Investigations into the Migration of Packaging Components into Food

In recent years, concern has arisen about the potential for compounds in food packaging, such as bisphenyl A (BPA), to migrate into food. Rafael Paseiro-Cerrato of the U.S. Food and Drug Administration (FDA) has conducted studies to investigate this concern. Specifically, those studies have addressed various types of can coatings in both short-term and long-term studies, as well as the question of whether short-term study protocols accurately simulate migration during longer-term storage. He used a range of analytical techniques, including gas chromatography (GC), high performance liquid chromatography (HPLC), and ultrahigh pressure liquid chromatography (UHPLC) combined with a diode array detector (DAD), charged aerosol detection (CAD), mass spectrometry (MS), and high-resolution MS (HRMS). He also investigated direct analysis in real time–MS (DART MS). Here, he talks to us about that work.

You have done a lot of research on the migration of polymeric compounds from food packaging, particularly can coatings, into food. You note that cross-linked polyester resins are being introduced on the market as alternatives to epoxy resins, which typically contain BPA as coatings for metal food cans. In one study, you looked at methods for identifying unknown compounds from polyester can coatings that could potentially migrate into food (1). Going into the study, how much a priori information did you have about the potential compounds in the liners? Are food manufacturers not required to disclose the compounds they use in can liners?

In the United States, any food contact material (FCM) must be approved by the Food and Drug Administration (FDA) before being placed into the market. Manufacturers must comply with the current legislation regarding pre-market approval of food contact materials. In the case of food can coatings approved prior to 2000, they are mainly represented by listings in Title 21 Code of Federal Regulations (CFR) Parts 170–199 as the result of agency actions on food additive petitions. In the CFR, lists of substances authorized to be used in the manufacturing of FCM are displayed, and it is available for public access. Since 2000, the way to introduce a FCM into the market is through the food contact notification (FCN) process. Manufacturers must file information to the FDA related to the FCM, including administrative, chemical, environmental, and toxicological information on substances. Migration of oligomers, residuals, and impurities must be determined, and estimated exposure is evaluated against toxicology information. A database that lists effective premarket notifications can be found on the FDA website (www.fda.gov). This information, as well as the information that can be accessed in the published literature, was used as a starting point for the identification studies.

How many different can coatings did you analyze? Do you think you have identified any compounds of major health concern? How large was the potential pool of compounds?

First, I would like to point out that the idea behind this study was to identify potential migrants and to track them in migration studies. In the real world, cans are processed and shipped to distributors. They may be sold immediately, or they may be warehoused for varying periods of time. In a modern distribution system, the storage time is very short. But some unpopular products can linger for months, or longer, in warehouses. FDA has designed a migration test to allow notifiers to develop migration data in a short (10 d) time period in the laboratory.

The goal of these experiments was to evaluate if current FDA migration tests accurately predict migration results into food and food simulants after a long-term storage. To perform these migration experiments, we needed first to identify potential migrants in the coatings, and, once identified, track them through migration experiments. In this first manuscript, we analyzed and identified potential migrants in polyester coatings. But we extended the food cans study to other coatings, including epoxy, acrylic-phenolic, and vinyl coatings. Related to the second question, as explained above, because any FCM needs to be approved by the FDA before being placed into the market, we did not expect any of these migrants to be a health concern. On the other hand, for the identification of potential migrants, we used internal data bases that contains hundreds of compounds.

Some of the samples in your study were subjected to a retort step. For those not familiar with food processing, could you briefly explain the retort process, and when it is used? In your estimation, how closely does the retort step in the analytical test mimic real-world conditions, and how did you arrive at that opinion?

A retort step, as we think of it, is a high-temperature sterilization step (typically around 121 °C for up to 2 h) to ensure the safety of canned foods, particularly those of low acidity. Dangerous microorganisms are eliminated, and the shelf-life of the canned food is extended. To mimic this common practice in food cans, to simulate real world conditions in our experiments, some of the employed cans were retorted in a commercial food retorting machine. In addition, the FDA guidelines for industry (see www.fda.gov) specifies that if cans are going to be subjected to a retort step, a thermal treatment at 121 °C for 2 h, followed by a migration test for 10 d at 40 °C, should be performed. Therefore, by retorting the food cans, we are simulating real food can processing conditions as well as complying with the FDA guidelines.

