Polychlorinated biphenyls (PCBs) exposure and cognitive function in adults: A systematic review of epidemiologic evidence and risk of bias assessment

Abstract

Background:

Polychlorinated biphenyls (PCBs) are persistent organic pollutants known for their environmental longevity and neurotoxicity. Although developmental exposure has been extensively studied, the cognitive impacts of PCB exposure in adult and aging populations remain less understood.

Objectives:

This systematic review evaluates the epidemiologic evidence linking PCBs exposure to cognitive function, with an emphasis on neurocognitive decline and dementia risk in adults.

Methods:

Following PRISMA guidelines, we searched PubMed and Web of Science for human epidemiological studies published in English up to November 2024. Inclusion criteria focused on empirical studies investigating associations between PCB exposure and cognitive outcomes in elderly population. Data were synthesized narratively, and a domain-based risk of bias assessment was conducted using the Navigation Guide methodology.

Results:

Twenty-four studies were included in the final analysis, spanning diverse populations and exposure contexts (occupational, environmental and dietary). Six studies examined the association between PCB exposure and dementia or Alzheimer’s disease, with overall mixed findings. Some studies reported associations in specific subgroups, particularly among individuals with higher exposure levels or among women, while others reported null relationships. For cognitive outcomes, memory, processing speed and executive functions were the frequently assessed domain; however, findings across studies were inconsistent and did not demonstrate a clear pattern of association. A smaller number of studies assessed cognitive decline, with similarly heterogeneous results. Risk of bias was generally low for exposure and outcome assessment but higher for confounding and other validity domains. Notably, inconsistent cognitive testing methods and limited longitudinal data constrained causal inference.

Conclusion:

Current epidemiologic evidence is insufficient to support a consistent association between PCB exposure and dementia or cognitive performance in aging populations. Further well-designed longitudinal studies using standardized cognitive assessments are needed to clarify potential relationships.

1. Introduction

Polychlorinated biphenyls (PCBs), along with 11 other organic compounds, were originally classified as persistent organic pollutants (POPs) under the Stockholm Convention due to their significant risks to both ecosystems and human health. PCBs have been classified as “carcinogenic to humans” (group 1) by the International Agency for Research on Cancer (IARC). PCBs are synthetic organic chemicals that were widely used throughout the 20th century in industrial applications, particularly in capacitor (~50%) and transformer (~25%) due to their stability and resistance to heat (Erickson and Kaley, 2011). Although their production was halted in the late 1970s due to mounting evidence of environmental persistence and adverse health impacts, PCBs continue to pose significant health risks globally because of their persistence in the environment and capacity to bioaccumulate in the food chain (ATSDR, 2002).

Despite production bans, PCBs remain ubiquitous in the environment due to their persistence (Melymuk et al., 2022), improper disposal (Pavuk et al., 2014), and legacy contamination (Fitzgerald et al., 2011). They are highly resistant to degradation, bioaccumulate in food webs (Saktrakulkla et al., 2020), and are detectable in soil (Vane et al., 2014), air (Othman et al., 2022), water (Sandy et al., 2012), and animal tissues (Holma-Suutari et al., 2016). Human exposure occurs primarily through dietary intake of contaminated fish, meat, and dairy products, as well as inhalation or dermal contact in contaminated areas (Ampleman et al., 2015; Saktrakulkla et al., 2020; Zhu et al., 2022). For instance, in a study analyzing 26 food items, the highest PCB concentrations were found in wild-caught salmon (380 pg/g WW), canned tuna (330 pg/g WW), and beef steak (290 pg/g WW) (Saktrakulkla et al., 2020). Airborne PCB concentrations in U.S. schools ranged from 0.5 to 194 ng/m³ indoors and from 0.03 to 3 ng/m³ outdoors, with indoor levels being 1–2 orders of magnitude higher, largely attributed to legacy sources such as Aroclors in building materials and modern pigments (Marek et al., 2017). PCBs’ lipophilic nature facilitates their storage in adipose tissues, contributing to chronic exposure even decades after initial contact.

As endocrine disruptors (Buha Djordjevic et al., 2020), PCBs interfere with hormonal regulation, PCBs pose significant risk factors for metabolic disorders such as type 2 diabetes (Gao et al., 2023) and obesity (Gao et al., 2024) through mechanisms like pancreatic β-cell dysfunction and insulin resistance (Hoyeck et al., 2022). According to a recent meta-analysis result, PCB exposure is significantly associated with an increased risk of diabetes mellitus (OR = 3.593; 95% CI 2.566–5.031), while weaker associations were observed for obesity (OR = 1.875; 95% CI 0.883–3.979), hypertension (OR = 1.335; 95% CI 0.902–1.976), and hypertriglyceridemia (OR = 1.611; 95% CI 0.981–2.643) (Mohd Efendy Goon et al., 2024). In addition, PCBs, widely recognized as persistent organic pollutants, are extensively studied for their neurotoxic effects on development. While perinatal exposure to PCBs is suggested to be associated with lower cognitive development and poor attention in middle childhood (Balalian et al., 2024) and autism spectrum disorders (Xu et al., 2023), PCBs also disrupt key biological pathways, including oxidative stress and calcium signaling, which are relevant to neurodegenerative diseases such as dementia (Pessah et al., 2019).

In this systematic review, we synthesize the epidemiologic evidence on the relationship between PCB exposure and cognitive outcomes, with a particular focus on adult and aging populations. While previous reviews have emphasized developmental neurotoxicity, the potential role of PCBs in age-related cognitive decline and dementia risk remains less well-characterized. To address this gap, we systematically identified, assessed, and evaluated studies reporting on PCB exposure in relation to cognitive performance, neuropsychological testing, or clinical cognitive outcomes. We also conducted a structured risk of bias assessment for each included study, following the Navigation Guide framework, to ensure transparency and methodological rigor. This review aims to clarify the strength and consistency of evidence linking PCB exposure to cognitive impairment, and to inform future research and public health strategies targeting environmental risk factors for neurodegenerative diseases.

2. Methods

This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, and was prospectively registered in the PROSPERO database (Registration number: CRD420251003803). This systematic review is limited to epidemiological studies of human participants that were published with full text in English. This review was guided by the PECO (Population, Exposure, Comparator, Outcome) framework:

• Population: Adult human populations, including older adults and aging cohorts.

• Exposure: Measured exposure to polychlorinated biphenyls (PCBs), via occupational, environmental, dietary, or general population sources.

• Comparator: Lower or no PCB exposure within the same or similar populations.

• Outcomes: Neurocognitive health outcomes, including clinically diagnosed dementia or Alzheimer’s disease, and performance-based cognitive assessments across domains such as memory, executive function, processing speed, and sensorimotor function.

This PECO structure informed our eligibility criteria, search strategy, and data extraction approach. We conducted comprehensive searches across two major scientific databases, Web of Science and PubMed, to capture literature examining the association between environmental pollutants, particularly PCBs, and neurocognitive disorders such as Alzheimer’s disease and dementia. The search was limited to peer-reviewed literature indexed in PubMed and Web of Science; gray literature sources (e.g., preprints, dissertations, and conference abstracts) were not systematically searched.

In the absence of standardized methods to synthesize the findings across the studies with insufficient data to conduct meta-analysis, we narratively presented the findings and adapted vote counting based on direction of effect as suggested by Cochrane handbook for Systematic Reviews of Interventions to summarize the findings for each domain within age groups (Cumpston et al., 2023).

2.1. Search Strategy

The search strategy was designed to identify peer-reviewed studies reporting empirical data on the effects of persistent organic pollutants (POPs), with a specific focus on PCBs, on cognitive function or neurocognitive disorders in human populations. The search terms included a combination of Medical Subject Headings (MeSH) and keywords, as follows:

Neurocognitive Terms: (“Alzheimer Disease”[MeSH Terms] OR Alzheimer*[Title/Abstract] OR “Dementia”[MeSH Terms] OR Dementia[Title/Abstract] OR “Cognitive Skills”[Title/Abstract] OR “Cognitive Function”[Title/Abstract] OR “Cognitive Impairment”[Title/Abstract] OR “Neurocognitive Disorders”[MeSH Terms])

AND Pollutant Terms: (“Environmental Pollutants”[MeSH Terms] OR “Persistent Organic Pollutants”[Title/Abstract] OR POPs[Title/Abstract] OR “Polychlorinated Biphenyls”[Title/Abstract] OR PCBs[Title/Abstract] OR “Polybrominated Diphenyl Ethers”[Title/Abstract] OR PBDEs[Title/Abstract] OR Pesticides[Title/Abstract] OR “Perfluoroalkyl Substances”[Title/Abstract] OR PFAS[Title/Abstract])

NOT Exclusion Criteria: (“Air Pollution”[MeSH Terms] OR “Air Pollutants”[MeSH Terms] OR “Air Quality”[Title/Abstract] OR Music[Title/Abstract] OR Historical[Title/Abstract] OR Literary[Title/Abstract])

NOT Publication Type Filters: (Review[Publication Type] OR “Meta-Analysis”[Publication Type] OR Editorial[Publication Type])

NOT Non-Human and Pediatric Terms: (Mice[MeSH Terms] OR Mouse[Title/Abstract] OR Zebrafish[MeSH Terms] OR “Danio rerio”[Title/Abstract] OR Rats[MeSH Terms] OR Rat[Title/Abstract])NOT Pediatric and Pregnancy Exclusions: (Children[MeSH Terms] OR Child[Title/Abstract] OR Pediatric[Title/Abstract] OR Pregnancy[MeSH Terms] OR Pregnant[Title/Abstract])

For Web of Science, the search was adapted using the Topic Search (TS) field, and Boolean operators were modified to align with the database syntax requirements. Searches were restricted to studies published in English up to November 2024.

2.2. Eligibility Criteria

Studies were eligible for inclusion if they: Investigated the effects of PCBs or other POPs on cognitive function or neurocognitive outcomes in adult human populations. Were empirical studies with primary data. Focused specifically on neurocognitive disorders, Alzheimer’s disease, or related cognitive functions in the context of PCB exposure.

Exclusion criteria were applied to exclude studies that: Examined pollutants unrelated to PCBs, such as general air pollution. Focused on pediatric or animal populations. Were review articles, meta-analyses, editorials, or case studies lacking primary empirical data.

2.3. Study Selection

A total of 755 records were identified across Web of Science and PubMed. After duplicate removal, 629 unique records remained (Figure 1). Two independent reviewers screened the titles and abstracts for relevance, excluding 552 records as they did not meet the inclusion criteria. Seventy-four full-text articles were assessed for eligibility. Following a detailed review, 53 studies were excluded after we decided to narrow our focus on the PCBs instead of other chemicals mentioned above. The final selection included 21 studies specifically examining the relationship between PCB exposure and neurocognitive health outcomes. To ensure the review reflects the most current evidence, we conducted an updated targeted search of the literature through March 2026 to identify recently published studies meeting the inclusion criteria. Three additional recently published studies were screened and included in the review.

