Associations between pyrethroid insecticide concentrations and dyslipidemia: Findings from the Dongfeng-Tongji cohort

Abstract

Pyrethroids are widely used insecticides worldwide, yet their effects on dyslipidemia remain incompletely understood. Our study quantified serum concentrations of seven parent pyrethroids in participants from the Dongfeng-Tongji cohort and evaluated both cross-sectional and prospective associations with lipid levels and dyslipidemia. In cross-sectional analyses, higher serum concentrations of fenvalerate and deltamethrin were consistently associated with increased odds of dyslipidemia and its subtypes, including hypercholesterolemia, hypertriglyceridemia, and hyperbetalipoproteinemia, whereas cyfluthrin was associated with higher odds of hypertriglyceridemia. Prospective analyses revealed that higher baseline serum concentrations of fenvalerate were associated with annual increases in total cholesterol and low-density lipoprotein cholesterol, as well as elevated risk of incident hypertriglyceridemia. Mixture analyses identified fenvalerate as an important contributor in several dyslipidemia subtype models. These findings suggest that pyrethroid exposure, particularly fenvalerate, may contribute to dyslipidemia and highlight the need for further mechanistic investigation and risk evaluation.

INTRODUCTION

Pyrethroids are a class of broad-spectrum synthetic insecticides widely used worldwide due to their high efficacy and relatively low mammalian toxicity (1). In recent years, they have largely replaced conventional insecticides, accounting for more than 30% of the insecticide market (2). Pyrethroids have been commonly detected in agricultural products and environmental media, including air, water, and dust (3). Human exposure occurs primarily through dietary intake and air inhalation (4). Despite their short half-life, the potential health effects of pyrethroids remain unclear.

Dyslipidemia is a highly prevalent metabolic disorder characterized by elevated levels of total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), and decreased levels of high-density lipoprotein cholesterol (HDL-C) (5). Dyslipidemia subtypes, such as hypercholesterolemia and hypertriglyceridemia, are common and clinically important lipid metabolism disorders (6). According to the Global Burden of Disease study, elevated LDL-C levels are now the third leading attributable risk factor for cardiovascular disease (CVD) in China (7). While genetic factors contribute to the development of dyslipidemia (8), environmental exposures, including endocrine-disrupting chemicals like pyrethroids, may also play a notable role in its pathogenesis.

Although several epidemiological studies have examined the association between pyrethroid exposure and lipid metabolism, the overall evidence remains limited and inconclusive. Most existing studies are cross-sectional and assess lipid parameters as continuous variables, which limits clinical relevance as dyslipidemia is typically defined on the basis of established clinical thresholds (9, 10). For instance, a cross-sectional study from Korea (n = 3692) reported positive associations between higher urinary levels of 3-phenoxybenzoic acid (3-PBA), a nonspecific pyrethroid metabolite, and serum levels of TC, TG, HDL-C, and LDL-C (10). Another cross-sectional study in India (n = 36) found that exposure to allethrin and prallethrin was significantly associated with increased levels of TG and very low-density lipoprotein levels (9). A cross-sectional study in China, including 2012 participants, also reported that serum concentrations of deltamethrin were correlated with various plasma lipid metabolites that may influence lipid metabolism (11). In contrast, a study conducted in Belgium and Luxembourg (n = 989) found no significant association between 3-PBA levels and dyslipidemia (12). These discrepancies may arise from differences in study design, population characteristics, and outcome definitions; in particular, variation in exposure assessment methods represents a key source of inconsistency across studies. Most prior studies relied on urinary 3-PBA, a short-lived and nonspecific metabolite that does not distinguish among individual pyrethroids and may introduce exposure misclassification (2, 13). By contrast, direct quantification of serum parent compounds enables compound-specific evaluation and may better reflect biologically relevant internal exposure (14). Such differences in exposure assessment likely contribute to heterogeneity in the existing literature.

Despite the gaps in human evidence, toxicological studies have provided biological plausibility for the role of pyrethroids in lipid dysregulation (15, 16). Experimental findings suggest that pyrethroid exposure can induce oxidative stress, trigger inflammatory responses, and disrupt endocrine signaling (15, 16). Given the increasing prevalence of dyslipidemia (17), identifying modifiable environmental risk factors, including pyrethroid exposure, has important public health implications. Nevertheless, high-quality epidemiological studies using prospective designs, clinically defined outcomes, and compound-specific exposure assessments are still lacking.

Therefore, this study combined cross-sectional and prospective analyses based on the Dongfeng-Tongji cohort to evaluate the association between pyrethroid exposure and dyslipidemia, annual average changes in lipid levels, and the risk of incident dyslipidemia. We hypothesized that serum pyrethroid concentrations are positively associated with dyslipidemia, with potentially stronger associations for specific compounds and the possibility of joint effects under mixed exposure.