Following extraction with acetonitrile, you used a variety of analytical techniques, including gas chromatography–mass spectrometry (GC–MS), high performance liquid chromatography (HPLC) with a diode array detector (DAD) and MS (HPLC–DAD/MS), HPLC with DAD and charged aerosol detection (CAD), and ultrahigh-pressure LC with high-resolution MS (UHPLC–HRMS). Why did you need to use so many techniques, and how did you determine the optimum technique for each of your experiments?

The aim of this experiment was to identify a large variety of potential migrants in the coatings. Migrants can be very diverse and can have different molecular weights (MW), polarities, belong to different chemical families (acids and esters, for example), and have different chemical properties (such as, for example, their boiling point). Therefore, migrants may have different responses depending on the analytical techniques. To capture a wide range of potential migrants present in the coating, the most logical way is to use a large battery of analytical techniques. We used GC for detection of volatile and semi-volatile compounds, and LC for non-volatile compounds. The various detectors captured migrants with different physicochemical properties. For example, DADs detects substances with chromophore groups, while charged-aerosols-detectors (CADs) may detect substances without chromophore groups. By using both detectors, we ensure that we are capturing both compounds with chromophore and those with no chromophore groups. In addition, each detector gives relevant information about migrants. For example, the HRMS supplies accurate mass of compounds, while the DAD gives spectrometric information. If during the migrant’s identification we obtain the accurate mass of an oligomer with a good ppm agreement (≤ 5 ppm) and an appropriate retention time, the technique gives enough confidence that this is likely the proposed compound. However, polyester oligomers may be formed with different isomeric monomers, such as isophthalic and terephthalic acid, which have the same exact mass, but completely different absorption in the ultraviolet (UV) spectra. The additional spectral data supplies valuable information on which one is the principal monomer present in that oligomer. The DAD detector will give this type of information, and we will know if the tentatively identified oligomers are mainly based on isophthalic or terephthalic acid based on the UV spectra. The use of several detectors complements each other. Data obtained using several analytical techniques also guides the researcher on which would be the best strategy to determine migrants for future analysis.

What compounds did you identify, and what did you find in terms of which techniques were best to identify which compounds?

Most of the identified compounds were polyester oligomers, but we also identified commonly used monomers used in polyester coatings. Each employed technique has relevance in the identification process, but probably the technique that supplied the most relevant information for the identification was, in my opinion, the UHPLC-HRMS, since it allows to separate compounds in the analytical column, and gives accurate mass and supply information about the fragments of a migrant. In any identification process, it is important to state that compounds remain tentatively identified until analytical standards are obtained and analyzed to experimentally prove that the identification was correct. In this study, to give confidence in the identification results, we were able to obtain standards of some tentatively identified migrants that supported that our identification was appropriate.

You also tested whether rapid, direct detection of the identified oligomers was practical, using direct analysis in real time–MS (DART–MS). What did you find? Could future studies be limited to a small set of techniques?

Using DART–MS, we tentatively identified 23 compounds directly from the coating (no sample preparation) in just a few minutes and with minimal method optimization. The DART–MS analysis was as simple as placing the coating between the DART source and the HRMS for a few seconds. In the end, we obtained in minutes very similar results as in the UHPLC-HRMS, where the analysis may take days. This means that the employment of DART–MS represents an improvement in the sample throughput for identification of substances, and it is a technique that has a lot of potential for food packaging applications.

Do you think you were not able to detect any important compounds?