Figure 1.

Flow chart showing papers included and excluded from the systematic review.

2.4. Data Extraction and Synthesis

Data extraction was performed using a standardized form to ensure consistency across studies. Extracted data included study characteristics (authors, year, study country), participant details (population characteristics, age range), exposure type and measurement, outcome measures (cognitive functions assessed), and key findings. Discrepancies in data extraction were resolved through discussion between the reviewers to reach consensus. Also, we extracted study design (cross-sectional, case–control, or longitudinal cohort) and explicitly considered temporality—whether PCB exposure preceded cognitive assessment—as a key element for interpreting causal inference.

The final selection of studies was synthesized narratively, with a focus on identifying common patterns and variations in findings related to PCB exposure and cognitive outcomes. This synthesis aimed to provide a comprehensive understanding of PCB exposure’s impact on neurocognitive health across various populations and exposure settings.

2.5. Risk and Bias assessment

We assessed risk of bias (ROB) across included studies using a domain-based approach aligned with the Navigation Guide methodology, as employed in the systematic review by (Lam et al., 2017) on PBDEs and neurodevelopmental outcomes. Each study was independently evaluated across domains including selection bias, exposure assessment, outcome assessment, confounding, and selective reporting. Ratings were assigned using standard categories: “low,” “probably low,” “probably high,” or “high” risk of bias. Consistent with (Lam et al., 2017) we did not use an overall numerical score or exclude studies based solely on a single high-risk domain. Our approach was further validated by findings from (Eick et al., 2022), which demonstrated that domain-based tools like the Office of Health Assessment and Translation (OHAT) and Navigation Guide preserve the evidentiary value of observational studies and avoid the over-exclusion of data that can occur with tools assigning overall study quality scores. This structured, transparent framework enhances the rigor and reproducibility of our risk of bias judgments in environmental epidemiology.

3. Results

3.1. Study characteristics

The 24 studies on PCB cognitive or AD assessments are distributed across several countries: the United States leads with 14 studies (58%), followed by Canada with three studies, Taiwan with two studies, Germany with two studies, Italy with one study, France with one study and China with one study (Table 1). In total, six studies used the diseases as the outcome while 18 studies assessed the cognitive functions (Figure 2). Exposure to PCBs occurs through several pathways, including occupational contact in industries like capacitor manufacturing and recycling (Cromberg et al., 2024; Fimm et al., 2017; Ruder et al., 2014; Seegal et al., 2013; Steenland et al., 2006), environmental contamination near polluted sites such as rivers and industrial areas (Fitzgerald et al., 2008; Haase et al., 2009; Fitzgerald et al., 2012; Raffetti et al., 2020; Tanner et al., 2020; Pan et al., 2022; Sasaki et al., 2023b, 2023a), dietary intake, particularly from consuming contaminated fish (Schantz et al., 2001) or accidental ingestion of PCB-contaminated cooking oil (Lin et al., 2010, 2008) and general environmental exposure (Bouchard et al., 2014; Lefèvre-Arbogast et al., 2025; Medehouenou et al., 2019, 2014; Parada et al., 2024; Przybyla et al., 2017; Zuo et al., 2024).

Table 1.

Characteristics of the included studies.

Authors and Year	Study Country	Study Design	Number of Participants	Sex	Age	Participants Characteristics	Exposure Event
Kilburn et al., 1989	USA	Case control	28 (14 exposure and 14 control)	Male	Exposed group: 33.6 ± 8.8 years Control group: 34.9 ± 8.5 years	Firemen exposed to PCBs during a transformer fire	Transformer fire leading to PCB exposure
Schantz et al., 2001	USA	Cohort study	179 (101 fish eater 78 non-fish eater)	42.5% male and 57.45% Female	49–86 median 64.3	Older adults who consumed Great Lakes fish	Consumption of PCB-contaminated fish
Steeland et al., 2006	USA	Retrospective cohort study (three cohorts)	16,906	50% male and 50% female	Average year of birth: 1934 Average first year of exposure: 1960	Occupational PCB-exposed workers, neurodegenerative mortality risk	Occupational PCB exposure
Fitzgerald et al., 2008	USA	Cross-Sectional Study	253	127 50% male and 126 50% female	55–74 years (mean: 63.9 years)	Older adults in Hudson River communities	Environmental exposure from Hudson River contamination
Lin et al., 2008	Taiwan	Retrospective cohort study	313 (162 exposed and 151 reference individuals)	Yucheng 81 (50% male):81 (50% female), reference 79 male (52%):72 female (48%)	60–91 years (mean age: 69.5 years)	Yucheng cohort members exposed to contaminated oil	Ingestion of PCB/PCDF contaminated rice oil
Haase et al., 2009	Canada	Cross-sectional Study design	283	67% female (189 females) 33% male (94 males)	17–79 mean 38	Native American population, adults	Environmental PCB exposure in St. Lawrence region
Lin et al., 2010	Taiwan	Retrospective cohort study	316 (exposed 165, control 151)	Exposed: 81 males (49.1%); 84 females (50.9%) Reference: 79 (52.3%); 72 (47.7%)	60–91 mean 69.5	Yucheng cohort, older adults with PCB exposure from oil contamination	Ingestion of PCB/PCDF contaminated rice oil
Fitzgerald et al., 2012	USA	Cross-sectional	144	67 male 46.5%, 77 female 53.5%	55–74 mean 63.5	Older adults in Hudson River communities	Environmental exposure from Hudson River contamination.
Seegal et al., 2013	USA	Cross-sectional	241	Male: 123 (53.5%), Female: 107 (46.5%)	50–89 years (mean age: 65 years)	Former capacitor workers, PCB and lead exposure	Occupational exposure
Ruder et al., 2014	USA	Retrospective cohort study	24865	Male: 11,788 (47.4%) Female: 13,077 (52.6%)	Median age at first employment was 22.4 years (interquartile range: 18.7–29.7 years).	Workers in U.S. capacitor manufacturing plants	Occupational exposure
Bouchard et al., 2014	USA	Cross-sectional	708	Male: 334 (47%) Female: 374 (53%)	60–84	Older adults in U.S., varying by age and sex from NHANES	Environmental PCB exposure
Medehouenou et al., 2014	Canada	Cross-sectional	2,023	Dementia group: 66.6% female, 33.4% male. No dementia group: 58.5% female, 41.5% male.	65+ years (mean age for dementia group: 85.8 years; no dementia group: 82.7 years)	Older adults from the Canadian Study of Health and Aging	General environmental PCB exposure
Przybyla et al., 2017	USA	Cross-sectional	498	Male: 43.45% (231 participants) Female: 56.55% (267 participants)	60–84	U.S. adults over 60, multiple neurotoxic exposures from NHANES	General environmental PCB exposure
Fimm et al., 2017	Germany	cross-sectional	237	Male: 87% (207 participants) Female: 13% (30 participants)	19–83 years (mean age: 44 years)	Former PCB-exposed workers from a recycling company	Occupational PCB exposure in recycling industry
Medehouenou et al., 2019	Canada	prospective cohort	669	Male: 40.7% Female: 59.3%	65+ years (mean age: 80.9 years, range not specified)	Older Canadian adults	General environmental PCB exposure
Raffetti et al., 2020	Italy	Prospective cohort	699	Male: 48.1% (336 participants) Female: 51.9% (363 participants)	Mean age of 63.2 years	Adults in highly polluted Northern Italy	Environmental exposure in a heavily polluted region
Tanner et al., 2020	USA	Longitudinal cohort	253 at baseline, with 116 completing follow-up assessments after 14 years	Male: 53.5% (baseline and follow-up).Female: 46.5%.	55–74 years at baseline; mean age at follow-up: 76.3 years	Older adults in New York State, community-dwelling	Environmental PCB exposure from river contamination
Pan et al., 2022	China	Cross-sectional	266	Male: 46.2% (123 participants) Female: 53.8% (143 participants)	61–90 years (median: 67 years, interquartile range: 63–74)	Older Chinese women aged 61–80	Environmental exposure in older women
Sasaki et al., 2023	USA	Cross-sectional	301	Male: 33.9% 102 Female: 66.1% 199 participants percentage	17–79	Mohawk adults, highly exposed via local environment	Chronic environmental exposure via St. Lawrence River
Sasaki et al., 2023b	USA	Cross-sectional	Number of Participants: 752 Mohawk Cohort: 23 NHANES Cohort: 729	Mohawk Cohort: Male: 30.4% (7 participants), Female: 69.6% (16 participants). NHANES Cohort: Not specified but includes diverse ethnic groups.	Mohawk Cohort: 60–79 years NHANES Cohort: 60–85 years	Akwesasne Mohawk cohort exposed to PCBs, pesticides	Environmental Exposure to PCB mixtures and pesticides in Akwesasne
Cromnerg et al., 2024	Germany	longitudinal cohort (HELPc B cohort)	Number of Participants: 116 Workers of the transformer recycling company: 59 Family members: 8 Workers of surrounding companies: 49	Male: 60.3% (70 participants) Female: 39.7% (46 participants)	19–69	Occupationally exposed workers, focus on verbal fluency	Occupational and environmental sources
Zuo et al., 2024	USA	Cross-sectional study (NHAN ES 2011–2014)	727	348 males (47.9%), 379 females (52.1%)	Mean 68.0 ± 6.7 years (⩾60 years)	Community-dwelling older adults participating in NHANES; individuals with available cognitive testing and biomarker measurements	Background environmental exposure
Parada et al., 2024	USA	Prospective population-based cohort	1837	Women: 1016 (55.3%); Men: 821 (44.7%)	Mean 55.8 years	Hispanic/Latino adults from four U.S. cities (Chicago, Bronx, Miami, San Diego)	Background environmental exposure
Lefèvre-Arbogast et al., 2025	France	Prospective population-based cohort	515	Women: 299 (58.1%); Men: 216 (41.9%)	Mean 72.5 years at baseline	Community-dwelling older adults free of dementia at baseline;	Background environmental exposure

Figure 2.

Risk of Bias and Study Quality Assessment for Included Studies in the PCB and Cognition Review.

