Determination of Diffusion Coefficients of Bisphenol A (BPA) in Polyethylene Terephthalate (PET) to Estimate Migration of BPA from Recycled PET into Foods

Abstract

Bisphenol A (BPA) is a known substance that is found in food contact materials as an intentionally added as well as a non-intentionally added substance. Traces of BPA were found as a non-intentionally added substance in recycled PET (rPET). In 2023, the EFSA proposed a new TDI of 0.0002 µg/kg bw/d, which is lower than the previous (temporary) TDI of 4 µg/kg bw/d by a factor of 20,000. The TDI of 0.0002 µg/kg bw/d would translate for a default 60 kg person eating one kilogram of food into a migration limit of 0.012 µg/kg in the food. This very low migration limit is a challenge to measuring BPA levels in food. A solution is to use migration modeling to establish maximum concentrations in rPET for different food contact applications. Precise diffusion coefficients for BPA in PET were determined within this study by use of migration kinetics. In June 2024, the European Commission proposed a new migration threshold limit for BPA of 1 µg/kg, which should be understood as a detection limit. From the results of this study, it can be concluded that a BPA concentration in the PET bottle wall of 297 mg/kg (3% acetic acid), 255 mg/kg (10% ethanol), and 192 mg/kg (20% ethanol) after storage for 365 d at 25 °C is in compliance with the migration threshold limit of 1 µg/kg. These maximum concentrations are far above the measured BPA concentrations on rPET bottles in Europe between 2019 and 2023. Therefore, the new proposed migration threshold limit for BPA cannot be exceeded.

1. Introduction

Bisphenol A (BPA), or 2,2-(4,4-dihydroxydiphenyl)propane, is a well-known organic substance with industrial–technical uses in the non-food area but also in contexts with food contact applications, where it might be intentionally used, e.g., as monomer in polycarbonate [1] and polysulfone plastics manufacture or in the internal coating of food cans [2,3], or where it may be non-intentionally present in food contact materials (FCMs) [4,5,6,7]. A review of BPA exposure and the levels of BPA released from consumer products as well as the levels measured in wastewater, drinking water, air, and dust is given by Vandenburg et al. [8]. One FCM sector where BPA may potentially come into contact with food is the area of PET recycling, where contaminants could enter the recycling feed stream and may remain in low amounts even after the applied obligatory cleaning steps towards the recycled PET (rPET) qualities for new food contact [9].

Concerning European food packaging legislation and toxicological evaluations, BPA was subject to a number of developments and fluctuations based on unrivaled attention in the scientific as well as the public domain, which most likely no other chemical substance has ever received. Back in 2002, BPA was authorized as an FCM substance in EU Directive 2002/72/EC [10] with a specific migration limit (SML) of 3 mg/kg food, which was based on the toxicological evaluation of the Scientific Committee on Food (SCF) at the time. In 2003, the SCF set a temporary tolerable daily intake (TDI) of 10 µg/kg bw/d based on an animal experimental no observed adverse effect level (NOAEL) of 5000 µg/kg bw/d and applying an uncertainty factor of 500. As a consequence, in 2004, BPA was listed in EC Directive 2002/72/EC with a reduced SML of 600 µg/kg food [11]. Following a re-evaluation by the European Food Safety Authority (EFSA), the NOAEL was still considered to be valid, but the uncertainty factor was revised from 500 to 100 based on new toxicological evidence. As a consequence, in 2006, the EFSA established a full tolerable daily intake (TDI) of 50 µg/kg bw/d [12]. However, the European (EU) Commission did not revise the SML back to 3 mg/kg food but left the restriction unchanged at 600 µg/kg food. Due to repeated criticism from European national agencies, the EFSA reviewed new scientific information several times but did not find sufficient evidence to revise the NOAEL of 5000 µg/kg bw/d until 2011 [13]. In 2012, the EFSA decided to undertake a full re-evaluation of BPA based on the most recent scientific evidence and, in 2015, published an opinion with a temporary TDI of 4 µg/kg bw/d [14]. In the follow-up in 2018, the EU Commission revised the SML of 600 µg/kg food to an SML of 50 µg/kg food, also taking an allocation factor of 20% to include exposure from other sources into account [15]. With this measure, BPA was also banned from being used in the manufacture of FCMs for infant formula as well as baby and young children’s foods. In 2020, the EU Commission, relying on the same temporary TDI, established a threshold limit of BPA in drinking water of 2.5 µg/L, set to take effect in January 2026 [16].

In the meantime, in 2016, based on a number of uncertainties related to the toxicological evaluation of BPA, the EU Commission mandated the EFSA to re-evaluate the risks to public health from the presence of BPA in foodstuffs and to establish a tolerable daily intake (TDI). For this re-evaluation, a pre-established protocol was used that had undergone public consultation. In this project, taking evidence from animal data into account and with support from human observational studies, the immune system was identified as most sensitive to BPA exposure. As a result, in 2023, a new TDI of 0.0002 µg/kg bw/d was established [17], which is lower than the previous (temporary) TDI of 4 µg/kg bw/d by a factor of 20,000. The TDI of 0.0002 µg/kg bw/d would translate for a default 60 kg person eating one kilogram of food into a migration limit of 0.012 µg/kg in food for intentionally added BPA. In this 2023 opinion, the EFSA concluded that there is a health concern for the consumer from dietary BPA exposure, as the dietary exposure estimates previously determined in 2015 [14] exceeded the new TDI in all population age groups by two to three orders of magnitude.