When analyzing a matrix with unknown chemical composition, no results are 100% definitive. We could potentially miss information during the process. When developing strategies for the identification of unknowns, there are different approaches to try to capture most of the compounds of interest. In this study, we used different techniques for identification (GC, LC), several databases, instrument settings that can capture large variety of molecules (extended chromatographic gradients), and detectors in full scan mode. To obtain confidence in the identification, and to prove that the employed identification methods work, we used a wide variety of standards with different MW, polarity, and chemical groups. We typically use a QC mixture to qualitatively calibrate the instruments (for example, a Grob mixture) to help ensure we get all compounds through the instrument and to the detector. However, that does not mean that we are detecting all compounds in the extracts. Migrants maybe extracted at a low concentration, have less sensitivity in a specific detector, or may simply not be included in any database. However, the identified compounds in this study were expected, and with the identified migrants, we were able to perform the long-term migration studies that were the main goal of  this investigation.

In a subsequent study, you evaluated long-term migration testing from polyester can coatings into food simulants (2). In that study, you examined the short-term protocols for such tests that are intended to mimic processing and long-term storage conditions. Why did you undertake this study in this way?

Current migration tests are intended to model what happens over those long storage times. After the retort step, migration tests usually go up to 10 d at a lower temperature. However, can coatings may be in contact with food for several years (approximately 2 to 5 years). Therefore, we wanted to evaluate if the current testing adequately simulates what may happen in food cans over the years, to ensure that food cans are
tested appropriately.

What were the food simulants you used, and why did you use simulants instead of real food samples? What are the challenges or caveats in extrapolating from your method to real food samples?

We used five types of food simulants: water, 3% acetic acid, 10% ethanol, 50% ethanol, and isooctane. These simulants either mimic or exaggerate migration that could occur in a wide range of foods, from aqueous to fatty foods. Food simulants are approved to be used under several national legislations, including the United States. Food simulants are usually simple solvent mixtures that facilitate the analysis of migrants (analyzing food is quite challenging), and tend to overestimate the migration that would occur in real foods. Depending on the type food intended to be in contact with a food contact material, there are specifically recommended food simulants. For aqueous foods, for example, a recommended food simulant is 10% ethanol. When extrapolating the data obtained in the food simulants to food, we will generally have worst case scenario estimates of how much mass transfer occurs into the real food. In addition, for this study, migration will supply relevant information on what may happen after a long-term exposure when the coating is in contact with food.

How did you determine what compounds to study? How did you handle the lack of standards for many of the oligomers studied?

For the monomers of interest (isophthalic acid, terephthalic acid, and nadic acid), we had standards available, and therefore they were selected for the analysis. For the oligomers, we did not. Selection of compounds with no available standards was based on their different polarities, structures, and concentrations in the samples. To monitor these substances throughout the experiment, we used some readily available oligomer standards that belonged to the same family of polyester oligomers as proxies. This approach has been used in the past by other authors. By using these available standards, we could track the relative changes in intensity during the migration experiment. In addition, the use of standards facilitates the method validation and gives confidence in the instrument performance.

What approach did you take to sample pretreatment?

We used a sample concentrator and an C18 SPE to achieve appropriate sensitivity for the analysis. In the case of monomers, I would point out the need of adding formic acid to the water simulant before loading the SPE C18 cartridge. This acidic condition allows the monomers to be retained in the cartridge, and then be eluted using acetonitrile. For some oligomer analysis, food simulants were injected in the HPLC without any sample pretreatment. This is a good example of the advantages of using food simulants instead of real food.

In this study, you monitored the migration of monomers and identified oligomers of polyester can coatings into food simulants during short-term and long-term migration experiments, from one day to roughly 1.5 years (515 d). How did you choose the time frame of the migration periods? What would happen after several years of storage?

Food cans are designed to have a shelf life between 2 to 5 years. Considering that traditional migration test consists in real time thermal processing followed by 10 d at 40 °C, we decided a that a migration test at 40 °C for 1.5 years monitoring multiple time points from 1 d to 1.5 years could model migration over long-term storage times.

Regarding the second question, the experimental data obtained in the study up to 1.5 years suggest that increases in concentration may occur beyond the 10 d at 40 °C during long-term migration experiment, particularly in ethanol-based food simulants.

What did you find in terms of the migration of the compounds into the food simulants? What do these results indicate about the accuracy of short-term studies?