Specifically, in the United States, several studies looked at the occupational exposure such as the firemen exposed to PCBs during a transformer fire (Kilburn et al., 1989), PCB-exposed workers from three cohorts (Steenland et al., 2006) and former capacitor workers (Ruder et al., 2014; Seegal et al., 2013). Also, in the USA, a few studies looked at the environmental exposure of highly polluted areas such as the Fort Edward and Hudson River area in New York where General Electric plants discharged nearly 1 million pounds of PCBs in to the Hudson River (Fitzgerald et al., 2012, 2008; Tanner et al., 2020), and Mohawk communities living along the St. Lawrence River that were highly exposed to PCBs because Akwesasne Reserve is downstream from a U.S. federal Superfund site and two NYS Superfund sites where PCBs were used as hydraulic fluids (Haase et al., 2009; Sasaki et al., 2023a, 2023b);Lastly, the NHANES (National Health and Nutrition Examination Survey) cohort has been essential in examining PCB exposure in the general population, with studies by (Bouchard et al., 2014) exploring variations in exposure by age and sex, and (Przybyla et al., 2017) assessing multiple neurotoxic chemicals, and more recent analyses incorporating mixture-based approaches to evaluate PCB exposure in relation to cognitive function (Zuo et al., 2024). A recent population-based cohort study in the United States (HCHS/SOL) examined associations between PCB exposure and 7-year cognitive change among Hispanic/Latino adults (Parada et al., 2024).

Two studies in Taiwan looked at the same participants who were exposed to PCBs by consumed toxic rice that contaminated by PCB/PCDF in 1979 (Lin et al., 2010, 2008). There are two Canadian studies (Medehouenou et al., 2019, 2014) looked at general population of men and women aged 65+ from the Canadian Study of Health and Aging. Also, another study from Canada looked at the Native American adults in the St. Lawrence region of upstate New York (Haase et al., 2009). German studies looked at the occupation PCB-exposed workers of a transformer and capacitor recycling facilities from the HELPcB cohort (Cromberg et al., 2024; Fimm et al., 2017). The Italian study investigated the PCB exposure at the highly polluted area in Brescia, a highly industrialized town in Northern Italy, produced PCBs from 1938 to 1984 where PCB contamination was found in soil, water and food (Raffetti et al., 2020). Additionally, a French population-based aging cohort (Three-City Study) evaluated long-term associations between plasma POPs, including PCBs, and cognitive decline, dementia risk, and brain atrophy over up to 17 years of follow-up (Lefèvre-Arbogast et al., 2025).

Study designs varied considerably and have important implications for interpreting the evidence. The majority of included studies were cross-sectional, particularly those based on general population datasets (e.g., NHANES) and environmentally exposed communities. These studies provide insight into associations between PCB exposure and cognitive performance but do not establish temporality. A smaller number of studies employed longitudinal or prospective cohort designs, including both community-based and occupational cohorts, while several retrospective cohort studies evaluated neurodegenerative disease mortality rather than cognitive trajectories.

Among longitudinal studies, follow-up duration varied substantially. Short-term follow-up was observed in occupational cohorts, such as the HELPcB study, which reassessed neuropsychological outcomes approximately one year after baseline (Cromberg et al., 2024). Intermediate follow-up periods were reported in population-based cohorts, including approximately 7 years in the HCHS/SOL (Parada et al., 2024) and approximately 8–9 years in the Italian Brescia cohort (Raffetti et al., 2020). Longer follow-up durations were observed in aging cohorts, including approximately 14 years in the Upper Hudson River cohort (Tanner et al., 2020) and up to 17 years in the French Three-City study (Lefèvre-Arbogast et al., 2025).

For the exposure assessment, PCB exposure levels varied substantially across studies, ranging from background environmental levels in general population cohorts to markedly elevated concentrations in occupational and highly contaminated settings. Studies also differed in reporting metrics (e.g., lipid-adjusted vs. wet-weight concentrations, mean vs. median vs. categorical exposure levels), limiting direct quantitative comparison across cohorts (see Appendix Table S1).

3.1.1. NHANES Cohort (U.S. General Population)

Two independent analyses of NHANES 1999–2002 examined serum PCB concentrations in relation to cognition measured by the Wechsler Adult Intelligence Scale (Third Edition)-WAIS-III Digit Symbol Coding (DSC) test (Bouchard et al., 2014; Przybyla et al., 2017).

In a regression analysis of 708 participants aged 60–84 years, lipid-standardized serum concentrations of 12 PCB congeners were used (Bouchard et al., 2014). A statistically significant interaction between age and dioxin-like PCB (five congeners) concentrations was observed (p=0.04). Among adults aged 70–84 years, higher dioxin-like PCB concentrations were associated with lower DSC scores (−2.7 points per 100 ng/g; 95% CI: −5.1, −0.2), whereas among those aged 60–69 years the association was positive but not statistically significant (2.9 points; 95% CI: −1.8, 7.7) (Bouchard et al., 2014). Sex-stratified analyses further indicated stronger inverse associations among women aged 70–84 years (−7.94 points comparing highest vs lowest exposure tertile; 95% CI: −15.12, −0.76), whereas corresponding estimates in men were smaller and not statistically significant (Bouchard et al., 2014).

Using the same NHANES cycles, a multipollutant path-analysis approach (final N=498) modeled individual congeners simultaneously (Przybyla et al., 2017). After adjustment for co-exposures and covariates, PCB 146 (a non-dioxin-like congener) was significantly inversely associated with DSC scores (β = −0.16; 95% CI: −0.29, −0.02; p=0.02), while PCB 153 was positively associated with DSC performance (β = 0.20; 95% CI: 0.05, 0.35; p=0.01) (Przybyla et al., 2017). Other congeners (e.g., PCB 74, PCB 118) were not significantly associated with cognitive performance in fully adjusted models. Stratified analyses suggested some variation in magnitude by sex and age, but formal interaction tests were not statistically significant. The authors concluded that non-dioxin-like PCB 146 was independently associated with lower cognitive functioning, while noting the unexpected positive association observed for PCB 153.

Later NHANES cycles of 2011–2014 incorporated broader cognitive assessment, including immediate recall (IRT), delayed recall (DRT), animal fluency (AFT), and the Digit Symbol Substitution Test (DSST) with standardized domain-specific and global z-scores (Zuo et al., 2024). Zuo et al., (2024) conducted a cross-sectional analysis of older U.S. adults (≥60 years) to examine associations between 24 serum PCB congeners and cognitive performance, evaluated both individually and within broader endocrine-disrupting chemical (EDC) mixtures. In covariate-adjusted generalized linear models with false discovery rate correction, PCB-199 was inversely associated with IRT and AFT scores, and PCB-206 was inversely associated with AFT performance(Zuo et al., 2024). Mixture analyses using least absolute shrinkage and selection operator (LASSO)-informed weighted quantile sum (WQS) regression showed that the overall EDC mixture index was negatively associated with AFT and global cognitive z-scores but positively associated with DRT. Bayesian kernel machine regression (BKMR) further indicated nonlinear and percentile-dependent associations, with mixture effects varying across cognitive domains (Zuo et al., 2024). In mixture contribution analyses, certain PCB congeners (including PCB-146 and PCB-114) contributed to observed associations, particularly for DSST performance; however, not all individual congeners demonstrated statistically significant independent effects. Overall, the study highlights domain-specific, congener-specific, and model-dependent associations rather than uniformly adverse effects of PCB exposure on cognition. In weighted quantile sum and Bayesian kernel machine regression models, PCBs contributed to mixture effects associated with poorer animal fluency and composite cognitive z-scores at higher mixture percentiles, whereas associations with immediate and delayed recall were less consistent (Zuo et al., 2024). Not all PCB components within mixtures demonstrated statistically significant independent effects.

Overall, findings from NHANES 1999–2002 indicate congener-specific and age-dependent heterogeneity, with evidence of adverse associations for certain dioxin-like and non-dioxin-like congeners, null associations for others, and even positive associations in some models. In NHANES 2011–2014, associations between PCB exposure and cognition were domain-specific, mixture-dependent, and not uniformly adverse across all tests or congeners.

3.1.2. Firefighter Transformer Fire Exposure Study

A case–control investigation evaluated neurobehavioral outcomes among firefighters exposed to PCBs and combustion by-products during a transformer explosion and fire (Kilburn et al., 1989). Fourteen firefighters who were exposed to smoke and fumes from a transformer containing PCBs during fire suppression activities were compared with 14 firefighters from the same department who were not involved in the incident. Neurobehavioral testing was conducted approximately six months after exposure.

Participants completed a comprehensive neurobehavioral test battery assessing memory, cognitive function, perceptual–motor speed, and mood, including tests of verbal and visual memory (Wechsler Memory Scale-WMS), digit span, block design, Culture Fair intelligence scale, trail-making tests (TMT), embedded figures, choice reaction time, and affective status using the Profile of Mood States (Kilburn et al., 1989). Initial testing showed that exposed firefighters performed significantly worse than unexposed controls on several cognitive measures, including verbal memory recall (VMR), visual memory, digit span backward, block design performance, TMT, Culture Fair scores, and choice reaction time, indicating impairments in memory, cognitive processing, and perceptual–motor speed (Kilburn et al., 1989).

Exposed firefighters then underwent a 2–3 week detoxification program involving supervised diet, exercise, and sauna therapy, and neurobehavioral testing was repeated six weeks later. Follow-up testing showed significant improvements in several cognitive outcomes, including verbal and visual memory, block design performance, embedded figures recognition, and TMT Part B scores. However, some measures, such as reaction time and balance, did not improve, and mood symptoms including fatigue and depression remained elevated (Kilburn et al., 1989).

Overall, this study suggests that acute occupational exposure to PCBs and related combustion products during a transformer fire may be associated with transient neurobehavioral impairments affecting memory, cognitive processing, and perceptual–motor function, with partial improvement observed after a detoxification intervention.

3.1.3. Great Lakes Fish Consumer Cohort (Michigan)

The Great Lakes fish consumer cohort includes older adults from Michigan with long-term exposure to PCBs through consumption of sport-caught Lake Michigan fish (Schantz et al., 2001). Participants were recruited from an existing cohort established in 1980–1982 that included individuals who consumed large quantities of Great Lakes fish (>24 lb/year) and a comparison group consuming little or none (<6 lb/year). The neuropsychological study included 179 community-dwelling adults aged 49–86 years (median age 64.3 years), of whom 101 were frequent fish consumers and 78 served as comparison participants. Blood samples were analyzed for PCBs and other contaminants including DDE, mercury, and lead.

Participants completed a comprehensive neuropsychological battery assessing memory and learning, executive function, and visuospatial function, including the WMS, California Verbal Learning Test (CVLT), Wisconsin Card Sorting Test (WCST), TMT, Stroop Color–Word Test, Hooper Visual Organization Test and DSST (Schantz et al., 2001). Multiple regression analyses were used to evaluate associations between serum PCB concentrations and cognitive outcomes while controlling for demographic, health, and lifestyle covariates.

