Persistent organic pollutants and cognitive decline among middle-aged or older adults in the Hispanic Community Health Study/Study of Latinos

Abstract

Persistent organic pollutants may negatively impact cognition; however, associations between persistent organic pollutants and changes in cognition among United States Hispanic/Latino adults have not been investigated. Herein, we examined the associations between 33 persistent organic pollutants and cognitive changes among 1837 Hispanic/Latino adults. At baseline (2008–2011; Visit 1), participants provided biospecimens in which we measured levels of 5 persistent pesticides or pesticide metabolites, 4 polybrominated diphenyl ethers and 2,2’,4,4’,5,5’-hexabromobiphenyl, and 24 polychlorinated biphenyls. At Visit 1 and again at Visit 2 (2015–2018), a battery of neurocognitive tests was administered which included the Brief-Spanish English Verbal Learning Test, Word Fluency Test, and Digit Symbol Substitution Test. To estimate the adjusted associations between changes in cognition and each POP, we used linear regression for survey data. Each doubling in plasma levels of polychlorinated biphenyls 146, 178, 194, 199/206, and 209 was associated with steeper declines in global cognition (βs range:−0.053 to −0.061) with stronger associations for the Brief-Spanish English Verbal Learning Test. Persistent organic pollutants, in particular polychlorinated biphenyls, were associated with declines in cognition over 7 years and may be a concern for Hispanic/Latino adults.

1. Introduction

Dementia results in substantial morbidity and mortality among adults 65 years or older in the United States (US). In 2023, close to seven million Americans were living with Alzheimer’s Disease (AD), the most prevalent type of dementia (Alzheimer’s Association, 2023), and AD ranks as the seventh leading cause of death (Xu et al., 2022). The number of people who will be diagnosed with dementia is projected to increase over the next 20 years due to the aging US population (Ortman et al., 2014), particularly among Hispanic/Latino adults. Given the increasing prevalence of dementia and the limited treatment options available, studies are needed that identify modifiable factors of preclinical or early-stage dementia including cognitive decline.

Environmental chemical exposures such as persistent organic pollutants (POPs) have been hypothesized to contribute to the pathology of neurodegenerative diseases including dementia (Weiss, 2011). POPs are synthetic organic chemicals known for their capacity to resist environmental degradation and bioaccumulate in adipose tissues (Qing Li et al., 2006). Groups of POPs include the organochlorine pesticides (OCPs), which were used extensively [>1.3 billion pounds of dichlorodiphenyltrichloroethane (US EPA, 2016)] in agriculture and vector control until the 1960s (Agency for Toxic Substances and Disease Registry (ATSDR), 2002a); polychlorinated biphenyls (PCBs), industrial chemicals used widely [>1.5 billion pounds (US Department of Commerce, 2024)] in electrical equipment and in hydraulic fluids, lubricants, and plasticizers from 1929 to 1979 (Agency for Toxic Substances and Disease Registry (ATSDR), 2002b); and polybrominated diphenyl ethers (PBDEs), chemicals used as flame-retardant in manufactured products since the 1970s until recently (Agency for Toxic Substances and Disease Registry (ATSDR), 2017). Following the 2001 Stockholm Convention, the commercial production and use or application of listed POPs is prohibited in most developed countries including the US. However, POPs exposure in humans continues through dietary sources including the consumption of animal fats, although measurable levels of POP residues are found in most foods including fruits and vegetables, dairy products, and pastries (Schafer and Kegley, 2002) and through environmental exposure through contaminated landfills and bodies of water (Banzhaf et al., 2019).

Several studies have explored biomarker measures of OCPs and PCBs in association with cognition. These previous studies, most of which were conducted in non-Hispanic White populations, reported lower performance on neurocognitive tests in association with higher levels of OCPs (K.-S. Kim et al., 2015; S.-A. Kim et al., 2015) and PCBs (Bouchard et al., 2014; Fitzgerald et al., 2008; Haase et al., 2009; Schantz et al., 2001) in cross-sectional studies, and with cognitive impairment (Lee et al., 2016) and cognitive decline (Medehouenou et al., 2019) in association with higher levels of OCPs and PCBs in case-control or prospective cohort studies. No studies, however, investigated POPs in relation to cognitive decline specifically among adults of Hispanic/Latino heritage, despite documented evidence of elevated exposure levels in this population (Nguyen et al., 2020), and no study has examined PBDEs in relation to cognition.

Understanding the relationships between POPs and cognitive decline will help to identify potential risk factors explaining underlaying causes of high burden of dementia among US Hispanic/Latino adults. Thus, our primary objective in this study was to examine the associations between POPs including OCPs, PCBs, and PBDEs measured in baseline plasma samples and 7-year changes in cognition among Hispanic/Latino adults who participated in the population-based Hispanic Community Health Study/Study of Latinos (HCHS/SOL).

