In the case of LC–MS methods, the matrix effect is associated with phenomena that occur in the ESI source. The ionization of analytes is reduced or favored by the presence of matrix components [125]. However, in the case of GC–MS methods, the matrix effect is usually caused by the retention/decomposition of analytes at the active sites (i.e., free silanol groups) in the injector, column, or detector. This is observed when standards are injected in pure solvents, but not when real samples are injected. In the latter case, matrix components can block the active sites; therefore, there are fewer active sites available to adsorb the analytes. Consequently, an enhancement in their signals is observed [126]. The determination of 341 and 315 pesticides in rice by LC–MS and GC–MS, respectively, represents a clear example of how signal suppression is usually observed in LC–ESI-MS, while signal enhancement is observed in GC–MS [127]. If the sample treatment proves ineffective in avoiding the matrix effect, it can be compensated for by applying two main strategies: (1) by using matrix-matched calibration curves (i.e., calibration curves of standards prepared in blank matrix extracts), or (2) by using isotopically labeled internal standards (IL-ISs). Both strategies are commonly applied for either LC–MS [25,128] or GC–MS methods [129,130]. It should be noted that although only one or a few IL-ISs are selected for the IL-IS approach, the quantitative performance of multiresidue methods increases when multiple IL-ISs covering different families of pesticides are used [131,132]. Other strategies to reduce the matrix effect include the possibility of applying standard addition or procedural calibration; however, matrix-matched calibration is still the preferred approach according to SANTE/2020/12830 [14]. The main challenge of using matrix-matched calibration is to find blank samples. In addition, this approach also implies the assumption that the matrix effect for a given product is the same for different samples, which is not necessarily true [133].

In GC–MS methods, the use of analyte protectants or coated inlet liners also reduces the matrix effect [126]. For example, a mixture of analyte protectants consisting of ethylglycerol, gulonolactone, D-sorbitol, and shikimic acid prepared in 4/1 (%, v/v) MeCN/water containing 0.05% (v/v) formic acid has been used for the analysis of 35 pesticides (i.e., organophosphorus pesticides, organochlorine pesticides, triazines, triazoles, carbamates, imidazoles, etc.) in tropical fruits [134]. In general, the matrix effect was significantly reduced when the analyte protectants were used. A reduction of more than three times was observed in the case of the compounds that showed a more significant matrix effect. Nevertheless, although the use of analyte protectants reduces the matrix effect and improves peak shape [135,136], these compounds must be dissolved in polar solvents such as water and MeCN. The injection of these solvents in GC systems has several drawbacks (i.e., limitations on injection volumes due to their high expansion coefficient, and poor focusing of chromatographic peaks due to their high polarity), which represents the main disadvantage of this strategy for reducing the matrix effect [69,126].

Current developments in mass spectrometers and data analysis have led to a growing interest in non-targeted methods for pesticide analysis; however, targeted methods are more established in routine food testing laboratories [137]. GC–MS-based targeted methods were the technique of choice for pesticide analysis until the late 1990s [91]. GC–MS methods were initially limited to the determination of volatile and thermostable analytes, but the implementation of derivatization strategies also allowed the analysis of non-volatile compounds. In general, early GC–MS methods involved the use of a single quadrupole (Q) operating in selected ion monitoring (SIM). Nevertheless, the development of hybrid mass spectrometers made GC–triple quadrupole (QqQ)-MS, in addition to GC–Q-linear ion trap (LIT)-MS, the preferred tool for pesticide residue control. GC–MS/MS works in single reaction monitoring (SRM) or multiple reaction monitoring (MRM). This acquisition mode is very selective, because it allows monitoring of at least two fragmentation transitions per analyte (one transition corresponds to a precursor/fragment ion pair), minimizing or avoiding matrix interference, and meeting current guidelines for proper pesticide identification [138]. It is also very sensitive, because it reduces the chemical noise in the chromatogram, allowing lower LODs to be achieved [139]. Current GC–QqQ-MS-based methods allow the simultaneous determination of hundreds of pesticides, as has been shown in the analysis of 220 pesticides in sesame seeds with a separation run of 22 min [140], or in the determination of 365 pesticides in grain, beans, fruits, and vegetables with a separation time of 36 min [141].

