Plasma concentrations of Lipophilic Persistent Organic Pollutants and Glucose Homeostasis in Youth Populations

Abstract

Background:

Exposure to lipophilic persistent organic pollutants (POPs) is ubiquitous. POPs are metabolic disrupting chemicals and are potentially diabetogenic.

Methods:

Using a multi-cohort study including overweight adolescents from the Study of Latino Adolescents at Risk (SOLAR, N=301, 2001–2012) and young adults from the Southern California Children’s Health Study (CHS, N=135, 2014–2018), we examined associations of POPs and risk factors for type 2 diabetes. SOLAR participants underwent annual visits for a median of 2.2 years and CHS participants performed a single visit, during which a two-hour oral glucose tolerance test was performed. Linear mixed models were used to examine associations between plasma concentrations of POPs [4,4’-dichlorodiphenyldichloroethylene (4,4’-DDE), hexachlorobenzene (HCB), PCBs-153, 138, 118, 180 and PBDEs-154, 153, 100, 85, 47] and changes in glucose homeostasis across age and pubertal stage.

Results:

In SOLAR, exposure to HCB, PCB-118, and PBDE-153 was associated with dysregulated glucose metabolism. For example, each two-fold increase in HCB was associated with approximately 2 mg/dL higher glucose concentrations at 30 minutes (p=0.001), 45 minutes (p=0.0006), and 60 minutes (p=0.03) post glucose challenge. Compared to individuals with low levels of PCB-118, individuals with high levels exhibited a 4.7 mg/dL (p = 0.02) higher glucose concentration at 15 minutes and a 3.6 mg/dL (p = 0.01) higher glucose concentration at 30 minutes. The effects observed with exposure to organochlorine compounds were independent of pubertal stages. PBDE-153 was associated with the development of dysregulated glucose metabolism beginning in late puberty. At Tanner stage 4, exposure to PBDE-153 was associated with a 12.7 mg/dL higher 60-minute glucose concentration (p = 0.009) and a 16.1 mg*dl⁻¹*hr⁻¹ higher glucose AUC (p = 0.01). These associations persisted at Tanner 5. In CHS, PBDE-153 and total PBDE were associated with similar increases in glucose concentrations.

Conclusion:

Our results suggest that childhood exposure to lipophilic POPs is associated with dysregulated glucose metabolism.

1. Introduction

Incidence of type 2 diabetes mellites (T2D) in youth populations has rapidly increased (Mayer-Davis et al., 2017). Between 2002 and 2011, the incidence of youth onset T2D in the United States increased from 9 cases per 100,000 to 12.5 cases per 100,000, and youth from health-disparate populations are disproportionately affected by this epidemic (Mayer-Davis et al., 2017). Although lifestyle factors such as diet and sedentary behavior play an important role in T2D risk, these factors cannot fully explain the current epidemic (Davis et al., 2007). Therefore, there is need to identify modifiable risk factors for T2D to develop public health measures to address this epidemic.

Lipophilic persistent organic pollutants (POPs) include a wide range of environmental pollutants, such as organochlorine pesticides (OCPs), polychlorinated biphenyls (PCBs), and polybrominated diphenyl ethers (PBDEs) (Kalantzi et al., 2004). These chemicals persist in the environment and bioaccumulate in adipose tissue of animals and humans (Kalantzi et al., 2004). It is because of these properties that despite the long-term discontinuation of their use low-level exposure is ubiquitous in most populations (Pumarega et al., 2016; Sjodin et al., 2014). Furthermore, due to the biological persistence and ability to bioaccumulate humans are exposed to not just a single POP but to mixtures of POPs which raises concerns for not only individual chemical exposures but also for chemical mixtures (Lee et al., 2018).

POPs are well-recognized as metabolic disrupting chemicals (Lee et al., 2014). Growing evidence suggests that the rising production and environmental presence of these chemicals may contribute to pathogenesis of T2D (Jaacks and Staimez, 2015; Kuo et al., 2013; Ngwa et al., 2015). In vitro and in vivo studies have shown that POPs can alter adipocyte function (Tung et al., 2014), decrease insulin sensitivity (Ruzzin et al., 2010), induce mitochondrial dysfunction (Ko et al., 2020b; Liu et al., 2017), and decrease oxidative phosphorylation (Ko et al., 2020a; Ko et al., 2020b; Marroqui et al., 2018), all of which are known risk factors for T2D (Prasun, 2020). Epidemiological studies have established associations between POPs and T2D (Beard et al., 2003; Montgomery et al., 2008). However, these studies have been performed in adults and are either cross-sectional or do not have detailed measures of glucose homeostasis outcomes (Everett et al., 2007; Everett and Matheson, 2010; Lee et al., 2006; Lim et al., 2008).

