Recent progress in the application of chromatography-coupled mass-spectrometry in the analysis of contaminants in food products

Abstract

An increasing concern has marked the past few years for food quality and safety, not only by the research community but also among the general public. Due to the increasing prevalence of emerging contaminants/analogues in various food matrices, significant efforts have been focused on developing green extraction, matrix preparation, multi-targeted, and potentially multi-matrix techniques that minimize waste generation, cost, and the use of organic reagents to optimize recyclability throughout the analytical procedure. This, therefore, not only guarantees consumer protection through food safety but also contributes to progress in green analysis, ultimately preventing the one-step forward and two-steps back phenomenon of ensuring food safety while damaging the environment. In this review, we provide an integrated, concise, and comprehensive timeline of the research endeavours in the application of chromatography coupled with mass spectrometry in the analysis of contaminants in plant-, aquatic-, and animal-derived food matrices.

Graphical abstract

Highlights

• •Food contaminant detection hinges on extraction methods and data analysis.
• •Green-multitargeted analytical strategies have been extensively researched.
• •AI-based data analysis resolves the traditional library searching coverage problems.
• •Optimisation of extraction techniques is pivotal to realising multi-matrix analysis.
• •Future research should not overlook investigating contaminant generation mechanisms.

1. Introduction

The projected global human population is anticipated to reach approximately ten billion individuals by 2050; with this ongoing growth, we are confronted with escalating obstacles in guaranteeing the widespread availability of safe, nourishing, and healthy food. By 2050, food production will need to surge by about 60 % of the levels recorded in 2012 to meet the rising demand (Nadathur et al., 2024). Additionally, as incomes in developing nations continue to rise and living standards improve, there has been a surge in the desire for new food sources and production systems (NFPS) (Tan et al., 2024) primarily arising from causes such as global trade, shifts in consumer preferences, the capacity for ecological benefits, and developments in Food production systems resilient to climate change (Y. Zhang et al., 2022). The food supply chain extends from the cultivation stage to the final consumption and involves a series of procedures, including processing, preservation, storage, packaging, transportation, and consumption; thus, various pollutants encompassing veterinary/pesticide/antibiotic drug residues (Liang et al., 2022; Qin et al., 2024; Wahab et al., 2022), heavy metals, emerging pollutants, illicit adulterants, and bacterial toxins, among various other substances (Sarker et al., 2022) can be incorporated into food items at several phases of the food supply chain that present severe health hazards if not identified before intake, making food safety an essential aspect of the developed world due to the significant risk of exposure to toxins. The integration of chromatography coupled with mass spectrometry is often regarded as the optimal analytical approach for identifying contaminants in food products. The hyphenated techniques described here combine chromatography's separation capacity with mass spectrometry's identification power. Where chromatography can separate compounds effectively, while mass spectrometry can determine the nominal or exact mass charge ratio (m/z) of both intact molecules and characteristic fragments. The sensitivity, selectivity, specificity, and speed of chromatography-mass spectrometry make it a highly effective method for analysing food pollutants, particularly for detecting toxins at trace levels (Ahuja et al., 2023).

Through the extensive exploration of the growing body of literature (2019–2024), we found several noteworthy reviews highlighting the application of chromatography coupled with mass spectrometry in the detection of contaminants in food products. However, there exists a gap in the literature with regard to an integration of the recent milestones in the implementation of gas chromatography, and liquid chromatography coupled with variations of mass spectrometry in the detection of both emerging and commonly detected contaminants encompassing plant-, aquatic, and animal-derived food products. The majority of reviews mainly focus on either the sample pretreatment or extraction techniques (Kalogiouri et al., 2024; Xu et al., 2021), one specific food contaminant (Vargas Medina et al., 2021; X. Li et al., 2022), or on only one specific form of chromatography (Nolvachai et al., 2023). Therefore, to supplement the knowledge gap in this sector, we highlight an integrated narrative on the recent trends in the development of chromatography-coupled mass-spectrometry strategies in food contaminant analysis with an emphasis on green analysis, multi-target strategies and the transition into artificial intelligence-based data analysis models. Over the past five years, many research efforts have been focussed on developing newer analytical strategies, especially in the nontargeted or multi-class analysis of emerging contaminants. Additionally, it can be observed that with the shifting trend in data processing and sample extraction techniques, analytical strategies are shifting from multi-contaminant or multi-residue analysis in a single run to multi-matrix detection techniques. This review will provide readers with valuable insights into the latest developments in analytical strategies, specifically in the analysis of emerging contaminants in various food substrates and contribute to improving food safety measures. Additionally, it highlights the potential shift towards more efficient data processing and sample extraction techniques in food contaminant analysis, reflecting the trend towards using multi-matrix detection methods, and consequently providing a possible pathway towards realising multi-matrix strategy development.

2. Plant-derived food products

People are being urged to increase their consumption of plant-derived foods to mitigate the adverse effects of the contemporary food system. Hence, the food industry is developing various novel assortments of plant-derived goods to cater to this growing need (McClements & Grossmann, 2021). According to a recent food system analysis, replacing only 50 % of animal-derived foods with plant-derived alternatives may lead to a 31 % decrease in agricultural greenhouse gas emissions (Gerber et al., 2024). Additionally, this change could fulfil up to 25 % of the global requirement for restoring biodiversity within a span of less than seven years. The plant-based food industry is experiencing significant growth to satisfy the increasing demand from consumers, which has been illustrated by the growing interest in NFPS mostly due to factors such as international trade, evolving consumer predilections, the potential for sustainability advantages, and advancements in climate-resilient food production systems (Tan et al., 2024). Nevertheless, NFPS can present novel obstacles for food safety organisations and the food industry. Most of the food safety problems associated with novel foods have been found in conventional foods. Nevertheless, new foods may present distinct food safety obstacles, thus unintentionally contributing to the generation of emerging contaminants with potentially fatal adverse effects. Consequently, traditional contaminant analytical approaches may prove inadequate in this new era.

2.1. Illicit alteration in herbal supplements

Consumers may be attracted to herbal nutritional supplements because they are natural remedies, devoid of any negative consequences (Steyn et al., 2018). Nonetheless, recent analytical data indicates that such preparations might present a risk of undesirable consequences due to concealed illicit additives. Chromatography-coupled mass spectrometry is a highly effective analytical technique used to detect and identify illegal chemicals that have been unlawfully added to food or health supplements especially in plant-derived matrices (Armenian et al., 2018; Sheng et al., 2024). Obtaining precise m/z values of the precursor ions and their fragments enables easy elucidation of the sought-after compounds by examining preestablished databases (Lieng et al., 2023). Nevertheless, identifying query compounds becomes difficult when they are not included in any database. Screening drug analogues, which undergo constant modification to evade the screening procedure while minimally affecting the supplement's pharmacological action, is difficult for food and drug regulating organisations globally. The strategy of library matching is commonly employed and valuable for identifying unknown drugs and derivatives related to erectile dysfunction (ED) for example. This method is however, hindered by the “coverage problem” caused by the lack of a query compound in the database, which hinders its accurate identification (Wei et al., 2019). To tackle this problem, several novel strategies have been proposed. Linear discrimination analysis (LDA) models have been developed to classify compounds based on their mass spectral properties and categorise them according to their structural classes. Consequently, the classification of new psychoactive chemicals, such as synthetic phenethylamines and tryptamines, was done using this model. The model was created using the gas chromatography–mass spectrometry (GC–MS) data of a training set of standard analogues and the categorization rate was determined to be more than 90 %. Further research demonstrated that the combination of principle component analysis (PCA) and LDA effectively distinguished between positional isomers of fluoromethcathinone and fluorofentanyl using GC–MS spectra (Bonetti et al., 2023; Kranenburg et al., 2020).

To address the issue of insufficient coverage, a significant breakthrough has been made in the form of a standalone artificial intelligence screener for illicit drugs and analogues (AI-SIDA) (Jang et al., 2019). This advanced technology has proven to be highly effective in identifying unknown erectile dysfunction drugs and similar substances. AI-SIDA is composed of three distinct layers: the first layer is referred to as liquid chromatography tandem mass spectrometry (LC-MS/MS) viewer (Fig. 1. a), the second layer is an AI classifier (Fig. 1. b), and the third layer is called Identifier (Fig. 1. c). The second layer of the AI classifier utilises an artificial neural network (ANN) model designed explicitly for classifying erectile dysfunction (ED) medicines and analogues. This layer successfully categorises the query MS/MS spectra into four distinct groups: three classes of ED drugs/derivatives and a separate category containing non-ED drugs. However, it is important to exercise caution when applying this AI classifier in the analysis of real samples. During an initial examination of the dataset obtained from several researchers, it was seen that the categorization rate for successful ED drugs/derivatives was not 100 %, despite being exceptionally high. The reason for this is that the ANN model utilized in their investigation was not fine-tuned to the specific mass spectrometric parameters employed in various facilities. Thus, it is strongly advised to construct the ANN classification model based on a particular dataset, which may even include data obtained at various collision energies. The third layer of identification comprised three search engines: Pick-count scoring (PCS), simple similarity search (SSS), and hybrid similarity search (HSS). Specifically, HSS demonstrated the ability to detect the subject compounds accurately, even when their LC − MS/MS spectra were not reported (Cooper et al., 2019). Recent demonstrations have shown that HSS is an effective search tool for identifying substances that are not included in the database. Additional research on the use of HSS for various emerging contaminants is anticipated in the next years, and recent reports have been made in non-targeted tandem mass spectrometry, enabling the monitoring of shifts in the chemical composition of organic materials in coastal saltwater (Eysseric et al., 2021; Petras et al., 2021). This presents a particularly exciting opportunity as there is a literature gap regarding food contaminant identification in plant-derived matrices.