Although LC separations have partially replaced GC separations in pesticide analysis in recent decades, QqQ-MS and QLIT-MS remain the gold standard detection tools used in analytical workflows for pesticide determination [15]. As in GC–MS/MS methods, SRM/MRM is applied as acquisition mode in LC–MS methods. Two or even three fragmentation transitions per pesticide are recorded to reduce the risk of false positives [142]. Recent multiresidue methods based on LC–MS workflows tend to cover the simultaneous analysis of a wide range of pesticides [143,144], which requires monitoring of a large number of fragmentation transitions. This is a real challenge for QqQ-MS instruments due to their slow acquisition rates. To ensure a minimum of 15 data points per chromatographic peak for satisfactory analysis reproducibility, transitions related to each analyte are only monitored within a time-limited window that covers the retention time of the compound. However, any unexpected shift in the retention time can partially or completely exclude an analyte from its detection window, resulting in truncated or undetected peaks [142]. Advanced targeted acquisition approaches for pesticide analysis should be explored to overcome this drawback following applications in other research fields [145]. For example, “Scout MRM” represents one of these advanced acquisition strategies that has already been applied to pesticide determination [142]. “Scout molecules” are compounds spiked in the samples and detected along the chromatographic separation. Once one of the “scout molecules” is detected, it triggers the monitoring of a transition group (i.e., a group of transitions from various analytes of interest). It stops upon detection of the next “scout molecule” and activation of the subsequent transition group acquisition.

Quadrupole instruments today are very robust, and probably provide the highest concentration sensitivity compared to other mass spectrometers, but only allow unit mass resolution. Consequently, their use is limited to targeted applications. The development of HRMS instruments has opened up new possibilities allowing semi-targeted (i.e., suspect screening) and non-targeted analysis of pesticide residues. Furthermore, high mass resolution is of great interest to avoid overestimating the amount of the target compound due to co-elution of the analyte with isobaric compounds in the matrix [146]. Depending on the HRMS technology, high mass accuracy (±0.001 Da), high mass resolution (ratio of mass to mass difference ≥ 20,000), and wide mass range (simultaneous acquisition of ions up to 2000 Da) can be obtained [147]. Mass resolving power of up to 1,000,000 can be achieved in routine analyses, specifically when using Fourier-transform ion cyclotron resonance (FT-ICR) mass spectrometers [148]. The term “resolving power” describes the ability to distinguish between ions that differ in the m/z by a small increment, and is equal to the full width at half-maximum (FWHM). FWHM is defined as m/Δm, where m represents the mass of the ion being measured, and Δm is the full width of the spectral peak at half-maximum peak height [149].

Due to the analytical performance and instrumentation cost, time-of-flight (ToF) and Orbitrap spectrometers, as well as hybrid Q-ToF and Q-Orbitrap instruments, are mainly used for pesticide determination in applications in which high mass accuracy is required [49]. More information on the capabilities of high-resolution mass spectrometers can be found in the literature [150,151]. One of the main reasons that have limited the implementation of HRMS in official pesticide controls has been the lower concentration sensitivity shown by high-resolution mass spectrometers compared to the low-resolution mass spectrometers traditionally used for this purpose. Despite QqQ-MS having been shown to be more sensitive and robust for pesticide determination at trace levels, Q-Orbitrap-MS operating in full scan can provide equal or even higher sensitivity for certain applications [146]. In fact, Orbitrap-MS has been shown to be applicable as an analytical tool in routine pesticide monitoring. Up to 210 pesticides have been determined simultaneously in fruits and vegetables under routine conditions (i.e., 102 samples), demonstrating these devices to be robust and efficient in terms of sensitivity (LOQs ≤ 10 µg/kg for the majority of pesticides and commodities) and mass accuracy (±1 mDa) [152]. Furthermore, Orbitrap-MS (140,000 FWHM, m/z 200) and ToF-MS (30,000 FWHM, m/z 556), both operating in full scan, have been compared for the quantification of pesticide residues (n = 146) in tea samples [153]. In general, similar results were observed for both mass spectrometers, but Orbitrap-MS was found to provide slightly better sensitivity and selectivity than ToF-MS at lower concentration levels (10 µg/kg), due to its higher mass resolution. Nevertheless, both mass spectrometers were demonstrated to be suitable for routine pesticide monitoring, not only because of the LODs achieved (i.e., detectable responses between 0.01 and 0.1 mg/kg, depending on the pesticide and the detector), but also in terms of intra- and inter-day precision (RSDs < 20%).