Although exposure to POPs may increase risk of T2D, evidence in children is lacking. No previous study has examined the individual or joint effects of multiple POPs exposures on longitudinal alterations of glucose metabolism and insulin secretion prior to disease development, a critical period in which interventions have the potential to stop or delay progression to T2D. Therefore, we aimed to characterize the association between POPs exposure burden and longitudinal changes in glucose homeostasis, insulin sensitivity, and beta cell function in two independent cohorts of youth and young adults.

2. Material and Methods

2.1. Study Populations

2.1.1. SOLAR cohort

The Study of Latino Adolescents at Risk of Type 2 Diabetes (SOLAR) is a prospective study including 328 Hispanic children recruited in two waves between 2001–2012 (Alderete et al., 2017). Participants underwent annual clinical visits at the University of Southern California General Clinical Research Center or Clinical Trials Unit for a median of 2.2 years. At the first visit, children were between 8–13 years old, had a BMI ≥ 85^th percentile based on their age and sex, were free of type 1 or type 2 diabetes, and were not taking medications known to influence insulin/glucose metabolism. All participants had a direct familial history of T2D. Participants were included in the study if there was sufficient fasting plasma sample volume from their 1^st or 2^nd visit, resulting in 301 participants (Figure S1.A.). Ethics approval for this study was provided by the University of Southern California Institutional Review Board. Prior to participation, written informed assent/consent were obtained from participants and their guardians.

2.1.2. CHS cohort

To test the generalizability of associations observed in SOLAR, samples were analyzed in a cohort of 136 young adults from the Meta-AIR study (Kim et al., 2019). Meta-Air participants were selected from the Southern California Children’s Health Study (CHS) cohort based upon a history of overweight or obesity in the 9^th or 10^th grade (ages 14–15), defined as an age and sex specific BMI ≥ 85^th percentile (McConnell et al., 2015). Meta-AIR participants completed a single clinical visit at the Clinical Trials Unit or the Diabetes and Obesity Research Institute (DORI) at the University of Southern California between 2014 and 2018. At the time of the clinical visit, participants were 18–23 years old, free from type 1 or type 2 diabetes, and not taking any medications known to influence insulin/glucose metabolism. One participant with fasting glucose > 200 mg/dL was removed from analysis, resulting in a total population size of 135 participants (Figure S1.B.). Ethics approval was provided by the University of Southern California Institutional Review Board, and all participants gave written informed consent.

2.2. Clinical Outcomes

At each clinical visit, participants from both the SOLAR and CHS cohort performed a 2 hour oral glucose tolerance test (OGTT). After baseline blood sampling, participants consumed a glucose load of the lesser of either 1.75 g per kg body mass or 75 grams. In the SOLAR cohort, blood was sampled at 15-, 30-, 45-, 60-, and 120-minutes post glucose challenge; in the CHS cohort, samples were collected at 30-, 60-, 90-, and 120-minutes post challenge. Samples were centrifuged to obtain plasma. Plasma obtained from blood collection tubes containing potassium oxalate and sodium fluoride were used to analyze glucose concentrations (mg/dL), while plasma obtained from sodium heparin tubes were frozen at −80°C for batch analysis of insulin concentrations (μU/mL). Fasting whole blood samples were used to measure glycated hemoglobin (HbA1c; (%)). Additional plasma aliquots were frozen at −80°C for future analysis.

Primary outcomes for this study included fasting glucose concentrations (time point 0 from the OGTT) and two hour OGTT glucose concentrations, as these are clinical markers used in the diagnosis of type 2 diabetes (Craig et al., 2009; Inzucchi, 2012). Impaired fasting glucose is defined as fasting glucose > 100 mg/dL, and impaired glucose tolerance is defined as a 2 hour glucose >140 mg/dL. HbA1c, another common clinical test for the diagnosis of diabetes, was also examined as an outcome. In adults, HbA1c values between 5.7–6.4 are considered prediabetes and values greater than 6.5 are considered diabetes (Inzucchi, 2012), although the diagnostic criteria of HbA1c in adolescents is less well defined (Craig et al., 2009). In addition to these outcomes, blood glucose and insulin responses from the OGTT were used to calculate glucose area under the curve (AUC; mg/dL * min) and insulin AUC (μU/mL * min) as markers of glucose homeostasis using the trapezoidal method (Le Floch et al., 1990). The homeostatic model assessment (HOMA-IR) was calculated to estimate insulin resistance using the formula (Glu₀ ∗ lns₀)/405 (Matthews et al., 1985), the Matsuda index (l𝑆l_{𝑀𝑎𝑡su𝑑𝑎}) was calculated to estimate insulin sensitivity using the formula $10,000/\sqrt{Gl{u}_0*In{s}_0*\overline{Glu}*\overline{Ins}}$ (Matthews et al., 1985), and the insulinogenic index (IGI) was calculated to estimate β-cell function using the equation (Ins₃₀ − Ins₀)/(Glu₃₀ − Glu₀) (Jensen et al., 2002), where Glu₀ and lns₀ are fasting glucose concentrations (in mg/dL) and insulin concentrations (in μU/mL), Glu₃₀ and lns₃₀ are glucose (in mg/dL) and insulin concentrations (in μU/mL) at 30 minutes, and $\overline{Glu}$ and $\overline{Ins}$ are the weighted average glucose (in mg/dL/min) and insulin concentration (in μU/mL/min) over the entire OGTT. Prior to analysis, clinical outcomes were assessed for significant deviations from a normal or log-normal distribution using q-q plots. In general, outcomes related to glucose concentrations were normally distributed (including fasting and OGTT glucose concentrations, glucose AUC, and HbA1C), while outcomes related to insulin concentrations were log-normally distributed (including insulin AUC, HOMA-IR, ISI_Matsuda, and IGI).