The EFSA opinion of 2023 [17] has been critiqued in the public domain and by national food safety and health protection bodies. The European Medicines Agency (EMA) is not in agreement with the change to the temporary TDI to the currently revised TDI of 0.0002 µg/kg bw/d [18]. Also, the Senate Commission on Food Safety (SKLM), which provides scientific advice on food safety issues to the Senate of German Research Foundation (DFG) as well as to federal/state governments and other authorities, published a critical discussion on the HBGV on BPA [19]. The German Bundesinstitut für Risikobewertung (BfR) commented critically on the outcome of the EFSA opinion and made a re-evaluation of the critical toxicological endpoints identified by the EFSA with the intention to provide an independently derived TDI [20]. The result was a TDI of 0.2 µg/kg bw/d, which is higher than the proposed new EFSA TDI of 0.0002 µg/kg bw/d by a factor 1000 [17]. The BfR considered their TDI as a health-based guidance value (HBGV) which is protective enough to exclude immunological effects in humans from BPA exposure. The UK Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (COT) concluded that the endpoint selected and approach applied by the BfR were more scientifically robust and appropriate than those used by the EFSA and agreed to adopt the TDI of 0.2 ug/kg bw per day derived by the BfR [21]. For the default consumer adult (60 kg person), the migration limit would then be at 12 µg/kg food, thus being in the range of the well-known 10 µg/L limit, which is still broadly considered to be an analytically manageable detection limit either for specific migrants such as vinyl chloride monomer or for NIAS [20]. In response to the outcome of the EFSA’s BPA re-evaluation, the European Commission quite recently published a draft regulation to ban the use of BPA in FCMs [22]. In this regulation, the European Commission stated that an FCM must not contain BPA or must not release BPA above the detection limit, which is set in Article 9(2)(b) to 1 µg/kg.

It is quite obvious that the new EFSA TDI of 0.0002 µg/kg bw/d (or 0.012 µg/kg for an adult person) will present analytical chemists—even the most experienced and best-equipped ones—with unresolvable tasks with regards to measuring BPA levels at such low concentrations in foods. It appears that control of such low exposure thresholds cannot be achieved even by modern instrumental analytical devices. The BfR HBGV, which is higher than the EFSA TDI or the EU Commission threshold limit of 1 µg/L by a factor 1000, is much more likely to be analytically verifiable in simple food simulants but still expected to be very challenging for measurements of BPA in foods due to their complex composition.

In the area of PET recycling for new food contact applications such as water bottles, a similar situation exists: in 2011, the EFSA established evaluation safety criteria for PET recycling processes when the intended use is the manufacture of FCMs [23]. As unknown contaminants may be introduced during the re-collection and recycling steps, the EFSA requires that it must be ensured that even potentially genotoxic substances are sufficiently removed in the recycling process. The cleaning efficiency is tested in a so-called challenge test, and the finally recovered PET material must not contain any unknown (and therefore potentially genotoxic) contaminants which could lead to consumer exposure above the lowest level of the threshold of toxicological concern (TTC) concept, which is 0.0025 µg/kg bw/d for genotoxic substances [24,25]. Depending on the population age groups (infants, toddlers, and adults) to be considered and the related food consumption specifications, the migration limit to be controlled will therefore be in the range from 0.017 µg/kg food (infants) to 0.15 µg/kg food (adults) [23]. Again, such low levels can hardly be analytically controlled in non-targeted measurements to seek for any potential contaminant of unknown identity in food simulants and in foods. Therefore, the EFSA applied an indirect “method of analysis”, more specifically “migration modeling”, to back-calculate from these low maximum allowable migration levels to their corresponding concentrations in the PET polymer, denoted C_mod [23]. For comparison with C_mod, in the challenge test, the cleaning efficiency is determined using surrogate substances by measuring their decrease in the PET material during the recycling process. The final residual concentrations in PET, denoted C_res, are compared with C_mod, and when C_mod > C_res, the reprocessed PET is considered to be safe [23].

The motivation for and objectives of this paper are the following: Although BPA is not a monomer used in the production of PET, it cannot be excluded that some BPA impurities may be non-intentionally present in rPET-based FCMs. This will trigger the analytically extremely challenging question of whether rPETs may contain BPA at levels exceeding the level implied by the new TDI or SML requirements. As mentioned above, verification of such low BPA residues in foods or food simulants is hardly any longer directly accessible by analytical means. We see a solution for this problem in proceeding in analogy to the migration modeling approach as described above for evaluation of PET recycling processes [23,26]. Starting from the level implied by the new TDI or SML, back-calculations can be made by migration modeling to establish maximum concentrations in rPET for different food contact applications [27,28]. These maximum concentrations can be compared with actual concentrations in rPET, which are analytically much more accessible than those in food (or food simulants). Such a procedure, however, requires precisely determined diffusion coefficients for BPA in the PET polymer matrix, which are not yet available in the scientific literature. Therefore, our first objective was to measure the diffusion coefficients and the activation energy of the diffusion of BPA in PET polymer. Our second objective was to describe an analytical method based on LC-MS (i) to verify the back-calculated levels in rPET and (ii) for sensitive BPA determination in rPET samples at very low limits of detection. This method was used to determine the concentration in various rPET samples from the European market between 2019 and 2023. Our working hypothesis is the following: by combining this analytical method with migration modeling parameters based on established BPA-related parameters, migration levels of BPA from rPET into foods will be accessible at, and even below, the new EFSA TDI [17] or the new proposed threshold limit of the European Commission [22].

