Contaminants of emerging concern in urine: a review of analytical methods for determining diisocyanates, benzotriazoles, benzothiazoles, 4-methylbenzylidene camphor, isothiazolinones, fragrances, and non-phthalate plasticizers

Human biomonitoring (HBM) frameworks assess human exposure to hazardous chemicals. In this review, we discuss and summarize sample preparation procedures and analytical methodology for six groups of chemicals of emerging concern (CECs), namely diisocyanates, benzotriazoles, benzothiazoles, 4-methylbenzylidene camphor, isothiazolinones, fragrances, and non-phthalate plasticizers, which are increasingly detected in urine, however, are not yet widely included in HBM schemes, despite posing a risk to human health. The sample preparation procedures depend largely on the chemical group; however, solid-phase extraction (SPE) is most often used due to the minimized sample handling, lower sample volume, and generally achieving lower limits of quantification (LOQs) compared to other extraction techniques. In terms of sample analysis, LC-based methods generally achieve lower limits of quantification (LOQs) compared to GC-based methods for the selected six groups of chemicals owing to their broader chemical coverage. In conclusion, since these chemicals are expected to be more frequently included in future HBM studies, it becomes evident that there is a pressing need for rigorous quality assurance programs to ensure better comparability of data. These programs should include the reporting of measurement uncertainty and facilitate inter-laboratory comparisons among the reporting laboratories. In addition, high-resolution mass spectrometry should be more commonly employed to enhance the specificity and selectivity of the applied analytical methodology since it is underrepresented in HBM. Furthermore, due to the scarcity of data on the levels of these CECs in urine, large population HBM studies are necessary to gain a deeper understanding of the associated risks.


Introduction
With the emergence of new consumer and personal care products, the gradual awareness of potentially hazardous chemicals entering the environment has also increased, especially when combined with improved analytical methods and instrumentation, such as chromatographic and highresolution mass spectrometric techniques that have allowed the detection of previously overlooked contaminants. Additionally, knowledge of the toxic effects of known compounds is increasing, many of which are now considered contaminants of emerging concern (CECs). However, human exposure to many compounds and even groups of compounds remains under-researched, especially in routine monitoring schemes, such as the National Health and Nutrition Examination Survey (NHANES), Human Biomonitoring for Responsible Editor: Philippe Garrigues Europe (HBM4EU), and the German Environmental Survey (GerES) (Yadav et al. 2021).
Human biomonitoring (HBM) measures potentially harmful chemicals and their metabolites in human samples to identify exposure sources, understand chemical risk, and ensure policy informing and efficiency of reduction strategies (Ganzleben et al. 2017). As such, it measures internal exposure that reflects the actual chemical uptake from oral, dermal, and inhalation exposure pathways (Vorkamp et al. 2021). Blood and urine are the most commonly investigated matrices. Although blood represents an optimal matrix since it is in contact with all tissues in the body and allows the partitioning of compounds into every organ, sampling blood is invasive which makes it less suitable for specific populations such as children and infants. In contrast, urine can be non-invasively collected as a 24-h composite or, more commonly, as a spot sample. The main disadvantage of collecting urine is the varying degrees of dilution due to an individual's hydration status, albeit adjustment methods like normalization to creatinine, specific gravity, and osmolarity can circumvent this issue (Esteban and Castaño 2009).
Although HBM schemes already include many chemicals, groups of potentially harmful compounds remain that have not yet been adequately addressed. These include, for example, pyrrolidones, isocyanates, benzotriazoles, benzothiazoles, ultra-violet (UV) filters, antimicrobials (MCI, MI), fragrances, and non-phthalate plasticizers, where evidence for their adverse health effects is mounting Salthammer 2020). In this paper, we review the analytical methods quantifying these compounds in human urine to compile data that could be used for future HBM studies.
Benzotriazoles (BTRs) are high production volume chemicals used as corrosion inhibitors and UV stabilizers for various photography and liquid coolant applications. Benzothiazoles (BTHs) are also used as corrosion inhibitors, industrial biocides, and vulcanisation accelerators in rubber production (Asimakopoulos et al. 2013b;Maceira et al. 2018;Naccarato et al. 2014). In the human body, they are biotransformed into more hydrophilic metabolites that are excreted as glucuronidated and sulphated compounds (Asimakopoulos et al. 2013b). Although data on their toxicological effects are limited, available evidence suggests that they can adversely affect the liver and kidney (Naccarato et al. 2014), while BTH is a suspected carcinogen (Li et al. 2018b).
Since phthalate-based plasticizers have been under criticism due to their adverse health effects, phthalate alternatives have entered the market with a constantly increasing share (Bui et al. 2016). Phthalate alternatives find applications as softeners in plastics and additives in personal care products and industrial materials. Although many have been well-studied, e.g. diisononyl ester of cyclohexane-1,2-dicarboxylic acid (DINCH), less is known about tri-(2-ethylhexyl) trimellitate (TEHTM), di-2-ethylhexyl adipate (DEHA), di-n-butyl adipate (DnBA), and di-(2-ethylhexyl) terephthalate (DEHTP). Most of these are alternatives for the traditional phthalate plasticizer di-2-ethylhexyl phthalate (DEHP), while DnBA is an alternative for di-n-butyl phthalate (DnBP). Their toxicity and migration rates are significantly lower than traditional phthalates; however, leaching into the surrounding media still occurs; for example, DnBA is restricted to low temperature applications due to high migration rates. This compound is often used in high concentrations (up to 8% of the final product) (Andersen 2006) 1 3 in personal care products, solvents, and cleaning products. It exhibits very low toxicity in humans with a no-observedadverse-effect level of 300 mg/kg bw (Ringbeck et al. 2020). However, estimating human exposure to DnBA is a challenge due to the lack of exposure data, intake rate estimates, and data on health effects. For instance, there is currently no data on the potential endocrine disruptive potential of DnBA (Bui et al. 2016). Monitoring is therefore needed as a preventive measure due to the increasing exposure of the population (Bui et al. 2016).