Fig. 1.

Illustration of the three layers of the Graphical User Interface (a) LC-MS/MS viewer (b) an AI classifier (c) Identifier. Adapted with permission from (Jang et al., 2019), Copyright © 2019, American Chemical Society.

The following innovative approaches have additionally been reported for the detection of compounds that are not present in the library or emerging contaminants. Two approaches, extracted common ion chromatograms (ECICs) and precursor ion scanning (PIS), were suggested for the rapid screening of both known and unknown compounds and illegal adulterants. This is based on the fact that common product ions generated from adulterants have a shared skeletal structure. Based on this premise, specific common ions, in combination with other chemometric approaches, were utilized to identify unfamiliar designer structures (L. Jian et al., 2021;Ki et al., 2019; Lee et al., 2020). However, in the procedures mentioned above, common ions were manually isolated based on the fragmentation pattern. As a result, certain classifications of compounds had a restricted number of common ions or neutral loss (Ki et al., 2019; Lee et al., 2020). This could conceivably elevate the incidence of false positives in detecting illicit adulterants that are not present in a reference database or are unanticipated. Classification fragment ion list characteristic (CFILC) was proposed as the solution to this challenge (Xue et al., 2021). CFILC can automatically and fully extract the set of characteristic ions directly from the MS/MS dataset, eliminating the need for manual extraction. Furthermore, the K-nearest neighbour (KNN) method was utilized to establish the decision boundary, which is the threshold for the similarity score. This was done to effectively manage false positives and false negatives. The preceding approaches are limited to screening a certain class of compounds, as they rely on common product ions with comparable structures. Therefore, a molecular networking method has been developed to detect unknown designer drugs by assessing the resemblance between MS/MS spectra. The accuracy of the algorithm depends on the representation of structural similarity between compounds (M. Wang et al., 2016; Yu et al., 2019). While this method excels in identifying compounds with a single chemical moiety, it cannot align spectra from compounds with multiple local modifications. Recent algorithms like MS2DeepScore (Huber, Van Der Burg, et al., 2021), Spec2Vec (Huber, Ridder, et al., 2021), and entropy similarity (Y. Li et al., 2021) have been developed to predict structural similarity between compounds. MS2DeepScore uses a Siamese neural network to predict structural similarity from MS/MS data. Spec2Vec, inspired by Word-2Vec, converts product ions to words to learn relationships between fragments. Spectral entropy was further proposed, as a measure for MS/MS spectral information content and entropy similarity by subtracting entropy from the mixed spectra. Based on these conclusions, an analysis was conducted to assess the performance of different spectral similarity algorithms and develop a molecular networking system for identifying chemicals not in the library. This analysis focused on rapidly detecting and identifying illicit substances added to dietary and herbal supplements.(Sheng et al., 2023). Four advanced spectral similarity algorithms were evaluated: modified cosine (shifted peak match), MS2DeepScore, Spec2Vec, and entropy similarity. Glucocorticoids were chosen as the potential illicit additive for comparison and application due to their widespread illegal use in the Chinese market. Spec2Vec was selected as the spectral similarity method for molecular networking after carefully evaluating its detection ability and false-positive rate. Spec2Vec's ability to detect unanticipated adulteration in several matrices and real-world samples was assessed. Spec2Vec spectral similarity-based molecular networking was found to be highly effective and competitive in detecting illegal adulteration. It also improved the ability to analyse large amounts of MS data. These reports concisely and explicitly demonstrate the significant potential of integrating machine or deep learning, molecular networking, and chromatography-based mass spectrometry in analysing both conventional and emerging contaminants. This opens up an avenue for prospective research where a single developed method can potentially have multi-matrix application. Table 1. Depicts an overview of the data analysis models for the analysis of mass spectra with a focus on artificial intelligence.

Table 1.

Data Analysis Models for Analysis Mass Spectra with a focus on Artificial intelligence.

Matrix	Extraction	Chromatography-Mass Spectrometry	Data Analysis Models (AI Model or Algorithm)	Application	Reference
Dietary Supplements	HLB-SPE, WAX-SPE, QuChERS, and pH-controlled LLE	LC-Q/TOF-MS	ECICs and NSLs	35 Sulfonamides	(Ki et al., 2019)
Men's Supplements	Methanol desolvation	LC-MS/MS	AI-SIDA	Identification of illicit Adulterants	(Jang et al., 2019)
Potentially Multi-matrix	Online	100,000 mass spectra of about 15,000 unique known compounds	MS2DeepScore	Retrieving chemically related compound pairs from large mass spectral datasets	(Huber et al., 2021)
Water	SPE	Q Orbitrap LC-MS/MS	Metfrag SPS GNPS	Identified 253 pharmaceutical and consumer additive contaminants	(Eysseric et al., 2021)
Sea Water	SPE	LC-MS/MS	Molecular Networking	Elucidation of Organic matter chemotype shifts and annotation of unknown compounds	(Petras et al., 2021)
Strawberries	QuEChERs	UHPLC-Q-E-MS	XGBoost, Random Forest, lightGBM Keras	Retention time prediction model was developed using 398 pesticides.	(Feng et al., 2021)
Herbal teas	QuEChERs	UHPLC-Q-E-HESI-MS	XGBoost, LightGBM, Keras, and Random Forest	Identified 122 samples containing pesticides	(Feng et al., 2022)
Potentially Multi-matrix	-	GC-EI-MS spectra, containing 306,622 GC-EI-MS spectra of 267,376 compounds	CSI:IOKR	Elucidation of silylated derivatives from mass spectra	(Ljoncheva et al., 2022)
Multi-Matrix	Online	GC–MS Spectra	PCA and LDA Chemometric Models	Differentiation of NPS positional isomers	(Bonetti et al., 2023; Kranenburg et al., 2020)
Herbal Dietary Supplements	Methanol desolvation	LC-HRMS	Molecular Networking MS2DeepScore Spec2Vec Entropy Similarity	Identification of contaminants not included in the library	(Sheng et al., 2023)
Wheat	Modified QuEChERS	GC-EI-MS/MS LC-ESI-MS/MS LC-ESI-HRMS UPLC-ESI/Qtrap-MS	PCA PLS-DA	Study the effects of pesticides on edible plants	(Pszczolińska et al., 2023)
Pork and Aquatic Matrices	Direct solvent with bead-beating disruption	LC-HRMS	ALHazards Finder	Identified 32 classes of pesticides, veterinary drugs and mycotoxins	(Chen et al., 2024)
Vials of Mixed and Single Standard Samples	-	LC-HRMS using 1051 CEC standards	Random Forest, XGBoost, Support Vector Regression, and Artificial Neural Network	Predicting CEC retention times with 579 pesticides, 415 pharmaceuticals, 28 industrial materials, 16 personal care products, 8 natural products, 3 food additives, and 2 other compounds.	(Song et al., 2024)

2.2. Pesticide residues

Pesticides have been widely employed in agricultural fields to safeguard crops from pests, insects, weeds, animals, and plant pathogens (Y. Huang & Li, 2024). Pesticides are categorised based on their target, which determines their use on a global scale. These groups encompass herbicides, insecticides, fungicides, rodenticides, molluscicides, nematicides, and plant growth regulators (Mahmood et al., 2016). Projections indicate that global pesticide usage in agriculture will experience a rise in the coming years, growing from around 4.3 million metric tonnes in 2023 to approximately 4.41 million metric tonnes in 2027 (Global agricultural use of pesticides 2023–2027, 2024). In 2021, the Americas accounted for over 50 % of the global pesticide utilization, representing the region with the highest usage of agricultural pesticides. The widespread use of pesticides denotes the inevitable possibility of their residues being found in numerous food commodities, which poses a significant threat to human health (Wahab et al., 2022). Prior research has demonstrated that exposure to pesticides frequently leads to both short- and long-term damage to human health including neurotoxicity, carcinogenicity as well as disruptions in the proper functioning of lipids, proteins, and carbohydrates (Hu et al., 2015; Rakitsky et al., 2000). Consequently, numerous governments and international organisations have established maximum residue levels (MRLs) for pesticides in various food products to protect consumers' health and rights. The occurrence of food safety events has, therefore, highlighted the necessity of implementing effective strategies that encompass a wide range of pesticide residues (Vargas-Pérez et al., 2020). This is crucial to ensure that all food products are subjected to standardised measures in order to mitigate any potential risks to human health.

Gas chromatography-tandem mass spectrometry (GC–MS/MS) has been extensively used for detecting various pesticide residues in recent years because of its great sensitivity and capacity to identify and quantify pesticides at very low concentrations (Ahammed Shabeer et al., 2018; Banerjee et al., 2012; X. Zhang et al., 2019). Various sample pretreatment techniques have been suggested to avoid contamination of the injection port and chromatographic column during GC analysis when determining pesticide residues in plant-derived matrices. This contamination produced by high boiling matrix compounds in the samples can impact chromatographic separation, the lifespan of the column, and the chromatographic signals.