2.3. Plasma Lipophilic POPs Concentrations

Concentrations of lipophilic POPs were measured in plasma samples by gas-chromatography coupled with high resolution mass spectrometry (GC-HRMS), as described previously (Hu et al., 2021). In the SOLAR cohort, plasma samples from the first or second visit were used to determine POPs concentrations. In CHS, plasma samples from the single clinical visit were used to determine POPs concentrations. Briefly, 16 ¹³C labeled chemical standards, each with 99% isotope enrichment, were spiked at final concentration of 1 ng/mL to assist chemical identification and quantification. Following this, 50 μL formic acid (Emprove^® Essential DAC, Sigma-Aldrich) was added to 150 μL plasma, which was immediately followed by addition of 200 μL hexane – ethyl acetate (2:1 v/v, ≥99% pure, Sigma-Aldrich). The chilled mixture was shaken vigorously on ice using multi-tube vortexer (VWR VX-2500) for 1 h and centrifuged at 1000 g, 4 °C for 10 min. The organic supernatant was transferred to a new tube with 25 mg MgSO₄ (≥99.99% pure, Sigma-Aldrich) and vortexed vigorously to remove water. After 10 min centrifugation at 1000 g, 80 μL of the final supernatant was analyzed. Samples were analyzed with three injections using GC-HRMS with a Thermo Scientific Q Exactive GC hybrid quadrupole Orbitrap mass spectrometer with 2 μL per injection. Data were collected from 3 to 24.37 min with positive electron ionization (EI) mode (+70 eV), scanning from m/z 85.0000 to 850.0000 with a resolution of 60,000. In addition to plasma samples, National Institute of Standards & Technology [NIST] Standard Reference Materials (SRM) 1958 and SRM-1957 were analyzed in every batch of 20 samples to support quality control and quantification using reference standardization, a protocol that was previously validated in high-resolution mass spectrometry data for LC (Go et al., 2015; Liu et al., 2020) and GC (Hu et al., 2021) methods.

Raw data was extracted using XCMS (Smith et al., 2006) and POPs were identified by comparing accurate mass m/z and retention time to standards run on the same platform using an error threshold of ±5 ppm m/z and 30 s retention time, which was further validated by co-elution of ¹³C isotopic standards. Quantification of targeted lipophilic POPs were performed by batch-wise reference standardization using NIST SRM-1958 and NIST SRM-1957. Peak intensities of the spectral m/z features were used to calculate mass fractions based on SRM certified values, and then median summarized to get the final concentration of each chemical. The limit of detection (LOD) in ng/mL for the corresponding lipophilic POP compounds is 0.005 for HCB, 0.005 for 4’4-DDE, 0.008 for PBDE-100, 0.015 for PBDE-153, 0.047 for PBDE-154, 0.013 for PBDE-85, 0.001 for PBDE-47, 0.004 for PCB-118, 0.013 for PCB-138, 0.001 for PCB-153, and 0.002 for PCB-180.

For lipophilic POPs detected in greater than 80% of samples in both cohorts (including HCB, 4’4-DDE, PBDE-47, and PBDE-154), non-detects were imputed using a Bayesian framework (see supplemental materials section 1.1 for details). Additionally, these POPs were log2 transformed prior to analysis to satisfy modeling assumptions. Lipophilic POPs which were detected in < 80% of samples in both cohorts (PBDE-85, 100, and 153, as well as PCB-118, 138, 153, and 180) were modeled as dichotomous (detect/non detect) variables. The correlation coefficients for the lipophilic POPs exposure variables were calculated using Pearson correlation for continuous to continuous, polyserial correlation for continuous-dichotomous, and tetrachoric correlation for dichotomous- dichotomous. Since previous research has shown that chemical congeners can have the same or similar biological mechanisms (Costa et al., 2014; Safe et al., 1985), we calculated the sum of organochlorine compounds (HCB and 4,4-DDE), the number of detected PCB compounds, and the number of detected PBDE compounds as variables to estimate the overall burden of each class of lipophilic POP.

