Mitochondrial dysfunction as a potential pathway linking DL-PCB exposure to intrauterine growth retardation

Abstract

Backgraound

Dioxin-like polychlorinated biphenyls (DL-PCBs) are persistent organic pollutants with known endocrine-disrupting and developmental toxicity. Increasing evidence suggests that prenatal exposure to DL-PCBs may adversely affect fetal growth; however, few studies have examined congener-specific and mixture-level effects, and the underlying biological mechanisms remain unclear. This study aimed to assess the relationship between umbilical cord DL-PCB concentrations and key fetal growth indicators including birth weight (BW), birth length (BL), and head circumference (HC), and to explore whether mitochondrial dysfunction mediates these effects.

Methods

A total of 5421 mother–infant pairs were recruited between 2022 and 2024 from a multicenter birth cohort in China. Twelve WHO-designated DL-PCB congeners were quantified in umbilical cord blood using gas chromatography–mass spectrometry (GC-MS). We employed multivariable linear regression, sex-stratified models, Bayesian kernel machine regression, generalized weighted quantile sum (gWQS), quantile g-computation (g-comp), and mediation analysis to evaluate individual and mixture effects.

Results

Eight DL-PCBs, particularly PCB-77 and PCB-126, were significantly associated with reduced BW (ranging from − 26.52 g to − 80.05 g per IQR increase). PCB-77 and PCB-126 were also negatively associated with HC, while no DL-PCBs showed significant associations with BL. Although inverse associations appeared stronger in boys, sex interactions were not statistically significant. Mediation analysis revealed that mitochondrial DNA content partially mediated the relationship between several DL-PCBs (e.g., PCB-77, PCB-81, PCB-118, PCB-126) and BW, with up to 10.76% of the total effect explained. Mixture models confirmed negative associations with BW and HC, with g-comp estimating a 161.95 g BW reduction per cumulative exposure unit. Bayesian kernel machine regression further identified PCB-77 and PCB-126 as dominant contributors.

Conclusion

These findings suggest that mitochondrial dysfunction may be a mechanistic pathway linking prenatal DL-PCB exposure to intrauterine growth restriction, especially reduced birth weight, highlighting potential developmental risks from environmental contamination.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12940-026-01263-1.

Introduction

Environmental pollution remains a critical public health concern globally, driven by widespread industrialization and the extensive use of synthetic chemicals [1, 2]. Among the most concerning groups of environmental contaminants are polychlorinated biphenyls (PCBs), a class of synthetic organic chemicals historically used in electrical equipment, paints, sealants, and other industrial applications [3]. Though banned or restricted in many countries since the late 1970 s, PCBs persist in the environment due to their chemical stability and lipophilicity. A specific subset, known as dioxin-like PCBs (DL-PCBs), structurally resembles dioxins and exerts toxicological effects through activation of the aryl hydrocarbon receptor (AhR) pathway [4]. DL-PCBs enter the human body primarily through the consumption of contaminated food, particularly high-fat animal products such as fish, meat, and dairy, as well as through inhalation of polluted air or dermal contact [5]. Once absorbed, they accumulate in adipose tissue and can cross the placenta, directly exposing the developing fetus [6].

Emerging evidence has linked DL-PCB exposure to a wide range of adverse health outcomes, especially due to their endocrine-disrupting properties. These chemicals are capable of interfering with hormonal signaling, immune responses, and normal cellular development [7–9]. In pregnant women, DL-PCBs can traverse the placental barrier, potentially altering fetal development during sensitive windows of growth [10]. Several toxicological and epidemiological studies have suggested associations between maternal exposure to DL-PCBs and increased risks of adverse reproductive and developmental outcomes, including fetal growth restriction, preterm birth, and neurodevelopmental delays [10–14]. However, the evidence remains inconclusive and varies depending on the specific congener of DL-PCBs, timing of exposure, and population characteristics. Fetal growth indicators, such as head circumference (HC), birth length (BL), and birth weight (BW), are critical metrics of prenatal development and are commonly used to assess the intrauterine environment [15]. These indicators not only reflect fetal well-being at birth but also serve as predictors of future health trajectories. For instance, low birth weight has been linked to increased risks of cardiovascular disease, diabetes, and cognitive impairments in later life [16]. Several environmental epidemiology studies have reported associations between prenatal exposure to various pollutants, such as fine particulate matter (PM2.5) [17], heavy metals, and per- and polyfluoroalkyl substances (PFAS) and impaired fetal growth metrics [18]. However, research specifically targeting DL-PCB exposure and its impact on fetal growth is still limited.

Mitochondrial DNA (mtDNA) plays a vital role in fetal development due to its involvement in cellular energy metabolism, oxidative stress regulation, and apoptosis [19]. Alterations in mtDNA content or function during gestation may disrupt placental and fetal cellular processes, potentially leading to impaired fetal growth [20, 21]. Specifically, alterations may include changes in mtDNA copy number, impaired mitochondrial biogenesis, or oxidative damage to the mitochondrial genome, all of which can reduce ATP production and weaken the cell’s ability to regulate oxidative stress [22]. Several studies have suggested that prenatal environmental exposures, including to PCBs, may influence mtDNA copy number [23, 24], which in turn could impact growth outcomes. Changes in copy number are often considered indicators of mitochondrial stress: lower mtDNA content may reflect mitochondrial dysfunction and reduced metabolic capacity, whereas higher copy numbers may represent compensatory upregulation in response to accumulated oxidative damage [22, 25]. Both patterns have been linked to altered fetal growth trajectories. However, the potential mediating role of mtDNA in the association between DL-PCB exposure and fetal growth has not been thoroughly investigated.

Although some studies have evaluated the relationship between environmental pollutant exposure during pregnancy and fetal development, there is a critical gap in understanding the specific effects of DL-PCBs on detailed birth outcomes. Given the persistence and bioaccumulation of DL-PCBs in human tissues, it is imperative to assess their developmental toxicity more rigorously (Letz, 1983). In this study, we hypothesize that mtDNA content acts as a biological mediator in the association between umbilical cord DL-PCB levels and fetal growth indicators. This study fills a critical knowledge gap by examining the relationship between dioxin-like PCB concentrations in umbilical cord blood and key measures of fetal growth within a large, forward-looking birth cohort. The use of advanced mixture modeling approaches allows for a more nuanced understanding of the cumulative and individual effects of DL-PCB congeners. Findings from this study will not only contribute to the scientific evidence base but also support policy initiatives aimed at minimizing prenatal toxicant exposures and improving birth outcomes.

Methods

Study area and population settings

The present study was carried out in Shenyang, China, from 2022 to 2024. As the capital of Liaoning Province and a major hub for heavy industry in northeastern China, Shenyang has undergone rapid urban expansion and economic growth in recent decades. However, this industrial boom has come with serious environmental consequences, including substantial air pollution linked to extensive manufacturing activities, particularly in the steel and machinery sectors. The city, home to over 9 million residents, has a humid continental climate characterized by sweltering, humid summers and prolonged, bitterly cold winters. Seasonal meteorological patterns, including wintertime temperature inversions, exacerbate the accumulation of airborne pollutants. These environmental stressors have raised significant public health concerns, especially for sensitive populations such as pregnant women and infants.