In the context of routine pesticide monitoring programs, LRMS-based methods operating in SIM or SRM/MRM have traditionally been applied to the simultaneous determination of 100–200 pesticides in a single run. However, analytical methods are evolving to cover a larger number of compounds. In this framework, HRMS-based methods have significantly expanded the testing scope of pesticides, monitoring almost 1000 analytes in less than 15 min [118]. Despite the fact that targeted analysis of pesticides can also be addressed by HRMS [22,154], HRMS-based workflows are primarily intended for non-targeted analysis of pesticides, including suspect screening approaches. The main objective of such methods is to determine pesticide residues in food products beyond a predefined scope of target pesticides. At the same time, these methods also make the identification of pesticide metabolites or (bio)transformation products feasible [155]. LC (or GC)–HRMS methods are mainly applied in two different contexts:

• Determination of “known unknowns” (i.e., pesticides with known chemical structure and suspected presence in the food sample, but without available reference standards). Suspect screening approaches are based on screening of large lists of priority compounds using their exact mass, and subsequent tentative identification by comparing MS/MS spectra against mass spectral databases [156]. For example, Bauer et al. developed a UPLC–Q-ToF-MS method for the suspect screening of pesticide metabolites in fruit and vegetables [35]. A database of 684 metabolites of 58 active compounds was created based on relevant pesticide metabolites reported in scientific literature and approval documents. This database was subsequently applied to the tentative identification of 47 pesticide metabolites in 96 representative fruit and vegetable samples selected from a daily routine pesticide analysis. This strategy is of high interest, since the detection of metabolites in the absence of the parent pesticide may provide evidence of its illegal use.
• Identification of “unknown unknown pesticides” without any a priori criteria, which requires a solid basic knowledge of chemistry and biochemistry to achieve an unambiguous structural elucidation of the relevant analytical signals [157]. Non-targeted analysis involves the detection of a large number of peaks, of which a significant number remain unidentified [155]. Since this analytical approach has several limitations, and does not always lead to satisfactory results to compensate for the time and effort involved in its execution, suspect screening approaches are the most commonly followed in applications related to pesticide analysis in food. In this regard, most HRMS methods for pesticide determination directly involve a list of high-interest pesticides. Consequently, they should be classified as suspect screening approaches rather than non-targeted approaches, as is often reported by many authors. Nevertheless, great efforts are currently being made to develop suitable non-targeted methods based on HRMS to detect unexpected contaminants and chemical residues in food, such as in the case of fipronil in Dutch eggs spread in Europe in 2017, much earlier [158].

In addition, HRMS has another important advantage over LRMS. Full-scan HRMS analyses provide complete information on the content of the sample, being limited only by the satisfactory ionization and detection of the molecules under the conditions of the analytical method carried out. Therefore, raw mass spectra can be retrospectively exploited to search for a specific substance if a specific concern arises. This approach is also applicable to the determination of pesticide metabolites or (bio)transformation products whose detection was not part of the primary study objective of the original analyses. A retrospective analysis of HRMS data has been carried out for the identification of pesticides in infant milk formulas. Pesticides are not the main contaminants or chemical residues expected in this type of matrix, but they were reported to be present in 33.3% of the samples [159]. In addition, retrospective data analysis has also been carried out for the identification of pesticide transformation products in food products (i.e., honey, meat, feed, and nutraceutical products such as ginkgo, soya, green tea, and royal jelly) that were previously found to content pesticide residues. One substance—3,5,6-trichloro-2-pyridinol—was identified as transformation product of alachlor and chlorpyrifos [160]. Although retrospective analysis opens up new possibilities for wide-scope screening and reuse of raw mass data, it is not without limitations. Its application in routine analysis may be ineffective due to the necessity of high experience of the operators and the time required for compound identification, as well as for extracting information of interest from HRMS data generated without a previous hypothesis.