Higher PCB exposure was associated with poorer performance on several measures of memory and learning. Specifically, PCB concentrations were inversely associated with verbal delayed recall on the WMS (p = 0.001), as well as with CVLT measures including CVLT semantic clustering (p = 0.006) and CVLT learning performance (List A, Trial 1; p = 0.037) (Schantz et al., 2001). In contrast, PCB exposure was not associated with performance on tests of executive function or visuospatial ability, including the WCST, TMT, Stroop test, DSST or Hooper Visual Organization Test. Other contaminants measured in fish consumers, including dichlorodiphenyl dichloroethene (DDE), mercury, and lead, were not consistently associated with neuropsychological outcomes.

Overall, findings from this cohort indicate that adult PCB exposure from contaminated fish consumption was associated with impairments in verbal memory and learning, while other cognitive domains such as executive and visuospatial function were largely unaffected.

3.1.4. Upper Hudson River Cohort (New York State)

The Upper Hudson River cohort includes community-dwelling adults aged 55–74 years residing near PCB-contaminated areas in New York State. Three publications evaluated cross-sectional and longitudinal associations between serum PCB concentrations and neuropsychological performance (Fitzgerald et al., 2012, 2008; Tanner et al., 2020). A 34-test battery evaluated memory, executive function, processing speed/visuospatial ability, motor coordination, affective symptoms, reaction time, and olfaction (including CVLT, WMS, TMT, Stroop, WCST, DSST, Grooved Pegboard, Beck Depression Inventory (BDI), and State-Trait Anxiety Inventory (STAI)).

In the baseline cross-sectional study (n = 253), Fitzgerald et al. reported that higher lipid-adjusted total serum PCB concentrations were associated with poorer performance on the CVLT Trial 1 (β = −0.576 per unit change in log adjusted lipid basis serum PCB, p = 0.035) and with higher BDI (β = 1.189, p = 0.007) (Fitzgerald et al., 2008). Associations were most apparent for verbal learning and memory; other cognitive domains and motor function were largely unaffected. Congener-specific analyses suggested stronger associations for more highly chlorinated congeners (Fitzgerald et al., 2012). A subsequent analysis in the same cohort (n=144) focused on Polybrominated diphenyl ethers (PBDE) exposure and PBDE × PCB interactions; it did not present new independent PCB–cognition effect estimates (Fitzgerald et al., 2012). Overall, PBDE concentrations were not consistently associated with neuropsychological performance, although interaction analyses suggested that higher PBDE levels were related to poorer verbal memory only among participants with higher PCB body burdens (Fitzgerald et al., 2012).

In a 14-year longitudinal follow-up of 116 participants, Tanner et al. observed that geometric mean serum ΣPCB concentrations (wet-weight) declined by approximately 22% over time (Tanner et al., 2020). Contrary to prior cross-sectional findings, higher PCB concentrations were not associated with accelerated cognitive decline. Significant ΣPCB × time interactions indicated that individuals with lower PCB levels exhibited expected age-related declines in verbal memory and increases in depressive symptoms, whereas those with higher PCB levels demonstrated small, non-significant improvements or stability over time (Tanner et al., 2020). The authors attributed these unexpected patterns to potential healthy survivor bias and residual confounding. Across publications from this cohort, cross-sectional evidence supports associations between higher PCB body burden and poorer verbal learning and increased depressive symptoms, whereas longitudinal results did not demonstrate progressive cognitive decline attributable to PCB exposure.

3.1.5. Yucheng Cohort (Taiwan Rice Oil Contamination)

The Yucheng cohort comprises adults exposed in 1979 to high levels of PCBs and polychlorinated dibenzofurans (PCDFs) through contaminated rice oil in Taiwan. Two studies evaluated neurocognitive outcomes among exposed individuals aged ≥60 years approximately 20–25 years after exposure (Lin et al., 2010, 2008).

In the primary cohort study (n = 162 exposed; 151 matched controls), Lin et al. administered a 10-test neuropsychological battery assessing cognition, memory modalities, learning, motor and sensory function, mood, and daily activity (Lin et al., 2008). Lin et al. reported that exposed women demonstrated significantly lower performance in VMR, attention and digit span (ADS), and visual memory span (VMS) compared with unexposed controls, after adjustment for age and education (Lin et al., 2008). Learning ability (derived from serial verbal recall trials) and delayed recall were also reduced among exposed women. In contrast, no significant differences were observed in exposed men (Lin et al., 2008). A significant dose–response relationship was observed across PCB exposure categories for VMR, ADS, and VMS in women (Lin et al., 2008). Motor function, sensory testing, depression scores, and activities of daily living did not differ between exposed and reference groups (Lin et al., 2008).

A subsequent analysis comparing the sensitivity of cognitive instruments in the same population (n=165 exposed; n= 151 controls) (Lin et al., 2010). The study’s a priori aim was to evaluate whether a global screen Mini-Mental State (MMSE) or a domain-specific short test (ADS) better detected PCB-related deficits. After adjustment for age, sex, and education, MMSE scores did not significantly differ between exposed and unexposed participants, whereas ADS-forward and ADS-total were significantly lower in the exposed group. Stratified analyses demonstrated dose-dependent reductions in ADS performance at medium (31–89 ppb) and high (≥90 ppb) historical PCB exposure levels (Lin et al., 2010). These findings suggest that domain-specific memory and attention measures were more sensitive than global cognitive screening tools in detecting PCB-related deficits.

Across publications from the Yucheng cohort, evidence indicates persistent, dose-dependent impairments in attention, working memory, and verbal learning among women exposed to high levels of PCBs/PCDFs in adulthood, with minimal effects observed in men.

3.1.6. Mohawk (Akwesasne) Cohort

The Akwesasne Mohawk cohort comprises adults residing near PCB-contaminated Superfund sites along the St. Lawrence River. In the primary investigation, 18 neuropsychological measures spanning memory, executive functioning, and motor performance were administered to 277 adults aged 18–79 years (Haase et al., 2009). Haase et al. identified a nonlinear (threshold) association between total serum PCB burden and neuropsychological performance across the domains of executive functioning, motor functioning, and memory. An inflection point was observed at approximately 2 ppb (ng/g) total PCBs, above which poorer performance emerged (Haase et al., 2009). The clearest threshold effects were observed for executive functioning—particularly the TMT and WCST—and for fine motor coordination as assessed by the Grooved Pegboard. Memory was also associated with PCB burden, driven primarily by logical (verbal) memory, although the nonlinear effect was weaker than for executive function. Significant age × PCB interactions were identified for Trail Making performance and Grooved Pegboard, indicating that threshold-related associations were largely confined to adults aged 37–79 years, whereas no clear exposure effects were observed among younger adults (Haase et al., 2009). No age interaction was detected for memory.

Subsequent analyses by Sasaki et al. focused on processing speed using the DSST in the same cohort (Sasaki et al., 2023a). Quantile-based mixture modeling incorporating low-chlorinated PCBs (inhalation-associated), high-chlorinated PCBs (ingestion-associated), hexachlorobenzene (HCB), DDE, and mirex demonstrated significant negative associations between the exposure mixture and DSST scores among adults aged 47–79 years (Sasaki et al., 2023a). In this age group, a one-quartile increase in the exposure mixture was associated with approximately a 4-point decrement in DSST performance. Low- and high-chlorinated PCB groups (and total PCBs in alternate models) were the dominant contributors to mixture effects, whereas HCB and DDE contributed modestly and mirex did not significantly contribute (Sasaki et al., 2023a). Single-pollutant models showed weaker or inconsistent associations, underscoring the importance of mixture-based approaches.

A comparative analysis between older Mohawks and NHANES participants demonstrated substantially higher PCB body burdens in the Mohawk cohort, with approximately three-fold higher mean total PCB concentrations and nearly five-fold higher low-chlorinated PCB levels among Mohawks (Sasaki et al., 2023b). Mixture modeling further revealed stronger overall cognitive effects in the Mohawk cohort, with broader contributions from both low- and high-chlorinated PCB congeners as well as selected organochlorine pesticides (DDE and HCB), whereas in NHANES older adults cognitive decline was driven primarily by high-chlorinated PCBs and DDE (Sasaki et al., 2023b). Across Mohawk cohort publications, findings consistently indicate that elevated PCB body burden—particularly at concentrations above approximately 2 ng/g—is associated with reduced executive function, processing speed, and selected motor outcomes (Sasaki et al., 2023b). These associations are most evident in older adults and are more clearly characterized when mixture modeling frameworks account for correlated exposures.

3.1.7. Occupational Cohorts

Occupational studies primarily involve electrical capacitor manufacturing workers with historically high PCB exposure. Steenland et al. assembled 17,321 workers (final analytic cohort 16,906 after exclusions) from three U.S. capacitor plants and examined neurodegenerative disease mortality through 1998 (Steenland et al., 2006). Overall mortality from non-cerebrovascular dementia was not consistently elevated in the cohort, but sex-stratified analyses revealed a significant excess of dementia mortality among highly exposed women (standardized mortality ratios = 2.04; 95% CI 1.12–3.43; 14 deaths) (Steenland et al., 2006). Exposure was classified using plant-specific job-exposure matrices, and outcomes were derived from death-certificate (underlying- and multiple-cause) data rather than longitudinal clinical or cognitive assessments (Steenland et al., 2006).

Neuropsychological investigations in the German HELPcB occupational cohort (mean age 43.96 ± 12.6 years; range 19–83) identified selective, domain-specific deficits rather than global cognitive impairment. At baseline (n = 237), participants completed a comprehensive battery spanning six core cognitive domains—intelligence, verbal fluency, attention, executive function, learning and memory, and sensorimotor performance—assessed through eleven standardized instruments/subtests (Fimm et al., 2017). Overall, PCB burden was associated with impairments in verbal fluency and sensorimotor processing, while no significant effects were observed for memory, attention, or other executive functions. Further analyses suggested congener-specific patterns. Lower-chlorinated PCBs were consistently associated with reduced verbal fluency across several task conditions. In contrast, adverse effects of higher-chlorinated PCBs on Aiming and of dioxin-like PCBs on Line Tracking were observed (Fimm et al., 2017).

In a one-year follow-up of a subsample (n = 116), PCB levels declined slightly but remained highly correlated with baseline concentrations, indicating stable exposure ranking. Deficits in verbal fluency persisted over time, whereas the previously observed sensorimotor processing effects were less consistent at follow-up (Cromberg et al., 2024). Taken together, findings from this cohort suggest modest and domain-specific associations, primarily affecting verbal fluency and motor-related functions rather than global cognition.