Participants in this study were pregnant women receiving antenatal care and delivery services at a major maternity hospital in Shenyang. Recruitment focused on women with singleton pregnancies who were expected to deliver spontaneously between 37 and 42 weeks of gestation. To minimize confounding factors, individuals with preexisting health conditions such as gestational diabetes, chronic hypertension, liver or kidney disorders, preeclampsia, and endocrine abnormalities were excluded. Additionally, women with obesity, smokers, and those taking medications known to influence glucose or lipid metabolism were not eligible. Out of nearly 7,000 women admitted for delivery during the study period, a total of 5421 met the eligibility criteria and were successfully included in the study. Trained interviewers collected sociodemographic and lifestyle information through structured, in-person interviews. The study protocol received ethical approval from the Institutional Review Board of Shengjing Hospital, China Medical University, and written informed consent was obtained from all participants before participation.

Sample collection and analysis of DL-PCBs

At delivery, ~ 10 mL umbilical cord blood was aseptically collected into sterile serum-separator tubes (clot activator). Samples clotted at room temperature for 20 min and were centrifuged at 2,500 rpm for 20 min. Serum was aliquoted into precleaned amber cryovials and stored at − 80 °C until analysis. We quantified 12 WHO dioxin-like PCB congeners (PCB-77, PCB − 81, PCB − 105, PCB-114, PCB-118, PCB-123, PCB-126, PCB-156, PCB-157, PCB-167, PCB-169, PCB-189).

For each sample, 2.0 mL serum was spiked with the surrogate/internal standard PCB-209 (J&K Chemical, Beijing, China). DL-PCBs were extracted by triple liquid–liquid extraction with 30 mL n-hexane/dichloromethane (2:1, v/v). Combined extracts were dried over anhydrous Na2SO4 and cleaned on an acidic silica gel column. Eluates were concentrated by rotary evaporation to ~ 50 mL and further reduced under nitrogen to the instrumental volume. Quantification was performed by GC–MS (Agilent 7890B/5977B) with an HP-5MS capillary column (30 m × 0.25 mm i.d., 0.25 μm film). Helium carrier gas flow: 1.5 mL/min. Oven: 90 °C start; 10 °C/min to 150 °C (1 min hold); 20 °C/min to 260 °C (5 min hold). Injector 250 °C; EI 70 eV; SIM mode. Method parameters were adapted from validated protocols [26, 27].

To account for inter-individual variation in serum lipid content, DL-PCB concentrations were expressed on both a wet-weight basis (ng/L) and a lipid-adjusted basis (ng/g lipid). Total serum lipids (TSL) were determined enzymatically: Total cholesterol (TC) and triglycerides (TG) were measured in serum aliquots (fresh-thawed) using standard enzymatic colorimetric assays on an automated analyzer following manufacturer instructions. TSL (mg/dL) was calculated using the Phillips formula:

1

where TC and TG are in mg/dL.

Lipid-adjusted DL-PCB concentrations were computed as:

2

with TSL converted from mg/dL to g/L (×0.1). Lipid-adjusted values were used in our furthere analyses.

Quality control and method validation

Each serum sample was extracted in triplicate with ≥ 3 instrumental injections per replicate. Method blanks (one per 2–3 samples), solvent blanks, and continuing calibration checks were included in every batch. Accuracy was verified with NIST SRM 1589a and SRM 1958. Surrogate (PCB-209) recoveries of 90–110% were accepted; samples outside this range were re-extracted. Precision (RSD) was < 13% across congeners. Limits of detection (LODs; µg/L) were: PCB-77 and − 81 (0.02); PCB-156, −157, −167, −123 (0.03); PCB-105 (0.07); PCB-114 and − 118 (0.05); PCB-169 and − 189 (0.04). Detection frequencies were high across congeners (e.g., PCB-126, 83.0%; PCB-77, 89.3%; PCB-81, 87.6%; PCB-167, 81.3%; PCB-105, 84.1%; PCB-114, 82.5%; PCB-169, 79.9%; PCB-156, 78.8%; PCB-123, 79.7%; PCB-118, 80.2%; PCB-157, 77.4%; PCB-189, 81.7%). Intraclass correlation coefficients for duplicate/triplicate sets were > 0.75, indicating excellent reproducibility. Values < LOD were imputed as LOD/√2. Reported concentrations include lipid-adjusted (ng/g lipid) metrics.

Quantification of mitochondrial DNA content

Genomic DNA was isolated from umbilical cord venous blood using the QIAamp DNA Blood Kit (QIAGEN, Hilden, Germany). To quantify relative mitochondrial DNA (mtDNA) content, we calculated the mtDNA-to-nuclear DNA (nDNA) ratio (mtDNA/nDNA), a widely accepted indicator of mitochondrial DNA copy number. This measurement was performed using a modified multiplex quantitative real-time PCR (qPCR) approach adapted from established protocols. Mitochondrial DNA amplification targeted the 12S rRNA gene using a TaqMan probe (Applied Biosystems, Waltham, MA; sequence: 6FAM-5′ TGCCAGCCACCGCG 3′-MGB) in combination with primers mtF805 (5′CCACGGGAAACAGCAGTGATT3′) and mtR927 (5′CTATTGACTTGGGTTAATCGTGTGA3′). For nuclear DNA quantification, the RNase P gene served as a single-copy reference, utilizing the TaqMan RNase P Control Reagents Kit (Applied Biosystems). All qPCR reactions were run on the Bio-Rad CFX384 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA) [28, 29].

To standardize mtDNA quantification across samples, a pooled reference DNA sample, consisting of equal aliquots from 300 individual specimens (20 µL each, final concentration 40 ng/µL), was used to generate a five-point standard curve ranging from 0.12 to 10 ng/µL. Each sample’s mtDNA/nDNA ratio was determined relative to this reference, with a ratio of 1 representing equivalent amounts of mtDNA. All assays were performed in triplicate. The method demonstrated strong reproducibility, with intra- and inter-assay coefficients of variation of 5% and 7%, respectively, confirming its high precision and reliability [30].

Fetal growth indicators

In this study, BW, BL, and HC were utilized as primary indicators of fetal growth. These parameters were routinely measured by trained medical staff immediately after delivery, following standard clinical protocols. BW was recorded to the nearest g using calibrated digital scales, while BL and HC were measured to the nearest 0.1 cm using standardized infantometers and non-stretchable measuring tapes, respectively. Birth outcome data were obtained directly from hospital medical records to maintain high accuracy and uniformity. These anthropometric indicators are well-established proxies for intrauterine development and are broadly acknowledged as key predictors of both immediate neonatal well-being and long-term health trajectories.