2. Methods

2.1. Study design

The HCHS/SOL is an observational study designed to investigate clinical, social, and behavioral determinants of diseases among Hispanic/Latino adults from diverse backgrounds (Sorlie et al., 2010). HCHS/SOL participant identification and recruitment procedures have been previously published (LaVange et al., 2010). In brief, households in four urban US cities (Chicago, Illinois; Bronx, New York; Miami, Florida; San Diego, California) were selected in two stages using probability-based sampling to increase the likelihood of selecting Hispanic/Latino households. The HCHS/SOL sampling strategy ensures that estimates generated from HCHS/SOL are representative of Hispanic/Latino adults residing in the cities from which they were recruited. HCHS/SOL baseline (Visit 1, V1) procedures took place in 2008–2011, during which 16,415 Hispanic/Latino adults were enrolled, completed interviewer-administered questionnaires, and underwent a clinical exam with blood draws for laboratory analyses. Participants who were 45 years or older (n=9714) completed a neurocognitive battery at V1 and of these 6377 completed the same battery seven years later in 2015–2018 at a second clinical assessment (Visit 2, V2) as part of the Study of Latinos–Investigation of Neurocognitive Aging (SOL-INCA) Ancillary Study (González et al., 2019). Of the 6377 who participated in SOL-INCA, 1972 also participated in the HCHS/SOL Persistent Organic Pollutants, Endogenous Hormones and Diabetes in Latinos Study ancillary study designed to investigate the associations among POPs, hormones, and diabetes among HCHS/SOL men and postmenopausal women without diabetes between the ages of 45–74. For each ancillary study, survey weights were generated which account for subsample participant enrollment into each study and for attrition from V1 to V2. In the present study, of the 1972 participants who participated in both ancillary studies, we excluded 83 who were missing one or more of the neurocognitive tests, 14 who scored less than 3 correct answers (n=13) or who were missing (n=1) the Six-Item Cognitive Impairment Screener (SIS) (Callahan et al., 2002), and 38 who had a history of stroke or transient ischemic attack. The analytic sample comprised 1837 participants after applying these exclusions. In Supplemental eTable 1 we provide a comparison of unweighted baseline characteristics for those included (n=1837) and not included (n=4540) in this study from among those who completed the neurocognitive assessments at V1 and V2 (n=6377).

The procedures of the HCHS/SOL and its ancillary studies were approved by the Institutional Review Boards (IRBs) of the participating institutions. HCHS/SOL participants provided written informed consent prior to participating in the study. The Centers for Disease Control and Prevention (CDC) laboratory’s involvement measuring deidentified human specimens did not constitute engagement in human-subjects research.

2.2. Neurocognitive assessments

Descriptions of the battery of neurocognitive tests are provided in Supplemental eTable 2, with additional details provided in a previous publication (González et al., 2019). In brief, Bicultural/bilingual technicians administered the following tests: the SIS (Callahan et al., 2002); the Brief-Spanish English Verbal Learning Test (B-SEVLT), which includes a measure of episodic learning consisting of three learning trials (B-SEVLT-Sum) and a measure of verbal memory consisting of a delayed recall trial (B-SEVLT-Recall) (González et al., 2001); the Word Fluency Test (WF), a measure of verbal functioning (Lezak, 2004); and the Digit Symbol Substitution Test (DSS), a measure of attention, processing speed, and executive function (Wechsler, 1981).

To facilitate comparisons across neurocognitive tests, we standardized all outcomes using survey linear regression: the scores of each cognitive test at V2 were regressed on scores at V1 with adjustment for the number of days between assessments. For each, test we calculated the difference between the V2 score and the predicted V2 score and divided by the root mean squared error [i.e., (Score_V2-Predicted Score_V2)/RMSE]. Consistent with our previous work, we also calculated a measure global cognitive change, which was the arithmetic mean of the individual standardized test scores (González et al., 2019).

2.3. Laboratory analysis

Levels of eight persistent pesticides or their metabolites [hexachlorobenzene, β-hexachlorocyclohexane (β-HCCH), oxychlordane, trans-nonachlor, p,p’-dichlorodiphenyldichloroethylene (DDE), o,p’-dichlorodiphenyltrichloroethane (o,p’-DDT), p,p’-DDT, and Mirex]; 11 brominated flame retardants [PBDEs 17, 28, 47, 85, 99, 100, 153, 154, 183, 209, and 2,2’,4,4’5,5’-hexabromobiphenyl (PBB-153)]; and 26 PCBs (28, 66, 74, 99, 105, 114, 118, 138, 158, 146, 153, 156, 157, 157, 167, 170, 178, 180, 183, 187, 189, 194, 196, 203, 199, 206, and 209) were assayed in stored V1 plasma samples at the Division of Laboratory Sciences in the National Center for Environmental Health at the CDC. The methodology for the extraction and quantification of the POPs, including quality assurance (QA) and quality control (QC) procedures, has been previously published (Centers for Disease Control and Prevention, 2006; Jones et al., 2012; Sjödin et al., 2004). In brief, 2 mL plasma samples were fortified with internal standards and extracted by automated liquid/liquid extraction using a Gilson 215 two liquid handler (Gilson Inc.; Middleton, WI). Gas chromatography isotope dilution high resolution mass spectrometry was used for final analytical determination of the target analytes. Co-extracted lipids were removed using a two layered silica/silica-H₂SO₄ column packed in a 3cc SPE cartridge, automated using the Rapid Trace SPE workstation (Biotage; Uppsala, Sweden). The concentrations of total triglycerides and total cholesterol were determined using test kits from Roche Diagnostics Corp. (Indianapolis, IN). Final determinations were made on a Hitachi 912 Chemistry Analyzer (Hitachi; Tokyo, Japan). Plasma levels of all POPs (Supplemental eTable 3) were reported as background corrected; every 24-sample set included three blanks. Of the 45 analytes measured, 12 including three persistent pesticides (o,p’-DDT, p,p’-DDT, and Mirex), six brominated flame retardants (PBDEs 17, 85, 99, 154, 183, and 209), and three PCBs (PCBs 66, 114, and 189), were detected in <80 % of participants and were not considered further. For the remaining 33 analytes, levels below the limit of detection (LOD) were imputed as the LOD divided by the square root of two (Hornung and Reed, 1990) and adjusted for lipids by dividing the analyte levels by total lipids (Phillips et al., 1989). Lipid-adjusted and imputed plasma POP levels were categorized into tertiles (see Supplemental eTable 4 for tertile cutpoints) for categorical analyses and log₂-tranformed for continuous analyses.