As specified in the previous question, increases in concentration may occur during the long-term experiment in food simulants. However, even in the worst-case migration scenario, the monomer concentrations in the simulants were well below the limits of concerns. This means that, even though the migration concentrations do not represent a concern, the short-term studies may underestimate the exposure of some compounds. That is why the experimental data obtained from this study suggest that changes in concentration may occur beyond the 10-d test. Therefore, migration protocols may need to be reviewed or modified to accurately predict migration after a long-term storage.

You also conducted a similar study analyzing short-term and long-term migration from epoxy resins and acrylic-phenolic coatings (3). Did the coatings addressed in this study present any analytical challenges different from the previous study of polyester coatings?

The analytical approach for the analysis of the coatings was similar. This is, in part, because food simulants simplify the analysis when compared with real food. In the method validation results, when comparing recoveries and stability in the food simulants, it can be observed that epoxy derivatives are more stable, and recoveries are better and more robust. These results are consistent with the chemistry of the analyzed compounds. This gives useful information about the challenges of analyzing different can coatings.

In both of these studies, did you investigate not only compounds present in the coatings, but also byproducts that might be formed by interaction of the coatings and the food simulants or real food samples?

During both studies, we tracked several migrants in the simulants over the experiment. Comparison of the migration tests and an acetonitrile extract of the coating showed that the migrant concentration in the extract was lower for certain compounds than in the simulants. This was particularly true for polyesters, but also for acrylic-phenolic coatings. In the case of epoxies, we found two compounds, probably non-epoxy derivatives, that also fit to this pattern. It was hypothesized that hydrolysis of certain compounds could occur during the migration experiments. In the case of polyesters, we observed similar trends in the stability study using available standards, which support this hypothesis. There is another explanation for this experimental observation, which is related with the simulant interacting with the coatings.

What did you find in this study in terms of the migration of the various compounds? Were there significant differences in the migration into the different types of food simulants?

We observed that in the water food simulant after the retort steps, the migrant concentration remained stable over the long migration experiments. In the case of 50% ethanol food simulant using non-retorted cans, the migrant concentration increased beyond the 10-d test until it reached a plateau and then remained stable. We have hypothesized that migration test at higher temperatures could give better estimates of migration. In terms of concentrations, they were higher in 50% ethanol as expected. This is because mass transfer rates of these compounds are usually higher in less polar food and food simulants.

Do you have any serious concerns about what you found?

No. As obtained in our previous studies related to can coatings, obtained results show that, even in the worst migration scenario, concentrations in the food simulants were below the limits of concern.

Based on the results of the study, how well do standard short-term study protocols simulate real conditions over longer-term storage?

Results obtained in this study align with our previous migration studies on can coatings, which suggest there might be a need for revision of the long-term storage simulation protocols to more accurately represent what would occur after actual long-term storage. Based on the result of these simulant studies, we are continuing the investigation by performing accelerated migration tests in food simulants, and studying the influence of long-term storage in real food in order to guide decisions on what migration test modifications might be necessary.

You followed up on your long-term migration studies of can coatings with an assessment of accelerated migration testing (4). What was the aim of this study? What did you find in this study?

The aim of this study was to evaluate if accelerated migration testing conditions (60 °C for shorter time periods) could simulate migration we observed in previous long-term migration studies in can coatings at 40 °C for 1.5 years, because it is not practical for those hoping to submit Food Contact Notifications to routinely conduct long term migration tests.

To perform the experiment, we placed the previously analyzed coatings, including vinyl, polyester, epoxy-resins, and acrylic-phenolic coatings, in contact with food simulants. We conducted a migration test from 4 h to 30 d at 60 °C. We sampled at multiple time points during the experiment. The accelerated migration testing condition in food simulants showed that migration at 60 °C for shorter times (days) produced similar results to the long-term study (up to 1.5 years) at 40 °C. This suggests how we might adjust our simulation protocol sometime in the future.

In another study, you examined the migration of three compounds from polyester food cans into actual food samples (5). How did you determine the compounds to study, and what food samples to test?