Most reported HRMS methods for pesticide analysis involve LC separation [49] and data acquisition in one of the following main MS modes: full scan, data-dependent acquisition (DDA), or data-independent acquisition (DIA) [161]. Full-scan acquisition involves only monitoring the accurate mass and isotopic distribution of ions reaching the detector, so full-scan data may not be sufficient for complete compound identification. Therefore, full-scan acquisition may be suitable for targeted analysis of pesticides when using HRMS [146,153], but it is not the most appropriate for suspect screening and non-targeted analysis. In these cases, product ion scan (MS²) data are needed for molecular structure identification. HRMS offers DDA and DIA modes as two data acquisition possibilities to simultaneously acquire information on the precursor ion and related fragmentation ions [162].

In DDA mode, data acquisition is carried out in full scan (MS¹), and when precursor ions meet predefined criteria such as a specific m/z or an intensity threshold, they are selected as candidates for fragmentation (MS²). Conversely, in DIA mode, all precursor ions are fragmented without applying any criteria, so maximum MS² is obtained. To date, various DIA strategies have been implemented [163], although the most commonly applied in pesticide analysis are those integrated into MS vendor software, including all-ion fragmentation (AIF) employed in Orbitrap instruments, or “all-in-one” analysis (MS^E) developed by Waters. Regardless of the applied DDA or DIA mode, the collected MS² is compared against mass spectral libraries for compound identification. DIA can potentially generate more MS² matches, resulting in a higher chance of identifying pesticides or related compounds of interest, but without the high selectivity of DDA that provides cleaner mass spectra [161,162]. The application of DDA mode can be recommended for pesticide residue determination, since most of these compounds are known and can be effectively identified by DDA [164]. On the other hand, DIA should be implemented in those workflows aimed at the discovery of pesticide metabolites and (bio)transformation products, since these are usually unknown compounds [165]. However, recent work has shown that pesticide identification rates can be increased when the DDA and DIA modes are combined, while also providing fewer false results for the target pesticides [166,167].

Although GC–HRMS methods have been reported for the analysis of pesticides in food, they represent a lower percentage than the LC–HRMS methods developed for this purpose. GC–HRMS workflows are not yet well defined, and the possibilities of this technique have not been fully explored [91]. The commercialization of hyphenated GC–Orbitrap-MS platforms is recent (2015) and, as previously mentioned, LC–MS methods cover a wide range of applications traditionally addressed by GC–MS without requiring derivatization, and involving soft ionization (ESI vs. EI). In this sense, the ionization source for GC–HRMS coupling can also raise several questions, as occurs in GC–LRMS methods. APCI produces molecular ions that facilitate compound identification in pesticide screening [124], and is a fairly generic ionization technique for GC-amenable compounds, but is not as generic as EI [168]. In addition, generic spectral libraries for APCI-MS are still lacking, while several mass spectral libraries exist for EI-MS (e.g., NIST spectral database), which are crucial for compound identification in non-targeted analyses. Although the latter refer to nominal spectral libraries (i.e., based on EI-LRMS measurements), they are instrument-independent, and can be used for library matching if the standard ionization energy of 70 eV is used. Minor differences between EI-LRMS and EI-HRMS spectra can be observed, such as a reduced relative abundance of ions < 90 m/z when working with GC–EI-Orbitrap-MS operated in full scan compared to the NIST library [168]. However, such differences do not prevent the observation of a high match for pesticide mass spectra, thus providing high confidence in compound identification. For example, non-targeted analysis of different food commodities—such as mushrooms and peppers, among others—by GC–EI-Q-Orbitrap-MS, along with further comparison of data against libraries containing more than 200,000 spectra, allowed the detection of several pesticides (i.e., lambda-cyhalothrin, triadimenol, imazalil, pyrimethanil, and tebuconazol as the most common compounds) without any prior hypothesis [169].