Recent studies have extensively applied the Quick, Easy, Cheap, Effective, Rugged, and Safe (QuEChERS) method, a sample preparation technique renowned for its precision, high recovery rate, and suitable matrix effect. It has been widely utilized to detect a multitude of contaminants in food matrices. Its primary parameters can be easily modified and improved to increase the efficiency of extraction and broaden the range of fields and matrices in which it may be applied. It is considered to be reliable for extracting a wide array of analytes from diverse chemical families and matrices (Perestrelo et al., 2019). Pretreatment of the sample prior to QuEChERS extraction is an important step, especially for solid materials like food matrices, and it should be simplified as much as feasible to allow for effective precipitation of the target analytes. As a result, mechanical procedures that aid in sample homogenization, such as grinding, microwave (Pan et al., 2003), or ultrasound, may be very beneficial. The clean-up phase, which can be skipped to simplify the QuEChERS technique without sacrificing analytical performance, is typically critical for improving recovery and minimizing the matrix effect. Finally, using two sequential clean-up processes in the QuEChERS approach can enhance the acquired results.

By integrating simplified QuEChERS extraction procedures with gas chromatography-tandem mass spectrometry (GC–MS/MS), a rapid method was devised in a recent study to identify 201 pesticides in various medicinal herbs (Fu et al., 2019). The method was verified following European Union regulatory criteria, with a limit of quantification (LOQ) of ≤10 ng mL⁻¹ and a recovery rate ranging from 70.0 % to 120.0 % for the majority of pesticides. The optimized technique was applied for evaluating 60 batches of medicinal plants from local markets in China, demonstrating pesticide contamination in 80.0 % of the samples. Pesticide residues were more concentrated in Citri reticulatae pericarpium, Crataegi fructus, and Lonicerae japonicae flos. The investigation found a significant presence of chlorpyrifos in samples of Citri reticulatae pericarpium, Crataegi fructus, and Cuscutae semen. It is worth pointing out that pesticide levels were found to be lower in root medicine and seed medicinal materials, such as Cuscutae and Ziziphi spinosae semen, than in flower or fruit medicinal materials. Previous literature has reported that contamination in medicinal plants can also come by absorbing pesticide residues in the soil through the root systems (H. Liu et al., 2014). This hypothesis may however need prospective investigation to confirm the amount of pesticide residues originating from root translocation and those from pesticide spraying as Fu et al. reported small amounts of pesticides detected in roots relative to flower or fruit medicinal materials (Fu et al., 2019). Future research could explore the potential health hazards linked to the identified substances or their effects on human intake. Fu et al. examined just a restricted selection of therapeutic herbs, consisting of 60 samples, and did not encompass a broad variety of herbal species (Fu et al., 2019). This could restrict the applicability of the results to other types of therapeutic herbs. The study lacked details on the precise cultivation and harvesting methods of the therapeutic herbs examined, which could impact the presence and concentrations of pesticide residues; additionally opened an avenue for further investigation into the lasting impacts of pesticide residues on the quality and safety of therapeutic plants.

A recent milestone in food contaminant analysis developed a novel technique utilising a modified QuEChERS technique coupled with HPLC-MS/MS to identify 102 pesticides in green tea, a complex matrix that may impact the chromatographic separation by co-extraction of compounds including but not limited to tea polyphenols, fatty acids, tea caffeine, and pigments (Y. Huang et al., 2019). This was the first study to evaluate pesticide residues in green teas from seven primary production regions in Jiangxi province, China. 67 % of the green tea samples were found to have pesticide residues, with a significant number carrying over five distinct pesticides. Acetamiprid exhibited the highest detection frequency among the pesticide residues identified in these samples. Certain tea samples had pesticide residual levels that exceeded the regulatory limits established by the European Union. The study demonstrated matrix effects, in the form of ion suppression, even after incorporating a clean-up procedure in the analysis. To account for matrix effects, external matrix-matched calibration curves were used for precise quantification. After modification, the approach demonstrated satisfactory specificity, accuracy, and precision. The developed method in this study showed promising potential for application in the multiclass detection of residues in other food matrices and, therefore, warrants further exploration. The developed method further lays a foundation for developing green analytical strategies that minimize the use of organic reagents.

To further illustrate the wide applicability of QuChERS in various matrices, a rapid, effective, and safe method was developed and validated for multi-residue determination of pesticides in sweet green pepper based on the optimized citrate (citrate salts used in extraction) version of QuEChERS for sample preparation and LC–MS/MS with electrospray ionization and triple quadrupole detection in the multiple reaction monitoring (MRM) mode (Da Costa Morais et al., 2018). Adding graphitized carbon to the clean-up stage enhanced the method's effectiveness. The approach demonstrated high accuracy and precision, with pesticide recoveries ranging from 70 % to 120 % and variation coefficients below 20 %. Acetonitrile was selected as the optimal solvent for extracting pesticides because of its equilibrium in recovery rates and negligible matrix effects. The researchers opted not to utilise dry ice in the partition process due to its negative impact on retrieving specific herbicides. The approach was selective, allowing it to distinguish between pesticides and other compounds present in the peppers. The pesticide acephate was present at a concentration of 1237 μg kg⁻¹, exceeding MRL. However, the study did not provide specific details on the number of pesticides found and reported that most sweet pepper samples had pesticide levels below the limits of quantification or below the MRL. This method can be considered environmentally friendly as it consumes fewer reagents, consequently generating less waste compared to other traditional methods and is therefore in line with the upcoming trends of developing green analytical techniques. However, the technique might not be immediately transferable to other matrices because it was explicitly developed for sweet green peppers.

3. Aquatic-derived food products

Providing food for an anticipated global population of 9 billion by 2050 is a formidable task that involves millions of farmers, food processors, dealers, researchers, technical specialists, and leaders worldwide (Mair et al., 2023). Aquaculture-derived fish and other aquatic products have the potential to significantly contribute to fulfilling the nutritional requirements of all individuals while also addressing the food security concerns of the most impoverished populations. Aquaculture, which involves the breeding, growing, and harvesting of animals and plants in various water habitats, is considered a very effective protein production method. It has significantly enhanced nutrition and food security in numerous regions across the globe. Aquaculture currently provides over 50 % of the total seafood produced for human consumption worldwide, and this proportion is expected to increase further (Aquaculture | NOAA Fisheries. (n.d.-a), 2024). The aquaculture industry has experienced tremendous growth due to market forces but has subsequently faced challenges like overcrowding, illnesses, and high mortality rates on a global scale (FAO, 2022). China incurs an annual loss of 100–200 billion Yuan (equivalent to around 15–30 billion USD) due to disease outbreaks in aquaculture. Bacterial infections account for over half of these diseases (X. Liu et al., 2017). Due to the complexity of aquatic ecosystems spanning across land and the increasing global urbanisation and industrialisation, pollution has emerged as a pressing concern for aquaculture development. The primary way organisms gather and store pollutants is from the water they live in (L. Huang et al., 2023), thus necessitating the development of environmentally friendly, efficient and rapid analytical techniques for the detection of contaminants in aquatic-derived food products. Multiple studies have indicated that liquid chromatography coupled with mass spectrometric detection is a preferred method for analysing contaminants or antimicrobial residues in fish. Since 2016, there has been a trend in analytical strategy towards using multi-residue and multiclass methods. These methods are time-saving and were anticipated to be the future trend in this sector (Santos & Ramos, 2016). The extraction procedure represents the main constraint in any multi-residual approach since it must ensure satisfactory recovery of all analytes with diverse physicochemical properties. Consequently, this phase has been identified as requiring further extensive investigation.

3.1. Ciguatera fish poisoning

Ciguatera fish poisoning (CFP) caused by ciguatoxins (CTXs) is a significant food-borne illness which has been reported to impact an estimated 50,000 people annually in tropical and subtropical regions. Fish are exposed to CTXs through grazing and predation (Soliño & Costa, 2020). These powerful neurotoxins are produced by the benthic microalgae Fukuyoa spp. and Gambierdiscus spp. The dynamics and transfer of toxins in fish and food webs are complicated, and the causes of CTXs are still largely unknown. CTX, can cross the blood-brain barrier and cause central nervous system involvement in human clinical cases. CFP causes a variety of symptoms, including gastrointestinal, neurological, neuropsychological, and cardiovascular symptoms, which can last for days, weeks, or months (Friedman et al., 2017). Fish recruitment and prevalence can be significantly impacted by CTXs, which may have an adverse effect on fisheries and fish populations. The identification of toxic fish and CFP incidences in non-endemic places, including Europe, have been caused by recent appearances of the Gambierdiscus spp. in temperate areas. As a response, monitoring systems and fishing limits have been put in place. There are dangers to fisheries and food safety as more than 400 fish species - many of which are high-value commercial species can carry the CTX virus. This, therefore, highlights the desperate need to ensure food safety and reduce the impact on fisheries, which requires efficient monitoring programs, prediction models for toxin occurrence, and risk area identification.

A milestone in the analysis of emerging toxins implemented liquid chromatography coupled with tandem mass spectroscopy (LC-MS/MS) for the analysis of ciguatera fish poisoning in contaminated fish samples from the Atlantic Coasts. This approach allowed for the confirmation and characterization of CFP toxins, particularly Caribbean Ciguatoxin-1 (C-CTX1), as the main toxin in the analysed fish samples (Estevez et al., 2019). LC-HRMS further established that C-CTX1 is the primary toxin causing ciguatera toxicity in fish from the European Atlantic coast (Estevez et al., 2020). It has widely been reported that the lipophilic nature of ciguatoxins (CTXs) makes it challenging to separate them from high levels of lipids and fatty acids in complex fish matrices, compromising the efficiency of CTX analyte recovery and chromatographic resolution. Thus, efforts were particularly focused on optimizing sample pretreatment, specifically in the sample clean-up step, using normal and reversed-phase solid-phase extraction (SPE) mechanisms to efficiently remove matrix interference and enhance detection. The study confirmed the presence of the new hydroxyl metabolite C-CTX1 in one of the analysed samples, indicating the potential for identifying novel CFP toxins (Estevez et al., 2019). The authors identified the need for the development of certified reference materials for CFP toxins to improve the validation and reliability of LC-MS/MS methods for CFP analysis. Owing to the limited number of pure standards and reference materials available, validation of the optimized method was restricted. The number of recovery tests conducted on spiked samples was further limited due to the scarcity of reference toxins, which could impact the method's accuracy. Although HRMS may lack sensitivity when compared to the MS/MS methods, it has proven to be very useful in the absence of pure standards, which was reflected in subsequent research efforts.

