Association between delivery-food consumption frequency and urinary bisphenol concentrations among Korean adults: a nationally representative study (Korean National Environmental Health Survey cycle 5, 2021–2023) Authors: Gwisang Park, Hyeyun Lee, Inchul Jeong, Eunhee Ha, Kihun Kim Published: 2026 ABSTRACT AbstractBackgroundBisphenol exposure remains a public health concern, particularly as the use of bisphenol analogues such as bisphenol S (BPS) has increased. In the post-pandemic period, delivery-food consumption has become routine in South Korea and may represent an important behavioral pathway for bisphenol exposure. This study examined whether delivery-food consumption in plastic containers is associated with urinary bisphenols among Korean adults.MethodsThis study used adult data from the Korean National Environmental Health Survey (KoNEHS) cycle 5 (2021–2023). Delivery-food consumption frequency (nine categories) was modeled per 1-time-per-month increase. Outcomes were urinary bisphenol A (BPA), BPS, and bisphenol F (BPF). Creatinine-adjusted concentrations were analyzed on the natural log scale using multivariable regression models. Model 1 adjusted for sociodemographic factors, and model 2 additionally adjusted for plastic-related behavioral variables.ResultsAmong 4,275 adults with complete bisphenol and urinary creatinine data, delivery-food consumption frequency was positively associated with creatinine-adjusted urinary BPS concentrations (1.49% increase per 1-time-per-month increase; 95% confidence interval [CI]: 0.74–2.24; p < 0.001) after sociodemographic adjustment, whereas associations with BPA (0.32% increase; 95% CI, −0.45 to 1.09; p = 0.416) and BPF (−0.14% change; 95% CI: −1.18 to 0.90; p = 0.787) were not statistically significant. After additional adjustment for plastic-related behavioral variables in model 2, the association remained statistically significant for BPS (1.54%; 95% CI, 0.76–2.33; p < 0.001), whereas the associations with BPA and BPF remained non-significant.ConclusionsMore frequent consumption of delivery foods packaged in plastic containers was independently associated with higher creatinine-adjusted urinary BPS concentrations. Given that common delivery container polymers in Korea are generally bisphenol-free, this association suggests that delivery-related behavior serves as a behavioral surrogate for a bundled exposure scenario. The observed BPS signal may partly reflect cumulative exposure from ancillary materials, such as thermal receipts and labels. Future studies should identify specific delivery-associated sources and migration pathways of bisphenols.Keywords: Bisphenols, Food packaging, Dietary exposure, Environmental exposure BACKGROUND Bisphenol A (BPA) has been extensively used in food-contact applications and consumer products, and is widely recognized as an endocrine-disrupting chemical.1 Bisphenols can interfere with multiple hormonal signaling pathways and affect diverse physiological systems,2-4 providing biological plausibility for effects that extend beyond reproductive outcomes5,6 to cardiovascular,7 metabolic regulation,8 neurobehavioral,9 and immunologic10 outcomes. In response to these health concerns, regulatory scrutiny has intensified,11,12 and the European Food Safety Authority established a markedly lower tolerable daily intake.13 However, these BPA-focused restrictions have inadvertently fostered “regrettable substitution”, whereby structurally similar analogues such as bisphenol S (BPS) and bisphenol F (BPF) are adopted as replacements, including in products marketed as “BPA-free”.14-16 Emerging evidence suggests that BPS and BPF may possess hormonal activities comparable to BPA,17 raising concerns that such substitutions may shift rather than eliminate toxicologic risk. For instance, studies suggest that BPS interferes with thyroid hormone signaling pathways,18 depresses cardiac contractility,2 and increases attention-deficit/hyperactivity disorder symptoms.19 As market-level bisphenol substitution accelerates, identifying behaviorally specific sources of these analogues in contemporary daily life has become a critical public health priority. The rapid expansion of app-based food delivery platforms, accelerated by the coronavirus disease 2019 (COVID-19) pandemic, has fundamentally reshaped dietary patterns20,21 and increased everyday opportunities for plastic-related exposures.22 Delivered meals are transported under sealed, hot, and humid conditions, which may promote chemical migration from packaging materials into food.23,24 Moreover, bisphenols may originate not only from the polymer matrix of containers but also from coatings, adhesives, printing inks, labels, and thermal paper receipts.25-27 Consequently, frequent delivery-food consumption may reflect a multi-source exposure pathway not fully captured by conventional measures of individual plastic use. Consistent with this exposure context, a recent Korean National Environmental Health Survey (KoNEHS) cycle 4 study reported differences in crude urinary BPA, BPS, and BPF concentrations across several lifestyle and dietary factors, including delivery foods packaged with plastic wrap.28 However, whether delivery-food consumption frequency is independently associated with creatinine-adjusted urinary bisphenol concentrations in the post-pandemic delivery environment remains unclear. Therefore, using nationally representative data from KoNEHS cycle 5 (2021–2023), we examined whether delivery-food consumption frequency was associated with urinary BPA, BPS, and BPF concentrations among Korean adults. We hypothesized that more frequent delivery-food consumption would be associated with higher urinary bisphenol concentrations, particularly BPA analogues in the context of BPA substitution. The findings may contribute to a better understanding of delivery-food consumption as a potential exposure-related behavior and support further research on packaging-related sources of bisphenol exposure. METHODS Study design and population This study was a retrospective secondary analysis of adult data from the nationally representative KoNEHS cycle 5. To ensure representativeness of the non-institutionalized South Korean adult population, KoNEHS cycle 5 employed a multistage stratified cluster sampling design based on enumeration districts from the Population and Housing Census. The dataset included information on delivery-food consumption frequency, urinary bisphenol concentrations (BPA, BPS, and BPF), and relevant covariates. Participants with missing BPA, BPS, BPF, or urinary creatinine data were excluded from the analytic sample. The final analytic sample included 4,275 adults. Exposure assessment Exposure was assessed using the survey item asking about the frequency of consuming delivery foods packaged in plastic containers. Responses were categorized into nine levels: rarely, 1/month, 2–3/month, 1/week, 2–3/week, 4–6/week, 1/day, 2/day, and 3/day. For descriptive analyses and categorical comparisons, the nine-level variable was used as collected. For regression analyses, the frequency categories were converted to a monthly frequency and modeled as the change per 1-time-per-month increase in exposure. Specifically, the categories were converted as follows: rarely = 0, 1/month = 1, 2–3/month = 2.5, 1/week = 4.345, 2–3/week = 10.863, 4–6/week = 21.725, 1/day = 30.415, 2/day = 60.830, and 3/day = 91.245 times/month. Laboratory analysis and quality control Outcomes were urinary concentrations (µg/L) of BPA, BPS, and BPF. According to the KoNEHS cycle 5 analytical manual, urinary environmental phenols were quantified using high-performance liquid chromatography coupled with tandem mass spectrometry. Urine specimens were protected from light and transported under refrigerated conditions (2–6°C) after collection and were stored at −70°C until analysis. The analysis was conducted by specialized laboratories designated and monitored by the National Institute of Environmental Research. Analytical quality control followed the KoNEHS protocol: calibration curve linearity was required to meet R² ≥ 0.995, and calibration check standards were measured at the end of each analytical batch with an acceptance criterion of ±10% of the target value. The method detection limit (MDL) was calculated as 3.14 times the standard deviation from seven replicate measurements (MDL = 3.14 × standard deviation). Values below the MDL were imputed as MDL/√2 and included in analyses (Supplementary Table 1). Statistical analysis To account for urine