2.4. Covariates

In both cohorts, height (m) and weight (kg) were measured at each visit and were used to calculate body mass index (BMI) as kg/m². Body composition was determined by dual-energy x-ray absorptiometry scan (DXA). Participants and their guardians also completed questionnaires detailing health history, familial health history, and sociodemographic characteristics. Since glucose homeostasis, insulin sensitivity, and β-cell function exhibit substantial changes throughout puberty (Kelly et al., 2011), we hypothesized that POPs concentration may have different effects depending on pubertal stage. In SOLAR, a physician performed a physical examination for determination of Tanner stage (Marshall and Tanner 1969, 1970). Tanner stages are an objective measure of developmental stage of secondary sex characteristics, where Tanner 1 is pre-puberty, Tanner 2 is early-puberty, Tanner 3 is puberty, Tanner 4 is late-puberty, and Tanner 5 is post-puberty. Socioeconomic status (SES) was assessed in SOLAR using a modified version of the Hollingshead Four-Factor Index (Alderete et al., 2017) and in CHS using parental education (Kim et al., 2019).

2.5. Statistical Methods

For SOLAR, linear mixed effects models were used to examine associations between childhood levels of lipophilic POPs and changes in glucose concentrations and homeostasis across visits allowing for separate intercepts for each individual. All models included adjustments for time varying (Age, BMI) and time invariant (SES, study wave) covariates. Significance of associations between lipophilic POPs concentrations and outcomes were assessed using likelihood-ratio tests between the final selected model with POPs exposure (as described below) to the null model (the model with no POPs exposure).

To examine whether POPs exposure may have differential effects depending on pubertal stage, we began by testing a Concentration*Tanner stage interaction term using a likelihood ratio test. If significant, the interaction term was retained in the model and effect estimates were calculated stratifying by tanner stage. When the Concentration*Tanner stage term was not significant, the term was removed, and a single effect estimate for exposure over all tanner stages was calculated. Effect estimates are provided as regression coefficients and 95% confidence intervals for each exposure estimate. For log2 transformed variables, these coefficients represent the change in the outcome per doubling of concentration. For dichotomous variables, these coefficients represent the difference in the outcome between detected and non-detected levels of the exposure. One of the outcomes, insulinogenic index, was log transformed prior to analysis in order to meet modeling assumptions.

In CHS, we used linear models to examine associations between childhood levels of lipophilic POPS and changes in glucose concentrations and homeostasis. Models included adjustments for Age, BMI, and SES, and Hispanic ethnicity.

To examine the overall effect of exposure to lipophilic POPs on glucose regulation, we performed a quantile-based g-computation analysis (Keil et al., 2020). This method estimates a single effect estimate (ψ) for the overall mixture by quantizing exposures. Quantizing ensures that all exposures are on the same scale, and the effect estimate for the mixture is interpreted as the change in the outcome associated with increasing each exposure by 1 quantized unit. In the current study, seven of the eleven exposures were detected in <66% of participants for at least one cohort. To incorporate all exposures into the mixture model, we defined categories of exposure as high versus low for all exposures. For exposures detected in >75% of participants, high versus low was defined as above or below the median, and for exposures detected in <75% of participants, high versus low was defined as above or below the limit of detection. In our analysis, the mixture effect ψ is interpreted as the estimated change in the outcome when increasing all exposures from low to high levels. Since quantile based g-computation is not designed to address complex longitudinal data (Keil et al., 2020), in the SOLAR cohort we performed this analysis using cross-sectional data from the first visit in late puberty (tanner stage 4). Tanner stage 4 was chosen because we observed the most consistent associations between exposure to single lipophilic POPs and glucose dysregulation at this developmental stage and because it had the largest sample size (n = 135) for a cross sectional analysis stratified by tanner stage. Models were adjusted for sex, age, and SES. In SOLAR models were also adjusted for study wave, and in CHS, models were adjusted for Hispanic ethnicity.

To address the issue of multiple comparisons, we corrected based on a false discovery rate (FDR) of 0.05. This correction was performed within each chemical, correcting for a total of 11 comparisons (Benjamini and Hochberg, 1995). See supplementary materials section 1.2 for additional details regarding statistical methods.

2.6. Sensitivity Analysis

To assess the robustness of our findings to various modeling assumptions, we performed the following sensitivity analyses. First, in the main analysis we treated sex as a potential confounder by including it as a variable in the statistical models. However, since POPs may have differential metabolic effects in males versus females (Le Magueresse-Battistoni, 2020), we reran our main analysis stratified by sex to examine effect modification. Second, since POPs accumulate in adipose tissue, differences in fat mass may alter the effects of POPs on glucose homeostasis outcomes (Valvi et al., 2020). Therefore, we performed a sensitivity analysis stratifying by high and low baseline percent fat mass. Thresholds for high and low percent fat mass were defined as the sex specific median percent fat mass at baseline. Data on percent fat mass was available in 289 participants from SOLAR and 86 participants from CHS. Third, since BMI may be an effect modifier in the association between POPs concentrations and type 2 diabetes endpoints (Kuo et al., 2013), we reran our main analysis excluding BMI from the statistical models.