The recent commercialization of IM–MS instruments has brought new possibilities to food analysis, and specifically to pesticide residue determination in the field of food safety [170,171]. The IMS dimension is located after the ionization source, so this analytical technique is applied to the separation of charged molecules based on their mass, charge, and shape under an electric field and in the presence of a neutral gas (e.g., typically N₂, He, or CO₂). In addition, chromatographic techniques (either LC or GC), ion-mobility spectrometry (IMS), and MS can be coupled, since the separation in these three dimensions occurs on different time scales (i.e., minutes/seconds, milliseconds, and microseconds, respectively). There are different IMS technologies coupled with MS on the market, including drift-tube ion-mobility spectrometry (DTIMS), traveling-wave ion-mobility spectrometry (TWIMS), trapped ion-mobility spectrometry (TIMS), high-field asymmetric-waveform ion-mobility spectrometry (FAIMS), and differential ion-mobility spectrometry (DMS). To date, pesticide determination has only been addressed by DTIMS–MS [172], TWIMS–MS [173,174,175], and DMS–MS [176]. Commercial DTIMS–MS and TWIMS–MS systems typically involve TOF mass spectrometers, because these types of analyzers offer a high acquisition rate. This instrumentation has been developed primarily for non-targeted analysis purposes. In contrast, DMS–MS systems involve QqQ-MS technology. Unlike DTIMS and TWIMS cells—where all ions entering the mobility cell are likely to reach the exit and, subsequently, be detected by MS—DMS cells act as filters of the analytes of interest [103,177]. Therefore, DMS–MS is mainly applied for targeted analysis purposes.

The high selectivity of DMS contributes to the removal of matrix compounds that co-elute in the LC dimension with target pesticides, as shown for the determination of phosphonic acid in samples that naturally contain a high concentration of phosphoric acid [23]. When applying LC–MS methods, the chromatographic separation of both substances can be compromised, while the most important qualifier mass transition of phosphonic acid is also the minor transition of phosphoric acid. Therefore, phosphoric acid can exert interference with phosphonic acid, leading to false positives. This can be avoided by integrating DMS into the LC–QqQ-MS workflow. In addition, the removal of interferences and matrix compounds in the DMS dimension not only improves method selectivity, but also improves method sensitivity by reducing chemical noise and removing adjacent chromatographic peaks from the matrix compounds. This improvement has been observed for the analysis of triazole fungicides at 0.01 mg/kg in different samples of plant origin (e.g., turnip and soy bean), which were barely identified as chromatographic peaks when applying an LC–MS method, but which could be integrated perfectly when using an LC–DMS–MS method [178].

The integration of DTMIS or TWIMS into non-targeted LC–MS workflows enhances the analytical performance of the method from three standpoints: selectivity, sensitivity, and confidence in compound identification [171]. This last aspect is of special importance for reducing the number of false positives reported in residue analysis. Both DTIMS and TWIMS (also TIMS) provide an additional identification parameter to chromatographic retention indices and mass spectra—the so-called collision cross-section (CCS). CCS is a molecular characteristic related to the spatial conformation of the ions in the gas phase. Although the drift time (i.e., the time required for ions to traverse the IMS cell) has also been used as an identification parameter in the analysis of pesticides in different fruits and vegetables [173], this parameter is instrument-dependent, and varies depending on the applied instrument settings. Nevertheless, CCS values have been shown to be reproducible between different IM–MS platforms when using the same IMS technology [179,180]. Consequently, it is possible to rely on CCS databases reported by different laboratories to use this information for identification purposes in pesticide analysis.

Several CCS databases for pesticides have been already reported [181,182]. They are also expected to be improved with CCS values for more pesticides, including transformation products and metabolites, due to the increasing implementation of LC–IM–MS instrumentation in residue analysis laboratories. In this context, the number of false positives in the screening analysis of pesticides is drastically reduced when the CCS is included in the identification criteria (i.e., threshold of ±2% from values in databases) [182]. This fact has been shown for the screening of 110 pesticides in cocoa beans at three different concentration levels (10, 50, and 150 µg/kg) [183], and for the screening of 156 pesticides at 10, 50, and 200 µg/kg in salmon feed (n = 20) [174]. In this last case, the number of false positives was reduced from 42 to 1 when the CCS was also selected as an identification criterion. Furthermore, the use of CCS values as identification criteria in the determination of pesticides by screening analysis has been shown to lead to a higher detection rate than the identification of at least one fragment ion, as this criterion is compromised for pesticides at low concentration levels [175].