A more recent milestone was chalked by Mudge et al., who identified Gambierdiscus silvae and G. caribeaus as sources of ciguatera poisoning in the Caribbean, with CTX-like activity (Mudge et al., 2023) since C-CTXs had not been identified from their presumed algal source in earlier studies. Due to the absence of pure standards, LC-HRMS was used for CTX confirmation. The study detected C-CTX5, a novel C-CTX analogue and confirmed its structure and its abundance in G. silvae, which is about 20 times higher than C-CTX1/2. The lack of reference standards, thus, made all data on C-CTXs qualitative, limiting the ability to provide quantitative analysis. Research demonstrated that the Caribbean isolates of G. silvae and G. caribaeus produce C-CTX5, a chemical that is a precursor to C-CTX1/2 in fish and is believed to be further converted to C-CTX3/4. Consequently, C-CTX5 is an early form of the CTX congeners that are prevalent in hazardous fish that induce CP in the Caribbean. In a NaV-specific in vitro experiment using mouse neuroblastoma cells, the novel Gambierdiscus toxin C-CTX5 generated a CTX-like reaction, confirming a toxicological effect of the same magnitude as C-CTX1/2. Compared to the two G. caribaeus strains, G. silvae 1602 SH-6 exhibited significantly higher NaV bioactivity (as assessed by N2a–MTT) and C-CTX5 content (as determined by LC− HRMS), indicating that G. silvae is also a prominent generator of C-CTX in Caribbean reef ecosystems.

However, the study did not provide detailed information on the specific methods used for chemical and enzymatic conversions of C-CTX5 into known fish metabolites associated with CFP and did not discuss the potential interactions between C-CTX5 with other present compounds in the environment, which may necessitate further research. Estevez et al. first emphasized the necessity for the future production of quantitative reference materials four years earlier as this prevented them from validating their developed analytical procedures and was further echoed by Mudge et al., with a view towards enabling assay development, toxicokinetic studies, and the development of diagnostic tests for C-CTXs (Estevez et al., 2019) This indicates slow progress in this field and a major drawback that requires urgent attention. Mudge et al., further suggested the use of isotopic labels, like oxygen-18, as tracers in C-CTXs to enhance future analytical and biochemical studies (Mudge et al., 2023). Moreover, the study highlighted the potential of using algal sources to produce analytical reference materials for C-CTXs, eliminating the need for large quantities of naturally incurred fish tissue which can hardly be sustained, providing critical and exciting avenues for future research that may be pivotal for analytical strategy development.

3.2. Antibiotic/veterinary drug residues

Aquaculture is in high demand for aquatic-derived food products, leading to the need for intensive practices. However, antibiotics are necessary to treat illnesses and mortality in order to maintain a consistent market supply and economic advantages. Antibiotics can inhibit bacterial growth or directly eliminate bacteria, effectively treating or preventing bacterial-induced disorders (Oliver et al., 2020). Nevertheless, the presence of antibiotic residues in animal-based food products, which humans ultimately consume, can result in antibiotic resistance, renal toxicity, allergic reactions, hearing impairment, arrhythmias, and the emergence of superbugs that are resistant to commonly used antibiotics. Consequently, this poses a significant risk to human health, potentially endangering lives. Therefore, it is crucial to evaluate the presence of antibiotic residues in animal-derived foods prior to their consumption (Y. Liu et al., 2022). Due to the significant harm caused by antibiotics in animal-based food products, numerous countries have implemented stringent MRLs for various aquatic-derived foods. Monitoring and detecting antibiotic residues in these food items poses a significant difficulty in the aquaculture and food businesses. Researchers have investigated different analytical methods to achieve precise, highly sensitive, and rapid detection of antibiotic residues over the recent few years.

Enrofloxacin (ENR), a fluoroquinolone antibiotic, inhibits bacterial DNA gyrase. Because of its broad antibacterial range, enrofloxacin has been approved for use in cattle, swine, poultry, and aquaculture. Enrofloxacin is a last-line treatment in aquaculture that effectively cures bacterial infections caused by F. columnare (Y. Jia et al., 2021), A. hydrophila (Yang et al., 2018), and A. sobria (Uney et al., 2021). In a recent study, ENR metabolites in actual fish samples were screened using ultrahigh-performance liquid chromatography coupled with Q-Orbitrap mass spectrometry (UHPLC-Q-Orbitrap MS) and Compound Discoverer software. Notably, another metabolite, deethylene-ENR, in addition to ciprofloxacin (CIP), was discovered and identified for the first time. Modifying the gradient elution's initial mobile phase ratio improved chromatographic conditions to enhance chromatographic separation. By varying the collision energy, mass spectrometry was tuned for fragmentation degree. Optimizing the settings for mass spectrometry enabled the subsequent structural derivation of ENR metabolites (Dai et al., 2023). The technique demonstrated excellent linearity, satisfactory recovery from fish samples, and good LOD and LOQ values. The five replicates of the ENR and CIP recoveries had acceptable RSDs, proving the accuracy and repeatability of the suggested procedure, indicating its reliability and precision for the determination of ENR and its metabolites in aquatic-derived products. Compound Discoverer software was used to identify unknown metabolites of ENR utilising the raw data obtained from ENR-positive and negative fish samples from UHPLC-Q-Orbitrap MS. According to the data, ENR was primarily metabolized to M1-ENR and CIP. The N-deethylation of ENR produced CIP, one of the recognized metabolites of ENR. The novel method was subsequently applied to fifteen fish samples gathered from local markets to identify ENR and its metabolites (CIP and deethylene-ENR). Owing to the lack of an official standard material for deethylene-ENR, semi-quantitative analysis was carried out using the ENR calibration curve, and the concentration of deethylene-ENR was determined using the ratio of the peak areas for ENR-D5 and deethylene-ENR.

The study found that 14 samples had ENR, 12 had CIP, and 12 had deethylene-ENR. 12 of the 14 ENR-positive samples had ENR metabolites, CIP, and deethylene-ENR, with only two samples lacking them. The level of CIP was generally higher than 2 % of ENR, although it was typically lower than ENR (Shan et al., 2018; W. Zhang et al., 2021). The study also found that 98 % of ENR was present in aquatic animals as the parent drug. The remaining metabolites were kept at a specific level, particularly when ENR was low. The residue levels of ENR residue markers are the sum of ENR and its main metabolite CIP, with the maximal residue limit (MRL) being 100 μg kg^{− 1} in fish. Three samples analysed exceeded the MRLs, indicating a somewhat frequent detection rate of ENR, implying its probable abuse in aquaculture (Guidi et al., 2018). Despite the significant contributions made by Dai et al., certain aspects warrant further investigation and present future research opportunities. The study did not investigate the possible toxicological implications of these metabolites on human health or the environment; instead, it concentrated on the screening and identification of ENR metabolites in fish samples. Only 14 fish samples tested positive for ENR, which may not accurately represent the total prevalence and distribution of ENR and its metabolites in aquatic goods. The possible build-up of ENR and its metabolites in various fish tissues as well as the possible movement of these residues up the food chain were not examined in this work. Additionally, the performance of the UHPLC-Q-Orbitrap MS approach was not compared to other analytical techniques that are frequently used to analyse fish samples for ENR and its metabolites. The possible breakdown or transformation products of ENR in aquatic ecosystems, which might offer additional insight into the fate and behaviour of this antibiotic in the environment, warrants further investigation.

In recent years, there has been an increasing number of publications on the multi-residue analysis of contaminants in aquatic-derived food products owing to multifaceted sources of contamination from both biotic and abiotic factors. Triple-quadrupole tandem mass spectrometry (QqQ-MS/MS) is the most widely used method for multi-residue analysis. QqQ-MS/MS provides superior quantitative performance for over 100 regulatory analytes in food matrices through enhanced acquisition methods and reaction monitoring modes (Cao et al., 2018). However, QqQ-MS/MS is unsuitable for high-throughput, non-targeted screening as it lacks scan speed and sensitivity in full-scan mode. HRMS is ideal for non-targeted food safety testing due to its high mass accuracy, resolution, scan speed, and sensitivity in full-scan mode. This has led to the development of rapid and wide-scope multi-residue screening methods like LC/Q-TOF-HRMS screening for over 600 multi-class compounds (Pérez-Ortega et al., 2017). Based on this prelude, a new optimized high-throughput screening technique for 756 multiclass chemical pollutants in aquaculture products using modified QuEChERS extraction coupled with LC/Q-TOF-HRMS was reported (Bai et al., 2022). This consequently led to establishing a mega-database with retention time/accurate mass data for 524 pesticides, 182 veterinary medications, 32 persistent organic pollutants, and 18 marine toxins allowing chemical identification using retrospective library searching. This database is superior to previous studies on HRMS-based multi-residue screening approaches for aquaculture products, which only addressed a limited number of contaminants (Turnipseed et al., 2019). Validation of the approach in four representative matrices (tilapia and grouper muscle tissues, as well as edible sections of oyster and scallop) demonstrated satisfactory recovery and reproducibility, with screening detection limits and quantification limits of less than 0.01 mg/kg for more than 90 % of the chemicals. The suggested method's applicability to high-throughput screening settings was proven by analysing 64 real-life samples from aquaculture farms and retail markets. It was demonstrated that fish muscle samples exhibited a stronger matrix effect than shellfish samples. It is important, however, to analyse the applicability of this method in other matrices to determine whether its performance is matrix-oriented. It would beneficial in prospective studies to compare the performance of this method with other existing or newly developed high throughput screening methods. Limited information was reported in this study on the possible rate of false negatives, and nothing was reported on the possible rate of false positives of the screening method.