2.4. Covariates

Potential confounders of the associations between plasma levels of POPs and cognitive function included the following demographic, behavioral, and clinical risk factors for dementia: age in years; sex (male or female); Hispanic/Latino heritage (Mexican, Puerto Rican, Cuban, Dominican, Central American, South American, or More than one/other heritage); country of birth [born in mainland 50 US States or in the District of Columbia (DC) or not born in mainland 50 US States/DC]; education (<high school, high school graduate, or >high school); body mass index (BMI) in kg/m²; smoking status (never smoker, former smoker, or current smoker); diet quality quantified using the Alternative Healthy Eating Index-2010 (AHEI-2010) and its individual components (Wang et al., 2014); and physical activity assessed using the Global Physical Activity Questionnaire (GPAQ; low physical activity, moderate physical activity, or high moderate-to-vigorous physical activity, MVPA) (Armstrong and Bull, 2006); pre-diabetes status (yes or no) defined by serum glucose levels adjusted for fasting time, post-oral glucose tolerance glucose levels, if available, hemoglobin A1c (HbA1c), self-reported diagnosis of diabetes, and use of antidiabetes medication(s); and hypertension (yes or no) defined using measured systolic and diastolic blood ≥140/90 mmHg or use of hypertension medication(s).

2.5. Statistical analysis

We characterized the target population at V1 using weighted descriptive statistics. We used survey-weighted means with standard errors (SEs) for continuous covariates and survey-weighted percentages for categorical covariates. We examined Spearman correlations between the 33 lipid-adjusted POPs (Supplemental eFigure 1), and used survey linear regression to estimate the associations [Betas (βs) or tertiles (T1 (reference), T2, or T3) and 95 % confidence intervals (CIs)] between each POP and longitudinal standardized neurocognitive test change scores.

All statistical analyses were conducted using SAS V.9.4 (SAS Institute Inc., Cary, NC, USA) and R V.4.3 (R Foundation for Statistical Computing, Vienna, Austria). For all analyses, we used a complete-case analysis and considered a two-tailed P-value of <0.05 as statistically significant.

3. Results

The HCHS/SOL target population demographics and characteristics at V1 are reported in Table 1. The average age was 55.8 (standard error, SE=0.4) years, over half (55.3 %) were women, and over a third (34.6 %) were of Mexican heritage followed by a quarter (26.1 %) who were of Cuban heritage. The majority (90.1 %) were born outside of the 50 US states/DC. One-fifth (19.9 %) were current smokers at baseline.

Table 1.

HCHS/SOL target population baseline characteristics (unweighted n=1837).

Characteristic	%weighted
Age in years, weighted mean [SE]	55.8 [0.4]
Sex
Female	55.3
Male	44.7
Hispanic/Latino heritage
Mexican	34.6
Cuban	26.1
Puerto Rican	14.8
Dominican	9.4
Central American	7.3
South American	5.0
Other	2.7
Country of birth
Not US-born	90.1
US-born	9.9
Education
<High school	35.5
High school graduate	20.0
>High school	44.5
Body mass index in kg/m2, mean [SE]a	28.8 [0.1]
Smoking
Never	53.1
Former	27.0
Current	19.9
AHEI-2010, mean [SE]b	50.7 [0.3]
Red or processed meats, mean [SE]b	3.9 [0.1]
Vegetables, mean [SE]b	4.2 [0.1]
Whole fruits, mean [SE]b	3.2 [0.1]
Whole grains, mean [SE]b	2.7 [0.1]
Physical activityc
Low MVPA	9.8
Moderate MVPA	43.5
High MVPA	46.7
Pre-diabetes
No	45.0
Yes	55.0
Hypertension
No	64.1
Yes	35.9

Hispanic Community Health Study/Study of Latinos (HCHS/SOL) participants completed all baseline assessments in 2008–2011.

AHEI-2010, Alternative Healthy Eating Index 2010; BMI, Body Mass Index; MVPA, moderate-to-vigorous physical activity; US, United States; SE, standard error

^aDerived from measured height and weight.

^bAssessed using the alternative healthy eating index (AHEI-2010). The AHEI- 2010 ranges from 0 (worst) to 110 (best). The individual AHEI-2010 components range from 0 (worst) to 10 (best).

^cAssessed using the Global Physical Activity Questionnaire (GPAQ).

3.1. Changes in global cognition

The results of the associations between levels of each POP and longitudinal changes in the global cognition score over the 7-year follow-up are reported in Table 2. Compared to the lowest tertiles, the highest tertiles of PCBs 146, 178, 194, and 209 were associated with the steepest declines in global cognition with β estimates below −0.10, although the confidence intervals for most of these associations included the null value and were not statistically significant. In analyses examining log₂-transformed levels, there were statistically significant steeper declines in global cognition in association with each doubling in levels of PCB146 (β=−0.053, 95 %CI=−0.103, −0.002); PCB178 (β=−0.055, 95 %CI=−0.104, −0.005); PCB194 (β=−0.052, 95 %CI=−0.102, −0.003); PCB199 (β=−0.053, 95 %CI=−0.100, −0.007); PCB206 (β=−0.053, 95 %CI=−0.106, −0.001); and PCB209 (β=−0.061, 95 %CI=−0.122, −0.001). Levels of the persistent pesticides and PBDEs were not strongly associated with changes in global cognition scores, with the exception of β-HCCH which was associated with a non-statistically significant increase (β_Log2=0.033, 95 %CI=−0.001, 0.067) in global cognition change scores.