3.1.8. Canadian Study of Health and Aging Cohort

The CSHA is a large, population-based cohort of adults aged ⩾65 years with plasma PCB and organochlorine (OC) pesticide measurements. In a cross-sectional analysis of 2023 clinically evaluated participants (574 dementia cases; 399 Alzheimer’s disease [AD]), plasma PCB concentrations were not associated with prevalent dementia or AD after multivariable adjustment (Medehouenou et al., 2014).

A subsequent prospective analysis of 669 nondemented participants followed for incident dementia (156 cases; 108 AD) found no association between baseline plasma PCB concentrations and risk of incident dementia or AD (Medehouenou et al., 2019). However, repeated-measures analyses using 3MS cognitive scores showed that higher concentrations of several PCB congeners (118, 153, 156, 163) were associated with cognitive decline as assessed with the Modified Mini-Mental State (3MS) (Medehouenou et al., 2019).

Across CSHA publications, PCB exposure was not associated with increased dementia incidence, although modest associations with global cognitive decline were observed for selected congeners.

3.1.9. Hispanic Community Health Study (HCHS/SOL)

The HCHS/SOL examined associations between plasma persistent organic pollutants and 7-year cognitive change among 1,837 middle-aged and older Hispanic/Latino adults (mean baseline age = 56 years) (Parada et al., 2024). Baseline plasma samples included 24 polychlorinated biphenyl (PCB) congeners, along with organochlorine pesticides and brominated flame retardants. Cognitive performance was assessed at baseline (2008–2011) and follow-up (2015–2018) using the Brief-Spanish English Verbal Learning Test (B-SEVLT), Word Fluency Test, and Digit Symbol Substitution Test, and standardized change scores were used to construct a global cognitive decline measure (Parada et al., 2024).

Each doubling (log₂ increase) in plasma concentrations of higher-chlorinated PCB congeners 146, 178, 194, 199/206, and 209 was associated with steeper declines in global cognition over 7 years (β estimates approximately −0.05 to −0.06 standard deviation units) (Parada et al., 2024). Associations were most consistent for episodic verbal learning and memory (B-SEVLT), with more limited or weaker effects observed for processing speed and verbal fluency (Parada et al., 2024). In contrast, organochlorine pesticides and brominated flame retardants were generally not associated with cognitive decline.

Overall, the findings indicate that higher PCB body burden—particularly for more highly chlorinated congeners—is associated with modest longitudinal cognitive decline in this diverse Hispanic/Latino population.

3.1.10. Italian Environmental Exposure Cohort (Brescia/Caffaro PCB-Contaminated Area)

A population-based cohort from the PCB-contaminated Brescia–Caffaro industrial area in northern Italy examined associations between baseline serum PCB concentrations and neurodegenerative disease outcomes among 699 adults aged >50 years (mean age 63.2 years) followed for a mean of 8.8 years (Raffetti et al., 2020). Serum concentrations of 24 PCB congeners were measured at baseline using gas chromatography–mass spectrometry, and total PCB exposure was analyzed both continuously and by tertiles. Incident dementia and Parkinson disease cases were identified through regional healthcare databases including hospital admissions, outpatient visits, and prescription records.

During follow-up, 36 cases of dementia and 20 cases of Parkinson disease were identified. In multivariable-adjusted Poisson regression models, higher serum PCB concentrations were associated with increased dementia risk (Raffetti et al., 2020). Compared with the lowest exposure tertile, dementia incidence was higher in both the second tertile (rate ratios RR = 2.30) and the highest tertile (RR = 4.35) (Raffetti et al., 2020). Trend analyses indicated approximately a twofold increase in dementia risk per tertile increase in PCB exposure, and spline models suggested a monotonic dose–response relationship between log-transformed PCB concentrations and dementia incidence.

Overall, this cohort study from a highly PCB-contaminated region provides evidence of a positive association between higher circulating PCB concentrations and increased dementia risk, although the relatively small number of cases limits precision and generalizability.

3.1.11. French Aging Cohort (Three-City Study)

The Three-City (3C) Bordeaux cohort investigated associations between baseline plasma persistent organic pollutants and dementia-related outcomes among 515 community-dwelling older adults (mean age 72.5 years) followed for up to 17 years (Lefèvre-Arbogast et al., 2025). Plasma concentrations of 15 PCB congeners, 12 organochlorine pesticides, and one brominated flame retardant were measured at baseline, and participants underwent 8 repeated neuropsychological assessments and magnetic resonance imaging assessments during follow-up.

In multivariable-adjusted analyses, neither individual POPs nor a composite POP exposure score were consistently associated with incident dementia, longitudinal cognitive decline, or medial temporal lobe atrophy. An isolated association was observed for PCB-52, which was linked to increased dementia risk; however, this finding was not supported by analyses of cognitive trajectories or neuroimaging outcomes. Significant interactions were detected between highly chlorinated PCB congeners (e.g., PCB-180, PCB-194, PCB-196–203) and APOE-ε4 carrier status, with adverse associations among APOE-ε4 carriers and opposite trends among non-carriers (Lefèvre-Arbogast et al., 2025).

Overall, this prospective population-based study did not provide consistent evidence that circulating PCBs concentrations in the general older population were associated with increased risk of dementia, cognitive decline, or brain atrophy.

3.1.12. Weitang Geriatric Diseases Cohort (China)

The Weitang Geriatric Diseases study examined associations between plasma PCB concentrations and cognitive dysfunction among community-dwelling older adults in eastern China. In this cross-sectional analysis, 266 participants aged 61–90 years (median age 67 years) were selected from a larger community-based cohort of residents aged ≥60 years. Plasma concentrations of six indicator PCB congeners (PCB28, PCB52, PCB101, PCB138, PCB153, and PCB180) were measured using gas chromatography–mass spectrometry, and cognitive function was assessed using the Abbreviated Mental Test (AMT) (Pan et al., 2022). Participants were classified as having normal cognitive function or cognitive dysfunction based on AMT scores.

Sequential logistic regression models adjusted for demographic, socioeconomic, and health-related factors showed no statistically significant associations between overall PCB exposure and cognitive dysfunction after Bonferroni correction (Pan et al., 2022). However, path analysis using structural equation modeling to evaluate co-exposure patterns indicated that PCB28 had a direct association with cognitive dysfunction among women aged ≤80 years, with a standardized factor load of 0.670 after controlling for other PCB congeners and potential confounders (Pan et al., 2022). Other PCB congeners did not demonstrate significant direct or indirect effects on cognitive outcomes in the final model.

Overall, findings from this cohort suggest that general PCB body burden was not consistently associated with cognitive dysfunction in the overall population, although congener-specific analyses indicated a potential adverse effect of lower-chlorinated PCB28 on cognitive impairment among older women.

3.2. Risk of Bias Assessment

Risk of bias was evaluated across nine domains using a domain-based approach modeled after Lam et al., (2017)(Figure 2). Most studies were rated as low or probably low for selection bias, although (Cromberg et al., 2024; Fimm et al., 2017; Lefèvre-Arbogast et al., 2025; Ruder et al., 2014; Sasaki et al., 2023b; Seegal et al., 2013) were rated probably high or high due to unclear or non-representative sampling strategies. Interpretation of selection bias should be considered in the context of study design. Occupational cohorts (e.g., capacitor workers and recycling workers) may be subject to healthy worker selection effects, whereby participants are generally healthier than the underlying population, potentially underestimating adverse associations. In contrast, community-based cohorts and environmentally exposed populations may reflect localized or high-exposure settings that are not representative of the general population, limiting generalizability. Cross-sectional studies, particularly those based on existing datasets such as NHANES, may also be affected by selective participation and survival bias, as individuals with higher exposure and poorer health may be less likely to be included. These design-specific selection processes should be considered when interpreting the direction and magnitude of observed associations.

Blinding was often insufficiently reported; (Cromberg et al., 2024; Fimm et al., 2017; Kilburn et al., 1989; Medehouenou et al., 2019; Pan et al., 2022; Sasaki et al., 2023a; Tanner et al., 2020) were rated probably high, suggesting potential for detection bias, whereby knowledge of participants’ exposure status could influence the administration, scoring, or interpretation of cognitive assessment.

Exposure misclassification was generally low, with most studies employing biomarker-based measurements of PCB levels. However, (Steenland et al., 2006) used estimated cumulative PCB exposure modeled from job-exposure matrices rather than measured serum levels, introducing uncertainty into individual-level exposure classification. Similarly, (Ruder et al., 2014) relied on historical exposure information based on work location and duration at a contaminated site without individual-level biomarkers, limiting precision and increasing the potential for misclassification.

Outcome misclassification was generally rated as low due to the use of standardized or well-established assessment procedures in most studies. However, (Steenland et al., 2006) was rated probably high because the study relied solely on death certificates to identify neurodegenerative disease outcomes.

Confounding was the most prevalent concern. (Cromberg et al., 2024; Fimm et al., 2017; Kilburn et al., 1989; Lin et al., 2010; Ruder et al., 2014; Sasaki et al., 2023a, 2023b; Steenland et al., 2006) were rated probably high or high for confounding due to inadequate adjustment for socioeconomic or environmental covariates. Incomplete outcome data were generally well reported, though (Lin et al., 2008; Pan et al., 2022; Sasaki et al., 2023b) were rated probably high or high. Selective reporting was mostly low risk, with (Przybyla et al., 2017) and (Pan et al., 2022) receiving probably high ratings due to insufficient reporting of null results. Conflict of interest was low or probably low in most studies, except (Medehouenou et al., 2014), which was rated as high risk due to sponsorship-related concerns. Other threats to validity were common; probably high or high risk was noted in 18 of 24 studies, including (Bouchard et al., 2014; Fimm et al., 2017; Fitzgerald et al., 2012, 2008; Kilburn et al., 1989; Lin et al., 2008; Medehouenou et al., 2019, 2014; Pan et al., 2022; Przybyla et al., 2017; Raffetti et al., 2020; Ruder et al., 2014; Sasaki et al., 2023a, 2023b; Seegal et al., 2013; Steenland et al., 2006; Tanner et al., 2020; Zuo et al., 2024). These ratings were often due to unclear quality control procedures, post hoc analyses, or insufficient reporting of participant exclusion and protocol deviations. Overall, while exposure and outcome assessments were robust, confounding and other bias domains were the most frequent sources of concern.

3.3. Dementia or Alzheimer diseases

Studies examining the relationship between PCB exposure and dementia or Alzheimer’s disease have produced mixed results. In population-based cohorts, findings are heterogeneous. (Raffetti et al., 2020) conducted a prospective cohort study in a highly polluted area of Northern Italy and found that higher total PCB serum levels were associated with an increased risk of dementia. This association showed a dose-response relationship, with subjects in the 2nd and 3rd tertile of total serum PCB distribution having an increased risk of dementia compared to the 1st tertile. In contrast, analyses from the CSHA did not identify an overall association between PCB exposure and dementia incidence or prevalence (Medehouenou et al., 2019, 2014), although higher concentrations of specific PCB congeners were associated with lower cognitive performance. More recent evidence from a large French population-based aging cohort with up to 17 years of follow-up also did not demonstrate consistent associations between PCB exposure and dementia risk overall, although subgroup analyses suggested potential effect modification by genetic susceptibility (e.g., APOE genotype) (Lefèvre-Arbogast et al., 2025).