Statistical analysis

Main analysis

Multivariable linear regression models were used to assess the association between umbilical cord blood concentrations of individual DL-PCB congeners and fetal growth indicators, including BW, BL, and HC. Each individual DL-PCB congener was analyzed as a separate exposure variable, with fetal growth parameters evaluated as the corresponding outcome measures. Potential confounding variables were selected based on previous research, biological plausibility, and statistical criteria. Covariates were included if they were associated with both exposure and outcome (p < 0.20) or if their inclusion altered the estimated exposure effect by more than 10% (Figure S1 in the Supplemental Materials). The final models were adjusted for key maternal and infant characteristics, including prenatal exposure to tobacco smoke (yes/no), maternal pre-pregnancy body mass index (BMI, kg/m²), maternal age (years), gestational age at delivery (weeks), infant sex (male/female), and household socioeconomic status (SES). SES was determined based on household income (categorized as low than median, or higher than median) and paternal and maternal education level (illiterate/primary school, high school, or university). Regression coefficients were reported per interquartile range (IQR) increase in DL-PCB concentration, indicating the magnitude of change in each fetal growth outcome. To ensure the reliability of the model estimates, standard regression diagnostics were conducted, including checks for linearity, independence and normality of residuals, homoscedasticity, and the influence of outliers. These steps ensured the robustness and validity of the linear modeling approach.

Stratified analysis by infant sex

To investigate whether the association between umbilical DL-PCB exposure and BW differed by infant sex, we conducted stratified linear regression analyses. Separate models were fitted for male and female newborns, with DL-PCB congeners as the exposure variables and BW as the outcome. All models were adjusted for the same set of covariates used in the main analysis (i.e. Multivariable linear regression). To formally test for effect modification by infant sex, we introduced an interaction term between DL-PCB concentration and infant sex into the regression model and applied the likelihood ratio test to compare the model fit with and without the interaction term. A p-value less than 0.05 from the likelihood ratio test was considered evidence of significant interaction, indicating that the association between DL-PCBs and BW differed by sex.

Bayesian kernel machine regression (BKMR)

To evaluate the combined effects and potential interactions among the 12 DL-PCB congeners on fetal growth outcomes, we applied Bayesian Kernel Machine Regression (BKMR). This method allows for flexible modeling of complex, non-linear relationships and potential interactions within a mixture of correlated exposures. In the BKMR framework, all DL-PCB congeners were modeled simultaneously as a mixture, and posterior inclusion probabilities were calculated to assess the relative importance of each congener. The same covariates used in the linear regression models were included in the BKMR models for adjustment. To ensure robust estimation, we implemented 10,000 iterations based on a Monte Carlo Markov Chain simulation. BKMR analysis was conducted using the “bkmr” package in R [31].

Generalized weighted quantile sum (gWQS) regression

To further explore the joint association of DL-PCB mixtures with fetal growth, we used generalized Weighted Quantile Sum (gWQS) regression. This approach is particularly suited for highly correlated exposures and provides a single index that summarizes the weighted contribution of each congener in relation to the outcome. DL-PCB concentrations were categorized into quantiles, and weights were assigned based on their relative contribution to the overall effect. Both positive and negative directionality were examined in separate models. The gWQS models were adjusted for the same covariates used in the linear regression. Analyses were performed using the “gWQS” R package with 100 bootstrap samples and 40% of the data reserved for validation [12, 32].

Quantile g-computation

We also applied quantile g-computation (g-comp); an emerging statistical method designed to estimate the joint effect of multiple exposures without assuming directional homogeneity among components. This approach estimates the change in the outcome per quantile increase in all exposures simultaneously, allowing for both positive and negative contributions within the mixture. The DL-PCB congeners were included as exposure variables, and the models were adjusted for the same covariates as described in the linear regression analyses. Quantile g-computation was implemented using the “qgcomp” R package, with models fitted separately for each fetal growth indicator (BW, BL, and HC) [33].

Mediation analysis

We investigated whether mtDNA content serves as a mediator in the association between umbilical DL-PCB levels and fetal growth outcomes, including BW, BL, and HC. In this mediation framework (exposure → mediator → outcome), DL-PCB concentrations were considered the exposure, mtDNA content the mediator, and the fetal growth indicators the outcomes. The direct effect was defined as the relationship between DL-PCB exposure and fetal growth independent of mtDNA, estimated by adjusting for mtDNA in the model. The indirect effect was determined by assessing how DL-PCB exposure influenced mtDNA levels, and in turn, how mtDNA levels affected fetal growth, controlling for DL-PCB concentrations. The total effect combined both direct and indirect pathways. Linear regression models were used for all analyses, adjusting for potential confounding factors same as main analysis. We estimated mediation effects using the “mediation” R package. The proportion of the total effect explained by mtDNA was calculated by dividing the indirect effect by the total effect and multiplying by 100. Confidence intervals for the mediation estimates were obtained through nonparametric bootstrapping with 5,000 replications.

Results

Population characteristics and DL-PCBs levels

Descriptive characteristics of the study population are presented in Table 1. A total of 5421 mother–infant pairs were included in the analysis. The median maternal age at delivery was 31 years (IQR: 13), and the median pre-pregnancy BMI was 22.32 kg/m² (IQR: 7.33). The distribution of infant sex was nearly equal, with 51.06% being female. Maternal and paternal education levels were similarly distributed, with roughly one-third of participants having completed an illiterate/elementary, high school, or university education. Passive tobacco smoke exposure during pregnancy was reported by 30.97% of women. The median gestational age at delivery was 39 weeks (IQR: 4). Regarding fetal growth indicators, the median BW was 3358 g (IQR: 856), the median BL was 50 cm (IQR: 6), and the median HC was 33 cm (IQR: 3).

Table 1.

Descriptive characteristics of the study population (N = 5421). Continuous variables are presented as median (interquartile range), and categorical variables are shown as counts (percentages)

Variables	Description
Maternal age at delivery (year); Median (IQR)	31 (13)
Fetal sex
Girl; N (%)	2768 (51.06)
Boy; N (%)	2653 (48.94)
pre-pregnancy BMI (kg/m2); Median (IQR)	22.32 (7.33)
Paternal education
Illiterate/elementary; N (%)	1802 (33.24)
High-school; N (%)	1828 (33.72)
University; N (%)	1791 (33.04)
Maternal education
Illiterate/elementary; N (%)	1794 (33.09)
High-school; N (%)	1829 (33.74)
University; N (%)	1798 (33.17)
Family Income
Lower that median; N (%)	2687 (49.57)
Higher than median; N (%)	2734 (50.43)
Passive tobacco smoke exposure
No; N (%)	3742 69.02
Yes; N (%)	1679 30.97
Gestational age (Week); Median (IQR)	39 (4)
Birth length (cm); Median (IQR)	50 (6)
Birth weight (g); Median (IQR)	3358 (856)
Head circumference (cm); Median (IQR)	33 (3)

Abbreviations:BMI Body Mass Index, IQR Interquartile Range, cm centimeter, g gram

The distribution of DL-PCB concentrations in serum cord blood samples is shown in Figure S2 of Supplemental Materials. Among the congeners, PCB118 exhibited the highest median concentration (3.12 ng/g lipid; IQR = 0.40), followed by PCB105 (1.55 ng/g lipid; IQR = 0.10) and PCB156 (0.51 ng/g lipid; IQR = 0.21). Conversely, PCB81, PCB169, and PCB126 showed the lowest median concentrations (≤ 0.05 ng/g lipid). Overall, higher-chlorinated congeners such as PCB118 and PCB105 were more prevalent compared to lower-chlorinated congeners.