Multiple techniques utilising self-constructed or commercially available databases have been developed to efficiently analyse large quantities of chemical pollutants (Bai et al., 2022). The databases typically contain multidimensional data on chemical pollutants, encompassing retention time, MS1, and MS2. Nevertheless, the data available in most databases about chemical pollutants is inadequate, particularly regarding MS2 spectra (Feng et al., 2022). Another approach is to evaluate probable chemical pollutants by analysing particular fragments of compounds (Liang et al., 2022) that are characteristic of their structure. Characteristic fragments can screen and identify unexpected or unknown chemical contaminants. Summarizing distinctive fragments is, however, a technique that requires an enormous amount of time and effort, thus making screening and recognition of unknown chemical contaminants in food a formidable challenge. As previously discussed under plant-derived matrices, several innovative techniques have been proposed for the artificial intelligence-based MS2 spectra classification of certain chemicals (Jang et al., 2019). Additionally, some studies have reported on multiple applications of supervised machine learning classification models (Koshute et al., 2022) to categorise MS2 spectra into fentanyl analogues and compounds that are not fentanyl. An unsupervised hierarchical clustering method was found to accurately group phosphodiesterase type 5 inhibitors found in dietary supplements (Tachi et al., 2022). These approaches were primarily founded on the notion that substances with analogous chemical structures exhibited comparable fragmentation patterns. A novel multi-matrix method for screening and identifying unknown chemical pollutants using LC-HRMS and machine learning (AIHazardsFinder) was recently reported (Chen et al., 2024). The primary objective was to develop an MS2 spectra classification model for predicting the spectrum classes of unidentified features. Initially, a training dataset was produced, which included MS2 spectra of several categories of chemical pollutants. Using the dataset, four classification models were tested using different algorithms, and the one with the highest performance was chosen, as illustrated in Fig. 2. In addition, the classification model's incidence of false positives was assessed using non-chemical pollutants. Subsequently, the screening technique was formulated utilising the classification model. The unidentified characteristics in LC-HRMS were forecast using the classification model and subsequently recognized by several databases, fragment analysis, and structural inference. The implemented technique was tested on 10 pork and 50 aquatic matrices to demonstrate its efficacy. The analysis revealed the presence of 19 chemical pollutants from 8 different classes in the actual samples. Notably, 8 of these contaminants were not previously included in their database. Table 2. illustrates an overview of the chromatographic-mass spectral trends in the analysis of food contaminants.

Fig. 2.

Overview of AIHazardsFinder workflow; a machine learning-based model for screening and identifying unknown chemical pollutants adapted with permission from (Chen et al., 2024), Copyright © 2024, Elsevier.

Table 2.

Chromatographic-Mass Spectral Trends in the Analysis of Food Contaminants.

Matrix	Extraction	Chromatography-Mass Spectrometry	Contaminants Identified	Method's Merits	Method's Demerits	Reference
Bovine muscle, Porcine muscle, Bovine liver, Bovine kidney, and Chicken liver	Solid-liquid extraction followed by EMR-Lipid cartridge clean-up	LC-MS/MS	Multi-class, multi-residue analysis of veterinary drugs including antibiotics, growth promoters, anti-parasitic/inflammatory agents	Wide applicability of method to diverse matrices Efficient matrix clean	Variable recovery rates in specific meat matrices	(Zhao et al., 2018)
Honey	PVA-coated Fe3O4	HILIC-MS/MS	AAs including streptomycin and dihydrostreptomycin	Efficient extraction with high recovery rates	Limited selectivity	(D. Li et al., 2018)
Herbs	QuEChERS	GC–MS/MS	201 pesticides including chlorpyrifos	High throughput sample preparation with simplified extraction	Variable recovery	(Fu et al., 2019)
Green Tea	QuEChERS	HPLC-MS/MS	102 pesticides	Green High efficiency	Complex matrix effects	(Huang et al., 2019)
Honey	Refined QuEChERS protocol with utilization of Z-Sep + and PSA	Combination of LC-MS/MS and GC–MS/MS techniques	207 pesticide residues	Comprehensive ability to analyse wide range of pesticide residues with high sensitivity and selectivity	Strong matrix effects for certain compounds	(Gaweł et al., 2019)
Fish samples	Standard MBA method used in Japan for ciguatoxin detection	LC-MS/MS LC-HRMS	C-CTX1 Characterization-main toxin in CFP	A dual analytical approach allows for both quantitation and confirmation of toxins Overcomes reference material resistance	Lack of reference materials Resource intensive & time consuming	(Estevez et al., 2019, Estevez et al., 2020)
Pork Samples	C18 SPE cartridges	HPLC-Q-TOF-MS	141 veterinary drug residues and their metabolites, including Sulfamethazine and its metabolites	Simultaneous detection and quantification Wide scope of analytes Advanced data analysis	Susceptible to matrix interferences	(X. Li et al., 2020)
Milk	QuEChERS technique modified using a melamine sponge	UPLC-MS/MS	Multiple residues of 57 veterinary drugs	Rapid and high matrix purification performance	Minimal interference from the milk matrix, could affect the accuracy of results	(B. Ji et al., 2021)
Milk	Hollow mesoporous molecularly imprinted nanoparticles DSPE Sorbent	LC-MS/MS	Seven MACs in real milk samples including azithromycin, spiromycin	Simplifies the preparation steps of HPMIPs with adsorption capacity and selectivity,	HMMIPDAs involves a complicated and lengthy process	(S. Ji et al., 2021)
Tilapia, Grouper, Oyster, Scallop	(Casey et al., 2021)	LC/Q-TOF-HRMS	756 multiclass pollutants including pesticides, marine toxins, veterinary drugs and persistent organic pollutants	High sensitivity and quantitative accuracy with a wide-Scope Screening capacity	Hampered by matrix effects Some compounds could not meet RT criterion	(Bai et al., 2022)
Dinoflagellate isolates Dictyota spp	Three rounds of freeze–thaw (−20 to 20 °C)	LC-HRMS	C-CTX5 (Novel Analogue) Gambierdiscus silvae and G. caribeaus - sources of CFP in the Caribbean	High sensitivity and precision in comprehensive molecular analysis	Absence of quantitative reference materials deems data qualitative	(Mudge et al., 2023)
Fish samples	Guidelines of the Ministry of Agriculture Announcement No. 1077 of China	UHPLC-Q-Orbitrap MS	Deethylene-ENR (Novel) Ciprofloxacin	Identified novel unknown metabolite	Costly and complex	(Dai et al., 2023)
Beef	QuChERS EMR-lipid and PSA	GC-Q-Orbitrap-MS	191 pesticide residues	Minimized false positives Retrospective Analysis	Possible isobaric interference	(Pang et al., 2023)
Animal-derived Matrices	Physical freeze drying for lipid elimination with TLC	HPLC-DAD/UHPLC-HRMS	Concurrent qualitative and quantitative analysis of 16 developing heterocyclic organic compounds	Simple and solvent-saving Efficient lipid removal	Method not applicable to other types of contaminants	(Y. Li et al., 2023)
Honey Milk Eggs	SPE with laboratory-made adsorbent (SiO2@PVA/HA)	LC-MS/MS	Trace Macrolide Antibiotics	Versatile in matrix applicability Environmentally friendly	May be specific to MACs Complex sample preparation	(Qin et al., 2024)
Honey	QuEChERS	LC-HRMS	694 different pollutants including including plasticizers, flame- Retardants, additives	Comprehensive sample preparation Applicable in Food Safety Monitoring and Authenticity	Limited applicability in food safety due to high-cost barrier	(Makni et al., 2024)

4. Animal-derived food products

In recent times, due to changes in dietary habits, there has been a growing focus on consuming different animal-based products, such as meat, eggs, honey, and milk. This has led to the ongoing growth of animal-based agricultural product production (Q. Jia et al., 2024). Consumption of animal-derived products on a yearly basis has risen in the past few decades, particularly for meat. In 2010, the average annual global consumption per person was 41.6 kilogrammes, whereas in 2020, it increased to 42.8 kilogrammes. Global animal farming has experienced a substantial expansion in response to the growing demand for meat and other animal-derived products. Approximately 25.9 billion chickens were reared in 2021; this represented an increase of nearly 79 % compared to the year 2001 (OECD & Food and Agriculture Organization of the United Nations, 2023). Thus, extensive activities associated with developing and producing animal-derived food products have led to an exponential increase in food contaminants. This section highlights the most recent reports (2019–2024) on the analysis of contaminants in meat, honey, and milk products.

4.1. Meat

Key issues related to meat consumption involve ensuring food safety and upholding high meat quality standards. Birds (chickens, duck, turkey) and mammals (beef, mutton, pork) are the main sources of meat, which can be either raw or processed. Chemical contaminants present in the environment result in a rise in foodborne diseases, particularly in processed foods (Azam et al., 2024). Meat, the main protein source in underdeveloped countries, is highly vulnerable to bacterial infection. Contamination arises at various stages of meat product production, including manufacture, packaging, shipping, and storage. These contaminants can have detrimental effects on the liver, kidneys, and nervous system, and can also induce mild cases of gastroenteritis (Font-i-Furnols, 2023). Chemical risk profiles have identified several stages in the meat supply chain where chemical pollutants can infiltrate, from meat production to retail. Millions of instances of foodborne illnesses worldwide, leading to hundreds of deaths annually, are most likely caused by animal products that have been contaminated; hence, driving the development of efficient and high throughput analytical strategies to ensure food safety.