dilution, creatinine-adjusted bisphenol concentrations were calculated by dividing urinary bisphenol concentrations in µg/L by urinary creatinine concentrations in g/L. Because the creatinine-adjusted concentrations were right-skewed, their natural logarithms were used as the outcome variables.29 These outcomes were analyzed using weighted multiple linear regression models with cluster-robust standard errors at the primary sampling unit level. Regression coefficients from the log-transformed models were back-transformed and expressed as percent differences using the formula (exp[β] − 1) × 100. Candidate covariates were identified a priori based on prior literature30 and subject-matter relevance, and were grouped into sociodemographic and plastic-related behavioral factors. As an exploratory step, Spearman rank correlations between each candidate covariate and adjusted urinary BPS were examined to understand their bivariate relationships (Supplementary Table 2). While these correlations provided preliminary insights, the final selection of covariates for model 2 was primarily based on subject-matter relevance and their roles as indicators of a plastic-intensive lifestyle, ensuring a comprehensive adjustment for potential confounding pathways regardless of their individual correlation strengths. Multicollinearity was assessed using the variance inflation factor (VIF), and all VIF values were <10. Specifically, model 1 was adjusted for key sociodemographic factors: sex, age group, household income, smoking, alcohol consumption, and education. Model 2 additionally adjusted for plastic-related behavioral variables, including use of coated containers (e.g., oven/air-fryer), use of coated frying pans, use of disposable paper cups, microwaved foods covered with plastic wrap, polyethylene terephthalate (PET)–bottled beverage consumption, placing hot foods into rigid transparent plastic containers, and cup ramen consumption. Model 2 was constructed to evaluate whether the association between delivery-food consumption and urinary BPS is independent of a general 'plastic-intensive' lifestyle. By adjusting for various plastic-related behaviors, we aimed to evaluate whether delivery-food consumption retained an independent association with urinary BPS beyond other common plastic-related behaviors. However, model 2 was not intended to identify the specific chemical source of BPS exposure, such as containers, labels, thermal receipts, or other delivery-related materials. As sensitivity analyses for BPS, we repeated model 2 after excluding participants with upper extreme BPS concentrations, defined as unadjusted urinary BPS concentrations greater than Q3 + 1.5 × interquartile range. We also repeated model 2 without applying sampling weights, while retaining cluster-robust standard errors at the primary sampling unit level. As an additional analysis, Spearman rank correlations were calculated between delivery-food consumption frequency and the plastic-related behavioral variables included in model 2 to evaluate the degree of interrelatedness among behavioral covariates. Correlation magnitudes were interpreted as very high (|r| ≥ 0.8), high (0.6–0.8), moderate (0.4–0.6), low (0.2–0.4), and negligible (<0.2).31 All statistical analyses were conducted using Python (version 3.12.12, Python Software Foundation, Beaverton, OR, USA). Ethics statement This study was granted exemption from Institutional Review Board review by the Institutional Review Board of Pusan National University Yangsan Hospital (IRB No. 55-2026-053) because it was a secondary analysis of de-identified data. Informed consent was obtained from all participants in the original KoNEHS. RESULTS General characteristics of the study population Table 1 summarizes the general characteristics of the included population and median creatinine-adjusted urinary bisphenol concentrations. After excluding participants with missing BPA, BPS, BPF, or urinary creatinine data, 4,275 adults were included in the final analytic sample. The study population comprised 1,942 men (45.4%) and 2,333 women (54.6%). The largest age group was 60–69 years (29.5%), followed by 50–59 years (21.4%), 40–49 years (16.9%), 30–39 years (11.8%), ≥70 years (11.6%), and 19–29 years (8.8%). Regarding body mass