2. Materials and Methods

2.1. Sample Materials

A total of 375 PET bottle samples were provided by European bottling companies and collected and analyzed between 2019 and 2023. The bottles came from companies in Austria, Belgium, the Czech Republic, Finland, France, Germany, Greece, Ireland, Italy, Norway, Lithuania, Poland, Romania, Switzerland, Spain, Sweden, the Netherlands, and the United Kingdom. The majority of the bottles were clear, but some were also green or blue. The companies also provided information on the material composition and recyclate content of the bottles (between 11% and 100%). The investigated PET bottles were all commercial materials that entered the European market during the period under investigation, and the materials used for the production of the bottles are commercially available PET materials. Therefore, no further characterization was carried out on the bottles. The bottles were analyzed before they were filled and came onto the market. The analyzed bottles were therefore not exposed to any environmental influences, contamination during handling, or UV radiation. Five bottles manufactured from virgin PET without any recyclate content were used as reference bottles.

In addition, a batch of 50 bottles manufactured from recycled PET with a high BPA concentration were provided by a research organization. These bottles were used for migration kinetics. The concentration of BPA in the PET bottles used for migration kinetics was determined to be 26.2 ± 0.4 mg/kg (mean value from 3 individual bottles).

2.2. Sample Preparation for the Determination of Bisphenol A in PET

The bottles were cut and milled with a centrifugal mill (RETSCH ZM-200, 750 µm sieve insert, cooling with liquid nitrogen). Approx. 3 g ground sample material was weighed into clean glass ampoules. The samples were spiked with 10 µL of ¹³C₁₂ bisphenol A internal standard (c_initial = 100 mg/L, Santa Cruz Biotechnology, Inc., Heidelberg, Germany). Subsequently, 20 mL dichloromethane (pest. grade, Honeywell, Charlotte, NC, USA) were added. The prepared samples were stored in a climate chamber for 3 days at 40 °C. After storage and cooling down, 1 mL aliquot of the extract was purged to dryness in nitrogen and re-uptaken in 1 mL acetonitrile (LC-MS grade, Carl-Roth, Karlsruhe, Germany): ultrapure water (1:1, v/v). Finally, the samples were transferred to a measuring vial via a 0.2 µm PTFE syringe filter (WICOM). A total of 375 PET bottles were analyzed over five years. All samples were prepared and analyzed in triplicate.

2.3. Migration Kinetics of Bisphenol A from PET into Food Simulants

For the kinetics, three samples per kinetic point were brought in contact with the simulant. An amount of 100 cm² of PET bottle was totally immersed with 50 mL of the corresponding food simulant (3% acetic acid, 10% ethanol, and 20% ethanol) and stored at temperatures of 30 °C, 40 °C, 50 °C, and 60 °C for 10, 20, 30, 40, 50, 60, and 90 days. Thus, including the blank values, 336 migration solutions were available for SPE sample preparation. The migration solutions were transferred quantitatively to clean glass ampoules and spiked with 10 µL of ¹³C₁₂ bisphenol A internal standard (c_initial = 5 mg/L, Santa Cruz Biotechnology, Inc.). In accordance with Machery-Nagel SPE-Application’s Note MN303211/303212 method, CROMABOND^® 3 mL SPE glass columns with a filling amount of 200 mg (Machery-Nagel) were used. The SPE columns were preconditioned with 5 mL methanol (LC-MS grade) followed by 5 mL ultrapure water. In the next step, the sample in the glass column was loaded with the migration sample (approx. 1–2 drops per second). Finally, BPA was eluted with 5 mL acetone (dest. grade): glacial acetic acid (99:1, v/v). The eluates were collected, purged to dryness in nitrogen stream, and reuptaken in 1 mL acetonitrile (LC-MS grade, Carl-Roth): ultrapure water (1:1, v/v). Each migration solution was measured twice. The six values (three samples each measured twice) were then averaged, and the standard deviation was determined.

2.4. Quantification of Bisphenol A in Dichloromethane Extracts and Food Simulants

An LC–tandem mass spectrometer (Thermo Finnigan TSQ Quantum Ultra AM) was used for chromatographic separation and detection. The separation was carried out on a Restek Raptor Biphenyl column, 2.7 µm, 50 × 2 mm (injection volume 10 µL, oven temperature 40 °C). The mobile phases consisted of ultrapure water (A) acetonitrile (LC-MS grade) (B). The constant flow rate was 500 µL/min, and the gradient program was as follows: initial conditions at 75% (A), 25% (B); gradient up to 0% (A), 100% (B) until 2.5 min; 25% (A), 75% (B) until 5 min. The ionization was achieved using atmospheric pressure chemical ionization (APCI) in negative polarity mode (settings: spray-voltage: 7.0 eV; vaporizer temperature: 425 °C; sheat gas flow: 15 arb; aux gas flow: 7 arb; capillary temperature: 275 °C; skimmer 0) with the multiple reaction mode (MRM) using the substance-specific mass transitions (parent ion (m/z): 227 daughter ion (m/z); 212 (quantifier); and 133 (qualifier).

The data were collected and processed by Thermo XCalibur 2.2. The specified working range of the LC-MS method is between 1 ng/mL and 1000 ng/mL. The limit of quantification (the lowest measurable standard) is 1 ng/mL, while the detection limit is calculated using the lowest measurable standard and is 7 µg/kg for a sample weight of approx. 3 g in the case of BPA in PET bottles and for approx. 50 g (according to 50 mL) 0.02 µg/L for food simulants. The recovery rate of BPA inside PET is 100.83% (±2.23%, n = 18) in the lower calibration range, 99.27% (±2.28%, n = 18) in the middle range, and 98.24% (±1.78%, n = 18) in the upper calibration range. The recovery rate of BPA in water is 99.79% (±3.0%, n = 6). The suitability of the method for the determination of the substance was confirmed in validation measurements and independent inter-laboratory tests. The modified extraction method for the food simulants 20% ethanol, 10% ethanol, and 3% acetic acid was tested with spiked solutions. The recovery rate of BPA is 97.85% (2.59%, n = 2) in 20% ethanol, 96.07% (±1.35%, n = 2) in 10% ethanol, and 96.77% (±2.48%, n = 2) in 3% acetic acid.