A recent study on multi-matrix contaminant analysis utilising solid-liquid extraction followed by enhanced matrix removal (EMR) EMR-Lipid cartridge cleanup coupled with LC-MS/MS was developed and validated for the analysis of veterinary drugs in various animal tissues including bovine muscle, porcine muscle, bovine liver, bovine kidney, and chicken liver (Zhao et al., 2018). The method was simple, reliable, and robust, allowing for the detection of multiple classes of veterinary drugs and residues. By employing a pass-through cleanup method in which an extract is loaded onto the cartridge and subsequently eluted by gravity, the EMR-Lipid cartridge could facilitate a pass-through of target analytes for subsequent analysis while implementing a highly selective and efficient interaction with lipids. The extraction process used in meat was meticulously refined to balance the recovery of several drug analytes and the elimination of co-extracted substances from the matrix. The results of the quantitative study demonstrated that more than 90 % of the veterinary drugs tested exhibited adequate recoveries, while over 95 % of the compounds displayed outstanding repeatability across all five meat matrices. The results indicated that the improved solid-liquid extraction method, followed by EMR-lipid cartridge cleanup, offered effective removal of impurities and achieved outstanding recovery and precision results for the analysis of various veterinary medications in meat. However, this method was not applied to real-world sample analysis, warranting further investigations, and the applicability of this method to other types of matrices should be done with intricate caution as certain solvents during the extraction may not be as effective.

In a more recent study, a unique organic-inorganic hybrid material, composed of polyvinyl alcohol and renewable humic acid (HA), was developed for the cost-effective and sustainable monitoring of macrolide antibiotics (including azithromycin, josamycin, roxithromycin, and clarithromycin) in animal-derived matrices (Qin et al., 2024) as is illustrated in Fig. 3. This material offered a low-cost solution with high enrichment efficiency and environmental friendliness. The sample's LOQ for macs was 0.008–0.500 μg kg⁻¹, demonstrating its high sensitivity and efficacy. The material demonstrated excellent reusability, as it could be recycled and utilized for a minimum of seven cycles, hence emphasising its sustainability and potential for long-term use. This study advances the development of bio-based materials used in the detection of animal-origin food, promotes sustainable practices and the recycling of resources in the monitoring of antibiotic residues with multi-matrix application potential.

Fig. 3.

Application of the bio-based organic-inorganic hybrid adsorbent in the sustainable detection of antibiotic residues in animal-derived matrices. (a) Preparation of samples and the SPE method. (b) The flowchart illustrating the process of preparing the sample for the developed method and the Chinese national standard method (Qin et al., 2024).

While low-resolution mass spectrometry (LRMS) exhibits excellent sensitivity and selectivity, it is unable to detect chemicals that are not specified in the multiple reaction monitoring mode. The utilization of high-resolution mass spectrometry in a comprehensive manner has emerged as a viable alternative technique. This approach enables the sensitive and extensive screening of analytes, thereby overcoming the constraints associated with LRMS, allowing the analysis of an unlimited number of compounds simultaneously, and detecting non-preselected and potentially undiscovered substances through “post-targeted” analysis. For a broad spectrum of authorised, prohibited, and unregulated veterinary drugs in pork, an untargeted screening method utilising HPLC-Q-TOF-MS was effectively developed and validated for the quantification of 141 veterinary drug residues and their metabolites, which belonged to eighteen different classes. The given approach of HRMS screening proved very selective due to its reliance on accurate mass measurement of fragment ions generated by CID, along with the retention times and correct masses. MS/MS data were utilized to identify stereoisomers and differentiate interferences (X. Li et al., 2020). To choose a suitable approach for preparing samples for a broad screening, the authors examined the general methodologies used for sample preparation, and the quantification capability of the Q-TOF-MS approach was further demonstrated. The effectiveness of this screening technology, which relies on a precise mass database, was validated by the analysis of actual pork samples. The analysis indicated the existence of sulfamethazine in a single porcine sample. In addition, using a generic sample extraction method and untargeted HRMS scanning allowed for identifying and detecting five metabolites of sulfamethazine in the analysed sample. Furthermore, the database can be expanded by incorporating precise theoretical mass information for novel compounds, eliminating the need for reference substances. This allows for the swift identification of new compounds and metabolites.

Pesticides are exponentially becoming a nuance and are present in several plants and soil, subsequently contaminating plants used for feed production. Consequently, farm animals consume this contaminated feed, leading to the accumulation of pesticides in their tissues (Cámara et al., 2020). Studies have demonstrated long-term exposure to pesticides to have significant health consequences for animals, including neurological, pulmonary, reproductive, and developmental toxicity (El-Nahhal, 2020; Ubaid Ur Rahman et al., 2021). Exposure to pesticides throughout the crucial developmental stages of animals can have enduring effects on both early and later stages of development, particularly on the development of the brain and endocrine system (Utembe, & Kamng'ona, A. W., 2021). Hence, it is crucial to establish high-throughput, extensive, sensitive, and precise multi-residue analysis techniques to thoroughly assess the presence of pesticides in animal-derived food matrices.

In an effort to tackle the increasing worry about pesticide contamination in animal-derived food products, a recent study utilized gas chromatography - quadrupole orbitrap mass spectrometry (GC-Q-Orbitrap-MS) to analyse the presence of 191 pesticide residues in beef (Pang et al., 2023). Due to the large amount of lipids and proteins present in beef muscle samples, the matrix effects can be significant. Additionally, preparing pesticide samples for multi-residue analysis is quite difficult. Therefore, the QuChERS extraction method was used. Typically, research on pesticide residues in animal-derived foods focuses on a limited number of pesticides that have comparable polarity. In such cases, the octanol/water partition coefficient (log Kow) can be utilized to aid in the characterization of these pesticides. However, in specific situations, particularly those involving the detection of multiple residues, the solvent must be able to extract residues of pesticides with different polarity, while preventing the extraction of interfering components from the matrix. When acetonitrile/acetone (1:1, v/v) was used, the pesticides were extracted, and the lipid interference was negligible, therefore, was selected as the extraction solvent. Employing an appropriate salting-out reagent can enhance the recovery of the compounds. The addition of the salting-out reagent stabilizes the pH and reduces co-extraction. The combination of MgSO₄, NaCl, C₆H₅Na₃O₇·2H₂O, C₆H₈Na₂O₈, and NaHCO₃ yielded the most favourable outcomes. The selection of C18 as the dSPE reagent was based on its superior performance compared to EMR-lipid and PSA (Pang et al., 2023). This is attributed to C18's high affinity for sterols, long-chain fat, and other non-polar interference components, making it very appropriate for meat matrices. As a result, the procedure was utilized in situations involving actual samples, allowing for an evaluation of its efficacy and suitability in regular analysis. Thirty beef samples were analysed, and the results indicated that none of the 191 pesticides were discovered.

Emerging contaminants are notorious for their ability to have negative effects on human health, even when present in trace amounts. Their durability, toxicity, bioaccumulation, and potential for long-range transmission have caused significant public concern (Li, Liu, et al., 2023). Halogenated organic contaminants (HOCs) are an example of highly significant and toxic pollutants of emerging concern due to their long-lasting nature and harmful impacts. It is important to note that food products are unavoidably polluted with HOCs throughout their production, use, and disposal (K. Jian et al., 2020). Exposure to HOCs has been linked to a wide range of health issues, such as reproductive, neurological, immunological, endocrine, behavioural, and carcinogenic effects in both wildlife and humans. Moreover, current research suggests that the presence of HOCs is associated with the development of obesity and type 2 diabetes (Kodavanti et al., 2023). The availability of target screening methods is restricted to a select few developing pollutants due to the unavailability of reference standards for these compounds (Badea et al., 2020). However, it is imperative to conduct a risk assessment of these micropollutants in different food matrices. Thus, combining chromatographic techniques with high-resolution mass spectrometry is increasingly becoming essential for screening new halogenated pollutants, both for suspect and non-target purposes. Nevertheless, utilising full-scan mode in quantitative analysis renders HRMS more vulnerable to matrix interference, impacting the sensitivity of quantitative and qualitative research techniques for trace unknowns. Therefore, the prerequisites for high-throughput screening include the efficient extraction and effective purification of materials. A novel approach combining physical freeze drying for lipid elimination with thin-layer chromatography (TLC) employed for sample purification and to reduce matrix effects coupled with HRMS was recently devised and verified for the concurrent qualitative and quantitative analysis of 16 developing heterocyclic organic compounds found in animal-derived food products (Li, Wang, et al., 2023). Following pre-treatment, the actual meat samples were analysed using HPLC-DAD/UHPLC-HRMS. The presence of HOCs was identified in actual food samples obtained from animals. The content of these compounds varied from not detected (ND) to 307.22 ng·g^{− 1} dw, or ND to 87.72 ng·g^{− 1} ww. These compounds included polyhalogenated carbazoles (PHCZs), halogenated phenols, and tetrabromobisphenol-A (TBBPA) analogs. In addition, the viability of employing HRMS for qualitative and semi-quantitative analysis of non-target halogenated pollutants in animal-derived meals was also verified. The most significant advantage of this method was the minimized use of solvents, which is a fundamental step towards developing green analytical strategies. The approach may not achieve complete separation of certain compounds, such as 2-monobromocarbazole and 3-monobromocarbazole, hence complicating their differentiation. It is recommended to optimize the developed method to fully separate these compounds and further develop new techniques for separation of fats.