Main analysis

Cord blood DL-PCB exposure and BW

In unadjusted linear regression analyses, eight DL-PCB congeners, PCB77, PCB81, PCB118, PCB123, PCB126, PCB157, PCB169, and PCB189, exhibited statistically significant inverse associations with birth weight (all p < 0.05). After adjustment for covariates, these associations remained statistically significant. Specifically, each 1 IQR increase in umbilical concentration of PCB77 corresponded to an average BW reduction of 80.05 g (β = − 80.05; 95% CI: − 102.39, − 57.71; p < 0.01), and PCB126 showed a similar association with a decrease of 69.45 g (β = − 69.45; 95% CI: − 92.30, − 46.61; p < 0.01). Other significant congeners, including PCB81, PCB118, PCB123, PCB157, PCB169, and PCB189, were associated with BW reductions ranging from 26.52 to 39.44 g per 1 IQR increase in exposure (all p < 0.05). Conversely, PCB105, PCB114, PCB156, and PCB167 were not significantly associated with BW in either crude or adjusted models (all p > 0.05). These findings are summarized in Fig. 1.

Fig. 1.

Association of exposure to umbilical DL-PCBs and birth weight based on crude and adjusted models. (Models were adjusted for maternal pre-pregnancy BMI, maternal age, gestational age at delivery, infant sex, prenatal tobacco smoke exposure, and household socioeconomic status)

Cord blood DL-PCB exposure and BL

The associations between umbilical DL-PCB concentrations and BL are presented in Fig. 2. In neither the unadjusted nor the adjusted linear regression analyses did any of the DL-PCB congeners exhibit statistically significant links with BL. Specifically, in the adjusted models, the beta coefficients per 1 IQR increase in DL-PCB exposure ranged from − 0.12 cm for PCB156 (β = − 0.12; 95% CI: − 0.27, 0.03; p = 0.11) to 0.11 cm for PCB81 (β = 0.113; 95% CI: − 0.032, 0.258; p = 0.13). Although PCB156 showed a marginal inverse association with BL in the crude model (p = 0.097), this did not reach statistical significance after adjustment (adjusted p = 0.11). Other congeners, including PCB77, PCB81, PCB105, PCB114, PCB118, PCB123, PCB126, PCB157, PCB167, PCB169, and PCB189, demonstrated no meaningful associations with BL, with all adjusted p-values exceeding 0.05. These findings indicate that prenatal exposure to individual DL-PCBs was not significantly related to BL in this cohort (Fig. 2).

Fig. 2.

Association of exposure to umbilical DL-PCBs and birth length based on crude and adjusted models. (Models were adjusted for maternal pre-pregnancy BMI, maternal age, gestational age at delivery, infant sex, prenatal tobacco smoke exposure, and household socioeconomic status)

Cord blood DL-PCB exposure and HC

The associations between umbilical DL-PCB concentrations and newborn HC are shown in Fig. 3. In both crude and adjusted linear regression models, only PCB77 and PCB126 were significantly associated with reduced HC. Specifically, each 1 IQR increase in maternal PCB77 was linked to a decrease of 0.12 cm in HC (adjusted β = − 0.12; 95% CI: − 0.20, − 0.05; p < 0.01), while PCB126 was associated with a 0.10 cm reduction (adjusted β = − 0.10; 95% CI: − 0.18, − 0.02; p = 0.02). The remaining DL-PCB congeners (PCB81, PCB105, PCB114, PCB118, PCB123, PCB156, PCB157, PCB167, PCB169, and PCB189) did not show statistically significant associations with HC, with all adjusted p-values exceeding 0.05.

Fig. 3.

Association of exposure to umbilical DL-PCBs and head circumference based on crude and adjusted models. (Models were adjusted for maternal pre-pregnancy BMI, maternal age, gestational age at delivery, infant sex, prenatal tobacco smoke exposure, and household socioeconomic status)

Sex-stratified associations between DL-PCB exposure and BW

Table 2 presents the results of sex-stratified linear regression analyses assessing the associations between umbilical DL-PCB exposure and BW. Overall, inverse associations were observed in both female and male infants, though the strength of association varied by congener and sex. Notably, for both girls and boys, exposure to PCB77, PCB81, and PCB126 was significantly associated with lower BW, with stronger effect estimates observed among male infants. For example, each 1 IQR increase in PCB77 was associated with a 65.81 g (95% CI: − 97.11, − 34.51; p < 0.01) reduction in girls and a 96.40 g (95% CI: − 128.62, − 64.19; p < 0.01) reduction in boys, though the p for interaction was 0.22, suggesting no statistically significant effect modification by sex. Similarly, for PCB105, a significant association was observed only in boys (β = − 34.17 g; 95% CI: − 65.95, − 2.39; p = 0.04), but not in girls (p = 0.73), with a p-interaction of 0.23. PCB123 also showed a stronger negative association in girls (β = − 56.44 g; p < 0.01) compared to boys (β = − 15.62 g; p = 0.33), but the interaction was not statistically significant (p-interaction = 0.08). A significant association with BW was found for PCB118 in girls (β = − 45.72 g; p < 0.01), but not in boys (β = − 20.78 g; p = 0.20), with no significant interaction by sex (p = 0.28). Other congeners such as PCB114, PCB156, PCB157, PCB167, PCB169, and PCB189 demonstrated similar patterns in both sexes, but none showed statistically significant sex interactions (all p-interaction > 0.05). These findings suggest that while some DL-PCBs may exert stronger effects on BW depending on fetal sex, there was no statistically robust evidence of sex-based effect modification in this cohort.

Table 2.

Sex-stratified associations between maternal serum DL-PCB concentrations (per 1 interquartile range [IQR] increase) and birth weight. Estimates (β-coefficients) and 95% confidence intervals (CIs) are derived from multivariable linear regression models adjusted for maternal pre-pregnancy BMI, maternal age, gestational age at delivery, infant sex, exposure to tobacco smoke during pregnancy, and socioeconomic status. P-interaction values reflect statistical tests for effect modification by infant sex

DL-PCBs congeners	Girl		Boys		p-interaction
	Beta-coefficient (95% CI)	p-value	Beta-coefficient (95% CI)	p-value
PCB77	−65.81 (−97.11, −34.51)	< 0.01	−96.40 (−128.62, −64.19)	< 0.01	0.22
PCB81	−64.70 (−96.56, −32.84)	< 0.01	−62.52 (−94.45, −30.59)	< 0.01	0.91
PCB105	−5.54 (−37.41, 26.34)	0.73	−34.17 (−65.95, −2.39)	0.04	0.23
PCB114	−14.17 (−45.92, 17.57)	0.38	−16.19 (−48.42, 16.04)	0.32	0.92
PCB118	−45.72 (−77.15, −14.29)	< 0.01	−20.78 (−52.61, 11.04)	0.20	0.28
PCB123	−56.44 (−87.80, −25.09)	< 0.01	−15.62 (−47.18, 15.94)	0.33	0.08
PCB126	−73.79 (−106.27, −41.31)	< 0.01	−64.56 (−96.82, −32.31)	< 0.01	0.76
PCB156	−16.50 (−48.60, 15.61)	0.31	−19.69 (−52.82, 13.44)	0.24	0.91
PCB157	−23.64 (−55.97, 8.70)	0.15	−30.17 (−61.94, 1.60)	0.06	0.70
PCB167	−5.89 (−37.07, 25.29)	0.71	−8.64 (−40.84, 23.55)	0.60	0.91
PCB169	−26.83 (−59.22, 5.58)	0.10	−24.82 (−56.70, 7.06)	0.13	0.97
PCB189	−31.05 (−62.70, 0.60)	0.05	−48.96 −81.32, −16.60)	< 0.01	0.44