index (BMI), 36.0% had a BMI ≥25 kg/m², and 2.6% had a BMI <18.5 kg/m². With respect to smoking status, 2,708 participants (63.3%) were never-smokers, whereas former and current smokers accounted for 19.8% and 16.8%, respectively. Delivery-food consumption in plastic containers was reported most commonly as “rarely” (1,991 participants, 46.6%); however, notable proportions reported consumption once per month (14.6%), 2–3 times per month (16.1%), once per week (11.0%), or 2–3 times per week (8.0%). High-frequency consumption was relatively uncommon, with 2.6% reporting 4–6 times per week and 1.0% reporting at least once per day. Detection rates were 92.2% for BPA, 96.6% for BPS, and 86.8% for BPF. Median creatinine-adjusted urinary concentrations were 0.972 µg/g Cr for BPA, 0.310 µg/g Cr for BPS, and 0.242 µg/g Cr for BPF. By delivery-food consumption frequency, creatinine-adjusted BPS and BPF concentrations showed statistically significant differences, whereas BPA did not. The distributions of creatinine-adjusted urinary BPA, BPS, and BPF concentrations according to the plastic-related behavioral variables included in model 2 are presented separately (Supplementary Table 3). Table 1.General characteristics and median creatinine-adjusted urinary bisphenol concentrations (KoNEHS cycle 5)CharacteristicUnweighted n (%)BPA/UCr (µg/g Cr)p-valueaBPS/UCr (µg/g Cr)p-valueaBPF/UCr (µg/g Cr)p-valueaTotal4,275(100)0.9720.3100.242Sex<0.001<0.001<0.001 Male1,942 (45.4)0.8710.2550.221 Female2,333 (54.6)1.0710.3620.261Age (years)<0.001<0.001<0.001 19–29377 (8.8)0.7300.2770.184 30–39503 (11.8)0.7870.2880.195 40–49723 (16.9)1.0050.3030.229 50–59914 (21.4)1.0340.3530.263 60–691,260 (29.5)1.0410.3200.266 ≥70498 (11.6)1.0550.2560.256BMI (kg/m²)0.5830.1300.419 <18.5111 (2.6)1.1080.2900.222 18.5 to <252,627 (61.5)0.9720.3140.242 ≥251,537 (36.0)0.9630.3040.240Household incomeb (million won/month)<0.0010.0080.197 <1 ($690)397 (9.3)1.0760.2460.259 1 to <2 ($1,380)736 (17.2)1.0730.3070.270 2 to <3 ($2,070)868 (20.3)1.0060.3240.258 3 to <5 ($3,450)1,116 (26.1)0.8990.2960.222 5 to <7 ($4,830)624 (14.6)0.9140.3310.221 ≥7446 (10.4)0.8330.3450.221 Unknown88 (2.1)1.1920.3370.277Smoking status0.044<0.0010.076 Never2,708 (63.3)0.9870.3330.241 Former847 (19.8)0.9240.2590.232 Current720 (16.8)0.9630.2840.257Delivery food in plastic packaging0.0790.0330.040 Rarely1,991 (46.6)1.0370.3030.252 1/month626 (14.6)0.9590.2940.232 2–3/month688 (16.1)0.9450.3070.252 1/week470 (11.0)0.8820.3000.204 2–3/week343 (8.0)0.8890.3470.233 4–6/week113 (2.6)0.8020.3740.218 1/day35 (0.8)0.8720.3940.125 2/day5 (0.1)0.7190.1850.183 3/day4 (0.1)1.1770.8660.031Values are presented as number (%) or median. Bisphenol concentrations are presented as creatinine-adjusted concentrations (µg/g Cr). All descriptive statistics in Table 1 are unweighted.KoNEHS: Korean National Environmental Health Survey; BPA: bisphenol A; UCr: urinary creatinine; Cr: creatinine; BPS: bisphenol S; BPF: bisphenol F; BMI: body mass index.ap-values from Kruskal-Wallis test;bAn exchange rate of 1,450 won per US dollar was applied. General characteristics and median creatinine-adjusted urinary bisphenol concentrations (KoNEHS cycle 5) Values are presented as number (%) or median. Bisphenol concentrations are presented as creatinine-adjusted concentrations (µg/g Cr). All descriptive statistics in Table 1 are unweighted. KoNEHS: Korean National Environmental Health Survey; BPA: bisphenol A; UCr: urinary creatinine; Cr: creatinine; BPS: bisphenol S; BPF: bisphenol F; BMI: body mass index. p-values from Kruskal-Wallis test; An exchange rate of 1,450 won per US dollar was applied. Urinary bisphenol concentrations and their association with delivery-food consumption In analyses modeling delivery-food consumption frequency as an increase of one time per month (Table 2), urinary BPS showed a statistically significant rise in creatinine-adjusted concentrations after sociodemographic adjustment, with an estimated 1.49% increase (95% confidence interval [CI]: 0.74–2.24; p < 0.001) for each additional time per month of delivery-food consumption. In contrast, the corresponding associations were not statistically significant for BPA (0.32%; 95% CI: −0.45 to 1.09; p = 0.416) or BPF (−0.14%; 95% CI: −1.18 to 0.90; p = 