4.2. Honey

Honey is a natural product formed from the nectar of flowers by honeybees; it is the only naturally occurring product generated from insects and serves applications in various industries such as cosmetics, medicine, and nutrition (Samarghandian et al., 2017). Within the context of our review, honey has been described as an animal-derived food product because the nectar and honey undergo honey bee regurgitation (n.d.), which results in the addition of enzymes. Enzymatic reactions in honey involve the action of the enzymes diastase and invertase, which convert oligosaccharides and disaccharides (sucrose and maltose) to glucose and fructose (Alaerjani et al., 2022). Therefore, as bee products become more widely used in our modern era, a new and urgent global health risk has surfaced: the contamination of honey with heavy metals, pesticides, antibiotics, and microbes. Consuming beekeeping products contaminated with pesticide residues has been connected to several health problems, such as allergic reactions, cellular deterioration, genetic abnormalities, and even possible cancerous impacts (Morariu et al., 2024). Regrettably, there are documented instances of infants contracting botulism after consuming tainted honey. Furthermore, the alarming rise in antibiotic resistance has been linked to the use of antibiotics in beekeeping procedures (Morariu et al., 2024; Pohl et al., 2017). Due to the imperative need to develop analytical techniques for various anthropogenic and naturogenic contaminants in honey, various facile and innovative strategies have been reported.

Given that the combination of LC and GC mass spectrometry allows for a comprehensive analysis while also producing higher levels of sensitivity and selectivity. A refined QuEChERS protocol was designed in a recent imperative study to analyse 207 pesticide residues in honey using LC-MS/MS and GC–MS/MS techniques (Gaweł et al., 2019). This study was the first to report on the utilization of Z-Sep + (a type of silica gel that has been modified with C18 and zirconium dioxide) either by itself or in conjunction with PSA for the purification of honey extracts. Z-Sep + serves as a complement to PSA because these two have distinct clean-up mechanisms. The PSA mechanism primarily relies on ionic interactions that are pH-dependent, whereas Z-Sep + operates through Lewis acid and hydrophobic interactions that are pH-independent. The study demonstrated that the effectiveness of removing matrices is enhanced by incorporating more sorbents (namely, C18 < Florisil < Z-Sep+) into the PSA process. Although the citrate buffered QuEChERS method is commonly preferred for analysing pesticide residues in honey (Shendy et al., 2016), it was found in this study that the weight of co-extractives in the honey matrix was minimized after performing d-SPE clean-up using PSA, Z-Sep+, and MgSO4, regardless of whether citrate, acetate, or non-buffered extraction conditions were used. A total of 155 honey samples collected over a span of three years were examined. Of these, 132 (85 %) of all the samples studied had at least one pesticide residue above the LOQ. It is important to highlight that nearly all pesticides detected in honey samples are permitted for use in the European Union as Plant Protection Products (PPPs) or veterinary medications and their concentrations were generally low. Residues of neonicotinoid pesticides with cyano substitution were detected in 77 % of honey samples, making them the most discovered pesticides. This category of extensively utilized insecticides is recognized for its detrimental impact on non-target pollinators and contributes to the excessive mortality of honeybee colonies (Andrione et al., 2016). Thiacloprid and acetamiprid were found in 68 % and 55 % of the total samples. The concentration of acetamiprid in two honey samples (0.13 mg/kg and 0.12 mg/kg) surpassed the MRL threshold (0.05 mg/kg). Amitraz metabolites, which are veterinary medications, were found in 35 % of the analysed honey samples. The amitraz concentration in one sample was 0.60 mg/kg, which exceeded the MRL when expressed according to the residue definition.

While conventional methods of routine analysis are quite sensitive, one limitation of LC/GC–MS/MS instruments used in MRM is that they can only identify a limited number of substances that have been prioritized and for which analytical standards are already available. Consequently, these methods cannot detect any other substances not specifically mentioned in the list. This is a concern since undisclosed, potentially hazardous compounds have the potential to contaminate food products without warning (Weng et al., 2020). To address this issue, a recent study demonstrated the efficacy and adaptability of LC-HRMS through a proof-of-concept investigation with 52 distinct honey samples obtained from various countries (Makni et al., 2024). A singular method utilising a QuEChERS procedure, previously refined in a prior investigation, was used to process honey samples. This method was specifically designed for the detection of anthropogenic contaminants in sweet items (Makni et al., 2023). The samples were analysed using UHPLC-HRMS. A single analytical batch was conducted using both ESI+ and ESI- modes, following rigorous QA and QC processes that met the quality criteria for metabolomics and NTA research. Using this approach, the authors effectively conducted a comprehensive study of 52 honey samples, specifically targeting 694 different pollutants for which analytical standards were accessible. Suspect screening analysis utilized MSDial software to identify endogenous compounds and additional anthropogenic contaminants to extend the analytical coverage. NTA employed multivariate and univariate statistical tools to conduct an exhaustive investigation of characteristics that might indicate the honey samples' floral and geographical origin. This study focused exclusively on honey as a subject of investigation, however, the methods and techniques employed have the potential for ready application to other types of matrices and thus warrant future investigation in other food matrices, especially in plant-derived food matrices such as fruit.

The key advantages of such multifaceted approaches include their ability to repurpose the same LC-HRMS data for several analytical purposes and employ alternative methodologies. The use of LC-HRMS is highly beneficial in identifying and quantifying anthropogenic pollutants and markers. In quality control laboratories, LC/GC-HRMS can be utilized alongside LC/GC–MS/MS to broaden the analysis and detect misuses, mixtures, and emerging contaminants. This is significant because LC/GC-HRMS is currently not fully utilized in quality control laboratories. Furthermore, this technology can be utilized to track honey's origin and identify fraudulent activity more effectively. In the context of food safety, this method is adept at identifying pollutants that contribute to the distinction of honey samples based on their floral or geographical source. Consequently, it could reveal the utilization of treatments exclusive to a particular plant species or limited to a certain geographic region. This can be especially advantageous in plant-based matrices where it might be extremely laborious to track the source of impurities. In the previous section, we mentioned that a recent study proposed the idea that certain pesticide residues could translocate from the soil to other parts of the plant (Fu et al., 2019). However, the specific source and translocation mechanisms of these contaminants were not definitively determined, therefore, this technology can be utilized to track the origin of those pesticide residues more effectively.

Lastly, hydrophilic interaction liquid chromatography (HILIC) coupled with mass spectrometry has proven to be a valuable method for analysing aminoglycoside antibiotics (AAs), addressing the challenges of poor separation and low response encountered when conventional reversed-phase liquid chromatography with UV detection is applied. These include challenges in the separation and detection of AAs that arise from the strong polarity and absence of chromophore groups (Moreno-González et al., 2017). Nevertheless, numerous substances in food samples exist that can interfere with the targets, directly impacting their responses. Therefore, it is crucial to employ an appropriate pre-treatment method to purify the food sample effectively. Previously, PVA-Sil nanoparticles were used to efficiently absorb AAs in honey samples (Y. Wang et al., 2015). However, this method required a significantly high amount of eluting solvent, materials, and time, which is not in line with green chromatography, thus emphasising the importance of a dependable and user-friendly pre-treatment technique.

When magnetic nanoparticles are evenly dispersed in biological samples, they can efficiently adsorb targets and complete the washing and elution procedures quickly, all under the control of an additional magnetic field. In addition to its easy operation, DSPE also has the advantages of requiring less sample and material, as well as being environmentally friendly (Jun et al., 2008). A recent study developed an effective extraction method using PVA-coated Fe₃O₄ (Fe₃O₄@PVA) microspheres via a facile way to absorb polar chemicals due to their high hydrophilic property and the advantages of functionalised magnetic nanoparticles (D. Li et al., 2018). Utilising HILIC-MS/MS, this method was used to enhance the identification of AAs in honey. The characterization results indicated that the Fe₃O₄@PVA microspheres had a uniform particle size, even distribution, and high magnetic strength. These outstanding qualities fully satisfied the demands of the dispersive phase extraction procedure. Notably, the entire extraction and detection procedure of AAs from honey samples could be completed within 40 min. The primary drawback of Fe₃O₄@PVA is restricted selectivity. However, this problem could be resolved via tandem mass spectrometry using a mass spectrometer that possesses a high level of selectivity.

Lastly, core-satellite magnetic nanosorbents (MNs) were effectively developed in a recent study to selectively extract macrolide antibiotics, functioning like a sieve that captures specific particles. These MNs were used as magnetic dispersive solid-phase extraction (MDSPE) sorbents to concentrate trace amounts of MACs from complex food matrices before quantification using LC-MS/MS (Fan et al., 2021). The nanocomposites exhibited exceptional magnetic characteristics, notable adsorption capacity, selectivity, and rapid binding kinetics for the antibiotics, rendering them highly efficient for extraction purposes. Five distinct honey samples were selected for the analysis of seven MACs, and the practicality of the MDSPE-HPLC-MS/MS method was further demonstrated by analysing milk, which further confirmed the effectiveness of MDSPE in selectively extracting MACs from complicated matrices, indicating the multi-matrix potential of the developed strategy.