3. Results

3.1. Concentations of Bisphenol A in Recyclate Containing PET Bottles in Europe

Overall, a total of 375 rPET bottles collected in Europe between 2019 and 2023 were analyzed regarding their BPA content. The bottles had varying contents of rPET ranging from 11% to 100%. The results of the quantitative determination of BPA in the rPET bottles are given in

Table 1. BPA concentrations in rPET bottles (n = 375) supplied from European bottling companies between 2019 and 2023.

	Concentration [mg/kg]
	2019	2020	2021	2022	2023	2019–2023
maximum	5.28	4.68	20.0	56.6	18.4	56.6
minimum	0.033	0.020	0.073	0.260	0.050	0.020
mean value	0.715	0.737	2.42	4.16	1.50	1.74
median	0.450	0.592	0.893	0.867	0.760	0.703
mean rPET content	71%	76%	98%	100%	100%	90%
analyzed samples	23	116	51	56	129	375

. In addition, five reference bottles made from virgin PET were analyzed. These reference bottles did not contain BPA above the detection limit of 0.007 mg/kg.

The average rPET content in the investigated PET bottles increased from 71% in 2019 to 100% in 2022 and 2023. The average content of rPET in the bottles over the last five years was 90%. The results show that the mean value of BPA concentration increases from 0.715 mg/kg in 2019 to 4.16 mg/kg in 2022, which correlates with the increasing rPET content in the analyzed PET bottles. In 2023, the mean BPA concentration drops significantly to 1.50 mg/kg. The median also increases between 2019 and 2021 but then drops to around 0.760 mg/kg in 2023. The highest BPA concentration was measured in 2022 at 56.6 mg/kg in one PET bottle. In 2021 and 2023, the highest concentrations were 20.0 mg/kg and 18.4 mg/kg, respectively. The median in 2022 was below that of 2021, which means that the high mean value in 2022 was triggered by the high outlier (and three further outliers with 34.4 mg/kg, 14.8 mg/kg, and 14.7 mg/kg). Without these four outliers, the mean value in 2022 would have been 1.99 mg/kg instead of 4.16.

3.2. Migration Kinetics into Food Simulants

Migration kinetics from PET into food simulants were used to determine the diffusion coefficients of BPA in PET. Similar approaches have been used in previous studies for antimony in PET as well as for the acetaldehyde-scavenging substance anthranilamide in PET [27,28]. The migration of BPA from PET bottles into food simulants was determined experimentally after contact up to 90 days and temperatures up to 60 °C in a migration kinetic experiment with seven kinetic points for each temperature and simulant. The concentration of BPA in the PET bottles used for the kinetics was determined to be 26.2 ± 0.4 mg/kg. An area of 100 cm² was then cut out of the PET bottles and totally immersed with the food simulants 3% acetic acid, 10% ethanol, and 20% ethanol, respectively. The samples were then stored at 30 °C, 40 °C, 50 °C, and 60 °C for up to 90 days in total. Below 30 °C, the migration is very slow, so that the concentrations of BPA are below the detection limits, and therefore no diffusion coefficient can be determined. The glass transition temperature of PET is between 60 and 70 °C. Therefore, 60 °C was chosen as the highest temperature for the migration experiments.

Samples of the food simulants were taken at regular intervals and analyzed for their BPA content. The results of the migration kinetics are given in

Table 2. Results of the migration kinetics of BPA into 3% acetic acid, 10% ethanol, and 20% ethanol as food simulants.

Food Simulant	Storage Time [d]	Concentration [µg/L]
		30 °C	40 °C	50 °C	60 °C
3% acetic acid	10	0.128 ± 0.010	0.242 ± 0.010	0.511 ± 0.022	1.11 ± 0.06
	20	0.188 ± 0.005	0.287 ± 0.007	0.716 ± 0.012	1.45 ± 0.08
	30	0.141 ± 0.005	0.310 ± 0.012	0.710 ± 0.029	1.57 ± 0.05
	40	0.155 ± 0.003	0.316 ± 0.005	0.841 ± 0.039	1.93 ± 0.11
	50	0.182 ± 0.006	0.341 ± 0.012	0.790 ± 0.007	1.89 ± 0.15
	60	0.159 ± 0.004	0.301 ± 0.009	0.634 ± 0.029	1.61 ± 0.04
	90	0.225 ± 0.008	0.460 ± 0.026	0.926 ± 0.106	2.42 ± 0.22
10% ethanol	10	0.154 ± 0.014	0.296 ± 0.026	0.688 ± 0.024	1.69 ± 0.044
	20	0.221 ± 0.016	0.349 ± 0.012	0.863 ± 0.033	2.02 ± 0.042
	30	0.172 ± 0.009	0.423 ± 0.035	0.969 ± 0.034	2.44 ± 0.07
	40	0.215 ± 0.002	0.409 ± 0.002	1.10 ± 0.01	2.64 ± 0.25
	50	0.235 ± 0.033	0.462 ± 0.009	0.990 ± 0.017	2.53 ± 0.13
	60	0.198 ± 0.013	0.424 ± 0.002	0.953 ± 0.009	2.59 ± 0.12
	90	0.260 ± 0.010	0.547 ± 0.029	1.22 ± 0.043	3.72 ± 0.070
20% ethanol	10	0.205 ± 0.005	0.383 ± 0.004	1.05 ± 0.02	2.88 ± 0.02
	20	0.286 ± 0.009	0.491 ± 0.004	1.30 ± 0.06	3.58 ± 0.06
	30	0.244 ± 0.005	0.603 ± 0.013	1.57 ± 0.03	4.25 ± 0.07
	40	0.259 ± 0.004	0.626 ± 0.018	1.74 ± 0.06	4.98 ± 0.10
	50	0.336 ± 0.010	0.700 ± 0.007	1.72 ± 0.07	4.26 ± 0.12
	60	0.298 ± 0.012	0.576 ± 0.026	1.39 ± 0.04	4.40 ± 0.17
	90	0.386 ± 0.019	0.736 ± 0.078	2.19 ± 0.10	6.20 ± 0.11