0.787). After additional adjustment for plastic-related behavioral variables in model 2, the association remained statistically significant for BPS (1.54%; 95% CI: 0.76–2.33; p < 0.001). Table 2.Percent difference in creatinine-adjusted urinary bisphenols per 1-time-per-month increase in delivery-food consumption frequency (KoNEHS cycle 5)OutcomeCrude model % diff (95% CI)p-valueModel 1 % diff (95% CI)p-valueModel 2 % diff (95% CI)p-valueBPA–0.41 (–1.12 to 0.32)0.2700.32 (–0.45 to 1.09)0.416–0.06 (–0.89 to 0.79)0.896BPS0.88 (0.14 to 1.63)0.0201.49 (0.74 to 2.24)<0.0011.54 (0.76 to 2.33)<0.001BPF–0.79 (–1.62 to 0.05)0.067–0.14 (–1.18 to 0.90)0.787–0.10 (–1.05 to 0.86)0.836Values are presented as percent difference (% diff) with 95% confidence intervals. Outcomes were natural log–transformed after creatinine adjustment [ln(bisphenol/UCr)]. The exposure was based on the survey item “Eating delivery foods packaged in plastic”, converted to times per month, and modeled per 1-time-per-month increase. The percent difference was calculated as (exp(β) − 1) × 100. p-values were obtained from weighted multiple linear regression models with cluster-robust standard errors at the primary sampling unit level. Model 1 was adjusted for key sociodemographic factors, including sex, age group, smoking, alcohol consumption, household income, and education. Model 2 additionally adjusted for use of coated containers (e.g., oven/air-fryer), use of coated frying pans, use of disposable paper cups, microwaved foods covered with plastic wrap, PET-bottled beverages, placing hot foods into rigid transparent plastic containers, and cup ramen consumption.KoNEHS: Korean National Environmental Health Survey; CI: confidence interval; BPA: bisphenol A; BPS: bisphenol S; BPF: bisphenol F; UCr: urinary creatinine; PET: polyethylene terephthalate. Percent difference in creatinine-adjusted urinary bisphenols per 1-time-per-month increase in delivery-food consumption frequency (KoNEHS cycle 5) Values are presented as percent difference (% diff) with 95% confidence intervals. Outcomes were natural log–transformed after creatinine adjustment [ln(bisphenol/UCr)]. The exposure was based on the survey item “Eating delivery foods packaged in plastic”, converted to times per month, and modeled per 1-time-per-month increase. The percent difference was calculated as (exp(β) − 1) × 100. p-values were obtained from weighted multiple linear regression models with cluster-robust standard errors at the primary sampling unit level. Model 1 was adjusted for key sociodemographic factors, including sex, age group, smoking, alcohol consumption, household income, and education. Model 2 additionally adjusted for use of coated containers (e.g., oven/air-fryer), use of coated frying pans, use of disposable paper cups, microwaved foods covered with plastic wrap, PET-bottled beverages, placing hot foods into rigid transparent plastic containers, and cup ramen consumption. KoNEHS: Korean National Environmental Health Survey; CI: confidence interval; BPA: bisphenol A; BPS: bisphenol S; BPF: bisphenol F; UCr: urinary creatinine; PET: polyethylene terephthalate. In sensitivity analyses restricted to BPS, the effect estimate was attenuated after excluding upper extreme BPS values but remained statistically significant (1.14%; 95% CI: 0.49–1.80; p < 0.001); the association also remained significant in the unweighted model 2 analysis (1.32%; 95% CI: 0.58–2.07; p < 0.001) (Supplementary Table 4). Correlation analysis of plastic-related behaviors To further examine these relationships, correlations between delivery-food consumption frequency and other plastic-related behaviors were examined (Table 3). Delivery-food consumption showed a moderate positive correlation with cup ramen consumption (r = 0.414). Correlations were low for use of coated containers (e.g., oven/air-fryer) (r = 0.253), and microwaved foods covered with plastic wrap (r = 0.218), and were negligible for placing hot foods into rigid transparent plastic containers (r = 0.196), use of disposable paper cups (r = 0.182), PET-bottled beverage consumption (r = 0.171), and use of coated frying pans (r = 0.104). Table 3.Correlations of delivery-food frequency with plastic-related behavior variables included in model 2Plastic-related