blue triangles indicate a significant increase in AST and red diamonds indicate a significant decrease in AST relative to control. Circles indicate no significant change in AST relative to control. Figure S2. Strip plots for PFOA and relative liver weight in animal studies. Blue triangles indicate a significant increase in relative liver weight relative to control. Circles indicate no significant change in relative liver weight relative to control. Plots are ordered by species and strain. Figure S3. Strip plots for PFOS and AST in animal studies. Blue triangles indicate a significant increase in AST and red diamonds indicate a significant decrease in AST relative to control. Circles indicate no significant change in AST relative to control. Plots are ordered by species and strain. Figure S4. Strip plots for PFOS and relative liver weight in animal studies. Blue triangles indicate a significant increase in relative liver weight relative to control. Black dots indicate no significant change in relative liver weight relative to control. Plots are ordered by species and strain. Table S5. Results for PFOA and ALT in animal studies. Table S6. Results for PFOS and ALT in rodent studies. Table S7. Results for PFOA and AST in animal studies. Table S8. Results for PFOA and relative liver weight in animal studies. Table S9. Results for PFOS and AST in animal studies. Table S10. Results for PFOS and relative liver weight in animal studies.

Description of domains in Office of Health Assessment and Translation (OHAT) Risk of Bias tool
Definitely low (++) Probably Low (+) Probably High (NR, -) Definitely High (--) Was administered dose or exposure level adequately randomized? (EA) Authors explicitly describe the randomization procedures and there is a concurrent control group.
Authors state that randomization occurred, but do not describe the randomization procedure. Inequalities in treatment group sizes are not explained.
There is no description of the procedures used to allocated animals to treatment groups (NR), or there is indirect evidence that the procedures were not random.
Animals were allocated to treatment groups using a non-random method, or there was no appropriate control group.
Was allocation to study groups adequately concealed? (EA) Authors state that allocation to study groups was concealed from study personnel at the time of randomization or treatment assignment, and that blinding was unlikely to be broken.
Though not stated explicitly it is implied that allocation to study groups was concealed from study personnel, or lack of concealment was unlikely to introduce bias.
There is no description of concealment procedures (NR), or lack of concealment may introduce bias.
Allocation to study groups was not concealed from study personnel, and this is likely to introduce bias.
Did selection of study participants result in appropriate comparison groups? (Co, CrSe) Exposed and nonexposed participants were similar (recruited from the same base population, similar demographics, similar response/missing rates, etc).
Exposed and nonexposed participants are likely to be similar, but no direct evidence is provided.
Exposed and nonexposed participants are not likely to be similar, or there is insufficient information (NR).
Exposed and nonexposed participants were not similar (did not arise from the same base population, different response rates, different health status and risks, etc).
Did the study design or analysis account for important confounding and modifying variables? (Co, CrSe) Authors provided comprehensive evidence-based justification for all confounding and modifying variables, and accounted for all risk factors (obesity, alcohol and smoking, type 2 diabetes, etc) The analyses adjusted for, at minimum, BMI and alcohol use, or provided evidencebased justification for lack of adjustment.
The analyses failed to adjust for either BMI or alcohol use, without scientific justification (ie. 'variable was not available in the dataset').
The analysis did not control for any confounding or modifying variables.
Were experimental conditions identical across study groups? (EA) Experimental conditions for treatment and control groups are described, and are identical.
Animal care conditions are not described in detail, but are stated to be identical across groups.
Animal care conditions are not described (NR) and are likely to differ between study groups.
Animal care conditions were not identical across study groups (ie., different vehicles).
Were the research personnel blinded to the study group during the study? (EA) Authors state that study personnel were blinded to the study group for the entire study, and that blinding was unlikely to be broken.
Though not stated explicitly, it is implied that study personnel were blinded to study groups, or lack of blinding was unlikely to introduce bias.
There is no description of blinding (NR), or it is implied that study personnel were not blinded.
Study personnel were not blinded, and this is likely to introduce bias.
Were outcome data complete without attrition or exclusion from analysis? (Co, CrSe, EA) There was no or insignificant loss of subjects and outcome data were complete. Any loss of animals was not related to the study conditions, or was treated as an outcome related to PFAS exposure.