. The migration kinetics are visualized in Figure 1. The results of the migration kinetics show that, as expected, migration increases over time and with increasing temperature. The kinetics follow Fick’s law, which is confirmed by the fact that the migration of BPA increases linearly with the square root of time. It is therefore possible to calculate the diffusion coefficient (D_P) from the experimental data.

3.3. Diffusion Coefficients and Activation Energies of Diffusion

The migration of BPA from the PET bottles into the simulants follows Fick’s law. Therefore, diffusion coefficients (D_P) can be calculated from the linear increase in migration over the square root of time (Figure 1) according to Equation (1).

$${\displaystyle \frac{m}{A}}={\displaystyle \frac{2}{\sqrt{\pi }}}{c}_{P,0}{\rho }_P\sqrt{D_{P}t}$$

(1)

with m/A as area-related migration, c_P,0 as concentration of BPA in the PET bottle wall, ρ_P as density of PET, D_P as the diffusion coefficient of BPA in PET, and t as storage time. The activation energies of diffusion (E_A) and the pre-exponential factor (D₀) were then determined from Arrhenius plotting of the temperature-dependent diffusion coefficients versus the reciprocal temperatures (Figure 2).

The lowest diffusion coefficients were measured in the migration kinetics in 3% acetic acid. The diffusion coefficients from the migration kinetics in 10% ethanol were slightly higher, and those in 20% ethanol were slightly higher again. It is very likely that the surface of the PET bottles is swollen by the ethanol, which increases the migration out of the PET and thus the diffusion coefficients (D_P). This swelling effect of the food simulants was also found in previous studies [29,30,31,32]. Real food does not show this source effect, so the migration of BPA into 20% ethanol can be regarded as a worst-case scenario. Migration into real beverages (instead of ethanolic food simulants) will therefore always be lower. Therefore, more realistic migration of BPA is expected for 3% acetic acid or 10% ethanol as food simulants at ambient temperatures. In addition, the activation energy from 3% acetic acid is adequate for the modeling of migration into water or carbonated beverages.

3.4. Migration Modeling

Migration modeling is a useful tool to predict the migration of BPA into food. Almost all applications of PET bottles and PET trays can be calculated with the experimentally determined activation energies (E_A) and pre-exponential factors (D₀) from

Table 3. Diffusion coefficients (D_P), activation energies of diffusion (E_A), and pre-exponential factors (D₀) from experimentally determined migration kinetics of BPA from PET bottles into food simulants.

Food Simulant	Temperature [°C]	Diffusion Coefficient DP [cm2/s]	Activation Energy EA [kJ/mol]	Pre-Exponential Factor D0 [cm2/s]	Correlation Coefficient r2
3% acetic acid	30	1.10 × 10−18
	40	4.14 × 10−18	131.2	3.87 × 104	0.9922
	50	2.08 × 10−17
	60	1.19 × 10−16
10% ethanol	30	1.67 × 10−18
	40	6.87 × 10−18	140.9	2.63 × 106	0.9901
	50	3.68 × 10−17
	60	2.60 × 10−16
20% ethanol	30	3.29 × 10−18
	40	1.37 × 10−17	153.6	7.71 × 108	0.9883
	50	9.84 × 10−17
	60	7.72 × 10−16

. The very low threshold values of BPA in particular are a big hurdle to determining the migration of BPA experimentally. Therefore, the migration of BPA can be predicted with the activation energies from this study. The experimental activation energies and pre-exponential factor (D₀) of the migration kinetics (Table 3) were used for the predictions. Furthermore, a partition coefficient of K_P,F = 1 is assumed, which simulates a good solubility of BPA in foodstuffs.

In addition to calculations with the activation energies (E_A) determined in this study, two prediction models were also used. Regulation 10/2011 [33] allows the use of prediction models in the conformity assessment of FCMs. The so-called A_P model was published in 2005 and is currently generally recognized for the prediction of diffusion coefficients [34,35]. However, the A_P model is highly overestimative. Therefore, another model was developed for the prediction of the diffusion coefficients of PET [36,37]. It is based on experimentally determined activation energies (E_A) and predicts migration much more realistically than the A_P model. The predicted diffusion coefficients are given in

Table 4. Predicted diffusion coefficients (D_P) of BPA in PET according to the A_P model [34,35] (predicted with a molecular weight of BPA of 228.3 g/mol with A_P’ = 3.1 and τ = 1577 K).

Temperature [°C]	Diffusion Coefficient [cm2/s]
25	8.45 × 10−15
30	1.65 × 10−14
40	5.84 × 10−14
50	1.92 × 10−13
60	5.87 × 10−13

(A_P prediction model) and

Table 5. Predicted diffusion coefficients (D_P), activation energies of diffusion (E_A), and pre-exponential factors (D₀) of BPA in PET according to the E_A-based prediction model [36,37] predicted with a molecular volume of BPA of 221.11 ų.