behavior variableSpearman rCup ramen consumption0.414Use of coated containers (e.g., oven/air-fryer)0.253Microwaved foods covered with plastic wrap0.218Placing hot foods into rigid transparent plastic containers0.196Use of disposable paper cups0.182PET-bottled beverage consumption0.171Use of coated frying pans0.104Spearman correlations were calculated between delivery-food consumption frequency (based on the survey item “Eating delivery foods packaged in plastic” and converted to times per month) and each plastic-related behavior variable included in model 2 (BPS). Values are Spearman’s rank correlation coefficients (r). All correlations were statistically significant (p < 0.001).PET: polyethylene terephthalate; BPS: bisphenol S. Correlations of delivery-food frequency with plastic-related behavior variables included in model 2 Spearman correlations were calculated between delivery-food consumption frequency (based on the survey item “Eating delivery foods packaged in plastic” and converted to times per month) and each plastic-related behavior variable included in model 2 (BPS). Values are Spearman’s rank correlation coefficients (r). All correlations were statistically significant (p < 0.001). PET: polyethylene terephthalate; BPS: bisphenol S. DISCUSSION This study used large-scale epidemiologic data to evaluate the association between consumption of delivery foods packaged in plastic containers and urinary bisphenol concentrations among Korean adults. In the fully adjusted models, delivery-food consumption frequency was significantly associated with higher urinary BPS. A major strength of this study is that it examined delivery-food consumption frequency as a specific exposure-related behavior after the COVID-19 pandemic, when food delivery had become widely established in daily life. This extends prior work by Shin et al. based on KoNEHS cycle 4 (2018–2020), which evaluated multiple plastic-related behaviors in relation to urinary bisphenol concentrations.30 By focusing on delivery-food consumption frequency, our study assessed its association with urinary bisphenol concentrations independently of other plastic-related behaviors. Notably, the association with urinary BPS persisted after adjustment for a broad range of plastic-related behaviors as well as sociodemographic factors. In our study, delivery-food consumption frequency was positively associated with urinary BPS, whereas no clear association was observed for BPA or BPF. This pattern is consistent with the broader regulatory and market context in which intensified restrictions on BPA have accelerated substitution toward structurally similar alternatives, particularly BPS. Consistent with this shift, recent human biomonitoring reports from some regions have noted rising urinary BPS levels that approach or even exceed BPA.32 Market-level substitution has been documented for specific exposure sources, most notably thermal paper, where BPS has been used as a developer and its use increased rapidly in some regions following BPA restrictions.17,33,34 Importantly, substitution does not necessarily imply risk reduction, with current studies concluding that BPS and BPF can be as hormonally active as BPA. Within this “regrettable substitution,”14-16 a signal concentrated in BPS rather than BPA may indicate that the important contributors within contemporary exposure pathways have changed, rather than that bisphenol-related exposure has disappeared. Notably, common polymers used in delivery containers are not those in which bisphenols are typically used as primary monomers. In Korea, polypropylene (PP) accounts for the majority of plastics used in disposable delivery packaging, followed by PET, polystyrene (PS), and polyethylene (PE)22 (Supplementary Table 5). This distribution is plausible given the Korean diet, in which hot soups and broths are frequently delivered and PP is favored for heat resistance.22 PET tends to be used for lower-temperature foods, PS for small sauce or side-dish containers, and PE for wraps and films. However, bisphenols are classically associated with epoxy resins27 (e.g., internal can coatings and some metal lids) and polycarbonate materials35-37 rather than PP, PET, PS, or PE. The association between delivery-food consumption