Table 2.

Associations [betas (βs) and 95 % confidence intervals (CIs) from survey-weighted linear regression] between plasma levels of persistent organic pollutants and longitudinal changes in global cognition.

Analyte	nunweighted	Tertile 2 vs. Tertile 1 β (95 % CI)a	Tertile 3 vs. Tertile 1 β (95 % CI)a	Log2(Analyte) β (95 % CI)a
Persistent pesticides
Hexachlorobenzene	1767	0.020 (−0.085, 0.126)	0.076 (−0.066, 0.217)	0.019 (−0.024, 0.062)
β-hexachlorocyclohexane	1782	0.052 (−0.046, 0.151)	0.094 (−0.029, 0.218)	0.033 (−0.001, 0.067)
Oxychlordane	1798	−0.100 (−0.202, 0.003)	0.026 (−0.087, 0.138)	−0.001 (−0.058, 0.056)
Trans-nonachlor	1783	−0.022 (−0.132, 0.089)	−0.015 (−0.122, 0.092)	−0.022 (−0.078, 0.034)
p,p’-DDE	1795	0.037 (−0.059, 0.134)	0.038 (−0.083, 0.159)	0.012 (−0.020, 0.044)
Brominated Flame Retardants
PBDE-28	1801	0.084 (−0.025, 0.193)	0.006 (−0.110, 0.121)	−0.002 (−0.034, 0.030)
PBDE-47	1803	0.049 (−0.064, 0.162)	−0.003 (−0.108, 0.101)	−0.001 (−0.026, 0.024)
PBDE-100	1804	−0.061 (−0.185, 0.063)	−0.055 (−0.147, 0.038)	−0.010 (−0.035, 0.016)
PBDE-153	1803	−0.033 (−0.158, 0.092)	−0.044 (−0.150, 0.062)	−0.012 (−0.040, 0.016)
PBB-153	1802	−0.061 (−0.178, 0.055)	−0.053 (−0.190, 0.083)	−0.031 (−0.078, 0.016)
Polychlorinated biphenyls
PCB28	1770	0.083 (−0.022, 0.187)	0.002 (−0.108, 0.113)	0.000 (−0.032, 0.032)
PCB74	1782	−0.030 (−0.146, 0.085)	0.020 (−0.093, 0.134)	−0.005 (−0.045, 0.036)
PCB99	1771	0.014 (−0.099, 0.128)	−0.003 (−0.106, 0.099)	−0.013 (−0.049, 0.023)
PCB105	1771	−0.001 (−0.097, 0.095)	−0.060 (−0.168, 0.048)	−0.015 (−0.050, 0.021)
PCB118	1777	−0.041 (−0.138, 0.056)	−0.068 (−0.183, 0.047)	−0.019 (−0.057, 0.019)
PCB138+PCB158b	1778	0.079 (−0.016, 0.174)	0.057 (−0.047, 0.162)	−0.011 (−0.055, 0.033)
PCB146	1777	−0.045 (−0.142, 0.053)	−0.108 (−0.239, 0.024)	−0.053 (−0.103, −0.002)*
PCB153	1802	−0.004 (−0.112, 0.103)	0.021 (−0.100, 0.141)	−0.026 (−0.076, 0.024)
PCB156	1800	0.037 (−0.068, 0.142)	0.084 (−0.037, 0.205)	−0.004 (−0.054, 0.045)
PCB157	1802	0.076 (−0.024, 0.176)	0.081 (−0.033, 0.196)	0.010 (−0.039, 0.059)
PCB167	1791	−0.025 (−0.130, 0.079)	0.014 (−0.093, 0.120)	−0.028 (−0.073, 0.016)
PCB170	1799	0.030 (−0.081, 0.140)	0.066 (−0.046, 0.177)	−0.035 (−0.098, 0.027)
PCB178	1763	−0.144 (−0.260, −0.028)*	−0.107 (−0.232, 0.018)	−0.055 (−0.104, −0.005)*
PCB180	1764	−0.050 (−0.147, 0.046)	−0.004 (−0.107, 0.099)	−0.036 (−0.092, 0.019)
PCB183	1772	−0.011 (−0.121, 0.099)	−0.038 (−0.152, 0.076)	−0.038 (−0.087, 0.011)
PCB187	1784	−0.009 (−0.124, 0.106)	−0.053 (−0.172, 0.067)	−0.042 (−0.094, 0.010)
PCB194	1792	−0.074 (−0.179, 0.031)	−0.103 (−0.215, 0.009)	−0.052 (−0.102, −0.003)*
PCB196+PCB203b	1791	−0.094 (−0.205, 0.016)	−0.073 (−0.192, 0.047)	−0.045 (−0.093, 0.004)
PCB199	1797	−0.047 (−0.146, 0.053)	−0.050 (−0.164, 0.063)	−0.053 (−0.100, −0.007)*
PCB206	1776	−0.113 (−0.223, −0.002)*	−0.079 (−0.183, 0.026)	−0.053 (−0.106, −0.001)*
PCB209	1798	−0.104 (−0.209, 0.001)	−0.108 (−0.225, 0.008)	−0.061 (−0.122, −0.001)*

Hispanic Community Health Study/Study of Latinos (HCHS/SOL) participants completed all baseline assessments in 2008–2011 (Visit 1, V1) and all follow-up assessments in 2015–2018 (Visit 2, V2). All regression analyses used HCHS/SOL survey weights.

CI, confidence interval; DDE, dichlorodiphenyldichloroethylene; PBB, poly-brominated biphenyl; PBDE, polybrominated diphenyl ether; PCB, polychlorinated biphenyl.