Temperature [°C]	Diffusion Coefficient [cm2/s]	Activation Energy [kJ/mol]	Pre-Exponential Factor [cm2/s]
25	1.09 × 10−18
30	3.33 × 10−18
40	2.72 × 10−17	165.8	1.22 × 1011
50	1.95 × 10−16
60	1.24 × 10−15

(E_A-based model).

Based on the experimentally determined activation energies of diffusion as well as of the two prediction models, the concentration of BPA in the bottle walls was predicted, which correlates with a migration of 1 µg/L as well as 0.012 µg/L. The storage conditions were assumed to be up to 365 days at 25 °C in contact with food with a surface/volume ratio of 6. These storage conditions were currently used by the EFSA for the evaluation of PET recyclates in beverage bottles [23,26]. The results are shown in

Table 6. BPA concentrations in the bottle wall, which correspond to a migration of 1 µg/L (EU cube, layer thickness 300 µm, diffusion coefficients calculated from the activation energies of diffusion from experimental migration kinetics or prediction models (E_A-based and A_P models)).

Temperature	Storage Time [d]	Concentration [mg/kg]
		3% Acetic Acid (EA exp)	10% Ethanol (EA exp)	20% Ethanol (EA exp)	DP Predicted EA Based Model	DP Predicted AP Model
25 °C	10	1792	1542	1167	1083	12.3
	20	1267	1108	825	765	8.67
	30	1033	892	672	624	7.09
	50	800	686	521	487	5.52
	90	597	514	387	361	4.12
	120	519	444	336	314	3.57
	150	462	397	300	279	3.19
	200	402	344	260	226	2.76
	300	327	281	212	198	2.25
	365	297	255	192	179	2.04
40 °C	10	542	395	263	218	4.68
60 °C	10	111	77.7	45.0	32.3	1.48

and

Table 7. BPA concentrations in the bottle wall, which correspond to a migration of 0.012 µg/L (EU cube, layer thickness 300 µm, diffusion coefficients calculated from the activation energies of diffusion from experimental migration kinetics or prediction models (E_A-based and A_P models)).

Temperature	Storage Time [d]	Concentration [mg/kg]
		3% Acetic Acid (EA exp)	10% Ethanol (EA exp)	20% Ethanol (EA exp)	DP Predicted EA Based Model	DP Predicted AP Model
25 °C	10	21.5	18.5	14.0	13.0	0.148
	20	15.2	13.1	9.90	9.18	0.104
	30	12.4	10.7	8.07	7.49	0.0851
	50	9.60	8.23	6.25	5.84	0.0663
	90	7.16	6.17	4.65	4.33	0.0494
	120	6.23	5.33	4.03	3.77	0.0428
	150	5.54	4.77	3.60	3.35	0.0383
	200	4.82	4.13	3.12	2.91	0.0331
	300	3.93	3.37	2.54	2.38	0.0270
	365	3.57	3.06	2.30	2.15	0.0245
40 °C	10	6.50	4.74	3.16	2.62	0.0562
60 °C	10	1.33	0.932	0.540	0.388	0.0178

and Figure 3 as a correlation of the bottle wall concentration and the storage time.

Regarding the new proposed threshold limit of 1 µg/L, maximum concentrations of 297 mg/kg (3% acetic acid), 255 mg/kg (10% ethanol), and 192 mg/kg (20% ethanol) in the PET bottle wall are predicted for 365 days of storage at 25 °C. Shorter storage times increase the above-mentioned maximum concentrations in the bottle wall. In addition to the storage conditions of up to 365 days at 25 °C, forced storage conditions of 10 days at 40 °C and 10 days at 60 °C were calculated. The forced contact conditions are used to accelerate the experimental migration testing and are part of the legal requirements according to Regulation 10/2011 [33]. Based on the experimentally determined activation energies (E_A) for 20% ethanol, the test conditions of 10 d at 40 °C simulate storage for 195 days at 25 °C. In contrast, testing for 10 d at 60 °C very unrealistically simulates 18.3 years of storage at 25 °C. Similar results were obtained for 3% acetic acid at 100 days (10 days at 40 °C) and 7.2 years (10 days at 60 °C), respectively. The test conditions of 10 days at 40 °C are therefore much more suitable for real food applications than the current test conditions of 10 days at 60 °C in accordance with European Regulation 10/2011 [33]. The literature data for the acetaldehyde-scavenging additive anthranilamide [27] in PET bottles as well as for the polymerization catalyst antimony [28] confirm that 10 days at 40 °C are better testing conditions than 10 days at 60 °C.

The current SML of 50 mg/L [15] allows bottle wall concentrations of BPA of 9620 mg/kg after storage of 365 d at 25 °C, 13170 mg/kg (10 d at 40 °C), and 2240 mg/kg (10 d at 60 °C), respectively. The implied migration level of 0.012 µg/L based on the EFSA TDI still allows 3.57 mg/kg (3% acetic acid), 3.06 mg/kg (10% ethanol), and 2.30 mg/kg (20% ethanol) of BPA in the bottle wall when stored for 365 days at 25 °C.

The results of the activation energy-based prediction model are very close to the results for 20% ethanol. This indicates that the migration is predicted realistically by the activation energy-based E_A model but that it is slightly overestimative. In contrast, as expected, the generally accepted A_P model predicted the migration of BPA from PET much more conservatively, which results in significantly lower concentrations of BPA, which corresponds to a migration of 1 µg/L and 0.012 µg/L.