RESULTS

Population characteristics

This study combined cross-sectional and prospective analyses based on the Dongfeng-Tongji cohort. The flowchart is presented in Fig. 1. Table 1 shows the basic characteristics of the participants in this cross-sectional study. The median age was 66.0 years [interquartile range (IQR): 60.9, 71.5] in participants with dyslipidemia and 65.8 years (IQR: 60.5, 71.3) in those without dyslipidemia. Compared with those without dyslipidemia, participants with dyslipidemia were more likely to be female (52.0%), nonsmokers (67.5%), had a higher body mass index (BMI) (median, 24.5 kg/m²), higher fruit intake (56.3%), and a higher prevalence of hypertension (73.0%) and diabetes (27.9%). Participants with dyslipidemia exhibited higher lipid levels and higher serum concentrations of fenvalerate and deltamethrin.

Fig. 1. Flow chart of the study.

Table 1. Basic characteristics of the study participants (cross-sectional, n = 3666).

Data are presented as median [interquartile range (IQR)] and number (percentages) for continuous and categorical variables. Pyrethroid concentrations are expressed on a wet-weight basis (nanograms per milliliter). Continuous variables were compared using the Mann-Whitney U test. Categorical variables were compared using the chi-square test or Fisher’s exact test, as appropriate.

	Dyslipidemia (n = 2339)	Nondyslipidemia (n = 1327)	P value
Age, years	66.0 (60.9, 71.5)	65.8 (60.5, 71.3)	0.31
Female, n (%)	1216 (52.0)	583 (43.9)	<0.001
Education, ≥12 years, n (%)	249 (10.7)	128 (9.7)	0.21
Smoking, ever, n (%)	759 (32.5)	479 (36.1)	0.031
Alcohol drinking, ever, n (%)	781 (33.4)	467 (35.2)	0.47
Physically active, yes, n (%)	1883 (80.5)	1065 (80.3)	0.98
Diet categories, ≥5 times/week, n (%)
Meat	875 (37.4)	486 (36.6)	0.54
Fish or seafood	266 (11.4)	127 (9.6)	0.21
Vegetables	2232 (95.4)	1259 (94.9)	0.41
Grains	666 (28.5)	349 (26.3)	0.12
Fruits	1316 (56.3)	685 (51.6)	0.023
BMI, kg/m2	24.5 (22.7, 26.7)	23.5 (21.2, 25.6)	<0.001
TG, mM	1.6 (1.1, 2.2)	1.0 (0.8, 1.2)	<0.001
TC, mM	5.2 (4.5, 5.8)	4.4 (3.9, 4.8)	<0.001
LDL-C, mM	3.0 (2.3, 3.6)	2.4 (2.0, 2.8)	<0.001
HDL-C, mM	1.3 (1.1, 1.7)	1.4 (1.3, 1.7)	<0.001
Total lipids, mg/dl	660.4 (597.4, 740.0)	531.2 (480.9, 572.7)	<0.001
Hypertension, yes, n (%)	1708 (73.0)	765 (57.7)	<0.001
Diabetes, yes, n (%)	653 (27.9)	212 (16.0)	<0.001
Pyrethroid concentrations
Fenvalerate, ng/ml	0.05 (0.02, 0.08)	0.04 (0.02, 0.07)	0.001
Deltamethrin, ng/ml	0.08 (0.05, 0.11)	0.07 (0.04, 0.11)	0.015
Cyfluthrin, ng/ml	0.01 (0.00, 0.01)	0.01 (0.00, 0.01)	0.29

Table S1 summarizes baseline characteristics of the overall cohort and the subset included in the prospective analysis. The prospective subset had a slightly lower proportion of females (43.9%), a higher proportion of ever smokers (36.1%), a lower BMI (median, 23.5 kg/m²), and a slightly lower concentration of fenvalerate. At the same time, other characteristics remained broadly comparable, suggesting the representativeness of the follow-up sample.

Serum pyrethroid concentrations, correlations, and longitudinal variability

Among the three pyrethroids with a detection rate higher than 70%, deltamethrin exhibited the highest median concentration (0.0903 ng/ml), followed by fenvalerate (0.0525 ng/ml) and cyfluthrin (0.0083 ng/ml) (table S2). Table S3 presents positive correlations among these three compounds. Longitudinal reproducibility analysis between 2013 and 2018 revealed a modest but statistically significant intraclass correlation coefficient (ICC) for fenvalerate (ICC = 0.18, P < 0.05), indicating notable within-person variability over time, as expected for nonpersistent environmental chemicals (table S4).

Association with lipid levels (cross-sectional)

The restricted cubic spline (RCS) analysis indicated significant exposure-response relationships between serum pyrethroid concentrations and lipid levels (fig. S1). For fenvalerate, the associations significantly deviated from linearity, showing a steeper increase at higher exposure levels for TC, TG, and total lipids (all P nonlinear < 0.05). Similar nonlinear patterns were observed for deltamethrin with TG and total lipids (both P nonlinear < 0.05). Cyfluthrin showed a positive association with TG, with an approximately linear trend (P overall < 0.05).