Mediation analysis

Figure 4 presents the estimated proportion of the effect of individual DL-PCBs on fetal growth indicators that is mediated through mtDNA. Mediation analysis revealed that mtDNA partially mediated the association between several DL-PCBs and BW. Significant mediation effects were observed for PCB-77 (6.35%; 95% CI: 3.57, 10.42; p < 0.01), PCB-81 (4.43%; 95% CI: 1.45, 8.64; p = 0.01), PCB-118 (10.76%; 95% CI: 4.09, 33.99; p < 0.01), and PCB-126 (5.57%; 95% CI: 2.41, 10.39; p < 0.01), indicating a modest but statistically significant role of mtDNA in the effect of these compounds on fetal weight. Marginal or non-significant effects were observed for PCB-105 (7.57%; p = 0.24), PCB-157 (7.13%; p = 0.07), and other congeners such as PCB-123, PCB-156, PCB-167, PCB-169, and PCB-189, which showed lower or negative mediation estimates with wide confidence intervals crossing zero. These findings suggest a congener-specific pattern, with only a subset of DL-PCBs influencing birth weight via mtDNA-related mechanisms.

Fig. 4.

Proportion mediated (%) with 95% confidence intervals for the mediating role of mitochondrial DNA (mtDNA) in the associations between individual dioxin-like polychlorinated biphenyl (DL-PCB) congeners and fetal growth outcomes, including birth weight (red), birth length (green), and head circumference (blue). Filled triangles represent statistically significant mediation effects (p < 0.05), while filled circles indicate non-significant effects. Negative and positive values indicate inverse and direct mediation contributions, respectively

In contrast to BW, the proportion of the effect of DL-PCBs on BL mediated by mtDNA was generally non-significant across all congeners. PCB-77 and PCB-126 showed relatively high mediation percentages (21.32% and 22.33%, respectively), but these estimates were statistically non-significant due to very wide confidence intervals (PCB-77: 95% CI: −279.35, 165.71; p = 0.54; PCB-126: 95% CI: −186.17, 257.79; p = 0.70). Other congeners, including PCB-81, PCB-105, PCB-114, PCB-118, PCB-123, and PCB-157, showed small or even negative proportions mediated, with p-values well above 0.05. These findings suggest that mtDNA does not substantially mediate the association between DL-PCB exposure and BL, and the large uncertainty around the estimates underscores the variability of this potential pathway.

For HC, no significant mediation effect via mtDNA was observed for any of the DL-PCBs examined. The proportion mediated ranged from − 0.73% for PCB-114 to 0.60% for PCB-118, with all confidence intervals crossing zero and p-values exceeding 0.90. These consistently null findings across all congeners, including PCB-77, PCB-81, PCB-105, PCB-126, and others, indicate that mtDNA likely does not play a mediating role in the relationship between prenatal DL-PCB exposure and neonatal HC.

Mixture analysis

BKMR analysis

Figure 5 presents the results of the BKMR analysis assessing the relationship between umbilical DL-PCBs exposure and fetal growth outcomes, including BW, BL, and HC. In the top panel, a clear inverse association is observed between DL-PCB exposure and BW. As exposure quartiles increase, the change in BW declines markedly, with the effect becoming significantly negative from the median exposure level onward. The 95% credible intervals do not overlap zero at higher exposure levels, indicating a statistically meaningful reduction in BW with increasing DL-PCB burden. In contrast, the middle panel shows a weaker and less consistent association between DL-PCB exposure and BL. While there is a downward trend in BL at higher exposure quartiles, the wider credible intervals overlap the null line across all exposure levels, suggesting greater uncertainty in this association. Similarly, the bottom panel shows the relationship between DL-PCB exposure and HC. Although the trend suggests a slight decrease in HC with increasing exposure, the credible intervals are wide and consistently include zero, indicating a non-significant effect. Overall, these results suggest that among the three fetal growth indicators, BW is most strongly and significantly affected by prenatal DL-PCB exposure, while the associations with BL and HC are weaker and statistically non-significant.

Fig. 5.

Estimated overall effect of umbilical DL-PCB exposure on fetal growth outcomes using BKMR. This figure shows the estimated change in (A) birth weight (g), (B) birth length (cm), and (C) head circumference (cm) across quartiles of DL-PCB mixture exposure. The results are derived from Bayesian Kernel Machine Regression (BKMR), adjusting for key maternal and infant covariates. Models were adjusted for maternal pre-pregnancy BMI, maternal age, gestational age at delivery, infant sex, prenatal tobacco smoke exposure, and household socioeconomic status

Figure 6 displays the univariate exposure–response functions derived from the BKMR model for individual DL-PCB congeners and their associations with BW (top panels), BL (middle panels), and HC (bottom panels). For BW, several congeners, notably PCB77, PCB81, PCB126, and PCB156, show clear inverse associations, with increasing concentrations associated with progressively lower BW. PCB77 and PCB126, in particular, demonstrate a consistent and strong negative trend, with effect estimates decreasing sharply across the exposure range. Conversely, congeners such as PCB114, PCB157, and PCB167 exhibit relatively flat or even slightly positive associations, suggesting minimal or no impact on birth weight. For BL, the exposure–response functions are generally flatter across all congeners, with only PCB105 showing a minor non-linear fluctuation, while the rest, including PCB77, PCB81, and PCB126, display negligible changes across the exposure range, indicating weak or no associations. Similarly, for HC, most congeners exhibit flat dose-response curves, with the exception of PCB77 and PCB156, which show subtle negative trends at higher concentrations. However, the credible intervals for BL and HC are wide and include the null for nearly all congeners, indicating a lack of statistically robust effects.

Fig. 6.

Univariate exposure–response functions for individual DL-PCB congeners and fetal growth indicators based on BKMR. Each panel illustrates the estimated change in (top row) birth weight (g), (middle row) birth length (cm), and (bottom row) head circumference (cm) across the concentration range (ng/mL) of individual DL-PCB congeners. Shaded areas represent 95% credible intervals

gWQS analysis

The gWQS regression analysis revealed a statistically significant inverse association between the overall mixture of DL-PCBs in umbilical cord blood and BW (Table 3). Specifically, a one-quartile increase in the mixture of DL-PCBs was associated with a reduction in BW by 47.32 g (95% CI: − 74.62, − 20.02; p < 0.01). However, no statistically significant associations were observed between DL-PCB mixture exposure and other fetal growth indicators, including BL (β = 0.099; 95% CI: − 0.104, 0.303; p = 0.34) or HC (β = − 0.019; 95% CI: − 0.124, 0.085; p = 0.71). Figure S3 of Supplemental Materials illustrates the relative contributions of individual DL-PCB congeners to the overall mixture effects on fetal growth indicators, as estimated by the gWQS models. For BW, PCB157, PCB114, and PCB156 emerged as the predominant contributors to the negative association observed, suggesting that these congeners played a central role in the reduction of fetal weight linked to DL-PCB exposure. In contrast, for BL, the highest relative weights were assigned to PCB81, PCB123, and PCB169, although the overall association in the gWQS model for this outcome was not statistically significant. Regarding HC, PCB114, PCB156, and PCB157 were identified as the major contributors, followed by PCB105 and PCB169, indicating their prominent role in the observed inverse association with this growth metric.