3. Results

3.1. Characteristics of the Study Population

Descriptive statistics of the study population at baseline are provided in Table 1. In total, 301 participants from the SOLAR cohort and 135 participants from the CHS cohort were included in this study with an approximately equal sex distribution between the two cohorts (SOLAR 42.2% female; CHS 44.4% female). Average age of adolescents in SOLAR was 11.3 years (SD, ± 1.7); average age for young adults in CHS was 19.4 years (SD, ± 1.3). SOLAR adolescents included in the analysis were of Hispanic ethnicity. Young adults were more ethnically diverse than the participants from the SOLAR cohort, with 58% identifying as Hispanic. The average BMI of participants from SOLAR and CHS was 28.2 and 29.5 kg/m², respectively.

Table 1.

General characteristics at baseline of the SOLAR (N=301) and CHS (N=135) cohorts

	SOLAR (N=301)	CHS (N=135)
Female, n (%)	127 (42.2)	60 (44.4)
Age, years	11.3 ± 1.7	19.4 ± 1.3
BMI, kg/m2	28.2 ± 5.8	29.5 ± 4.4
Hispanic, n (%)	301 (100%)	78 (58%)
Number of visits per subject	3.8 ± 2.6	1
Puberty status (% per stage)	32/62/6	-
Fasting glucose, mg/dL	89.9 ± 6.6	90.3 ± 7.8
2-Hour glucose, mg/dL	124 ± 18 (299)	119 ± 27 (123)
Glucose AUC, mg/dL * min	266 ± 32 (296)	262 ± 45 (123)
Insulin AUC, μU/mL * min	305 ± 189 (291)	226 ± 151 (85)
HOMA-IR	3.3 ± 2.3 (293)	2.9 ± 4.0 (85)
ISIMatsuda	1.4 ± 0.8 (288)	1.9 ± 1.3 (85)
IGI	3.4 ± 2.3 (291)	2.4 ± 2.3 (85)
HbA1c, %	5.5 ± 0.3 (168)	5.2 ± 0.3 (135)

Data presented as mean ± SD (n). Pubertal status presented as pre-puberty (Tanner stage 1), puberty (Tanner stages 2–4), and post puberty (Tanner stage 5). AUC: Area under the curve. HOMA-IR: Homeostatic model of insulin resistance; ISI_Matsuda: Matsuda Insulin Sensitivity Analysis; IGI: Insulinogenic Index (index of β-cell function).

3.2. Characteristics of lipophilic POPs concentrations

A summary of POPs concentrations is shown in Table 2. Concentrations of POPs split by recruitment wave are provided in Table S1 and compared to the National Health and Nutrition Examination Survey (NHANES 2003–2004) in Table S2. We observed a large percentage (> 30%) of non-detected concentrations for individual PCBs (Table 2); whereas each OCP was detected in > 90% of participants. There was a wide range in the percentage of non-detection for the various PBDE congers (1.7 to 76.7%, Table 2). Correlation of lipophilic POPs is shown in Figure 1. POPs compounds were largely uncorrelated. In SOLAR, the median observed correlation coefficient was 0.05 (Min: −0.10; Max: 0.47). In CHS, the median observed correlation coefficient was −0.02 (Min: −0.24; Max: 0.29).

Table 2.

Concentrations (ng/mL) of lipophilic persistent pollutants in SOLAR and CHS

	SOLAR (N=301)					CHS (N=135)
Chemical	GM (95% CI)	50th percentile	75th percentile	90th percentile	% below LOD	GM (95% CI)	50th percentile	75th percentile	90th percentile	% below LOD
4,4’-DDE	0.62 (0.53, 0.72	0.72	0.96	2.15	3.7	0.26 (0.19, 0.36	0.65	0.8	0.97	7.4
HCB	0.02 (0.02, 0.02)	0.02	0.03	0.06	3.7	0.02 (0.02, 0.03)	0.03	0.03	0.04	0
PBDE-47	0.18 (0.16, 0.20)	0.19	0.32	0.5	1.7	0.06 (0.05, 0.08)	0.09	0.16	0.22	7.4
PBDE-100	*	0.05	0.06	0.07	28.2	*	0.05	0.06	0.07	44.4
PBDE-153	*	0	0	0.08	76.7	*	0	0.07	0.1	71.9
PBDE-154	0.05 (0.04, 0.05)	0.05	0.08	0.13	3.3	0.05 (0.04, 0.06)	0.07	0.1	0.14	14.1
PBDE-85	*	0	0	0.05	75.1	*	0	0.03	0.92	67.4
PCB-118	*	0.02	0.02	0.03	41.9	*	0.02	0.02	0.03	28.9
PCB-138	*	0.04	0.04	0.04	36.5	*	0	0.04	0.04	50.4
PCB-153	*	0.04	0.05	0.06	39.9	*	0.03	0.05	0.06	36.3
PCB-180	*	0	0.05	0.06	50.8	*	0.04	0.05	0.06	48.9

Figure 1.