Occupational cohort studies similarly show inconsistent patterns regarding the relationship between PCB exposure and dementia-related outcomes. In a large multi-cohort analysis of capacitor workers, Steenland et al., (2006) found no overall excess of dementia mortality associated with PCB exposure; however, subgroup analyses indicated elevated risks among highly exposed women, suggesting potential effect modification by sex or exposure level. In contrast, Ruder et al., (2014), using a separate cohort of capacitor workers, also reported no consistent increase in dementia-related mortality at the population level.

Together, these findings indicate that while some subgroups—particularly highly exposed individuals or women—may exhibit increased risk, the overall evidence linking PCB exposure to dementia in both occupational and general populations remains limited and inconsistent.

3.4. Cognition assessment tools

Across the reviewed studies, neuropsychological assessments targeted a wide range of cognitive domains (Table 2, Table S1 Appendix), with memory and learning being the most frequently evaluated. Other commonly assessed domains included executive function, attention and processing speed, and, to a lesser extent, motor or sensorimotor performance. Global cognition was also evaluated in several studies, typically using screening instruments or composite scores.

Table 2.

cognitive test domain and disease outcome used in studies.

Cohort	Studies	Global Cognition	Memory	Attention	Executive Function	Processing Speed	Language	Affective (Mood)	Cognitive Decline	Dementia
PCB-Exposed Firemen	Kilburn et al. (1989)		✔		✔	✔		✔
Great Lakes Fisheaters	Schantz et al. (2001)		✔		✔
NHANES (1999–2002)	Bouchard et al. (2014);Przybyla et al. (2017)					✔
NHANES (2011–2014)	Zuo et al. (2024)	✔	✔			✔	✔
Upper Hudson River (NY, USA)	Fitzgerald et al. (2008);Fitzgerald et al. (2012);Tanner et al. (2020)		✔		✔	✔		✔	✔
NIOSH Capacitor Workers (U.S.)	Steenland et al. (2006); Ruder et al. (2014)									✔
NY Capacitor Workers (GE)	Seegal et al. (2013)		✔	✔	✔	✔
Mohawk (Akwesasne)	Haase et al. (2009); Sasaki et al. (2023a);Sasaki et al. (2023b)		✔		✔	✔
HCHS/SOL (U.S. Hispanic/Latino)	Parada et al. (2024)		✔			✔	✔		✔
Canadian Study of Health & Aging	Medehouenou et al. (2014); Medehouenou et al. (2019)	✔							✔	✔
Taiwan Yucheng Cohort	Lin et al. (2008);Lin et al. (2010)	✔	✔	✔
HELPCB Cohort (Germany)	Fimm et al. (2017); Cromberg et al. (2024)		✔	✔	✔		✔
Brescia/Caffaro Cohort (Italy)	Raffetti et al. (2020)									✔
Three-City Study (France)	Lefèvre-Arbogast (2025)	✔	✔	✔	✔		✔		✔	✔
Older Chinese Females Cohort	Pan et al. (2022)	✔

Less frequently examined domains included visual-spatial/perceptual function, verbal fluency or language-related measures, and affective state, which were often assessed as secondary outcomes within broader neuropsychological batteries. In addition, a smaller subset of studies evaluated cognitive decline and dementia-related outcomes, reflecting the limited availability of longitudinal data with clinical endpoints. Because many studies employed comprehensive test batteries spanning multiple domains, these categorizations are not mutually exclusive.

In studies examining the neurocognitive effects of PCB exposure, intelligence was not typically treated as a primary outcome, but rather assessed or approximated as a covariate to account for differences in baseline cognitive ability, education, and cognitive reserve. This approach helps reduce confounding and improve the interpretation of associations between PCB exposure and cognitive performance.

The neuropsychological instruments used varied across studies, although several commonly used tools were identified. Measures of processing speed and attention, such as the Digit Symbol Substitution/Coding Test (DSST/DSCT) and TMT, were frequently employed. Memory and learning were commonly assessed using the WMS and CVLT, while executive function was evaluated using instruments such as the WCST and Stroop Color-Word Test.

Assessments of motor and sensorimotor function were less consistently included and were typically limited to occupational or neurobehavioral studies, using tools such as Grooved Pegboard, finger tapping or oscillation tasks, and reaction-time measures. Affective state was evaluated in some studies using instruments such as the BD) and STAI, generally as secondary outcomes. Other specialized measures, including olfactory testing and intelligence-related assessments, were rarely used and were not consistently reported across studies.

Overall, the diversity of cognitive domains assessed and the variability in neuropsychological instruments contribute to challenges in comparing findings across studies and highlight the need for greater standardization in future research.

3.5. Cognitive domains outcome

Cognitive outcomes were categorized into domains based on commonly used neuropsychological frameworks, including global cognition, memory and learning, executive function, attention, processing speed, motor function, language, and affective symptoms. Although these domains are conceptually distinct, considerable overlap exists in the cognitive processes assessed by specific test. When studies reported composite scores or used tests spanning multiple domains, outcomes were categorized based on the primary cognitive construct assessed. This approach was used to ensure consistency while acknowledging the inherent overlap between cognitive domains.

3.5.1. Global Cognitive Function

Studies assessing global cognitive function in relation to PCB exposure employed heterogeneous measurement approaches, which can be broadly grouped into (1) global cognitive screening instruments, (2) composite or repeated-measure cognitive scores, and (3) processing speed–based proxy measures.

Global cognitive screening instruments, including the MMSE, AMT, and 3MS, were used in several cohorts. Across these studies, associations between PCB exposure and global cognition were generally weak or non-significant. For example, studies in the Yucheng cohort and other community-based populations found no significant differences in MMSE scores between exposed and reference groups after adjustment (Lin et al., 2010). Similarly, analyses using AMT and 3MS reported either null or modest associations that were attenuated after controlling for confounders (Medehouenou et al., 2019; Pan et al., 2022). These findings suggest that global screening tools may have limited sensitivity to detect PCB-related neurocognitive effects.

Composite and longitudinal measures of global cognition, derived from multiple neuropsychological tests, have been increasingly used in recent studies and provide a more sensitive assessment of cognitive change over time. In the HCHS/SOL cohort, higher concentrations of several PCB congeners were associated with greater declines in global cognitive scores over approximately seven years (Parada et al., 2024). In contrast, a large French population-based cohort with repeated cognitive assessments over up to 17 years did not observe consistent associations between PCB exposure and global cognitive decline (Lefèvre-Arbogast et al., 2025). These longitudinal findings highlight substantial heterogeneity across populations despite the use of more robust cognitive measures.

Processing speed–based measures, particularly the DSC, were frequently used as proxies for global cognitive function due to their sensitivity to early cognitive impairment. In NHANES-based studies, higher PCB exposure—especially for certain dioxin-like and non-dioxin-like congeners—was associated with lower DSC performance, with stronger effects observed in older adults and women (Bouchard et al., 2014; Przybyla et al., 2017). More recent mixture-based analyses further suggest that PCB-containing chemical mixtures may be associated with lower global cognitive z-scores and DSST performance, although results vary across statistical models and cognitive domains (Zuo et al., 2024).

Overall, evidence linking PCB exposure to global cognitive function is inconsistent and appears to depend strongly on the method of assessment. Studies using screening instruments generally report null findings, whereas more sensitive measures—including composite scores and processing speed tests—suggest modest associations in some populations. However, these findings are not consistent across cohorts, and the overall strength of evidence for global cognitive impairment remains limited.

3.5.2. Memory and learning

Thirteen studies assessed memory and learning outcomes in relation to PCB exposure, including multiple analyses from several cohorts (Table 2). Memory was evaluated using a range of neuropsychological instruments, most commonly standardized verbal learning and memory tests such as the WMS and CVLT, as well as domain-specific measures including verbal recall, visual memory span, and attention-based working memory tasks (e.g., ADS). These measures capture both verbal memory (e.g., verbal learning and delayed recall) and visual memory, which were considered together due to variability in reporting across studies.

Across studies, inverse associations between PCB exposure and memory performance were most consistently observed for verbal learning and delayed recall. Environmentally exposed cohorts and populations with higher PCB body burdens frequently demonstrated poorer performance on WMS- and CVLT-based measures (Fitzgerald et al., 2008; Schantz et al., 2001). In the Yucheng cohort, significant deficits in verbal and visual memory, as well as working memory assessed by ADS, were observed only among women exposed to PCBs, with evidence of dose-response relationships across exposure categories (Lin et al., 2010, 2008). Notably, in this cohort, global cognitive screening (MMSE) did not detect differences, whereas domain-specific memory measures did, suggesting greater sensitivity of targeted memory assessments.

However, findings were not consistent across all studies. Several occupational cohorts and studies using similar neuropsychological instruments reported no significant associations between PCB exposure and memory outcomes (Fimm et al., 2017; Seegal et al., 2013). Longitudinal analyses further highlight this heterogeneity. In the Upper Hudson River cohort, memory trajectories over time did not show consistent decline associated with higher PCB exposure (Tanner et al., 2020), and a large French aging cohort with repeated cognitive assessments similarly reported no consistent association between PCB exposure and memory decline over extended follow-up (Lefèvre-Arbogast et al., 2025). In contrast, more recent population-based longitudinal evidence suggests that memory-related domains may be among the most sensitive to PCB exposure, with stronger associations observed for episodic verbal learning and recall compared to other cognitive domains (Parada et al., 2024). Additionally, NHANES-based analyses using immediate and delayed recall measures (IRT and DRT) found that certain PCB congeners were associated with differences in memory performance, although the results were inconsistent across regression, WQS, and BKMR models and did not show a uniform direction of effect (Zuo et al., 2024).

Overall, the evidence linking PCB exposure to memory and learning is suggestive but inconsistent. Associations are more frequently observed in studies using sensitive domain-specific measures and in populations with higher exposure levels, whereas findings from occupational and longitudinal studies are more variable. These discrepancies likely reflect differences in exposure profiles, study design, and the sensitivity of cognitive assessment tools.

3.5.3. Executive Function

Several studies assessed executive function in relation to PCB exposure across occupational, environmental, and population-based cohorts. Executive function was primarily evaluated using neuropsychological tests of cognitive flexibility, set shifting, and inhibitory control, most commonly the WCST, TMT, and Stroop Test.