Table 3.

Estimated associations between DL-PCB mixture exposure in umbilical cord blood and fetal growth indicators using gWQS and g-computation models

Fetal growth indicators	Model	Beta-coefficient (95% CI)	p-value
Birth weight	gWQS	−47.323 (−74.623, −20.024)	< 0.01
	g-comp	−161.945 (−197.95, −125.93)	< 0.01
Birth length	gWQS	0.099 (−0.104, 0.303)	0.34
	g-comp	−0.012 (−0.245, 0.221)	0.92
Head circumference	gWQS	−0.019 (−0.124, 0.085)	0.71
	g-comp	−0.188 (−0.314, −0.062)	< 0.01

Beta coefficients represent the change in birth weight (g), birth length (cm), and head circumference (cm) per one interquartile range (IQR) increase in the DL-PCB mixture. Models were adjusted for maternal pre-pregnancy BMI, maternal age, gestational age at delivery, infant sex, prenatal tobacco smoke exposure, and household socioeconomic status

gWQS generalized weighted quantile sum regression, g-comp g-computation

G-computation analysis

Results from g-comp supported the adverse impact of DL-PCB mixtures on fetal growth. The estimated effect of the DL-PCB mixture on BW was notably stronger in this model, with a significant reduction of 161.95 g (95% CI: − 197.95, − 125.93; p < 0.01) for each one-unit increase in cumulative exposure. Additionally, g-comp analysis revealed a significant negative association with HC, with a decrease of 0.188 cm (95% CI: − 0.314, − 0.062; p < 0.01). No significant effect was observed for BL (β = − 0.012; 95% CI: − 0.245, 0.221; p = 0.92). Figure S4 of Supplemental Materials presents the estimated weights of individual DL-PCB congeners derived from the g-comp models for each fetal growth indicator. For BW, all congeners contributed negatively to the overall mixture effect, with PCB167, PCB77, and PCB81 demonstrating the highest negative weights, suggesting their substantial roles in reducing BW. In contrast, for BL, both negative and positive weights were observed. PCB81 and PCB156 showed the most prominent positive weights, whereas PCB114 and PCB157 exhibited moderate negative contributions. Regarding HC, the model identified PCB114 as the primary positive contributor, followed by PCB77 and PCB126. Notably, the majority of DL-PCBs showed mixed directional weights, highlighting the complex nature of congener-specific effects in the mixture (Table 3).

Discussion

This study is among the first large-scale investigations to comprehensively evaluate the individual and joint effects of DL-PCBs measured in umbilical cord blood on multiple fetal growth indicators, including BW, BL, and HC, while also exploring potential sex-specific associations and mixture effects using advanced statistical models. Our results demonstrate that increased prenatal exposure to specific DL-PCB congeners, especially PCB77, PCB81, PCB118, PCB123, PCB126, PCB157, PCB169, and PCB189, was strongly linked to reduced birth weight, remaining significant even after controlling for important maternal and infant factors. For HC, only PCB77 and PCB126 showed significant inverse associations, while no DL-PCBs were significantly associated with BL. In sex-stratified analyses, these inverse associations generally persisted in both male and female infants, with slightly stronger effects observed in boys, although no statistically significant sex interactions were detected. Our mediation analysis revealed that mtDNA partially mediated the relationship between several DL-PCB congeners and birth weight, with significant effects observed for PCB-77, PCB-81, PCB-118, and PCB-126, suggesting a modest yet statistically robust biological pathway. In contrast, mtDNA did not significantly mediate the effects of DL-PCBs on birth length or head circumference, indicating potential outcome-specific mechanisms in fetal development. Mixture analysis using gWQS confirmed a significant inverse relationship between the combined DL-PCB exposure and BW, with PCB157, PCB114, and PCB156 contributing most to the mixture effect. G-computation models further supported these findings, identifying a stronger inverse effect on BW and additionally detecting a significant reduction in HC, with PCB167, PCB77, and PCB81 as the dominant negative contributors. Overall, these results underscore the potential of both individual and combined DL-PCB exposures to adversely affect fetal growth, particularly birth weight, and suggest the need for further research into the mechanisms and critical windows of susceptibility.

Available evidence

Notably, PCB-118 showed the highest median concentration among the DL-PCBs measured in our cohort. This pattern is consistent with previous reports indicating that PCB-118 is one of the most abundant mono-ortho dioxin-like congeners in several technical PCB mixtures, such as Aroclor 1254, where it typically accounts for 7–13.5% by weight. Although detailed documentation of the specific PCB formulations historically used in China is limited, available evidence suggests that commercial mixtures introduced before the national ban also contained relatively high proportions of mono-ortho congeners, including PCB-118. These historical patterns may partially explain the dominance of PCB-118 in contemporary human samples and underscore its importance in evaluating PCB-related health effects [34].

Our study provides novel and comprehensive evidence linking prenatal exposure to individual and mixed DL-PCBs with adverse fetal growth outcomes in a large birth cohort. Using both single-congener models and advanced mixture approaches (BKMR, gWQS, and g-comp), we found that elevated umbilical cord blood concentrations of specific DL-PCBs, particularly PCB77, PCB126, and PCB157, were significantly associated with lower BW, and to a lesser extent, reduced HC. These findings are broadly consistent with prior epidemiological and toxicological studies, and they add important new insights by identifying congener-specific risks and dose–response patterns in a population with environmental exposure levels. Our findings are in agreement with Patel et al., who reported inverse associations between maternal serum levels of PCB-118, PCB-153, and PCB-187 and daughters’ BµW in the Avon Longitudinal Study, with the strongest effects among mothers with lower education levels a key marker of socioeconomic status also adjusted for in our study [35]. Similarly, Kofoed et al., using registry data from Denmark, found that maternal airborne exposure to lower-chlorinated PCBs was associated with a 32 g reduction in BW and an increased risk of small-for-gestational-age (SGA) births, aligning with our results despite differences in exposure route and assessment methods [10]. Furthermore, our findings resonate with the meta-analysis by Zou et al., which confirmed a significant negative association between prenatal PCB exposure and neonatal BW across several cohorts, with stronger effects observed for umbilical cord serum measurements supporting our use of cord blood as a sensitive biomarker of in utero exposure [36]. Toxicological evidence also substantiates our results. For example, Ge et al. (2023) demonstrated in a mouse model that prenatal exposure to PCB118 disrupted placental angiogenesis and reduced fetal body and placental weights, suggesting mechanistic plausibility for our observed associations with impaired BW and HC [37]. Our work also aligns with the findings of Kezios et al., who observed that prenatal exposure to mono- and di-ortho-substituted PCBs was associated with shorter gestational length, which could partially mediate the relationship between PCB exposure and restricted fetal growth [38]. Additionally, Murphy et al. (2010) found that higher preconception concentrations of anti-estrogenic PCB congeners were associated with substantial reductions in BW, underscoring the relevance of specific biological activity patterns and critical exposure windows [39]. While several studies, such as Berghuis et al. and Balalian et al., have explored longer-term developmental effects and associations with growth outcomes from PCB exposure via seafood consumption, their results were more mixed or of low certainty. However, the negative correlations they reported with BW and head circumference are in line with our findings [13, 40]. Finally, the foundational study by Patandin et al., one of the earliest to link background-level PCB and dioxin exposure to lower BW and postnatal growth in Dutch infants, further corroborates our results by showing that prenatal PCB exposure, even in non-accidentally exposed populations, can hinder early growth [41]. Our study not only supports but extends the existing body of evidence by offering a congener-resolved and mixture-informed analysis of DL-PCB exposure and its associations with fetal growth parameters, while adjusting for a broad range of maternal and perinatal covariates. The consistency of our results with both observational and mechanistic literature highlights the public health relevance of reducing maternal PCB exposure, even at low background levels, to protect fetal growth and long-term developmental outcomes. In interpreting the congener distribution in maternal serum, it is important to recognize that the specific pattern of PCB congeners in a woman’s body reflects not only current exposures but also historical lifetime exposures linked to regional use and environmental sources. Studies have shown that congener profiles in human biological samples, such as serum or breast milk, often resemble the congener patterns found in local environmental sources or consumed foods, indicating that exposure history contributes to the “fingerprint” of PCBs retained in the body [42]. For example, research in the Akwesasne community demonstrated that breast milk PCB congener patterns of women who consumed locally contaminated fish more closely matched the congener profile of those fish than did patterns of women with lower fish consumption, consistent with diet‑driven exposure signatures. Similarly, analyses of serum PCB patterns among Mohawk men revealed that individuals with occupational or dietary exposure histories had serum congener distributions resembling specific commercial Aroclor mixtures previously used in the region. These findings underscore that maternal congener patterns can carry information about past exposures and mixture composition, which in turn may influence internal PCB burdens relevant to fetal and infant exposure [43, 44].