There was attrition, but it was unlikely to introduce bias. Authors may state that there was no attrition/mortality without providing the initial sample size.
Sample sizes for outcomes vary without justification or explanation.
No sample sizes/survival data are provided (NR).
There was significant loss of subjects from loss to follow up (humans) or death (animals), or large numbers of subjects were excluded from analysis.
Can we be confident in the exposure characterization? (Co, CrSe, EA) EA: Purity of the experimental compound is described and is at or above 95%, confirmed by independent testing.
The experimental compound was consistently administered throughout the study. Co, CrSe: Exposure was evaluated using the same method for all subjects. Exposure assessment method is the 'gold-standard' or directly measures exposure (in body fluids, environment). EA: Purity is at or above 95%, with no independent testing, or is below 95% with independent testing.
The experimental compound was consistently administered throughout the study. Co, CrSe: Exposure was assessed using well-established methods with high validity.
EA: Purity is not described (NR), or is below 95%. Co, CrSe: Exposure assessment method is not described (NR), or exposure was assessed using indirect measures (ie. questionnaires) that are not validated or well-established.
EA: the experimental compound was highly contaminated and/or administered inconsistently. Co, CrSe: Exposure was assessed using methods known to have poor validity.
Can we be confident in the outcome assessment? (Co, CrSe, EA) Outcomes were assessed using the 'gold-standard' method, at the same time for all study groups, and Outcomes were assessed using an acceptable but not 'gold-standard' method.
Outcomes were Outcome assessment and/or blinding not adequately described (NR).
Outcomes were assessed using an insensitive instrument, after different lengths of time, and were not outcome assessors were blinded to study group.
assessed at the same time for all study groups and assessors were blinded, or deviations from these criteria were not expected to introduce bias.
Methods used in outcome assessment were insensitive, or outcomes were not assessed at the same time for all study groups.
blinded, and these were expected to introduce significant bias.
Were all measured outcomes reported? (Co, CrSe, EA) All outcomes described in the methods are completely reported in the results.
All outcomes described in the methods are reported in the results. EA: Histopathological results are lacking some detail, but this is unlikely to be due to selective reporting.
Not enough information is provided to evaluate the potential for selective reporting (NR). Outcomes described in the methods are not presented in the results. EA: Specific outcomes (ie., histopathological findings) not described in methods but are reported in the results.
Outcomes were reported for some study groups but not others.
Outcomes were selectively reported for different study groups.
Were there other potential threats to internal validity?
Note concerns about the choice of statistical methods, adherence to study protocol, study design, or undue influence of study sponsors.  Legend: definitely low risk of bias (++); probably low risk of bias (+); probably high risk of bias (-); definitely high risk of bias (--); not reported (NR) Two elements did not apply to animal studies and were excluded from the table.     90 Mice  Table S8.  87 Rats 87 Rats 47 Rats 47 Rats 47 Rats 47 Rats 46 Rats 46 Rats 46 Rats 34 Rats 34 Rats 34 Rats 34 Rats 34 Rats 34 Rats 34 Rats 34 Rats  Table S9.   47 Rats 47 Rats 47 Rats 47 Rats 46 Rats 46 Rats 46 Rats 34 Rats 34 Rats 28 Rats 28 Rats 28 Rats 28 Rats 36 Rats 36 Rats 36 Rats 36 Rats 36 Rats 36 Rats 36 Rats 36 Rats . Strip plots for PFOS and relative liver weight in animal studies. Blue triangles indicate a significant increase in relative liver weight relative to control. Black dots indicate no significant change in relative liver weight relative to control. Plots are ordered by species and strain. Abbreviations: End of treatment (EOT); marginal methionine/choline-deficient diet (mMCD); choline supplementation (CS); naringin (Nar); Sprague Dawley (SD); grape seed proanthocyanidin extract (GSPE); high fat diet (HFD); low fat diet (LFD); initial high fat diet followed by standard diet (HFD/STD). *Atmospheric exposure occurred for 5 hours/day, 5 days/week. An accessible version of this figure is available in Table S10.