Association with dyslipidemia and its subtypes

In multivariable logistic regression models (Table 2), after adjusting for potential confounders, higher serum fenvalerate concentrations (T3 versus T1) were significantly associated with increased odds of dyslipidemia, hypercholesterolemia, and hyperbetalipoproteinemia, with odds ratios (ORs) [95% confidence intervals (CIs)] of 1.33 (1.12, 1.58), 1.47 (1.24, 1.75), and 1.53 (1.25, 1.86), respectively. In addition, each unit increase in inverse normal transformed (INT) fenvalerate corresponded to 11 to 21% higher odds of dyslipidemia and major subtypes (all P < 0.05).

Table 2. Association of serum concentrations of fenvalerate, deltamethrin, and cyfluthrin with dyslipidemia and its subtypes.

Participants were categorized into tertiles on the basis of the distribution of each serum pyrethroid concentration in the total population. P trend was derived from a similar model, in which tertiles were treated as continuous variables and tested for linear trend across tertiles. Models were adjusted for age, gender, BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes. For each pyrethroid, T1 (the lowest tertile) was used as the reference group. T, tertile.

	Tertiles of serum pyrethroid concentrations			P trend	Linear model
	T1	T2	T3
Dyslipidemia
Fenvalerate	<0.029	0.029–0.063	≥0.063
N (cases/total)	753/1222	762/1222	824/1222
OR (95% CI)	1.00	1.02 (0.86, 1.21)	1.33 (1.12, 1.58)	0.001	1.14 (1.06, 1.23)
Deltamethrin	<0.056	0.056–0.098	≥0.098
N (cases/total)	752/1222	781/1222	806/1222
OR (95% CI)	1.00	1.14 (0.96, 1.35)	1.25 (1.06, 1.49)	0.010	1.12 (1.04, 1.21)
Cyfluthrin	<0.005	0.005–0.009	≥0.009
N (cases/total)	779/1222	764/1222	796/1222
OR (95% CI)	1.00	0.95 (0.80, 1.13)	1.08 (0.91, 1.28)	0.41	1.05 (0.97, 1.13)
Hypercholesterolemia
Fenvalerate	<0.029	0.029–0.063	≥0.063
N (cases/total)	356/1219	404/1219	457/1219
OR (95% CI)	1.00	1.21 (1.02, 1.44)	1.47 (1.24, 1.75)	<0.001	1.21 (1.12, 1.30)
Deltamethrin	<0.056	0.056–0.098	≥0.098
N (cases/total)	349/1219	413/1219	455/1219
OR (95% CI)	1.00	1.28 (1.07, 1.53)	1.47 (1.24, 1.76)	<0.001	1.20 (1.11, 1.29)
Cyfluthrin	<0.005	0.005–0.009	≥0.009
N (cases/total)	403/1219	384/1220	430/1218
OR (95% CI)	1.00	0.94 (0.79, 1.11)	1.11 (0.94, 1.32)	0.26	1.05 (0.97, 1.13)
Hypertriglyceridemia
Fenvalerate	<0.029	0.029–0.063	≥0.063
N (cases/total)	349/1219	328/1219	379/1219
OR (95% CI)	1.00	0.90 (0.75, 1.09)	1.16 (0.97, 1.39)	0.11	1.11 (1.03, 1.20)
Deltamethrin	<0.056	0.056–0.098	≥0.098
N (cases/total)	342/1219	327/1219	387/1219
OR (95% CI)	1.00	0.94 (0.78, 1.13)	1.23 (1.03, 1.48)	0.032	1.09 (1.01, 1.18)
Cyfluthrin	<0.005	0.005–0.009	≥0.009
N (cases/total)	334/1219	342/1220	380/1218
OR (95% CI)	1.00	1.06 (0.88, 1.27)	1.25 (1.04, 1.50)	0.017	1.10 (1.02, 1.20)
Hyperbetalipoproteinemia
Fenvalerate	<0.029	0.029–0.063	≥0.063
N (cases/total)	219/1219	263/1219	303/1219
OR (95% CI)	1.00	1.25 (1.02, 1.53)	1.53 (1.25, 1.86)	<0.001	1.19 (1.09, 1.29)
Deltamethrin	<0.056	0.056–0.098	≥0.098
N (cases/total)	228/1219	284/1219	273/1219
OR (95% CI)	1.00	1.33 (1.09, 1.62)	1.26 (1.03, 1.54)	0.017	1.12 (1.02, 1.21)
Cyfluthrin	<0.005	0.005–0.009	≥0.009
N (cases/total)	253/1219	250/1220	282/1218
OR (95% CI)	1.00	1.00 (0.82, 1.22)	1.17 (0.96, 1.42)	0.12	1.07 (0.98, 1.16)
Hypoalphalipoproteinemia
Fenvalerate	<0.029	0.029–0.063	≥0.063
N (cases/total)	125/1219	87/1219	98/1219
OR (95% CI)	1.00	0.67 (0.50, 0.90)	0.80 (0.60, 1.07)	0.11	0.88 (0.78, 1.00)
Deltamethrin	<0.056	0.056–0.098	≥0.098
N (cases/total)	119/1219	110/1219	81/1219
OR (95% CI)	1.00	0.94 (0.71, 1.24)	0.68 (0.50, 0.92)	0.016	0.88 (0.77, 1.00)
Cyfluthrin	<0.005	0.005–0.009	≥0.009
N (cases/total)	114/1219	106/1220	90/1218
OR (95% CI)	1.00	0.91 (0.68, 1.21)	0.78 (0.58, 1.04)	0.097	0.93 (0.82, 1.06)