Correlation of lipophilic compounds in the SOLAR and CHS cohorts. Correlations between two continuous variables were calculated using the Pearson correlation coefficient; correlations between two binary variables were calculated with polychoric correlation coefficients; and correlations between a continuous and binomial variable were calculated with polyserial correlation coefficients. * indicates chemicals detected in less than 75% of samples and which were analyzed as binary (detect/non-detect); ** indicates correlations which were statistically significant based upon a p<0.05.

3.3. Associations between Lipophilic Persistent Organic Pollutant Concentrations and Type 2 Diabetes Risk Factors, Insulin Sensitivity and Beta Cell Function

3.3.1. Organochlorine compounds.—

In youth, concentrations of HCB and PCB-118 were independently associated with impaired glucose homeostasis, including higher glucose levels during OGTT independent of age or Tanner stage.

In SOLAR, each two-fold increase in HCB was associated with approximately 2 mg/dL higher glucose concentrations at 30 minutes (95% CI: [0.84, 3.5]; p = 0.001), 45 minutes (95% CI: [1.3, 4.6]; p = 0.0006), and 60 minutes (95% CI: [0.22, 3.9]; p = 0.03) post glucose challenge. Youth concentrations of HCB were also associated with a 3.0 mg*dL⁻¹*hour⁻¹ (95% CI: [0.68, 5.4]; p =0.01) increase in glucose AUC per doubling of HCB (Figure 2). After adjusting for multiple comparisons, HCB remained significantly associated with glucose at 30 minutes (q = 0.009), glucose at 45 minutes (q = 0.007), and glucose AUC (q = 0.045).

Figure 2.

Point estimates and 95% confidence intervals for associations of (A) Hexachlorobenzene (HCB); (B) polybrominated diphenyl ether (PBDE)-153, and (C) Total PBDEs with risk factors for type 2 diabetes in SOLAR and CHS. Associations of HCB with glucose concentrations during the Oral Glucose Tolerance Test were not time dependent, so panel A represents the change in glucose concentration and glucose AUC per doubling of HCB across all developmental stages for SOLAR. Panels B and C represents the difference in outcome for (B) detected vs. non-detected levels of PBDE-153 and (C) the 80^th vs. 20^th percentile of total PBDE exposure across developmental stages. * indicates chemicals detected in less than 75% of samples and which were analyzed as binary (detect/non-detect). Notes: OGTT Time point 0 corresponds to fasting glucose; Glu: Glucose; AUC: Area under the curve; ISI Matsuda: Matsuda Insulin Sensitivity Index.

PCB-118 exposure was also associated with higher glucose levels in the SOLAR cohort. Compared to individuals with low levels of PCB-118, individuals with high levels of PCB-118 exhibited a 4.7 mg/dL (95% CI: [0.8, 9.6]; p = 0.02) higher glucose concentration at 15 minutes and a 3.6 mg/dL (95% CI: [0.9, 6.3]; p = 0.01) higher glucose concentration at 30 minutes post glucose challenge (Figure S2). These associations did not remain significant after correcting for multiple comparisons. In CHS, there were no associations between HCB or PCB-118 and glucose outcomes. No organochlorine compounds showed consistent associations between cohorts (Figure S2).

3.3.2. Brominated compounds.—

Youth exposure to PBDEs was associated with glucose dysregulation, including higher OGTT glucose levels and lower insulin sensitivity, and these relationships were tanner stage dependent (Figure S3). In youth from SOLAR, PBDE-153 was associated with higher OGTT glucose concentrations and PBDE-100 was associated with lower insulin sensitivity; however, these effects were only present at late and post puberty (Figure S3). Specifically, at Tanner stage 4, exposure to PBDE-153 was associated with a 12.7 mg/dL higher 60-minute glucose concentration (95% CI: [3.26, 22.1], p = 0.009) and a 16.1 mg*dl⁻¹*hr⁻¹ higher glucose AUC (95% CI: [3.9, 28.2]; p = 0.01). At Tanner stage 5, these associations persisted, where youth exposure to PBDE-153 was associated with 9 mg/dL higher glucose at 60 minutes (95% CI: [0.6, 17.4]; p = 0.04), 6.7 mg/dL higher glucose at 2 hours (95% CI: [0.46, 13.0]; p = 0.04), and 11.3 mg*dl⁻¹*hr⁻¹ higher glucose AUC (95% CI: [0.05, 22.1]; p = 0.04). Additionally, at tanner stage 5, PBDE-100 exposure was associated with −0.40 lower ISI_Matsuda index (95% CI: [−0.63, −0.16]; p = 0.001; Figure 2).