^aModel is adjusted for age (continuous in years), sex (female or male), Hispanic/ Latino heritage (Mexican, Puerto Rican, Cuban, Dominican, Central American, South American, or other), Nativity (US-born or not US-born), education (<high school, high school graduate, or >high school), body mass index (BMI; continuous in kg/m²), smoking status (never, former, current), Alternative Healthy Eating Index-2010 (AHEI-2010; continuous); physical activity (low, moderate, or high moderate-to-vigorous physical activity); pre-diabetes (no or yes); and hypertension (no or yes);

^bThese analytes are quantitated and reported as the sum of their concentrations because they coelute during GC/MS analysis.

^*P < 0.05

3.2. Changes in neurocognitive test scores

The results examining the associations between plasma levels of each POP and longitudinal changes in each of the neurocognitive tests over the 7-year follow-up are reported in Tables 3 and 4 for B-SEVLT-Sum and B-SEVLT Recall, respectively, and Supplemental eTables 5 and 6 for Word Fluency and DSS, respectively. For B-SEVLT-Sum (Table 3), T3 versus T1 of PCB209 was associated with a 0.214 (95 %CI=−0.393, −0.034) standard deviation (SD) decline in change scores, and a doubling in levels of PCB99 and PCB178 were associated with 0.065 (95 %CI=−0.123, −0.007) SD and 0.076 (95 %CI=−0.149, −0.004) SD declines in change scores, respectively. A doubling in levels of β-HCCH was associated with a 0.071 (95 %CI=0.014, 0.128) SD increase in change scores.

Table 3.

Associations [betas (βs) and 95 % confidence intervals (CIs) from survey-weighted linear regression] between plasma levels of persistent organic pollutants and longitudinal changes in performance on the Brief-Spanish English Verbal Learning Test learning trials (B-SEVLT-Sum).

Analyte	nunweighted	Tertile 2 vs. Tertile 1 β (95 % CI)a	Tertile 3 vs. Tertile 1 β (95 % CI)a	Log2(Analyte) β (95 % CI)a
Persistent pesticides
Hexachlorobenzene	1767	0.035 (−0.129, 0.199)	0.188 (−0.053, 0.430)	0.062 (−0.014, 0.138)
β-hexachlorocyclohexane	1782	0.067 (−0.095, 0.229)	0.171 (−0.022, 0.364)	0.071 (0.014, 0.128)*
Oxychlordane	1798	0.003 (−0.162, 0.167)	0.184 (−0.027, 0.395)	0.063 (−0.028, 0.154)
Trans-nonachlor	1783	−0.066 (−0.237, 0.105)	−0.002 (−0.178, 0.175)	0.006 (−0.083, 0.096)
p,p’-DDE	1795	0.013 (−0.136, 0.162)	0.081 (−0.111, 0.273)	0.036 (−0.010, 0.082)
Brominated Flame Retardants
PBDE-28	1801	0.100 (−0.073, 0.273)	0.035 (−0.150, 0.220)	0.011 (−0.050, 0.072)
PBDE-47	1803	0.031 (−0.147, 0.209)	−0.028 (−0.197, 0.141)	0.004 (−0.044, 0.052)
PBDE-100	1804	−0.074 (−0.257, 0.109)	−0.081 (−0.252, 0.090)	−0.008 (−0.055, 0.039)
PBDE-153	1803	−0.023 (−0.213, 0.167)	−0.093 (−0.261, 0.074)	−0.027 (−0.072, 0.018)
PBB-153	1802	−0.014 (−0.220, 0.192)	−0.015 (−0.252, 0.221)	−0.021 (−0.087, 0.045)
Polychlorinated biphenyls
PCB28	1770	0.170 (0.019, 0.321)*	0.001 (−0.173, 0.176)	−0.022 (−0.082, 0.037)
PCB74	1782	−0.128 (−0.292, 0.036)	−0.026 (−0.215, 0.163)	−0.038 (−0.106, 0.030)
PCB99	1771	−0.114 (−0.292, 0.065)	−0.074 (−0.246, 0.098)	−0.065 (−0.123, −0.007)*
PCB105	1771	−0.026 (−0.180, 0.128)	−0.100 (−0.258, 0.058)	−0.032 (−0.089, 0.025)
PCB118	1777	−0.140 (−0.297, 0.017)	−0.108 (−0.281, 0.066)	−0.050 (−0.112, 0.012)
PCB138+PCB158b	1778	0.010 (−0.158, 0.179)	0.044 (−0.141, 0.230)	−0.037 (−0.111, 0.037)
PCB146	1777	−0.160 (−0.340, 0.020)	−0.107 (−0.315, 0.102)	−0.076 (−0.149, −0.004)*
PCB153	1802	−0.082 (−0.260, 0.095)	0.020 (−0.187, 0.227)	−0.058 (−0.140, 0.023)
PCB156	1800	−0.028 (−0.206, 0.150)	0.103 (−0.118, 0.325)	−0.018 (−0.111, 0.075)
PCB157	1802	0.013 (−0.150, 0.175)	0.138 (−0.070, 0.346)	0.008 (−0.083, 0.099)
PCB167	1791	−0.116 (−0.279, 0.047)	0.035 (−0.143, 0.212)	−0.040 (−0.115, 0.035)
PCB170	1799	−0.051 (−0.234, 0.132)	0.111 (−0.104, 0.326)	−0.036 (−0.143, 0.070)
PCB178	1763	−0.212 (−0.395, −0.029)*	−0.093 (−0.302, 0.116)	−0.053 (−0.128, 0.021)
PCB180	1764	−0.121 (−0.303, 0.061)	0.023 (−0.190, 0.237)	−0.043 (−0.140, 0.055)
PCB183	1772	−0.010 (−0.186, 0.165)	−0.032 (−0.229, 0.165)	−0.066 (−0.132, 0.001)
PCB187	1784	−0.054 (−0.238, 0.129)	−0.084 (−0.291, 0.124)	−0.066 (−0.140, 0.007)
PCB194	1792	−0.097 (−0.292, 0.098)	−0.134 (−0.356, 0.087)	−0.058 (−0.150, 0.033)
PCB196+PCB203b	1791	−0.124 (−0.298, 0.050)	−0.117 (−0.322, 0.088)	−0.064 (−0.145, 0.017)
PCB199	1797	−0.085 (−0.267, 0.097)	−0.073 (−0.290, 0.144)	−0.057 (−0.137, 0.023)
PCB206	1776	−0.144 (−0.320, 0.032)	−0.040 (−0.252, 0.173)	−0.055 (−0.132, 0.023)
PCB209	1798	−0.281 (−0.437, −0.124)*	−0.214 (−0.393, −0.034)*	−0.060 (−0.140, 0.021)