Similarly, the T3 group of deltamethrin had higher odds of dyslipidemia, hypercholesterolemia, hypertriglyceridemia, hyperbetalipoproteinemia, and lower odds of hypoalphalipoproteinemia compared with the T1 subgroup, with ORs (95% CIs) of 1.25 (1.06, 1.49), 1.47 (1.24, 1.76), 1.23 (1.03, 1.48), 1.26 (1.03, 1.54), and 0.68 (0.50, 0.92), respectively. In the linear model, deltamethrin was associated with a 12, 20, 9, and 12% higher odds of dyslipidemia, hypercholesterolemia, hypertriglyceridemia, and hyperbetalipoproteinemia, respectively (all P < 0.05).

The T3 group of serum cyfluthrin concentration exhibited an association with hypertriglyceridemia, with OR (95% CI) of 1.25 (1.04, 1.50). Cyfluthrin was associated with 10% higher odds of hypertriglyceridemia in the linear model (P < 0.05).

Among compounds with lower detection frequencies, fenpropathrin was associated with increased odds of dyslipidemia, hypercholesterolemia, and hyperbetalipoproteinemia, whereas permethrin was linked to increased odds of hyperbetalipoproteinemia (table S5).

Sensitivity analyses excluding participants receiving lipid-lowering therapy (n = 522) yielded comparable results, supporting the robustness of the primary findings (tables S6 and S7). Stratified analyses indicated that associations between deltamethrin and dyslipidemia were more pronounced in females and nonsmokers. Positive associations for fenvalerate were observed in most subgroups, with evidence of effect modification by hypertension status (fig. S2).

Association of pyrethroid mixture with dyslipidemia and its subtypes

In the mixture analysis, both quantile g-computation (Q-gcomp) (OR = 1.22; 95% CI: 1.08 to 1.37; Fig. 2A) and weighted quantile sum (WQS) (OR = 1.24; 95% CI: 1.08 to 1.42; Fig. 3A) models indicated that joint pyrethroid exposures were significantly associated with dyslipidemia, with fenvalerate contributing the greatest weight across hypercholesterolemia, hyperbetalipoproteinemia, and hypoalphalipoproteinemia in WQS models. Figures 2 and 3 show similar associations for dyslipidemia subtypes, with consistent mixture effects across hypercholesterolemia, hypertriglyceridemia, hyperbetalipoproteinemia, and hypoalphalipoproteinemia. Given the dominant contribution of fenvalerate in the mixture models, sensitivity analyses excluding fenvalerate were conducted. The overall mixture association with dyslipidemia was attenuated, with reduced effect estimates and some associations no longer reaching statistical significance (figs. S3 and S4).

Fig. 2. Associations between pyrethroid mixtures and dyslipidemia and its subtypes estimated using the Q-gcomp model.

(A) dyslipidemia; (B) hypercholesterolemia; (C) hypertriglyceridemia; (D) hyperbetalipoproteinemia; and (E) hypoalphalipoproteinemia. Component weights were obtained from the nonbootstrap quantile g-computation (Q-gcomp) model, whereas the overall mixture ORs and 95% CIs were estimated using 1000 bootstrap samples. Bars represent the weights of individual pyrethroids contributing to the overall mixture effect for each outcome. Positive and negative component weights were normalized separately within each direction and should not be compared directly across directions. ORs and 95% CIs indicate the total effect of the mixture on dyslipidemia and its subtypes. Models were adjusted for age, gender, BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes.

Fig. 3. Associations between pyrethroid mixtures and dyslipidemia and its subtypes estimated using the WQS model.

(A) dyslipidemia; (B) hypercholesterolemia; (C) hypertriglyceridemia; (D) hyperbetalipoproteinemia; and (E) hypoalphalipoproteinemia. Weighted quantile sum (WQS) regression was used to evaluate the associations between serum pyrethroid mixtures and dyslipidemia and its subtypes. [(A) to (D)] Positive overall mixture effects. (E) A negative overall effect; a distinct color scheme was used in (E) to highlight this opposite direction of association. Models were adjusted for age, gender, BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes.

Association with the annual change in lipid levels

In prospective analysis, compared with the lowest exposure group, individuals in the middle-to-high fenvalerate exposure group had an average annual increase of 0.027 mM in TC and 0.022 mM in LDL-C (P = 0.027 and 0.028, respectively; Table 3). Stratified analyses suggested effect modification by hypertension status for the associations of fenvalerate with ΔTC, ΔTG, and ΔLDL-C (fig. S5).

Table 3. Association of serum concentrations of fenvalerate, deltamethrin, and cyfluthrin with annual average changes in lipid levels.

Annual average change in blood lipid levels (Δlipid) was calculated as: [follow-up lipid level − baseline lipid level)/follow-up duration (years)]. Serum pyrethroid concentrations were divided into tertiles, with the second and third tertiles combined as the middle-to-high exposure group and the lowest tertile as the reference group. Models were adjusted for age, gender, baseline lipid level, BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes.