Across studies, associations between PCB exposure and executive function were generally weak and inconsistent. While several cohorts included executive function assessments, many did not observe significant associations. For example, studies in environmentally exposed populations that incorporated tests such as WCST and TMT did not report clear associations between PCB exposure and executive function outcomes (Fitzgerald et al., 2008; Haase et al., 2009). Similarly, other studies using comparable neuropsychological instruments reported null associations (Seegal et al., 2013).

Longitudinal evidence also remains limited and heterogeneous. Cohort analyses with repeated cognitive assessments have not demonstrated consistent changes in executive function over time in relation to PCB exposure (Lefèvre-Arbogast et al., 2025). In addition, population-based studies suggest that associations with executive-related domains are generally weaker than those observed for memory outcomes (Parada et al., 2024).

Overall, the evidence linking PCB exposure to executive function is limited and inconsistent, with most studies reporting null or weak associations. Compared with memory-related outcomes, executive function appears to be less consistently affected, suggesting potential domain-specific differences in sensitivity to PCB-related neurotoxicity.

3.5.4. Attention and processing Speed

Several studies assessed attention and processing speed in relation to PCB exposure using both domain-specific and composite neuropsychological measures. Attention was commonly evaluated using tasks such as the ADS, while processing speed was primarily measured using the DSST/DSCT and similar timed tasks. Although these domains are conceptually distinct, they are closely related and frequently assessed together due to their shared reliance on rapid information processing and sustained attention.

Across studies, associations between PCB exposure and attention and processing speed were mixed but more consistently observed than for executive function. In the Yucheng cohort, PCB exposure was associated with deficits in attention and working memory, particularly among women, with evidence of dose-response relationships (Lin et al., 2010, 2008).

Population-based studies also suggest associations with processing speed. In NHANES analyses, higher PCB concentrations were associated with lower DSC performance, particularly among older adults and in analyses of specific congeners or mixtures (Bouchard et al., 2014; Przybyla et al., 2017). More recent mixture-based models similarly identified associations between PCB-containing mixtures and DSST performance, although findings varied across statistical approaches (Zuo et al., 2024). Comparable results were observed in environmentally exposed populations using mixture-based analyses (Sasaki et al., 2023a, 2023b).

Some studies also assessed verbal fluency using tasks such as the AFT or category-based word generation. These measures reflect semantic retrieval and processing efficiency and are often analyzed alongside attention and processing speed. Associations between PCB exposure and verbal fluency were reported in selected cohorts, particularly in occupational and NHANES-based analyses, although findings were not consistent across models or populations (Cromberg et al., 2024; Zuo et al., 2024). In other cohorts, verbal fluency was included within composite scores without clear domain-specific effects (Lefèvre-Arbogast et al., 2025; Parada et al., 2024).

However, not all studies observed consistent associations. Some cohorts reported null or weak findings for attention and processing speed measures (Lefèvre-Arbogast et al., 2025; Seegal et al., 2013). Overall, the evidence linking PCB exposure to attention and processing speed is suggestive but inconsistent, with stronger signals observed in studies using sensitive, timed measures.

3.5.4. Motor Function

Few studies have examined motor function in relation to PCB exposure, and available evidence is primarily limited to occupational cohorts. Motor outcomes were assessed using neurobehavioral tests of coordination, sensorimotor performance, and psychomotor function, often as part of broader test batteries rather than as standalone outcomes.

Early findings from an occupationally exposed cohort of firemen suggested impairments in motor coordination and sensorimotor performance among exposed individuals compared with controls (Kilburn et al., 1989). However, subsequent studies have not consistently replicated these findings. In an occupational cohort of capacitor workers, Seegal et al., (2013) included multiple motor and coordination measures but did not observe any associations between PCB exposure and motor performance. In the HELPcB cohort, PCB exposure was associated with impairments in specific sensorimotor tasks, although motor effects were not consistent across all measures (Fimm et al., 2017). Longitudinal evidence from the same cohort suggests that motor-related effects are less consistent over time. In follow-up analyses, previously observed associations with motor performance were not robustly maintained across all measures, whereas deficits in verbal fluency persisted, indicating differential stability of domain-specific effects (Cromberg et al., 2024).

Overall, the evidence linking PCB exposure to motor function is limited and inconsistent, with few studies available and largely null findings in more recent analyses. Compared with cognitive domains such as memory and processing speed, motor outcomes appear to be less sensitive to PCB-related effects.

3.5.6. Cognitive decline

Evidence on cognitive decline associated with PCB exposure remains limited but has improved with the inclusion of longitudinal studies. Findings across cohorts are heterogeneous. Some studies reported associations between higher PCB concentrations and greater decline in global cognition or specific domains (Medehouenou et al., 2019; Parada et al., 2024), while others did not observe consistent relationships over extended follow-up periods (Lefèvre-Arbogast et al., 2025; Tanner et al., 2020).

Recent population-based studies incorporating repeated cognitive assessments suggest that PCB exposure may be associated with modest declines in cognitive function over time, particularly in memory-related domains (Parada et al., 2024). However, inconsistent findings across cohorts highlight the complexity of interpreting temporal relationships.

Overall, while longitudinal evidence strengthens the assessment of temporality, the current literature provides limited and inconsistent evidence for an association between PCB exposure and trajectories of cognitive decline.

4. Discussion

This study synthesizes findings from multiple epidemiological investigations examining the association between PCBs exposure and neurocognitive outcomes. The evidence across various populations—including occupationally exposed workers, community residents, and specific demographic groups such as the elderly—suggests a complex relationship between PCB burden and cognitive function. The majority of studies support the hypothesis that PCB exposure is associated with adverse effects on cognitive function, particularly in domains such as memory, processing speed, and sensorimotor coordination. However, the findings on global cognitive decline, executive function, and dementia risk remain mixed. While several high-exposure cohorts—such as occupationally exposed workers and environmentally burdened communities—exhibited clear cognitive impairments, others, including general population studies, presented more nuanced or null associations. These inconsistencies suggest that the neurotoxicity of PCBs is influenced by a complex interplay of factors, including sex, age, exposure dose and duration, PCB congener type, and co-exposure to other pollutants. This underscores the importance of applying sensitive neuropsychological measures and considering vulnerable subgroups in future research aimed at elucidating the mechanisms and modifiers of PCB-related cognitive decline.

4.1. Biological mechanisms

The mechanisms underlying PCB-induced cognitive impairment remain an area of ongoing investigation. PCB neurotoxicity is mediated through several well-characterized biological mechanisms, including: (1) altered dopamine (DA) signaling; (2) disruption of thyroid hormone signaling; (3) perturbation of intracellular calcium (Ca²⁺) dynamics; and (4) induction of oxidative stress (Pessah et al., 2019). Among these, dopaminergic disruption is the most frequently proposed and extensively studied pathway underlying the association between PCB exposure and cognitive impairments, particularly in the memory domain.

Animal studies have consistently demonstrated that elevated PCB levels are associated with reduced dopamine concentrations in the brain, likely through interference with dopamine neurotransmission(Brun et al., 2021; Faroon et al., 2000; Mariussen et al., 1999). PCBs inhibit dopamine transport by affecting both the vesicular monoamine transporter 2 (VMAT2) and the dopamine transporter (DAT), ultimately resulting in reduced synaptic dopamine availability in brain regions critical for motor and cognitive function (Fonnum et al., 2006). In a notable occupational study of former capacitor workers, Seegal et al., (2010) reported an inverse relationship between lipid-adjusted total serum PCB concentrations and DAT density in women, suggesting a sex-specific vulnerability of the dopaminergic system to PCB exposure.

Dopaminergic signaling is essential for both executive function and memory, primarily through its modulatory effects on the prefrontal cortex (PFC) and hippocampus (Ott and Nieder, 2019; Takahashi et al., 2008). Disruption of dopamine signaling in these regions impairs cognitive performance, as demonstrated by both clinical and preclinical studies (Narayanan et al., 2013; Takahashi et al., 2008). The PFC is particularly dependent on balanced dopamine levels to support attention regulation, cognitive flexibility, and working memory, with performance following an inverted U-shaped function relative to dopamine availability (Cools and D’Esposito, 2011). In the hippocampus, dopamine enhances long-term potentiation and facilitates novelty detection—mechanisms fundamental to memory encoding and consolidation (Lisman and Grace, 2005). Therefore, it is biologically plausible that chronic PCB exposure—through disruption of dopamine signaling—can impair these cognitive domains. Supporting this, both human epidemiologic and animal experimental studies have reported deficits in working memory and executive functioning associated with early-life or occupational PCB exposure, reinforcing the role of dopaminergic dysfunction in PCB-induced cognitive impairments.

4.2. PCB exposure assessment

The assessment of PCBs in human blood has advanced significantly over the decades, transitioning from total PCB measurements to targeted congener analyses and toxic equivalency approaches. Early studies, such as Kilburn et al., (1989) and Schantz et al., (2001), used wet-weight serum measurements to evaluate exposure in occupational and dietary contexts, linking high PCB burdens to neurobehavioral and cognitive impairments. These foundational studies, however, lacked congener-specific analyses and lipid adjustment. By the 2000s, researchers embraced congener-specific targeting and lipid-standardized metrics, enabling more precise associations between exposure and health outcomes. Fitzgerald et al., (2008) advanced this methodology by quantifying 30 congeners in lipid-adjusted serum samples, linking exposure to cognitive decline.

Subsequent studies expanded on this approach by categorizing PCBs into LPCBs, HPCBs (Cromberg et al., 2024; Fimm et al., 2017; Sasaki et al., 2023b, 2023a), and dlPCBS-dioxin-like congeners (Bouchard et al., 2014; Fimm et al., 2017; Przybyla et al., 2017), revealing their unique toxicological profiles. Fimm et al., (2017) demonstrated distinct effects, with LPCBs impairing verbal fluency and HPCBs and dlPCBs affecting sensorimotor processing, while Sasaki et al., (2023a) and Cromberg et al., (2024) further highlighted the cumulative effects of PCB mixtures on verbal fluency across chlorination levels. Studies such as Bouchard et al., (2014) and Medehouenou et al., (2014) utilized toxic equivalency methods to quantify the impact of dioxin-like PCBs. Bouchard et al., (2014) linked dioxin-like congeners to significant cognitive declines, particularly in older women, while Medehouenou et al., (2019) found associations between dioxin-like PCB 118 and cognitive impairments in aging populations. Together, these studies underscore the importance of distinguishing among PCB subgroups—LPCBs, HPCBs, and dioxin-like congeners—while leveraging lipid-adjusted and toxic equivalency methods to elucidate the nuanced relationships between PCB exposures and health outcomes.