Hispanic Community Health Study/Study of Latinos (HCHS/SOL) participants completed all baseline assessments in 2008–2011 (Visit 1, V1) and all follow-up assessments in 2015–2018 (Visit 2, V2). All regression analyses used HCHS/SOL survey weights.

CI, confidence interval; DDE, dichlorodiphenyldichloroethylene; PBB, poly-brominated biphenyl; PBDE, polybrominated diphenyl ether; PCB, polychlorinated biphenyl.

^aModel is adjusted for age (continuous in years), sex (female or male), Hispanic/ Latino heritage (Mexican, Puerto Rican, Cuban, Dominican, Central American, South American, or other), Nativity (US-born or not US-born), education (<high school, high school graduate, or >high school), body mass index (BMI; continuous in kg/m²), smoking status (never, former, current), Alternative Healthy Eating Index-2010 (AHEI-2010; continuous); physical activity (low, moderate, or high moderate-to-vigorous physical activity); pre-diabetes (no or yes); and hypertension (no or yes);

^bThese analytes are quantitated and reported as the sum of their concentrations because they coelute during GC/MS analysis.

^*P < 0.05

Table 4.

Associations [betas (βs) and 95 % confidence intervals (CIs) from survey-weighted linear regression] between plasma levels of persistent organic pollutants and longitudinal changes in performance on the Brief-Spanish English Verbal Learning Test recall (B-SEVLT-Recall).

Analyte	nunweighted	Tertile 2 vs. Tertile 1 β (95 % CI)a	Tertile 3 vs. Tertile 1 β (95 % CI)a	Log2(Analyte) β (95 % CI)a
Persistent pesticides
Hexachlorobenzene	1767	0.122 (−0.036, 0.281)	0.009 (−0.230, 0.248)	−0.040 (−0.133, 0.053)
β-hexachlorocyclohexane	1782	0.125 (−0.049, 0.299)	0.106 (−0.095, 0.307)	0.030 (−0.028, 0.088)
Oxychlordane	1798	−0.125 (−0.320, 0.070)	−0.072 (−0.311, 0.167)	−0.019 (−0.115, 0.076)
Trans-nonachlor	1783	0.063 (−0.104, 0.230)	−0.020 (−0.215, 0.174)	−0.031 (−0.120, 0.058)
p,p’-DDE	1795	0.173 (0.006, 0.341)*	0.133 (−0.085, 0.351)	0.017 (−0.032, 0.067)
Brominated Flame Retardants
PBDE-28	1801	0.016 (−0.169, 0.202)	−0.048 (−0.238, 0.143)	−0.022 (−0.076, 0.031)
PBDE-47	1803	0.010 (−0.180, 0.201)	−0.094 (−0.288, 0.099)	−0.019 (−0.062, 0.024)
PBDE-100	1804	−0.112 (−0.314, 0.090)	−0.118 (−0.289, 0.052)	−0.035 (−0.080, 0.009)
PBDE-153	1803	−0.047 (−0.241, 0.146)	−0.121 (−0.309, 0.067)	−0.039 (−0.087, 0.009)
PBB-153	1802	−0.165 (−0.347, 0.017)	−0.165 (−0.381, 0.051)	−0.066 (−0.130, −0.001)*
PCB28	1770	0.087 (−0.109, 0.283)	0.099 (−0.091, 0.289)	0.044 (−0.014, 0.103)
PCB74	1782	−0.043 (−0.229, 0.142)	0.053 (−0.140, 0.245)	0.011 (−0.060, 0.081)
PCB99	1771	0.024 (−0.147, 0.194)	−0.016 (−0.195, 0.164)	−0.006 (−0.073, 0.061)
PCB105	1771	0.072 (−0.102, 0.245)	−0.047 (−0.244, 0.150)	0.000 (−0.066, 0.067)
PCB118	1777	0.002 (−0.157, 0.162)	−0.045 (−0.241, 0.151)	−0.003 (−0.074, 0.068)
PCB138+PCB158b	1778	0.074 (−0.104, 0.251)	0.051 (−0.144, 0.246)	−0.018 (−0.095, 0.059)
PCB146	1777	−0.082 (−0.260, 0.096)	−0.151 (−0.371, 0.070)	−0.065 (−0.148, 0.018)
PCB153	1802	0.013 (−0.180, 0.206)	0.007 (−0.202, 0.215)	−0.040 (−0.127, 0.047)
PCB156	1800	0.115 (−0.082, 0.313)	0.061 (−0.165, 0.287)	−0.018 (−0.113, 0.078)
PCB157	1802	0.121 (−0.063, 0.305)	0.104 (−0.110, 0.317)	0.001 (−0.094, 0.096)
PCB167	1791	0.032 (−0.145, 0.210)	0.032 (−0.146, 0.210)	−0.030 (−0.106, 0.046)
PCB170	1799	0.083 (−0.127, 0.293)	0.010 (−0.209, 0.229)	−0.066 (−0.170, 0.038)
PCB178	1763	−0.128 (−0.327, 0.071)	−0.184 (−0.400, 0.031)	−0.071 (−0.151, 0.010)
PCB180	1764	−0.038 (−0.231, 0.155)	−0.103 (−0.316, 0.111)	−0.070 (−0.170, 0.030)
PCB183	1772	−0.012 (−0.202, 0.179)	−0.055 (−0.251, 0.140)	−0.047 (−0.124, 0.030)
PCB187	1784	−0.024 (−0.229, 0.180)	−0.131 (−0.351, 0.088)	−0.063 (−0.155, 0.029)
PCB194	1792	−0.105 (−0.296, 0.086)	−0.264 (−0.484, −0.045)*	−0.104 (−0.198, −0.009)*
PCB196+PCB203b	1791	−0.112 (−0.306, 0.082)	−0.156 (−0.370, 0.058)	−0.085 (−0.176, 0.005)
PCB199	1797	−0.073 (−0.263, 0.117)	−0.136 (−0.360, 0.088)	−0.088 (−0.173, −0.002)*
PCB206	1776	−0.206 (−0.402, −0.011)*	−0.141 (−0.351, 0.069)	−0.064 (−0.153, 0.025)
PCB209	1798	−0.060 (−0.262, 0.143)	−0.118 (−0.328, 0.093)	−0.047 (−0.139, 0.045)