and urinary BPS—despite the prevalence of BPS-free container polymers—suggests that our independent variable functions as a behavioral proxy. In the current delivery ecosystem, the act of consumption is inextricably bundled with the handling of ancillary materials, including thermal receipts and labels that may contain bisphenol analogues. Therefore, the observed BPS signal may partly reflect cumulative exposure from these ancillary materials rather than chemical migration from the container matrix itself. First, label- and receipt-related pathways may be important, because BPS is widely used as a color developer in thermal paper. Delivery foods commonly involve receipts, barcode labels, and printed stickers, which may increase opportunities for dermal contact and subsequent hand-to-mouth transfer. Furthermore, some label-associated chemicals may permeate packaging films into food under certain conditions.25 Second, low-level contamination and migration from plastics that do not intentionally use bisphenols as monomers may still occur via non-intentionally added substances38 and cross-contamination during manufacturing, or the incorporation of recycled plastics. Experimental studies have reported trace-level bisphenol migration from polymers such as PP,39 PET,38,40 and PS.41,42 Even trace residues may become consequential in delivery settings, where high temperatures, lipophilicity and acidity could synergistically promote chemical migration.43-46 Alternative explanations could also be considered. Delivery orders may co-occur with additional packaged items, such as canned beverages or canned foods with epoxy resin linings,27 or adhesives that introduce additional exposure reservoirs.26 Moreover, the potential for residual confounding should be considered, as delivery-food consumption frequency showed positive correlations with several other plastic-related behaviors, such as cup ramen and PET-bottled beverage consumption (Table 3). However, the association between delivery frequency and urinary BPS remained robust in model 2 even after adjusting for these lifestyle factors (Table 2). This persistent significance suggests that delivery services represent a distinct exposure pathway that may not be fully addressed by conventional plastic-related risk-reduction efforts. Delivery-food consumption may function as a behavioral proxy for a bundled exposure scenario that includes non-container elements such as thermal receipts and labels. However, because these sources were not separately measured, their specific contributions to urinary BPS concentrations cannot be determined in the present study. From a toxicological perspective, bisphenols may exhibit non-monotonic dose–response relationships,3,47 and evidence of adverse effects has been reported even at low doses.4 Reflecting these concerns, the European Union has recently tightened controls on bisphenols in food-contact materials, moving beyond migration-limit management toward broader use restrictions. Specifically, Regulation (EU) 2024/3190 notes that the previous assessment of BPS may be insufficient, and correspondingly removes BPS from the authorized substances list for plastics.13 In Korea, the Food Standards Codex of the Ministry of Food and Drug Safety has mainly focused on bisphenol-based resins such as polycarbonate and epoxy, whereas common disposable delivery containers are generally made of other polymers. BPS is also regulated only for a limited subset of materials, such as polyethersulfone, which is not commonly used for mainstream disposable delivery containers. In this context, the observed association between delivery-food consumption and urinary BPS suggests that relevant exposure sources may extend beyond the container body itself and may involve other components of the delivery process, such as labels, thermal-paper receipts, adhesives, and multi-component packaging. Although causal interpretation is limited by the cross-sectional design, the findings may inform future research and more comprehensive monitoring and risk-management strategies for delivery-related exposures. This study has several limitations. First, our exposure indicator inherently captures a bundled exposure scenario. It does not allow for the statistical isolation of chemical