4.3. Milk

India emerged as the largest global consumer of cow milk in 2023, with a staggering consumption of over 87 million metric tonnes. The European Union had the second-highest milk consumption, amounting to 23.7 million metric tonnes (Global consumption of milk per year by country, 2023), making milk one of the most consumed products globally. Milk is an emulsion that has the ability to effectively retain a wide variety of pollutants that have diverse physicochemical properties (Souza et al., 2021). Milk can be contaminated through several means. Feedstuffs can serve as carriers for remnants of phytosanitary substances, including herbicides, insecticides, and fungicides. Administration of veterinary medications to animals through oral, parenteral, percutaneous, or intramammary routes can lead to milk contamination, particularly as a result of poor farming practices such as non-compliance with prescribed withdrawal periods (Souza et al., 2021).

The LC-HRMS approach offers clear benefits in terms of retrospective analysis and non-target screening, as discussed in the preceding sections. Nonetheless, the detection of analytes at trace levels may, however, result in a slight compromise of sensitivity and quantitative accuracy (Hou et al., 2020). This compromise may be addressed by the LC-MS/MS, which offers enhanced selectivity, sensitivity, and quantitative precision. Major research efforts have recently been applied in sample preparation as this is a critical determinant of method performance and applicability. QuEChERS technique modified using a melamine sponge to simultaneously analyse multiple residues of 57 veterinary drugs in milk samples using UPLC-MS/MS was recently reported (Ji et al., 2021). To the best of our knowledge, this was the first study to report on the application of the melamine sponge in the purification for matrix adsorption. For the confirmatory analysis of each drug, it was necessary to have a minimum of two pairs of MRM ions. Precise and sensitive determination of multi-residues of veterinary drugs, particularly for structural isomers of sulfonamides and quinolones with identical precursors and fragments, requires sufficient chromatographic separation and effective electrospray ionization. In comparison to the typical QuEChERS method, matrix removal could be readily accomplished within a matter of seconds through a straightforward process of soaking and squeezing, without the requirement of additional phase separation procedures, as shown in Fig. 4. The melamine sponge surface possesses unique hydrophilic and lipophilic qualities, allowing various extraction solvents to easily penetrate its many cross-linking micropores. This facilitates efficient matrix adsorption by supporting a comparatively large specific surface area. Furthermore, the melamine sponge used was readily accessible in the market at an affordable price, resulting in a substantial decrease in the expenses associated with regular monitoring. The study's findings also indicated that the developed method, which utilized a melamine sponge as an exceptional adsorbent for matrix purification, exhibited notable effectiveness and adequate reliability.

Fig. 4.

Application of melamine sponge in matrix purification for determination of multi-class veterinary drugs in milk samples (B. Ji et al., 2021). Reuse rights obtained from Elsevier copyright clearance center 5,796,921,065,265.

Using microextraction techniques has proven to be an important approach for sample preparation with regard to environmental sustainability. The significant decrease in both sample size and quantity of organic solvents and other hazardous reagents, along with a minimal generation of waste, are key environmentally friendly aspects of these techniques (Płotka et al., 2013). In addition, the straightforward and user-friendly operation, as well as the ability to work well with various substances, make it convenient to combine different microextraction techniques with other traditional sample treatments. This is particularly useful for isolating, purifying, and concentrating specific analytes.

A recent study described a novel approach to create hollow mesoporous molecularly imprinted nanoparticles (HMMINs) inspired by pomegranates (Ji, Huang, Li, Luo and Zheng, 2021). This was achieved by using dopamine to self-assemble the nanoparticles through an organic-organic process. The one-pot synthesised HMMINs, with their distinctive chemical composition and architecture, offered several advantages. They exhibited high imprinting efficiency and binding capacity, making HMMIPDAs a promising choice as a sorbent to enhance the selectivity of analytes from complex matrices. As a result, the synthesised nanomaterial was initially used as a selective nano sorbent DSPE for seven MACs in real milk samples, exhibiting minimal matrix effects and exceptional recyclability, which aligns with and promotes the development of green analytical techniques. The DSPE-HPLC-MS/MS techniques demonstrated strong linear relationships for all seven MACs, and the limit of quantification was within the range of 0.030–0.144 μg·kg⁻¹. Fig. 5. illustrates the synthesis HMMIPDAs and workflow of the DSPE-LC-MS/MS procedures.

Fig. 5.

a) One-pot synthesis scheme of hollow mesoporous molecularly imprinted nanoparticles, b) Application as a selective nano sorbent DSPE for seven MACs in real milk samples (S. Ji et al., 2021).

To achieve green analysis, simultaneous analysis of three significant contaminant groups (phthalates, polycyclic aromatic hydrocarbons, and pesticides) in complex infant formula matrices was achieved, which represents a breakthrough in the field (Henrique Petrarca et al., 2024). A precise and meticulous analytical method utilising gas chromatography–mass spectrometry was devised to detect and quantify 45 different food contaminants. Combining dispersive solid-phase extraction and low-density solvent-based dispersive liquid-liquid microextraction techniques allows for effective clean-up and analyte enrichment. The most optimal analytical responses were achieved at elevated temperatures in pulsed splitless mode with high pulse pressures. There were minimal matrix effects observed for most of the analytes, suggesting a potential correlation between these effects and the physicochemical properties of the analytes. The method performance characteristics were achieved to ensure accurate monitoring of regulated compounds in infant formula, with limits of detection and quantification set at the maximum allowable levels. In addition, the proposed method was evaluated for its environmental impact and usefulness using the green assessment metric tools; AGREE and BAGI respectively.

To enhance the analytical performance of the method employed for detecting veterinary drug residues, it is imperative to boost the analytical efficiency of the method while simultaneously minimizing the associated cost and time. Multi-class methods are one of the most effective approaches for identifying veterinary drug residues in animal-derived foods. These methods enable the detection of multiple compounds through a single analysis. Given the diverse physio-chemical properties of various classes of veterinary drugs, analysing multiple classes of these drugs presents a formidable challenge. This task necessitates the simultaneous extraction and enrichment of different compounds from animal matrices. In addition, the protein and fat levels in the matrices and the low concentrations of analytes in animal samples pose challenges in developing multi-class veterinary drug analysis methods.

5. Future perspectives and conclusions

As the world is making major efforts to accelerate its food production capacity to keep up with the anticipated population growth, traditional food cultivation, processing and preserving methods are proving to be inadequate, leading to the surge in novel food production techniques that present the potential to introduce unforeseen pollutants with potentially fatal adverse effects. This has necessitated the development of simple, environmentally friendly, multiclass, multi-matrix food contaminant analytical techniques that ensure food safety in this new era and protect consumer rights and overall health. In this review, we have provided an integrated, concise, and comprehensive overview of the research efforts in the application of chromatography coupled with mass spectrometry in the analysis of contaminants in plant-, aquatic-, and animal-derived food matrices.

The recent five years have experienced a trend in the development of multiclass/multi-residue detection strategies, and it can be projected that the next five years will gravitate towards the development of green analytical techniques for multiclass, multi-matrix analysis with the integration of artificial intelligence for both data analysis and prediction of analytical strategy design. The probable key to achieving this is optimizing extraction techniques to find the balance between the recovery of analytes and the elimination of co-extractives.

Although HRMS may lack sensitivity compared to the MS/MS method, it has proven useful in the absence of pure standards. It has been widely echoed that the characteristic fragments can be used to screen and identify unexpected or unknown contaminants. However, summarizing distinct fragments requires an enormous amount of time and effort. And the process of identifying query compounds becomes difficult when they are not included in any database which introduces coverage problems. Artificial intelligence, molecular networking, classification fragment ion list characteristic (CFILC), and other machine/deep learning-based models are showing tremendous potential in bridging the gap of the coverage problem in the analysis of emerging contaminants and illicit adulteration of food supplements. Future research can explore the application of unsupervised hierarchical clustering for screening antibiotic or veterinary drug residues across animal, aquatic and plant-derived matrices. And additionally, not only focus on incorporating machine learning in data analysis but also the development of predictive models for analytical strategy design.

Finally, using LC-HRMS, Gambierdiscus sylvae and G. caribeaus (Mudge et al., 2023) were recently identified as sources of ciguatera poisoning in the Caribbean, with CTX-like activity. This points to the possibility of analysing contaminants in food products and simultaneously elucidating the underlying mechanism and source of contamination, which could be invaluable in furthering our understanding of the emerging contaminants in food commodities. The dynamics and transfer of toxins in fish and food webs are complicated, and the causes of CTXs are still largely unknown, thus presenting another avenue for future research. Lastly, the chemical and enzymatic conversions of C-CTX5 into known fish metabolites associated with CFP and the potential interactions between C-CTX5 and other present compounds in the environment may necessitate further research.

Combinatorial analytical platforms of GC/LC-MS/MS or MS/MS with HRMS are anticipated to be reported more in these coming years as this allows for enhanced qualitative and quantitative analyses and provides versatility in complex matrices, especially in the coming era of multi-matrix strategies.

Furthermore, it is important to investigate further the potential health risks associated with the identified food contaminants and their impact on human consumption. Currently, studies tend to focus solely on detecting contaminants or elucidating health hazards but rarely do both simultaneously. Establishing a solid evidence base through comprehensive research would be valuable in guiding regulatory decisions, implementing targeted measures to reduce the presence of harmful substances in the food supply chain, and shaping food safety policies.

CRediT authorship contribution statement

Zhuzi Chen: Writing – review & editing, Writing – original draft, Visualization, Conceptualization. Zamar Daka: Writing – review & editing, Writing – original draft, Visualization, Conceptualization. Liying Yao: Investigation. Jiamin Dong: Validation. Yuqi Zhang: Visualization. Peiqi Li: Data curation. Kaidi Zhang: Software. Shunli Ji: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

No data was used for the research described in the article.