Hispanic Community Health Study/Study of Latinos (HCHS/SOL) participants completed all baseline assessments in 2008–2011 (Visit 1, V1) and all follow-up assessments in 2015–2018 (Visit 2, V2). All regression analyses used HCHS/SOL survey weights.

CI, confidence interval; DDE, dichlorodiphenyldichloroethylene; PBB, poly-brominated biphenyl; PBDE, polybrominated diphenyl ether; PCB, polychlorinated biphenyl.

^aModel is adjusted for age (continuous in years), sex (female or male), Hispanic/ Latino heritage (Mexican, Puerto Rican, Cuban, Dominican, Central American, South American, or other), Nativity (US-born or not US-born), education (<high school, high school graduate, or >high school), body mass index (BMI; continuous in kg/m²), smoking status (never, former, current), Alternative Healthy Eating Index-2010 (AHEI-2010; continuous); physical activity (low, moderate, or high moderate-to-vigorous physical activity); pre-diabetes (no or yes); and hypertension (no or yes);

^bThese analytes are quantitated and reported as the sum of their concentrations because they coelute during GC/MS analysis.

^*P < 0.05

For B-SEVLT-Recall (Table 4), T3 versus T1 of PCB194 was associated with a 0.264 (95 %CI=−0.484, −0.045) SD decline in change scores. Additionally, each doubling in levels of PBB-153, PCB194, and PCB199 were associated with 0.066 (95 %CI=−0.130, −0.001), 0.104 (95 %CI=−0.198, −0.009), and 0.088 (95 %CI=−0.173, −0.002) declines in B-SEVLT-Recall change scores. For word fluency (Supplemental eTable 5), only PBDE-153 was associated with a statistically significant increase in WF scores (β_Log2=0.049, 95 %CI=−0.005, 0.094), and for DSS (Supplemental eTable 6), PCB178 (β_Log2=−0.072, 95 %CI=−0.142, −0.001), PCB199 (β_Log2=−0.074, 95 %CI=−0.142, −0.007), and PCB206 (β_Log2=−0.092, 95 %CI=−0.164, −0.021) were associated with statistically significantly decreases in DSS scores.

4. Discussion

Our primary objective in this study was to examine whether plasma levels of 33 POPs were associated with longitudinal changes in cognitive function among middle-aged or older Hispanic/Latino adults from diverse backgrounds. Some PCBs, and in particular highly chlorinated PCBs, were associated with declines in cognitive function over the 7-year follow-up, while the persistent pesticides and the brominated flame retardants were not. When we examined the POPs in association with each of the individual tests, the PCBs were more strongly associated with changes in scores on the B-SEVLT-Sum or the B-SEVLT-Recall, measures of episodic verbal learning and memory than with global cognition or with the other neurocognitive tests. Most POPs were weakly associated with lower verbal fluency scores.