The selectivity improvement provided by DTIMS or TWIMS in LC–IM–MS methods is directly related to their resolving power, which obviously depends on the IMS technology. In general, it has been estimated that the resolving power is still very limited in the LC–IM–MS instrumentation used in multiresidue analysis [103]. However, IMS has already been shown to be effective for the separation of isomeric pesticides, such as the protonated ions of ipconazole and tebufenpyrad (C₁₈H₂₄ClN₃O)[181]. The use of current IMS technology with increased resolving power, such as cyclic-TWIMS, also allows the separation of charged isomers (e.g., protomers). This helps to understand why the relative abundance of characteristic product ions of a pesticide can vary depending on whether they are produced from analytical standards or chemical residues in food matrices [184]. Although according to SANTE 11312/2021 it is not necessary to take into account the ion ratio for pesticide identification when using HRMS, the ion ratio is a parameter to consider for this purpose when using LRMS [138]. Information on the formation of promoters for different pesticides could justify non-detected compounds present in the sample in targeted analyses when using unit mass spectrometers—due to a missing quantifier/qualifier ion, or to the alteration of their ion ratio [175]. Therefore, it could be useful to avoid false negatives when the ion ratio criterion is not fulfilled but pesticide residues are present in the sample.

Finally, DTIMS and TWIMS in LC–IM–MS workflows also isolate targeted compounds from background noise, providing cleaner mass spectra, facilitating mass spectral interpretation, and improving confidence in pesticide identification [172]. This effect also provides extracted ion chromatograms of greater quality, resulting in increased concentration sensitivity due to the improvement of the signal-to-noise ratio (S/N). Although this effect has not been extensively explored for the determination of pesticides in complex food matrices, improvements of S/N from 2- to 7-fold have been reported for the determination of small molecules when TWIMS has been integrated into an LC–MS method [185].

Ambient-ionization mass spectrometry (AIMS) is an emerging technology in the field of pesticide analysis. Compared to traditional analytical techniques used in this field, AIMS provides higher sample throughput as it involves direct, real-time, in situ analysis with minimal or no sample preparation. However, although the performance of AIMS has been studied with great interest by many routine food safety laboratories, the applications evaluated so far have not met the monitoring demands in field trials for pesticide screening. The main limitations of AIMS observed in the initial trials for routine implementation are related to the narrow analytical scope and the high rate of false positives and negatives [186]. AIMS can be considered as a promising approach for application in pesticide analysis—especially as an initial strategy to decide which samples should be sent to the laboratory for further testing. However, technical advances are still required to improve the poor reproducibility currently observed in sample analysis due to the sample heterogeneity and ion collection. These advances should also increase the concentration sensitivity, which is crucial in residue analysis but is quite limited due to the high matrix effect attributed to AIMS [76].

There is a wide range of AIMS techniques currently available for pesticide analysis, but these are mainly classified into ESI-related and APCI-related techniques. Although different AIMS approaches have been evaluated for the determination of pesticides [187], desorption electrospray ionization (DESI) and direct analysis in real time (DART) represent the most commonly used AIMS techniques in residue analysis. DESI is based on the spraying of charged solvent microdroplets at the sample surface, desorbing and ionizing the analytes that subsequently enter into the mass spectrometer [188]. The high irreproducibility of the analytical results provided by DESI is mainly associated with instrument setup and operation [189], as well as the fact that only a part of the sample surface is examined even when the pesticide residues are not homogeneously distributed on it. In this sense, the use of IL-ISs improves the reproducibility of DESI methods, and (semi)-quantification of pesticide residue levels is feasible [190]. On the other hand, DART involves the generation of plasma via the ionization of gas molecules (i.e., He, Ar, or N₂), which can be consequently heated, and are impinged on the sample to cause desorption and ionization of the analytes [191]. Despite AIMS having generally been applied to the qualitative determination of pesticides in food, DART in combination with the QuPPe extraction method has been shown to be efficient for quantification purposes in the analysis of highly polar pesticides in lettuce and celery [192].