	β (95% CI)	P value
Fenvalerate
ΔTC (mM/year)	0.027 (0.003, 0.050)	0.027
ΔTG (mM/year)	0.005 (−0.012, 0.021)	0.58
ΔLDL-C (mM/year)	0.022 (0.002, 0.042)	0.028
ΔHDL-C (mM/year)	0.001 (−0.008, 0.010)	0.82
ΔTotal lipids (mg/dl per year)	2.724 (−0.143, 5.591)	0.063
Deltamethrin
ΔTC (mM/year)	0.003 (−0.021, 0.026)	0.83
ΔTG (mM/year)	−0.004 (−0.020, 0.013)	0.67
ΔLDL-C (mM/year)	−0.007 (−0.027, 0.013)	0.48
ΔHDL-C (mM/year)	−0.002 (−0.012, 0.007)	0.66
ΔTotal lipids (mg/dl per year)	−0.076 (−2.963, 2.811)	0.96
Cyfluthrin
ΔTC (mM/year)	0.019 (−0.004, 0.043)	0.11
ΔTG (mM/year)	0.002 (−0.015, 0.018)	0.84
ΔLDL-C (mM/year)	0.015 (−0.004, 0.035)	0.13
ΔHDL-C (mM/year)	0.004 (−0.005, 0.013)	0.42
ΔTotal lipids (mg/dl per year)	1.868 (−0.996, 4.733)	0.20

No significant associations were observed for deltamethrin and cyfluthrin, whereas permethrin was associated with an average annual decrease of 0.021 mM in LDL-C (Table 3 and table S8, P = 0.023), and results were largely unchanged after excluding participants receiving lipid-lowering therapy (tables S9 and S10).

Association with the risk of incident dyslipidemia and its subtypes

The prospective analyses revealed significant associations between higher serum fenvalerate concentrations and the risk of incident hypertriglyceridemia (OR = 3.44; 95% CI: 1.51 to 7.82). Additionally, higher concentrations of cyfluthrin were associated with an increased risk of hypercholesterolemia (OR = 2.41; 95% CI: 1.38 to 4.21) and hyperbetalipoproteinemia (OR = 4.49; 95% CI: 1.47 to 13.70) (Fig. 4).

Fig. 4. Forest plot of serum concentrations of fenvalerate, deltamethrin, and cyfluthrin with incident dyslipidemia and its subtypes.

(A) dyslipidemia; (B) hypercholesterolemia; (C) hypertriglyceridemia; (D) hyperbetalipoproteinemia; and (E) hypoalphalipoproteinemia. Low group refers to the reference group (the lowest tertile of serum pyrethroid concentrations), and high group refers to the middle-to-high exposure group (combined the second and third tertiles of serum pyrethroid concentrations). Models were adjusted for BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes. *P < 0.05. Arrows indicate that the upper confidence limit extends beyond the plotted range.

DISCUSSION

In this large prospective cohort, serum concentrations of several pyrethroid insecticides were associated with adverse lipid profiles and dyslipidemia. Among the measured compounds, fenvalerate showed the most consistent associations across both cross-sectional and prospective analyses, whereas deltamethrin and cyfluthrin showed more selective associations with specific lipid outcomes (table S11). Together, these findings suggest that pyrethroids may have both shared and compound-specific associations with lipid dysregulation, with fenvalerate emerging as the most reproducible signal in the present study.

Participants were selected from the Dongfeng-Tongji cohort, an ongoing prospective cohort of retired workers in Hubei Province. Compared with previous reports, parent pyrethroid concentrations in this study were generally lower than those reported in some occupational settings and broadly consistent with low-level exposure in general populations. In this cohort, exposure is likely derived from common sources such as dietary intake, residential pesticide use, and environmental contamination through air or dust (2, 18, 19). For example, nonoccupational populations across 15 provinces in China showed geometric mean concentrations of 0.1548 ng/ml for fenvalerate (n = 451) and 0.7432 ng/ml for cyfluthrin (n = 232) (20), whereas another study reported markedly higher deltamethrin concentrations (65.0 ng/ml; n = 100) among occupational flower farm workers (21). Our findings indicate that pyrethroid-lipid associations may be detectable under relatively low background exposure conditions.

The different patterns observed across cross-sectional and prospective analyses likely reflect differences in exposure timing, temporal variability, and the biological time scale of dyslipidemia development. Pyrethroids are rapidly metabolized, and a single serum measurement may therefore primarily capture recent exposure rather than longer-term burden (22, 23). Short-term perturbations in hepatic lipid synthesis, lipoprotein secretion, oxidative stress, or endocrine signaling may be detectable in cross-sectional analyses, whereas sustained dyslipidemia development involves longer-term transcriptional regulation, metabolic adaptation, and homeostatic control (24–26). This framework may help explain why some compounds, particularly deltamethrin, showed clearer cross-sectional than longitudinal associations. In addition, within-person temporal variability may attenuate prospective associations when exposure is assessed only once. Behavioral changes after diagnosis, including dietary modification, altered pesticide contact, or medication use, may also contribute to weaker longitudinal signals. Heterogeneity across specific lipid fractions further suggests that different pyrethroids may influence lipid metabolism through partially distinct biological pathways (27–29). These considerations indicate that partial divergence between cross-sectional and prospective findings should not necessarily be interpreted as inconsistency, but rather as reflecting the complex interplay between short-lived exposures and long-term metabolic phenotypes.