Several studies have examined POP biomarkers in relation to cognitive function, but results have been mixed results. However, to our knowledge, we are the first to examine POP biomarkers and cognitive decline among Hispanic/Latino adults. In contrast to our findings reported here indicating no association between OCPs and cognitive decline, previous studies using National Health and Nutrition Examination Survey (NHANES) data, a US nationally representative sample, reported worse executive cognitive functioning (i.e., DSS) with increasing serum levels of six OCPs including DDT, trans-nonachlor, and β-HCCH (K.-S. Kim et al., 2015; S.-A. Kim et al., 2015). Our findings on PCBs, however, are consistent with previous studies, though most have used a cross-sectional study design. A study of 180 non-Hispanic White adults from Michigan and reported inverse associations between PCB levels and memory and learning (Schantz et al., 2001). The second study included 253 predominantly (97 %) non-Hispanic White adults from the Hudson River in New York, and reported that PCBs 105, 118, 138, 170, 180, and 194 were associated with poorer memory performance (Fitzgerald et al., 2008). Using data from NHANES, Bouchard and colleagues reported lower executive function in association with increasing levels of dioxin-like PCBs (PCBs 74, 118, 156, 126, and 169), but only among older adults (Bouchard et al., 2014). In the last cross-sectional study, among 277 Native American adults, higher levels of total PCBs predicted worse executive and motor functioning and memory (Haase et al., 2009). Additionally, PCBs were also examined in relation to cognition in a retrospective cohort study (Lin et al., 2008) and in a prospective cohort study (Medehouenou et al., 2019). In the retrospective study of Taiwanese adults, women with higher levels of PCBs and polychlorinated dibenzofurans showed declines in verbal and nonverbal attention, concentration, and verbal memory, although these associations were not observed among men (Wechsler, 1981). At least one prospective study has examined PCBs in relation to cognition. In the Canadian Study of Health and Aging, PCBs 118, 153, 156, 163 and the organochlorine pesticide p,p’-DDT and the DDT metabolite p,p’-DDE were associated with declines in mental status scores (Medehouenou et al., 2019). To our knowledge, we are the first to report on PBDEs and PPB-153 in association with cognitive function, though they have been hypothesized to be neurotoxicants (Fonnum and Mariussen, 2009). Altogether, previous studies and the findings reported here, conducted across various racial or ethnic groups, support the hypothesis that PCBs may be neurotoxic and may thus contribute to cognitive decline, and potentially dementia.

The biological mechanisms underlying the associations between PCBs cognition are not fully elucidated; however, several hypotheses have been proposed (Fonnum and Mariussen, 2009). First, PCBs can disrupt the dopaminergic and serotonergic systems resulting in decreased neurotransmitter concentrations in parts of the brain including the frontal cortex, caudate nucleus, and striatum (Seegal et al., 1986). As dopamine and serotonin are key neurotransmitters, disruption of these systems may result in dysfunction across multiple cognitive domains including memory, attention, learning, and decision making (Bhatia et al., 2023). Second, PCBs are known endocrine disrupting chemicals with the ability to mimic hormones (Wolff et al., 1997). Hormones are vital in the development and function of the central nervous system. Estrogens, and in particular estradiol, have important actions in the hippocampus directly through nuclear and membrane estrogen receptor (ER)-mediated mechanisms and indirectly through exertion of neurotrophic effects (Luine, 2014). Therefore, estrogenic PCBs could potentially positively affect cognition, while anti-estrogenic and anti-androgenic PCBs could negatively affect cognition. While we did not explore whether these relationships vary by sex, this may be an important consideration for future studies. Third, although the evidence is mixed, PCBs can also disrupt thyroid hormone homeostasis (Hagmar, 2003), alterations of which impact cognition. Lastly, PCBs could impact cognition through effects on oxidative balance (Liu et al., 2020) or energy metabolism including glucose homeostasis (Tan et al., 2023), and thus be potentially mediated by chronic conditions including hypertension, diabetes, or other metabolic conditions. Future studies, however, will need to be designed with careful attention to the temporality of the exposures, mediators, and outcomes. Importantly, however, given that humans are exposed to POPs as a mixture with varying biological activities and POPs have high correlations with each other and with other non-persistent chemicals, identifying which chemicals are responsible for a certain biological effect in epidemiological studies remains a challenge. Nonetheless, future studies should consider chemical mixtures, as the POPs may have synergistic or antagonistic effects and such studies may help characterize the importance of each POP, and the net effect of the chemical mixture.

Our study had multiple strengths including the prospective design, the use of a relatively large and well-phenotyped cohort of diverse Hispanic/Latino adults, the use of a cognitive battery of neuropsychological tests and assessment of multiple cognitive domains, and biomarker assessments of POPs. However, several limitations should be noted. We had only one measure of POPs and therefore are unable to examine changes in POP levels; however, POPs have long half-lives ranging from years to decades (Ashraf, 2017), and so a single measurement may reflect long-time exposure levels. Future studies should consider the potential of these environmental chemicals to vary over time. Also, plasma POP levels do not provide information about the source of exposure, which will be important to understand to inform primary prevention efforts. Multiple comparisons could have resulted in false-positive associations; however, adjustment of multiple comparisons, which reduces Type I errors at the expense of Type II errors, in hypothesis driven epidemiologic research is not recommended (Rothman, 2014). Lastly, the effect sizes reported here are modest. However, at the population level, the negative cognitive effects of these persistent pollutants would be substantial. With longer follow-up and as this cohort ages, however, the effects of POPs on cognition may be more apparent.

5. Conclusions

We found that PCBs were associated with worse cognitive function over time in a representative sample of Hispanic/Latino adults followed for over 7 years. Given that PCBs have pleiotropic effects, additional epidemiological studies further investigating these associations are warranted, particularly among Hispanic/Latino adults who are expected to face a dramatic increase in the burden of dementia in coming years (Matthews et al., 2019). Although POPs are banned in many countries, their continued use ensures continued exposure both directly through exposure during use or application of these chemicals or indirectly through diet and environmental sources. This may be particularly relevant for immigrant populations who may be exposed years or decades prior to immigrating to the US. Last, although there are few strategies to reduce the body burden of POPs, these results may be important for identifying those who may be at increased risk of cognitive decline, and may thus benefit from interventions designed to preserve or promote cognitive function. However, primary prevention, though challenging given the ubiquitous nature of POPs, may be the best approach for reducing the impact of these chemicals on brain health.

Data availability

Data are available through an approved request with the HCHS/SOL Study