In young adults from the CHS, there were similar associations between the total number of detected PBDEs and glucose homeostasis (Figure S3). In this cohort, individuals with >4 detected PBDE compounds, compared to individuals with <2 detected PBDEs, had 8.5 mg/dL higher 60-minute glucose (95% CI: [1.5, 15.4]; p = 0.02), 7.6 mg/dL higher 90-minute glucose (95% CI: [1.8, 13.4]; p = 0.01), and 10.6 mg*dl⁻¹*hr⁻¹ higher glucose AUC (95% CI: [2.1, 19.1]; p = 0.017). In both the SOLAR and the CHS cohorts, no associations between brominated compounds and glucose homeostasis outcomes were statistically significant using a false discovery rate of 0.05.

3.4. Associations of Lipophilic POPs mixtures with Type 2 Diabetes Risk Factors

In adolescents from the SOLAR cohort, exposure to the mixture of lipophilic POPs was associated with altered responses to an oral glucose tolerance test in late puberty (Figure 3). Specifically, increasing all lipophilic POPs exposures from low to high was associated with a 23 mg/dL higher glucose concentration 15 minutes post glucose challenge (95% CI: [2, 43]; p = 0.03), although this association was not significant after correcting for multiple comparisons. The top four compounds driving this association were PCB-118, PBDE-153, PBDE-47, and PBDE-154, which together accounted for 85% of the observed positive association. In CHS, increasing all lipophilic POPs exposures from low to high was associated with a 45 mg/dL higher glucose concentration 60 minutes post glucose challenge, though this association did not meet the threshold for statistical significance (95% CI: [−1, 91]; p = 0.06). The top four compounds driving this association included PBDE-85, PBDE-47, PBDE-100, and PBDE-154, which together accounted for 76% of the observed positive association. In both the SOLAR and CHS cohorts, no other glucose homeostasis parameters, including Insulin AUC, HOMA-IR, ISI_Matsuda, IGI, or HbA1c were associated with the lipophilic POPs mixture.

Figure 3.

Associations of a mixture of eleven lipophilic POPs with glucose responses to an oral glucose tolerance test in adolescents in late puberty (SOLAR) and young adults (CHS). Effect estimates (ψ) and 95% confidence intervals indicate the change in glucose levels (A) or glucose area under the curve (B) when increasing all lipophilic POPs levels from low to high. For POPs detected in >75% of participants, low vs. high levels were determined by the median; for POPs levels detected in <75% of participants, low vs. high levels were defined as detected versus non-detected. Glu: Glucose; AUC: Area under the curve.

3.5. Sensitivity Analysis

Adolescent males and females from SOLAR both exhibited similar positive associations for HCB and OGTT glucose outcomes (Figure S4, Figure S5). Adolescent males from SOLAR exhibited similar associations of PCB-118, PBDE-100, and PBDE-153 with glucose homeostasis outcomes, although these associations were generally null in females. In CHS, associations between total PBDEs and glucose concentrations were positive in females and null in males. When stratified by baseline percent fat mass, results remained similar in individuals with higher percent fat mass, but associations were generally null in individuals with lower percent fat mass (Figure S6, Figure S7). When excluding BMI from the statistical models, results generally remained unchanged (Figure S8).

4. Discussion

In this multi-cohort study, we evaluated the association between lipophilic POPs and longitudinal changes in glucose homeostasis, insulin sensitivity, and beta cell function in youth and young adults at risk of T2D. Plasma concentrations of lipophilic POPs, including chlorinated and brominated compounds, were associated with glucose dysregulation in individuals undergoing an OGTT. Associations with the chlorinated compounds were found to not be influenced by Tanner stage, whereas we observed differences across pubertal stages (Tanner 1–5) for associations with the brominated compounds. Our results suggest that POPs with properties of persistence in adipose tissue may perturb glucose regulation.

Epidemiological studies examining the association between PBDE exposure and T2D are limited, and findings from previous studies on PBDEs and diabetes have had mixed results (Han et al., 2020; Lim et al., 2008; Zhang et al., 2016). PBDE-153 has been positively associated with diabetes and/or metabolic syndrome in studies in China (Han et al., 2020; Zhang et al., 2016) and in the US, (Lim et al., 2008) but other studies have reported null or small protective effects (Airaksinen et al., 2011; Turyk et al., 2015). In the present study, PBDE-153 was associated with higher glucose at 1- and 2- hour OGTT timepoints as well as higher glucose AUC in the SOLAR cohort. The effects were strongest post puberty. These findings are in concordance with results from Lim et al., who used data from NHANES, 2003–2004 and found that PBDE-153 showed an association with T2D, while associations between other PBDE congeners failed to reach statistical significance (Lim et al., 2008). Notably, specimens used to measure PBDE concentrations from Lim et al. were collected during an overlapping time window as the collection of specimens from the SOLAR cohort. This is an important component as exposure to these persistent pollutants is largely dependent on year of production, and the use of these chemicals has evolved over time. In the CHS cohort we observed significant associations with ΣPBDE and 1-hour glucose and glucose AUC, but we did not see significant associations with individual PBDE congeners. These differences could be due in part to differences in the patterns of exposure in the SOLAR and CHS cohorts, which are reflected in the differences in the correlation structures of PBDE congeners in the two cohorts. Further, given that invidivuals are exposed to mixtures of exposures rather than single exposures, we explored the effects of chemical mixtures on changes in glucose concentrations during the OGTT. In this analysis, we observered similar associations as observed with the single exposure models. We also observed consistent, positive associations in both the SOLAR and the CHS cohorts, although most associtaions did not meet the threshold for statistical significance.