migration from containers versus exposure from ancillary materials like thermal receipts. Consequently, findings should be interpreted as the impact of the delivery ecosystem as a behavioral cluster, rather than a risk specific to plastic polymers alone. Given the multi-source nature of these exposures, the observed associations could be substantially driven by unmeasured confounding factors inherent to delivery behaviors. Second, the cross-sectional design precludes causal inference, and temporality between delivery-food consumption frequency and urinary bisphenol levels cannot be established. The exposure indicator based on the frequency of consuming delivery foods packaged in plastic containers may also not fully represent overall delivery-food intake, particularly when some deliveries arrive in non-plastic packaging. Nonetheless, plastic containers are common in the Korean delivery market, and KoNEHS provides no survey item that directly captures overall delivery frequency, making this measure a pragmatic proxy for delivery-related packaging exposure. Third, self-reported frequency may introduce recall and interpretation variability across participants, such as whether beverages or desserts are counted as delivery foods. Fourth, because bisphenols have relatively short biological half-lives, a single urine measurement is more likely to reflect recent exposure than long-term average intake. Even so, urinary biomarkers are widely used in environmental epidemiology because they integrate multiple exposure routes over a relevant short time window and can be implemented at scale in biomonitoring surveys; moreover, delivery consumption has become a habitual behavior for many Korean adults, allowing population-based analyses to inform average exposure patterns and their behavioral correlates. Future studies could help reduce these uncertainties with designs that better address temporality and with clearer, more standardized survey measurement of delivery-related packaging behaviors. CONCLUSIONS Based on the KoNEHS cycle 5 data collected after the COVID-19 pandemic, delivery-food consumption in plastic containers was independently associated with higher urinary BPS concentrations. In contrast, associations with BPA and BPF were not statistically significant. As container polymers are generally BPS-free, our findings suggest that non-container elements may be important contributors to this association. While our exposure indicator was based on delivery food in plastic packaging, this association may reflect a bundled behavioral cluster—including exposure to thermal paper receipts, labels, and non-intentionally added substances—and the potential for unmeasured confounding within this ecosystem should be explicitly considered. Given the endocrine-disrupting potential of bisphenols and the possibility of non-monotonic low-dose effects, future research and policy should target the reduction of bisphenols in the entire delivery chain, rather than focusing solely on plastic containers. Abbreviations body mass index bisphenol A bisphenol F bisphenol S confidence interval coronavirus disease 2019 creatinine Korean National Environmental Health Survey method detection limit polyethylene polyethylene terephthalate polypropylene polystyrene urinary creatinine variance inflation factor Footnotes Funding This research was supported by the Environmental Health Center for Training Environmental Medicine Professionals at Inha University Hospital, funded by the Ministry of Environment, Republic of Korea. This study was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2025-25467811). This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (RS-2026-25476099). Competing interests Inchul Jeong, a contributing editor of the Annals of Occupational and Environmental Medicine, was not involved in the ed­itorial evaluation or decision to publish this article. All remaining authors have declared no conflicts of interest. Author contributions Conceptualization: Park G, Lee H. Data curation: Park G. Methodology: Park G, Jeong I, Ha E, Kim K. Formal analysis/validation: Park G. Project administration: Jeong I, Ha E, Kim K. Funding acquisition: Park G. Writing - original draft: Park G, Lee H. Writing - review & editing: Jeong I, Ha E, Kim K.