Fenvalerate represented the strongest and most consistent signal in this study. It was associated with dyslipidemia and several clinically relevant subtypes in cross-sectional analyses, contributed the largest weight in mixture models of major subtypes, and was also prospectively associated with annual increases in TC and LDL-C levels, as well as incident risks of lipid disorder. Its dominant contribution in mixture analyses likely reflects both exposure characteristics and biological properties. In this cohort, fenvalerate showed a high detection frequency and the greatest temporal reproducibility among the measured pyrethroids, which may enhance its contribution in regression-based mixture models. Some attenuation of associations in mixture analyses relative to single-pollutant models is expected because single-pollutant models do not fully account for correlated coexposures, whereas mixture models redistribute shared variance across compounds. Thus, these differences may reflect modeling strategy rather than conflicting results (30). From a clinical and public health perspective, the annual increases in TC and LDL-C associated with higher fenvalerate exposure were modest at the individual level; however, persistent higher exposure over time could shift lipid trajectories in an unfavorable direction. Even small rightward shifts in population-level lipid distributions may have meaningful implications for long-term dyslipidemia and cardiovascular risk (31–33), particularly in populations with widespread and continuous environmental exposure.

Several mechanisms may plausibly underlie the consistent associations observed for fenvalerate. Experimental studies suggest that pyrethroids can induce oxidative stress, inflammatory responses, and disturbances in nuclear receptor-mediated lipid regulation, including pathways involving peroxisome proliferator–activated receptor (PPAR) signaling (29, 34–36). However, these shared pathways do not by themselves explain the more cholesterol-related pattern observed for fenvalerate. A more specific explanation may involve its endocrine activity. Experimental evidence indicates that fenvalerate can interfere with steroid hormone signaling and estrogen-responsive pathways, which may influence hepatic cholesterol homeostasis through effects on LDL receptor expression, cholesterol synthesis, and lipoprotein turnover (27, 37–39). This mechanism is more consistent with the observed associations of fenvalerate with TC- and LDL-C–related outcomes. Although these mechanisms were not directly examined in the present study, the consistency of our epidemiological findings and experimental evidence collectively supports fenvalerate as a dominant component within the measured pyrethroid mixture. Future mechanistic research is warranted to separate the effects of exposure patterns from inherent toxicological potency.

By contrast, the associations for deltamethrin and cyfluthrin appeared more selective. Deltamethrin was mainly associated with dyslipidemia and related subtypes in cross-sectional analyses, including TG-related outcomes, with stronger associations evident among females and nonsmokers. These findings suggest that deltamethrin exposure may be more strongly linked to short-term metabolic perturbations than to sustained longitudinal changes. Experimental studies suggest that deltamethrin may affect lipid metabolism through oxidative stress, inflammatory responses, and disruption of thyroid and steroid hormone signaling (40–43). These pathways may contribute to broader lipid dysregulation, although current evidence does not support a clearly TG-specific mechanism. The more pronounced associations in nonsmokers may reflect the absence of nicotine-related metabolic perturbations, which could, otherwise, obscure the effects of environmental exposures on lipid regulation (44, 45). Cyfluthrin showed fewer associations overall, but its cross-sectional association with hypertriglyceridemia and prospective associations with selected incident lipid disorders suggest that it should not be dismissed. Human evidence for cyfluthrin remains limited, but experimental studies implicate oxidative stress and lipid peroxidation as plausible mechanisms (36, 46). Together, these findings support the view that heterogeneity across pyrethroids may reflect compound-specific toxicokinetics and biological activity rather than a single shared mechanism.

This study has several strengths. First, it combined both cross-sectional and prospective analyses within a well-characterized cohort. This enhances the robustness of the findings and contributes to a more comprehensive understanding of the potential effects of pyrethroid insecticides on lipid metabolism. Second, this study measured the serum concentrations of seven parent pyrethroid compounds, allowing for a more comprehensive assessment of exposure to individual pyrethroid types and reducing the risk of exposure misclassification caused by the overlapping metabolic pathways of different pyrethroids. Several limitations should be acknowledged. First, exposure assessment relied on a single serum measurement, and only fenvalerate showed temporal reproducibility. Although its ICC was comparable to those reported for other nonpersistent environmental chemicals (typically, 0.1 to 0.3) (47–49), it still indicates within-person variability over time, which may introduce nondifferential exposure misclassification and attenuate the true associations. Notably, the consistent associations observed for fenvalerate across both cross-sectional and prospective analyses support the robustness of our findings. Second, specific exposure routes, including detailed dietary intake and inhalation, were not quantitatively assessed. Although major dietary categories were adjusted using standardized variables, residual confounding from more detailed dietary patterns or specific food items cannot be entirely excluded. Third, lipid-related molecular biomarkers, including markers of LDL receptor activity or PPAR-related signaling, were not measured, which precluded direct validation of the proposed mechanisms at the population level. Last, unmeasured confounding from coexposures, genetic susceptibility, or other unknown factors cannot be ruled out. Despite these limitations, converging experimental evidence supports our interpretations, and future studies should replicate these findings in diverse populations and adopt cohort or nested case-control designs incorporating lipid-related molecular biomarkers to further clarify the underlying mechanisms.