Multiple potential mechanisms may link exposure to brominated compounds and dysregulated glucose metabolism. In the SOLAR cohort, we observed significant Tanner-dependent associations between the Matsuda index, a marker of insulin sensitivity, and both PBDE-100 and total PBDEs. In contrast, in the CHS cohort, there were no associations between PBDEs and the insulinogenic index, a marker of β-cell function. This raises the possibility that childhood exposure to PBDEs may decrease insulin sensitivity, which could lead to glucose dysregulation and may predispose individuals to T2D later in life.

Epidemiological studies of organochlorine exposure and T2D are much more abundant than those of the brominated compounds (Lee et al., 2018; Magliano et al., 2014). However, fewer longitudinal studies examining the associations of organochlorine compounds and diabetes have been completed. In the present study, we observed that among the chlorinated compounds, HCB and PCB-118 were significantly associated with higher glucose levels during an OGTT in the SOLAR cohort. While not statistically significant, the results trended in the same direction in the CHS cohort with more limited power due to smaller sample size. We observed little overlap with congener-specific findings for associations of exposures to OC compounds and T2D (Lee et al., 2018; Singh and Chan, 2017; Zong et al., 2018). Wu et al. demonstrated in the Nurses’ Health Study that plasma concentrations of HCB were associated with increased risk of developing T2D. These findings were further supported with a meta-analysis of pooled data from six other prospective analyses and demonstrated positive associations for HCB with incident T2D (Wu et al., 2013). In a more recent and larger prospective study in the Nurses’ Health Study, associations for DDE and other POPs with T2D were observed (Zong et al., 2018). Other epidemiologic studies have concluded that POPs disrupt glucose homeostasis but have found associations with discordant congeners including PCBs 138, 153 (Lee et al., 2018; Wolf et al., 2019), ΣPCB (Lee et al., 2018; Singh and Chan, 2017), 4,4’-DDE (Lee et al., 2006; Lee et al., 2014; Turyk et al., 2015).

The underlying mechanisms between exposure to organochlorine compounds and type 2 diabetes pathophysiology are not fully understood. Proposed mechanisms include 1) dysregulation of glucose and insulin (Ruzzin et al., 2010), 2) perturbation to nuclear receptor function (AhR, CAR, and PXR), which increases chronic low-grade inflammation, decreases mitochondrial function and fatty acid oxidization, and increases lipogenesis to ultimately produce insulin resistance syndrome (Hernandez et al., 2009), and 3) dysregulation of peroxisome proliferator-activated receptor (PPAR) and AhR. PPAR plays a role in glucose homeostasis and translation of GLUT4, the insulin-regulated glucose transporter (Wang et al., 2008).

Our study has limitations. The follow up period of our study was rather short compared to the latency period for most chronic diseases. Previous studies have found that low level exposures to POPs is associated with the development of metabolic disorders including T2D and the latency period for the development of disease after exposure to POPs is unknown. It is possible that this underestimates true effect and that we would see an increased incidence of disease status with an increased follow up time. Also, our findings might be susceptible to unmeasured confounding as we were not able to extensively evaluate the impacts of diet-related exposures. However, we performed several sensitivity analyses to examine potential effect modification or confounding from several variables including sex, fat mass, and BMI, and found that our results were robust to a variety of different modeling assumptions.

Our study has several strengths. First, we included longitudinal clinically relevant outcome measures. Numerous studies have previously investigated associations between POPs and T2D; however, none to our knowledge have used longitudinal measures of OGTT. Secondly, the longitudinal design of our study allowed prospective assessment of lipophilic POPs exposure using stored participant blood samples. By using biomarkers of exposure and outcome, we improve the validity of our research and reduce the potential for biases in the measurement of exposure and outcomes. Additionally, we had detailed information on pubertal status based on Tanner staging, a key covariate for exposure outcome association studies within adolescent populations. We also included two independent cohorts and found similar exposure outcome associations in both cohorts. The generalizability of findings between cohorts suggests a consistent association between lipophilic POPs and risk factors for type 2 diabetes in youth and young adults.

5. Conclusions

To our knowledge, this multi-cohort study was the first to evaluate the associations between youth lipophilic POPs concentrations and longitudinal measures of OGTT across pubertal stages and compare findings to another young adult cohort. The study addresses an important gap in our understanding of the contribution of early-life environmental exposures to the T2D epidemic in youth. Regulatory and behavioral interventions aimed at lowering POPs contamination and early-life exposure may help prevent onset of youth T2D.