In this study, we analyzed food cans that can be purchased in the U.S. market. We selected food cans lined with polyester coatings. We wanted to collect data on migrant concentrations in food. In addition, we also wanted to use this study to investigate the influence storage time on migrant concentrations. To perform the experiment, we selected two of the previously tentatively identified oligomers and monomers used in the manufacture of polyester that was found in the food simulant migration studies. We selected a wide variety of canned foods, such as coconut milk, chicken noodle soup, and mushrooms, among others.

What analytical techniques did you use in this study, and why?

We used Fourier transform infrared (FT-IR) spectrometry for the coating identification, allowing us to select the products manufactured with polyester coatings for the analysis. In the sample preparation process, food was homogenized, migrants were extracted with acetonitrile containing 0.1% formic acid, centrifuged, concentrated, precleaned, conditioned in appropriate solvent, filtered, and finally injected in the chromatographs. LC was selected as the analytical technique for the analyses of migrants, based on the results from the previous work. For monomer quantification, we used a LC–MS/MS detector, which is selective, robust, and provides good sensitivity. For oligomer determination, we selected LC-DAD, which, based on our previous experience, supplies good and reliable results for oligomers determination. In addition, we also used UHPLC-HRMS for the confirmation of these compounds.

What did you find overall? And what did you find specifically in terms of the influence of storage times, store, and lot in the concentrations of packaging migrants?

In this study, we measured the monomers and we estimated oligomer concentrations in different food samples. Obtained results were in the part-per-billion concentration range. We also investigated the influence of storage time. In general, we observed that, although some migrants increased in concentrations based on the best-by-date of the products, others decreased in concentration in the food samples. However, because products could not be traced back, we don’t have information on multiple factors that could impact migrant concentrations. These include temperature variations during storage or transport, lot variations, and overall food consistency. As a result, because conditions were not controlled, it is not possible to attribute observed changes to storage time in the studies samples. The method we used for this study was validated in the different food samples following the FDA FVM Guidelines for Single Laboratory Validation.

Do these results suggest the need for significant additional studies of a wide range of real food samples? What are your next steps in this work?

A controlled experiment on real food samples would enhance our understanding of the impact long-term storage has on real food. However, although it would be interesting to continue investigating the effects of long-term storage in real canned foods, that would require significant help and input from the food industry. A controlled experiment like this would require a diverse set of food cans, coated with the same coating, (polyesters, vinyl, acrylic-phenolic or epoxy) and placed in contact with different types of food (from aqueous food to fatty food), in a set of controlled time-temperature conditions (20, 30, and 40 ^oC for times up to 1.5 years) over a long period of time.

References

1. R. Paseiro-Cerrato, S. MacMahon, C.D. Ridge, G.O. Noonan, and T.H. Begley, J. Chromatogr. A 1444, 106–113 (2016). DOI: 10.1016/j.chroma.2016.03.038
2. R. Paseiro-Cerrato, G.O. Noonan, and T.H. Begley, J. Agric. Food. Chem. 64, 2377–2385 (2016). DOI: 10.1021/acs.jafc.5b05880
3. R. Paseiro-Cerrato, J. DeVries, and T.H. Begley, J. Agric. Food. Chem. 65, 294–2602 (2017). DOI: 10.1021/acs.jafc.7b00081
4. R. Paseiro-Cerrato, L. DeJager, and T.H. Begley, Molecules 24, 3123 (2019). DOI: 10.3390/molecules24173123
5. R. Paseiro-Cerrato, L. DeJager, and T.H. Begley, Food Control 101, 69–76 (2019). DOI: 10.1016/j.foodcont.2019.02.033

Rafael Paseiro-Cerrato is an expert in the field of food packaging materials, with a strong background in analytical chemistry, pharmacy, nutrition and food science. He actively leads research projects involving analytical methods development, validation and data analysis resulting in numerous publications. Since 2012, he has been working at the Center for Food Safety and Applied Nutrition within the US Food and Drug Administration in College Park, Maryland.