Overall, our findings suggest that pyrethroid exposure, particularly fenvalerate exposure, may contribute to dyslipidemia. Given the widespread presence of pyrethroids in the environment, continued biomonitoring, exposure reduction efforts, and further mechanistic investigations are warranted. Females may represent a potentially susceptible subgroup and warrant particular attention in future risk assessment and mechanistic research.

MATERIALS AND METHODS

Study population

This study was conducted in the Dongfeng-Tongji cohort, a prospective study initiated in 2008 among retired employees of Dongfeng Motor Corporation, a large state-owned enterprise in Shiyan, Hubei, China (50). In 2013, after the first follow-up, 38,295 participants were enrolled. They completed standardized questionnaires, underwent physical exams, and provided fasting blood samples. After excluding individuals with preexisting CVD, cancers, and insufficient blood samples, 24,415 relatively healthy participants remained at baseline in 2013. Among them, 3841 participants underwent serum pyrethroid measurements, and 3666 passed quality control and were included in the present analysis (51).

For the cross-sectional analysis at baseline in 2013, after excluding 9 individuals with missing lipid data, we assessed nonlinear relationships between serum pyrethroid concentrations and lipid levels among 3657 participants. Among the 3666 participants with complete diagnostic information, 2339 were identified as having dyslipidemia. Subtypes of dyslipidemia were further classified among the 3657 participants with complete lipid data, including 1217 cases of hypercholesterolemia, 1056 cases of hypertriglyceridemia, 785 cases of hyperbetalipoproteinemia, and 310 cases of hypoalphalipoproteinemia.

For the prospective investigation in 2018, we included 1327 participants after excluding those who had been diagnosed with dyslipidemia at baseline in 2013 (n = 2339). Among the 1327 participants, 822 with complete lipid data were included in the annual lipid change analysis. We examined incident dyslipidemia and its subtypes using a nested case-control design. A total of 381 case-control pairs of dyslipidemia matched 1:1 by gender and age (±2 years) were selected from the 1327 participants with complete diagnostic information. Similarly, for the dyslipidemia subtypes, 159, 102, 74, and 66 matched pairs for incident hypercholesterolemia, hypertriglyceridemia, hyperbetalipoproteinemia, and hypoalphalipoproteinemia, respectively, were included using the same matching criteria among the 822 participants with complete lipid data. We further assessed the long-term stability of serum pyrethroid concentrations by calculating ICCs over a 5-year interval in 125 participants who attended both the 2013 baseline and 2018 follow-up surveys and had sufficient serum volume (51, 52).

The research protocol was approved by the Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology (approval number 2012-10). All participants provided written informed consent prior to participation.

Laboratory measurements

We assessed seven serum pyrethroids, including bifenthrin, fenpropathrin, cyfluthrin, cyhalothrin, permethrin, fenvalerate, and deltamethrin, using gas chromatography–triple quadrupole mass spectrometry (Agilent 8890 GC-7010B) according to previously published protocols (53). Sample preparation included two rounds of liquid-liquid extractions and one round of solid-phase extraction. Each analytical batch included four quality control samples: one blank sample (negative control), two positive controls at low (0.5 ng/ml) and high (2 ng/ml) concentrations, and one pooled sample composed of mixed serum from study participants. The intra- and inter-assay coefficients of variation of the quality control samples were all below 15% (54). The limits of detection (LODs) for seven pyrethroids ranged from 0.003 to 0.2 ng/ml (table S2). Concentrations below the LOD were imputed as LOD/$\sqrt{2}$. Among them, cyfluthrin, fenvalerate, and deltamethrin were detected in over 70% of the samples and were included as continuous variables. Fenpropathrin and permethrin were dichotomized on the basis of detectability. Although fenpropathrin was detected in more than 50% of samples, its distribution was highly skewed, with many values clustered near the limit of quantitation. Continuous or multicategory modeling yielded unstable estimates, so we dichotomized this compound to improve model stability. Bifenthrin and cyhalothrin had low detection rates (0.16 and 0.08%, respectively) and were excluded from the subsequent analysis.