A key finding across studies is the variability in cognitive effects depending on exposure levels, congener composition, and individual susceptibility factors. The prospective cohort study by Raffetti et al., (2020) in Northern Italy demonstrated a dose-dependent relationship between PCB exposure and dementia risk, with a more than sevenfold increased risk among individuals in the highest tertile of exposure. Similarly, Lin et al., (2008) and Lin et al., (2010) found significant dose-dependent impairments in verbal memory recall and attention, particularly among women, highlighting potential sex differences in susceptibility to PCB neurotoxicity.

4.3. Congener-specific evidence

PCB exposures were characterized using different classification frameworks across studies, including analyses of individual congeners, dioxin-like versus non-dioxin-like groupings, and classifications based on degree of chlorination (e.g., low- versus high-chlorinated PCBs). These approaches reflect distinct toxicological and exposure-related properties but are not directly comparable, as individual congeners may fall into multiple categories. This heterogeneity complicates cross-study comparisons and contributes to inconsistencies in observed associations.

At the individual congener level, several studies identified adverse associations with specific PCB congeners. In NHANES-based analyses, PCB-146 was associated with lower cognitive performance (Przybyla et al., 2017), while PCB-199 and PCB-206 were linked to poorer performance in selected cognitive domains in later survey cycles (Zuo et al., 2024). Similarly, in the HCHS/SOL cohort, higher concentrations of PCB-146, 178, 194, 199/206 and 209 were associated with greater longitudinal declines in global cognition over approximately seven years (Parada et al., 2024). Evidence from the CSHA further suggested that congeners such as PCB-118, 153, 156, and 163 were associated with declines in global cognitive function (Medehouenou et al., 2019). However, findings were not uniform; for example, PCB-153 showed inconsistent or even positive associations with cognitive performance in some cross-sectional analyses (Przybyla et al., 2017), raising the possibility of residual confounding or reverse causation.

When grouped by structural characteristics, patterns suggest differential effects by degree of chlorination. Higher-chlorinated PCBs, which are more persistent and primarily associated with dietary exposure, were more consistently linked to cognitive decline and dementia-related outcomes (Parada et al., 2024). In contrast, lower-chlorinated PCBs, which are more volatile and often associated with inhalation exposure, were more strongly associated with specific neurobehavioral domains such as verbal fluency in occupational (Cromberg et al., 2024; Fimm et al., 2017). Mixture-based analyses also indicated that both low- and high-chlorinated PCB groups contribute to cognitive deficits, particularly in older adults, although their relative importance varied by population and modeling approach (Sasaki et al., 2023a, 2023b).

Evidence based on dioxin-like versus non-dioxin-like classifications was more limited but suggests that both classes may be relevant through distinct biological pathways. Dioxin-like PCBs were associated with lower cognitive performance in population-based studies, particularly among older adults (Bouchard et al., 2014), whereas non-dioxin-like congeners such as PCB-146 were independently associated with poorer cognitive outcomes in multipollutant models (Przybyla et al., 2017).

Overall, while congener-specific patterns are biologically plausible and supported by some consistent findings—particularly for higher-chlorinated PCBs—the epidemiologic evidence remains heterogeneous. Differences in exposure classification, high inter-correlation among congeners, variability in cognitive assessments, and limited longitudinal data constrain the ability to clearly distinguish neurotoxic profiles across individual PCB congeners or congener groups.

4.3. Study design

A major limitation of the current evidence base is the limited incorporation of temporality, as most studies are cross-sectional and assess PCB exposure and cognitive outcomes at a single time point (e.g., (Bouchard et al., 2014; Fitzgerald et al., 2008; Przybyla et al., 2017)). Such designs preclude determining whether exposure preceded differences in cognitive performance, thereby limiting causal inference. Accordingly, these findings should be interpreted as associations with cognitive performance rather than evidence of cognitive decline.

Although longitudinal evidence is emerging, it remains relatively limited and heterogeneous. Several recent studies have begun to address temporality by examining changes in cognition over time. For example, in a large U.S. Hispanic/Latino cohort, higher concentrations of several PCB congeners (e.g., PCB-146, 178, 194, and 209) were associated with steeper declines in global cognition over approximately seven years (Parada et al., 2024). In contrast, findings from a French population-based cohort with up to 17 years of follow-up did not provide consistent evidence linking PCB or broader POP exposure to cognitive decline, dementia risk, or brain atrophy, although subgroup-specific associations (e.g., by APOE genotype) were observed (Lefèvre-Arbogast et al., 2025).

Evidence from occupational cohorts is similarly mixed. In a longitudinal study of PCB-exposed workers, Cromberg et al., (2024) reported that impairments in verbal fluency persisted in one year follow up, whereas other cognitive domains did not show consistent associations with PCB exposure. Earlier longitudinal work in environmentally exposed populations also reported inconsistent exposure–time relationships. For example, in a 14-year follow-up of the Upper Hudson River cohort, Tanner et al., (2020) observed significant exposure–time interactions, but higher PCB concentrations were not associated with greater cognitive decline and were, in some cases, associated with small, non-significant improvements over time.

Taken together, these longitudinal studies provide important evidence supporting temporality, but do not yet yield a coherent pattern of associations across populations, cognitive domains, or PCB congeners. Differences in exposure levels, congener profiles, population characteristics, follow-up duration, and susceptibility factors (e.g., age or genetic risk) likely contribute to these inconsistencies.

Furthermore, the long biological half-lives of PCBs complicate temporal interpretation, as measured serum concentrations may reflect cumulative lifetime exposure rather than exposure during etiologically relevant windows.

Overall, while longitudinal studies strengthen causal inference compared to cross-sectional analyses, the current body of evidence remains insufficient to clearly establish temporal and causal relationships between PCB exposure and cognitive trajectories. Future research should prioritize prospective cohort designs with repeated measures of both PCB exposure and cognitive function to better characterize exposure timing, trajectories of cognitive change, and potential critical windows of vulnerability.

4.4. Neurocognitive assessments

The variability in cognitive assessment across studies is not simply a matter of different tests being used, but rather how similar tests are interpreted and applied. Several studies included widely used instruments such as the WCST, TMT, and Stroop test, yet these were not consistently assigned to the same cognitive domains. For example, in the Upper Hudson River cohort, TMT and Stroop were used to assess executive function (Fitzgerald et al., 2012, 2008), whereas in other studies TMT was treated as a measure of attention (Seegal et al., 2013). Similarly, the DSST, which was the only cognitive measure in NHANES analyses, has been variably interpreted as reflecting processing speed and motor ability or even memory and executive function (Bouchard et al., 2014; Przybyla et al., 2017). This inconsistency complicates interpretation, as studies may appear to assess different domains while relying on overlapping measures.

Differences in how cognitive outcomes were constructed further contribute to heterogeneity across studies. Many studies administered multi-test neuropsychological batteries but reported results at the level of individual test scores—for example, presenting findings separately for CVLT subtests, WMS components, or executive measures such as TMT and Stroop—rather than combining them into a single summary measure (Fimm et al., 2017; Fitzgerald et al., 2012, 2008; Haase et al., 2009; Schantz et al., 2001; Seegal et al., 2010; Tanner et al., 2020). In contrast, other studies used composite scores or global indices, such as DSST-based cognition in NHANES, the 3MS in CSHA, or composite cognitive scores in HCHS/SOL and the Three-City study (Bouchard et al., 2014; Lefèvre-Arbogast et al., 2025; Medehouenou et al., 2019; Parada et al., 2024).

These approaches are not directly comparable. Composite measures may improve statistical efficiency but can obscure domain-specific effects, whereas test-level analyses provide greater specificity but may be more variable and less comparable across studies. In the current literature, studies using domain-specific memory measures, particularly CVLT and WMS-based assessments, more often reported associations in environmentally exposed populations (Fitzgerald et al., 2008; Schantz et al., 2001), whereas studies relying on global or composite measures frequently reported null or inconsistent findings (Lefèvre-Arbogast et al., 2025; Medehouenou et al., 2014). However, these differences likely reflect variation in measurement approach as much as differences in underlying neurotoxicity.

4.5. Other limitations

Notably, several recent studies published after the initial search period were incorporated through an updated literature review. These newer studies, including longitudinal analyses with repeated cognitive assessments, strengthen the evidence base but also highlight ongoing inconsistencies across populations and study designs.

Despite these findings, several limitations should be acknowledged. Many of the included studies are cross-sectional, limiting causal inference. Differences in exposure assessment methods, including the use of lipid-adjusted vs. wet-weight PCB measurements, may also contribute to inconsistencies across studies. Additionally, some studies did not control for potential confounders such as socioeconomic status, diet, or comorbid health conditions, which may influence cognitive performance. The lack of longitudinal studies remains a critical gap, as cross-sectional studies can only provide snapshots of cognitive function at a single time point and cannot establish whether PCB exposure accelerates cognitive decline over time. Further longitudinal research with standardized exposure assessment and rigorous adjustment for confounders is needed to better characterize the long-term impact of PCB exposure on cognitive decline and dementia risk.

Other sources of bias were a recurring concern across studies, with 18 out of 24 studies rated as probably high or high in this domain. These elevated ratings largely stemmed from the retrospective nature of the study designs, which relied on historical PCB exposures in occupational or environmentally exposed populations. In such contexts, researchers faced inherent limitations in documenting exposures, controlling for confounders, and standardizing data collection procedures. Many studies lacked detailed reporting on participant recruitment, eligibility screening, and adherence to analytic protocols, making it difficult to assess consistency and transparency. Additionally, the long latency between exposure and outcome measurement compounded the risk of bias. Although these studies provide valuable insights into long-term effects of PCB exposure, their retrospective design constrained the ability to fully mitigate bias, particularly in domains that rely on prospective control, documentation, and standardized quality assurance procedures.

In conclusion, the evidence suggests that PCB exposure is associated with deficits in multiple cognitive domains, with higher exposure levels linked to greater neurocognitive impairment. However, the lack of consistency in cognitive testing methodologies and the dearth of longitudinal studies limit comparability across studies and weaken causal inferences. Future research should adopt standardized neurocognitive assessments, conduct longer follow-up studies, and establish clearer domain definitions to facilitate cross-study comparisons and improve our understanding of PCB-related neurotoxicity. Given the persistence of PCBs in the environment, continued monitoring of exposed populations and further investigation into potential intervention strategies to mitigate neurotoxic effects remain crucial. Finally, it is important to note that this review did not include gray literature and may therefore be subject to publication bias, particularly if studies reporting null associations are less likely to be published. Inclusion of gray literature in future reviews may provide a more comprehensive assessment of the evidence base.

Figure 3.

Timeline of studies and cohorts on polychlorinated biphenyls (PCBs) and neurocognitive outcomes.