4. Discussion

The concentration of BPA in PET bottles containing rPET for drinking water and soft drinks was monitored over a period of five years, with a total of 375 analyzed bottles (Table 1). The results showed mean annual values between 0.715 mg/kg and 4.14 mg/kg. The concentrations increase significantly from 2021, which can be explained by a higher proportion of post-consumer recyclate in the PET bottles. The median was between 0.450 mg/kg and 0.893 mg/kg, which is significantly lower than the mean value. This indicates that individual (high) outliers trigger an increase in the mean value. The median is therefore more suitable for evaluating the development of BPA in rPET bottles. However, the high concentrations in the outliers should be considered critical and should be observed in further monitoring programs.

The results of the quantification of BPA in rPET are in good agreement with the literature data. Dreolin et al. found BPA concentrations in rPET pellets in the range between 0.394 mg/kg and 10.12 mg/kg (n = 13), while virgin PET contained significantly lower concentrations of BPA (between 0.025 mg/kg and 0.432 mg/kg, n = 10) [9]. Nunez et al. found 0.014 ± 0.00005 mg/kg in virgin PET and 0.25 ± 0.082 mg/kg, 0.33 mg/kg, and 2.35 ± 0.08 mg/kg in three rPET samples [38]. Nuygen et al. found 0.24 ± 0.04 (n = 5) and 0.38 ± 0.02 mg/kg (n = 5) in washed post-consumer colored and transparent PET flakes [39]. Fan et al. determined BPA in 16 PET bottles from the Chinese market. The BPA concentration was determined to be between 0.0641 ± 0.00582 mg/kg and 0.0377 ± 0.00573 mg/kg (n = 16). It was not described in the study whether the PET bottles contained recyclates [40].

The annual median values of BPA in the investigated PET bottles are all below 3.57 mg/kg (3% acetic acid), 3.06 mg/kg (10% ethanol), and 2.30 mg/kg (20% ethanol), which are concentrations in the PET bottle wall that correlate with a migration of 0.012 µg/L after storage for 365 days at 25 °C (Table 6). The concentrations found in rPET bottles will therefore not be exceeded, and the non-intentionally introduced concentrations of BPA do not pose a problem for PET recycling. Assuming that most of the mineral water and soft drinks packed in PET bottles are consumed within around 90 days, the mean values are below the corresponding maximum concentrations in PET, which are 7.16 mg/kg (3% acetic acid), 6.17 mg/kg (10% ethanol), and 4.65 mg/kg (20% ethanol), and are therefore below the TDI-implied migration level of 0.012 µg/L. However, the very high outliers are far higher. It is therefore meaningful and recommendable to check the BPA in the individual batches used in the production of rPET bottles. In addition, the entry route of BPA into the rPET cycle should be clarified.

Fan et al. [40] also determined the migration at different temperatures. The migrated amount of BPA after storage for 7 days at 4 °C and 25 °C were determined to be between 0.26 ± 0.07 and 18.7 ± 2.58 ng/L and between 0.62 ± 0.10 and 22.6 ± 4.97 ng/L (n = 16), respectively. When stored at 70 °C for 7 days, the BPA concentration ranged from 2.89 to 38.9 ng/L. The migration values show a high scatter range, but based on the concentrations in the PET bottles of 0.0641 ± 0.00582 mg/kg and 0.0377 ± 0.00573 mg/kg, the predicted migration from the activation energies determined in this study is within these concentration ranges.

Dreolin et al. also measured the migration of BPA into 3% acetic acid and 20% ethanol [9]. For a bottle wall concentration of 0.263 mg/kg, the results of migration into 20% ethanol after storage for 10 days at 60 °C were 3.4 ± 0.7 µg/L. Based on migration kinetics into 20% ethanol from this study, a migration of 5.85 µg/L is predicted, which is in good agreement with the experimental result of Dreolin et al. Another bottle with 1.26 mg/kg BPA results in migration value of 3.5 ± 0.5 µg/L and 4.2 ± 0.6 µg/L for 3% acetic acid and 20% ethanol, respectively. Based on the kinetic results of this study, 12.2 µg/L (3% acetic acid) and 30.3 µg/L (20% ethanol) were predicted after 10 days at 60 °C. The experimental values measured by Dreolin et al. and the predicted values based on the results of our migration kinetics are approximately one order of magnitude above the suggested migration limit of 0.012 µg/L. However, this is due to the fact that 10 days at 60 °C greatly overestimates the migration and is unsuitable for a migration assessment for room temperature storage of up to 1 year.

Some studies in the literature determined the concentration of BPA in water packaged in PET bottles. In almost all cases, however, only the water is measured, not the PET bottle. This can lead to false conclusions. For example, Guart et al. [41] detected BPA in PET-bottled water. However, BPA was not detected in the PET bottles cut in pieces but was detected in HDPE caps. These results show the migration of BPA is most likely from the caps. Amiridou and Voutsa [42] found 0.0046 µg/L of BPA in water bottled in PET. After exposure for 15 and 30 days at around 40 °C, BPA does not increase in concentration, which is an indication that BPA does not migrate from the bottle wall but has another source. It is therefore very important that the PET bottle wall concentration is always determined in such migration studies. The diffusion coefficients determined in this work can also be used to validate the migration of BPA from the bottle wall into the beverage.

Some studies investigated the influence of temperature or sunlight on the amount of BPA released into water [7,43,44,45]. The concentrations of BPA ranged from 0.001 µg/L to about 0.02 µg/L. Increasing the temperature and sunlight increased the concentration of BPA in the packaged water. Unfortunately, none of the studies determined the concentration of BPA in PET materials, so no diffusion coefficients can be derived. A comparison with these results is also difficult as the migration is directly related to the concentration of BPA in PET, which is not known. However, the magnitude of the migration is in a similar range as in our study. Khanniria et al. [46] determined concentrations of BPA in packaged drinking water to be higher by a factor of approximately 10. However, the concentrations of BPA in the PET bottles were not measured here either, so the cause of the higher concentrations remains unclear.