Our sex-stratified analysis revealed notable differences in the strength of association between prenatal DL-PCB exposure and BW by infant sex, although statistical tests for interaction did not indicate significant effect modification. Specifically, PCB77, PCB81, and PCB126 were inversely associated with BW in both male and female infants, with consistently stronger effect estimates observed among males. For example, PCB77 was associated with a 96.40 g reduction in BW among boys compared to a 65.81 g reduction in girls, suggesting potential heightened vulnerability in male fetuses. Similar sex-specific patterns were also evident for PCB105 and PCB123. These findings are biologically plausible given known sex-specific differences in fetal growth regulation, placental function, and susceptibility to environmental stressors [45]. Prior studies have proposed that male fetuses may exhibit a greater growth trajectory but lower resilience to intrauterine toxic exposures, potentially leading to more pronounced impacts from endocrine-disrupting chemicals like DL-PCBs [46, 47]. However, the lack of statistically significant interaction terms in our study (all p-interaction > 0.05) suggests that these apparent differences should be interpreted cautiously and may reflect variation in exposure-response relationships rather than definitive sex-based susceptibility. Nonetheless, our results highlight the importance of considering fetal sex as a potential modifier in environmental epidemiology studies and underscore the need for further research with larger sample sizes or meta-analyses to more clearly delineate sex-specific effects of DL-PCB exposure on fetal growth.

Our study identified a significant mediating role of mtDNA in the association between select DL-PCBs and BW, providing novel evidence that mitochondrial dysfunction may represent a biologically relevant mechanism in fetal growth impairment following prenatal exposure to dioxin-like compounds. Specifically, PCB-77, PCB-81, PCB-118, and PCB-126 showed statistically significant indirect effects mediated through mtDNA, with mediated proportions ranging from approximately 4.4% to 10.8%. Among these, PCB-118 demonstrated the highest proportion mediated (10.76%), highlighting a potential congener-specific influence of DL-PCBs on mitochondrial pathways related to intrauterine growth restriction. These findings are consistent with and extend prior research indicating that mitochondrial biomarkers are sensitive to environmental toxicants and may serve as intermediaries in fetal programming [48–50]. Previous studies have linked higher levels of prenatal DL-PCBs to reduced BW, with possible mechanisms involving oxidative stress, endocrine disruption, and inflammation [51]. Our results align with these reports, yet uniquely suggest that mitochondrial dysfunction, as reflected by altered mtDNA, may be a measurable intermediate step in this process. Notably, PCB-77 and PCB-126, which are among the most potent ligands of the aryl hydrocarbon receptor (AhR), also exhibited significant mtDNA-mediated effects on BW. This supports the hypothesis that AhR activation may impair mitochondrial integrity and biogenesis, which in turn could interfere with placental function and nutrient transfer to the fetus [52]. Experimental studies have demonstrated that prenatal exposure to dioxin-like PCBs impairs mitochondrial respiration and promotes placental oxidative stress, biological events that plausibly reduce fetal growth potential [53, 54]. Conversely, no meaningful mtDNA mediation was observed for BL or head HC, with wide confidence intervals and non-significant p-values across all congeners. This divergence suggests that mtDNA alterations may exert a more pronounced influence on fetal weight rather than skeletal or cranial development. The distinction may reflect differing developmental trajectories and sensitivities of these fetal parameters to mitochondrial function or energy availability during gestation. Overall, our study adds to the growing body of literature highlighting the vulnerability of fetal development to mitochondrial toxicity induced by environmental pollutants. The identification of a measurable mediation effect strengthens the plausibility of mtDNA as a mechanistic biomarker in the developmental origins of health and disease framework, particularly for growth-related outcomes like BW.