Outcome assessment

Dyslipidemia was diagnosed according to the 2023 Chinese guidelines for lipid management (6). Participants were classified as having dyslipidemia if they met any of the following criteria: (i) TC ≥ 5.2 mM, (ii) TG ≥ 1.7 mM, (iii) LDL-C ≥ 3.4 mM, (iv) HDL-C < 1.0 mM, (v) self-reported physician-diagnosed dyslipidemia, or (vi) receiving lipid-lowering therapy in the past 2 weeks. The four subtypes of dyslipidemia were diagnosed as follows: hypercholesterolemia (TC ≥ 5.2 mM), hypertriglyceridemia (TG ≥ 1.7 mM), hyperbetalipoproteinemia (LDL-C ≥ 3.4 mM), and hypoalphalipoproteinemia (HDL-C < 1.0 mM). Annual changes in lipid levels were calculated as (lipid levels at follow-up − lipid levels at baseline)/follow-up duration (years).

Covariate definition

During the baseline survey in 2013, information was collected regarding sociodemographic characteristics (gender, age, and education level), lifestyle factors (smoking status, alcohol drinking status, dietary intake frequency, and physical activity), and medication use in the past 2 weeks. Physical examinations, including measurements of weight, standing height, blood pressure, and blood lipid levels, were conducted. Total lipids were calculated using TC and TG concentrations following the formula: total lipids (milligrams per deciliter) = 2.27 × TC × 38.6 + TG × 88.5 + 62.3 (55). BMI was calculated as weight (kilograms) divided by the square of height (square meters). Smoking status was categorized into ever (smoking at least one cigarette per day during the past 6 months or having quit >6 months) and never smoking. Alcohol drinking status was categorized as ever (drinking alcoholic beverages at least once per week during the past 6 months or having quit >6 months) and never drinking. Dietary intake frequencies of meat, fish or seafood, vegetables, grains, and fruits were classified as yes (≥5 times/week) or no (<5 times/week) according to weekly intake. Physical activities with a metabolic equivalent of task (MET) value of 3.0 to <6.0 are classified as moderate intensity, while those with a MET value of 6.0 or higher are classified as high intensity. Being physically active was defined as at least 150 min of moderate intensity activities weekly and/or at least 75 min of high intensity activities weekly (56). Physically inactive refers to not meeting the physical activity criteria. Education levels were categorized as high school or lower (primary, middle, and high school) and college or higher (college, university, and postgraduate). Hypertension was defined as measured blood pressure of ≥140/90 mmHg, taking medications for hypertension, or self-reported physician-diagnosed hypertension. Diabetes was defined as fasting glucose of ≥7.0 mM, glycated hemoglobin of ≥6.5%, taking antidiabetic medications, using insulin, or self-reported physician-diagnosed diabetes. Missing values for categorical variables were coded as a separate category. Missing continuous variables were imputed with the median value. The missing data rates for both types of variables were below 2%.

Statistical analysis

We summarized the baseline characteristics of the study participants using medians (IQR) for continuous variables and numbers (percentages) for categorical variables. Differences between groups were compared using the Mann-Whitney U test for continuous variables and the chi-square test or Fisher’s exact test for categorical variables. Spearman correlation coefficients were calculated to assess pairwise correlations between serum concentrations of pyrethroids with detection frequencies of >70% (fenvalerate, deltamethrin, and cyfluthrin). Because all serum pyrethroid concentrations were right skewed, values were normalized using the INT-transformed method before regression analyses.

For the cross-sectional analysis, we used RCS models to explore the nonlinear associations between individual pyrethroid concentrations and lipid levels, with knots placed at the 10^th, 50^th, and 90^th percentiles and the median used as the reference. Multivariable logistic regression models were used to evaluate the associations between pyrethroid tertiles and the prevalence of dyslipidemia and its subtypes, estimating ORs and 95% CIs. Covariates included age, gender, BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes. Subgroup analyses were conducted by demographic and lifestyle factors.

To evaluate the joint effects of pyrethroid mixtures on dyslipidemia and its four subtypes, we applied two mixture analysis approaches: Q-gcomp and WQS regression models. The Q-gcomp model allows for component exposures to have different effect directions, offering greater flexibility (57). The WQS model assumes the same effect direction and is appropriate for addressing correlations among exposures, with results that are easier to interpret (58). The complementary use of both methods provides a more robust evaluation of mixture effects. In addition, to assess whether the overall mixture association was primarily driven by a single compound, we conducted a sensitivity analysis excluding the pyrethroid that showed the strongest association in the primary analysis and reestimated the joint effects of the remaining compounds.

For the prospective analysis, we used multivariable linear regression models to assess the associations between serum pyrethroid concentrations and annual changes in lipid levels. Exposure levels were dichotomized by merging the second and third tertiles as a middle-to-high exposure group and comparing them with the lowest tertile. In addition to adjusting for the covariates included in the cross-sectional models, baseline levels of each lipid parameter were further controlled to account for individual differences at baseline. We used conditional logistic regression to evaluate the association between pyrethroid exposure and incident dyslipidemia among participants without dyslipidemia at baseline. Conditional logistic regression models were adjusted for BMI, education level, smoking status, alcohol drinking status, physical activity, diet categories (meat, fish or seafood, vegetables, grains, and fruits), hypertension, and diabetes. All analyses were conducted using R software (version 4.4.2). A two-sided P value of less than 0.05 was considered statistically significant.

Supplementary Materials

This PDF file includes:

Figs. S1 to S5

Tables S1 to S11