The experimentally determined diffusion coefficients of BPA in PET from this study can be compared with those from another published predictive diffusion model. Using the so-called A_P model [34,35] and the E_A-based model [36,37], diffusion coefficients for BPA in PET at different temperatures can be calculated, as shown in Table 4 and Table 5. The A_P model works with a fixed activation energy of diffusion (E_A) of 100 kJ/mol applicable to any molecule in PET. This value is significantly lower than the experimentally determined E_A values of 131.2 kJ/mol (3% acetic acid), 140.9 kJ/mol (10% ethanol), and 153.6 kJ/mol (20% ethanol). The E_A-based model predicts an E_A value of 165.8 kJ/mol. Together with the pre-exponential factor (D₀), the E_A is important for correctly including the influence of temperature on migration.

At 30 °C, the E_A-based model [36,37] predicts almost exactly the same migration as experimentally determined in 20% ethanol (factor 1.0). The migration in 3% acetic acid is overestimated by a factor of 1.7). At 60 °C, slight overestimation by factors of 1.3 (20% ethanol) and 3.3 (3% acetic acid) is found. This is in good agreement with the results from a validation study with a factor 1.3 on average based on 263 diffusion coefficients in PET for 66 substances at temperatures between 40 °C and 120 °C [37] and confirms that migration prediction from PET using an E_A-based model works well. The E_A-based predictive model calculates migration to be slightly higher than the experimentally determined value. This is reasonably required by the legislation [33].

The A_P model with A_P = 3.1 and τ = 1577 K [34,35], on the other hand, overestimates the migration at 30 °C by a factor of 71 compared to the experimental results from the 20% ethanol migration kinetics. In comparison, the experimental 3% acetic acid kinetics at 30 °C migration is overestimated by a factor of 122. At 60 °C, overestimation factors of 71 (20% ethanol) and 70 (3% acetic acid) are obtained. These comparison figures show that the A_P model is obviously more overly conservative than necessary for polymers such as PET with low diffusion characteristics and activation energies above 100 kJ/mol.

5. Conclusions

PET does not contain intentionally added BPA. However, trace contaminations may be introduced via the rPET content because BPA is ubiquitously present. Diffusion modeling is a tool to predict the migration of BPA into PET-packed food. Diffusion coefficients and activation energies of diffusion were derived from migration kinetics into food simulants at different temperatures. From this, the migration of BPA into food can be calculated down to levels in the low ng/kg food range or even below if the concentration in the rPET is known, which, in fact, can be measured. Furthermore, conversely, maximum concentrations in PET can be calculated, which correlate with such low levels, thus opening up an option to indirectly control these low levels in the PET itself.

The current SML of BPA of 50 µg/kg food can never be exceeded because the corresponding concentrations in rPET would need to be unrealistically high (around 9600 mg/kg assuming storage for 365 days at 25 °C, predicted with the experimentally determined E_A and D₀ from 20% ethanol kinetic). The new proposed threshold limit of 1 µg/L [22] will allow maximum concentrations of BPA in the bottle wall of 296 mg/kg (3% acetic acid), 254 mg/kg (10% ethanol), and 191 mg/kg (20% ethanol). These maximum concentrations are far above the BPA concentrations measured between 2019 and 2023. Therefore, also, the new proposed threshold limit cannot be exceeded. On the other hand, however, the new TDI of 0.0002 µg/kg bw/d and a corresponding maximum migration level of 0.012 µg/kg food for a 60 kg person is more challenging. From the results of this study, it can be concluded that maximum BPA concentrations of 3.57 mg/kg (3% acetic acid), 3.06 mg/kg (10% ethanol), and 2.30 mg/kg (20% ethanol) in PET bottle walls intended for food contact for up to 365 d at 25 °C are in compliance with the maximum migration level of 0.012 µg/kg. These maximum values are borderline with the experimentally determined BPA concentrations in rPET. However, the planned prohibition on intentionally added BPA in food packaging will probably have a positive effect on the concentrations of BPA in rPET, most probably resulting in (significantly) lower concentrations of BPA in rPET-containing beverage bottles. The monitoring of BPA in rPET over the next few years will show whether this reduction of the BPA concentration in rPET bottles was unique or is a trend. The monitoring of BPA in rPET initiated with this study can now be continued after the ban on BPA and will show to what extent legal measures have an impact on non-intentionally added BPA in rPET.

The applied method for the determination of BPA described in this paper was found suitable for quantification in rPET. The detection limits are low enough, and the method can be used for the quality assurance control of recyclates.

The BPA concentrations determined in this study provide an overview of how much BPA can be determined in commercial beverage bottles with recyclate content. As expected, the concentrations in the investigated 375 bottles show a broad variation. Under given migration conditions, the migration of BPA into the beverage is directly proportional to the concentration of BPA in the PET bottle wall. In order to minimize the migration of BPA, the concentration in the bottle wall can be minimized.

The study confirms that prediction of the diffusion coefficients according to the E_A-based model and the resulting migration calculations of BPA from PET into food match very well with the experimental results. In contrast, the A_P model was confirmed to be highly overestimative in comparison with the experimentally determined migration into food simulants.

Finally, BPA should be minimized in rPET, which would require precise knowledge of the entry route of BPA into the rPET cycle. So far, the scientific literature provides little information on the contamination pathways of BPA into rPET, which deserve further explorative studies.