Possible biological mechanisms

Our study provides compelling evidence that mitochondrial dysfunction, particularly alterations in mtDNA, may mediate the detrimental effects of prenatal exposure to DL-PCBs on fetal growth, especially BW. Several congeners (e.g., PCB-77, PCB-81, PCB-118, and PCB-126) demonstrated significant indirect effects on BW through changes in mtDNA, suggesting that disruption of mitochondrial integrity is a key mechanistic pathway. These findings are consistent with the hypothesis that DL-PCBs impair placental and fetal mitochondrial function by inducing oxidative stress, leading to compromised energy production and altered metabolic programming in utero [55]. Mitochondria play a central role in energy generation, cellular differentiation, and nutrient exchange across the placenta, all of which are critical for supporting rapid fetal growth [56]. DL-PCBs, as persistent organic pollutants, are known to increase reactive oxygen species (ROS) production and compromise mitochondrial membrane potential, resulting in mtDNA damage or copy number alterations [57]. Such changes have been shown to interfere with mitochondrial biogenesis, respiration, and ATP production, processes that are particularly vital for weight gain during late gestation [58]. Our findings, which quantify the proportion of the BW effect mediated by mtDNA (e.g., 10.76% for PCB-118), extend previous epidemiological observations and point to mitochondrial damage as a quantifiable intermediary between toxicant exposure and fetal growth restriction. A mechanistically plausible contributor to these effects is activation of the aryl hydrocarbon receptor (AhR) by DL-PCBs [59]. AhR is a ligand-activated transcription factor highly expressed in placental and fetal tissues, where it regulates genes involved in xenobiotic metabolism, oxidative stress response, and mitochondrial homeostasis. Dysregulated AhR signaling has been shown to inhibit mitochondrial biogenesis and promote apoptosis, which may further compromise placental efficiency and fetal nutrient supply [52]. Additionally, AhR activation may suppress vascular endothelial growth factor (VEGF) signaling and impair placental angiogenesis, thereby limiting oxygen and nutrient delivery to the fetus [60]. While our mediation analysis supports mitochondrial involvement in BW outcomes, it is notable that no significant mediation via mtDNA was observed for BL or HC. This finding suggests that the effects of mitochondrial disruption may be more pronounced in energy-dependent processes that drive weight accumulation (e.g., fat deposition, protein synthesis) rather than structural development or cranial growth. This aligns with existing literature indicating that different fetal growth parameters may be governed by distinct biological pathways and have variable sensitivity to mitochondrial dysfunction. In addition to mitochondrial damage, several complementary mechanisms likely contribute to DL-PCB-induced growth impairment. Endocrine disruption is well-documented for DL-PCBs, including interference with thyroid hormone signaling, estrogen/androgen pathways, and the insulin-like growth factor (IGF) axis, all of which are critical for fetal development [61]. Thyroid hormones, in particular, regulate metabolism, neurodevelopment, and tissue growth, and their disruption by DL-PCBs has been associated with reductions in BW and HC. Similarly, DL-PCB-induced oxidative stress can trigger inflammatory responses in the placenta, damaging trophoblasts and reducing vascularization, while also altering the expression of nutrient transporters and growth-regulating genes via epigenetic modifications such as DNA methylation and histone acetylation [53]. Collectively, these mechanisms, mitochondrial dysfunction, AhR activation, oxidative stress, endocrine disruption, impaired placental transport, and epigenetic alterations, interact in a complex, synergistic manner to disrupt fetal growth trajectories. Our findings provide novel mechanistic insight into the central role of mitochondria in mediating environmental insults during pregnancy and support the need for future research to further unravel the interplay between these pathways. To better elucidate these mechanisms, future studies should incorporate longitudinal designs with repeated biomarker assessments to track mitochondrial and endocrine alterations across gestation. Experimental animal models, particularly AhR-responsive rodent models, have already demonstrated DL-PCB induced oxidative stress, mitochondrial dysfunction, and fetal growth restriction, and they may provide an important platform for testing mechanistic pathways under controlled exposure conditions. In vitro placental trophoblast systems could also help clarify whether mtDNA damage precedes or follows changes in nutrient transport, angiogenesis, or inflammatory signaling.

Strengths and limitations

This study benefits from several important strengths that support the robustness of the findings. First, the large sample size of 5421 mother–infant pairs provide substantial statistical power to detect associations between individual DL-PCB congeners and fetal growth indicators, as well as to conduct sex-stratified and mixture analyses. The measurement of 12 specific DL-PCB congeners in umbilical cord blood offers an objective and precise assessment of prenatal exposure, capturing the internal dose delivered to the fetus at birth. Importantly, the study comprehensively evaluates multiple fetal growth outcomes, BW, BL, and HC, allowing for a nuanced understanding of how DL-PCBs may differentially affect various aspects of fetal development. The use of advanced statistical approaches such as BKMR, gWQS regression, and g-comp to analyze mixtures adds depth to the analysis by considering the combined and potentially synergistic effects of multiple congeners, which better reflect real-world exposure scenarios. Additionally, adjusting for a wide range of maternal and paternal sociodemographic factors and accounting for infant sex through stratified analyses enhances the validity and interpretability of the associations observed. Nevertheless, some limitations must be considered. Despite the large overall sample, the relatively modest effect sizes observed for individual congeners and the lack of statistically significant sex interactions suggest that residual confounding or measurement variability may influence the results. The reliance on a single time-point measurement of DL-PCBs in cord blood at delivery may not fully capture exposure variability across pregnancy, potentially leading to exposure misclassification. Furthermore, while BW and HC showed significant associations with certain congeners and mixtures, BL did not, which may be due to differences in measurement precision or biological sensitivity among these growth indicators. The complex and sometimes mixed directional weights observed in mixture models highlight the challenges in disentangling the specific roles of individual congeners, pointing to possible non-linear or interactive effects that remain difficult to fully characterize. Finally, as the study population’s sociodemographic and environmental context may differ from other regions, caution is warranted when generalizing findings to broader populations. Overall, these strengths and limitations should be carefully weighed when interpreting the implications of prenatal DL-PCB exposure on fetal growth.

Conclusion

In summary, this large-scale cohort study provides compelling evidence that prenatal exposure to DL-PCBs, measured in umbilical cord blood, is significantly associated with reduced fetal growth, particularly BW and HC. Several individual DL-PCB congeners, including PCB77 and PCB126, showed robust inverse relationships with these fetal growth indicators, while BL appeared less affected. Mixture analyses using advanced statistical models further confirmed the detrimental impact of combined DL-PCB exposures on BW and HC, underscoring the importance of considering real-world exposure scenarios involving multiple congeners. Although some variation in effects by infant sex was observed, no statistically significant sex-based effect modification was identified. Based on our mediation analysis, we conclude that mitochondrial DNA alterations significantly mediate the association between prenatal DL-PCB exposure and reduced birth weight, highlighting mitochondrial dysfunction as a key biological pathway through which environmental toxicants may impair fetal growth. These findings highlight prenatal DL-PCB exposure as a modifiable risk factor for impaired fetal growth, which is a critical determinant of newborn health and long-term developmental outcomes. From a public health perspective, the results emphasize the urgent need for continued monitoring and reduction of DL-PCB exposure among pregnant women to protect fetal development and prevent adverse birth outcomes. Policymakers should consider stricter regulations on persistent organic pollutants like DL-PCBs and promote interventions to minimize environmental contamination and human exposure, particularly in vulnerable populations. Furthermore, public health initiatives should focus on education and awareness campaigns targeting pregnant women regarding potential sources of DL-PCBs and ways to reduce exposure. Future research is warranted to elucidate the underlying biological mechanisms linking DL-PCB exposure to fetal growth impairment and to explore the potential long-term health consequences for exposed children. Longitudinal studies with repeated exposure measurements throughout pregnancy would improve exposure assessment accuracy and help clarify critical windows of susceptibility. Furthermore, exploring the joint impact of DL-PCBs alongside other environmental contaminants is crucial to comprehensively grasp the intricate interactions of chemical mixtures affecting fetal development. Such evidence will be vital to inform targeted interventions and refine regulatory standards aimed at safeguarding maternal and child health.

Authors’ contributions

Huisheng Yao, Chao Jiang, Hongyan Zhang, and Si Si contributed equally to this work. Xingqiang Li conceived and designed the study, while Huisheng Yao and Chao Jiang were responsible for data collection. Hongyan Zhang and Si Si carried out data analysis and interpretation. Yue Zhang provided clinical insights and contributed to critical revisions of the manuscript. Xingqiang Li supervised the entire project. All authors participated in drafting and revising the manuscript and approved the final version for submission.

Funding

This research was supported by Shengjing Hospital of China Medical University (grant number:202237589).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of China Medical University (Protocol Number: CMU/2022/IRB-045). Written informed consent was obtained from the parents or legal guardians of all participating children prior to inclusion in the study.

Written informed consent was obtained from the parents or legal guardians of all participants prior to study participation.

Competing interests

The authors declare no competing interests.