DEHA is a high production volume adipate, and human exposure is continuously increasing (Bui et al. 2016). Animal studies show evidence of adverse health outcomes in rats following exposure to high doses of DEHA (Silva et al. 2013). DEHTP is a structural isomer of DEHP that has not shown any of the reported adverse health effects of DEHP, with a reported NOEL of 500 mg/kg bw/day. It is produced in quantities similar to DINCH in the European Union (> 10,000 t/year) and has been frequently detected in household dust and indoor air and humans are widely exposed to this compound (Bui et al. 2016). TEHTM is a tri-ester structurally similar to DEHP but has lower toxicity. It is, however, classified as a Category 2 reproductive toxicant, a Category 3 specific target organ toxicant, and is relatively environmentally persistent (half-life 16-60 days). As such, TEHTM has a set derived no effect level of 1.13 mg/ kg bw/day within the European Chemicals Agency (ECHA) registration (Höllerer et al. 2018a).
These phthalate alternatives fall under the high molecular-weight compounds (> 6 C atoms in the side chain) and undergo similar metabolic pathways in humans. The tri-ester compound TEHTM can form three diester isomers during hydrolysis, followed by further hydrolysis to monoesters. TEHTM, DEHA, DnBA, and DEHTP also undergo biotransformation to secondary metabolites containing hydroxy, oxo or carboxy groups on the side chain. We included them in the list of phthalate alternatives due to their expected increasing market share, limited exposure data and limited toxicological data.
Organic ultra-violet (UV) filters are a group of compounds capable of filtering UV radiation due to their degree of conjugation and are used in personal care products such as sunscreens and other cosmetic products to protect the skin. They can be roughly classified into benzophenone (BP), cinnamate, crylene, camphor and salicylate derivatives. The latter consists of other organic UV filters that do not fall into the abovementioned groups (Huang et al. 2021). For example, although organic UV filters are highly studied in HBM, an exception is the camphor derivative 4-methylbenzylidene camphor (MBC). MBC is currently detected at relatively low concentrations (<LOD -13.93 ng/mL) and frequencies, with most studies focusing on aquatic systems, where it is known to be highly lipophilic and accumulates in the fatty tissue of biota (Ao et al. 2018b;Huang et al. 2021). The metabolic pathway of MBC in humans has been studied after dermal application. 3-(4′-carboxybenzylidene)camphor (CBC) and four isomers of 3-(4′-carboxybenzylidene) hydroxycamphor (CBC-OH) were identified as the main metabolites of MBC and are the main biomarkers of exposure (León-González et al. 2013).
Human exposure to MBC can result from either direct dermal application or ingestion of contaminated food and drinks, for instance, fish consumption or tap water (Li et al. 2019). Health concerns associated with MBC have increased due to its suspected endocrine activity, while in vitro studies suggest that MBC can enhance the process of apoptosis and hence directly affect nerve cells (Broniowska et al. 2016). Other studies report adverse effects on the immune system (Ao et al. 2018a), and it has been suggested that MBC facilitates the migration of human breast cancer cells (Alamer and Darbre 2018). Studies investigating the toxicological profile of MCB report mild endocrine disruptive effects on the thyroid gland and the potential to delay tissue growth and placenta formation during early pregnancy (HBM4EU 2020) Exposure to fragrances is widespread as they find applications in personal care products and scented non-cosmetic products. The main pathways of exposure are inhalation and dermal absorption. The most common health endpoint of fragrances is an allergic response, with some studies reporting endocrine disruptive effects (Dodson et al., 2012). Musk xylene (MX) and musk ketone (MK) are nitro musks, which have now been replaced in most countries due to their estrogenic activity and neural toxicity (Taylor et al., 2014) with polycyclic musks, such as celestolide (ADBI), traseolide (ATII), galaxolide (HHCB), and tonalide (AHTN). MX is classified under REACH as a chemical of high concern, while in the case of MK, a warning about its use has been placed (Taylor et al. 2014). Polycyclic musks are under scrutiny because of their persistence and adverse health effects, such as estrogenic activity. Along with musks, other synthetic fragrances of concern include lysmeral with the known metabolites lysmerol, lysmerylic acid, hydroxylysmerylic acid, tert-butyl hippuric acid, (TBHA), and tert-butyl benzoic acid (TBBA)  and 7-hydroxycitronellal (7-HC) with the main metabolite 7-hydroxycitronellic acid (7-HCA). Known health endpoints of 7-HC are skin irritation and skin sensitization (Stoeckelhuber et al. 2018), whereas endocrine disruptive effects are suspected (Scherer et al. 2021).
Frequently used antimicrobials in personal care products are isothiazole and its derivatives. The most extensively used among them are based on isothiazolinone. The knowledge of their potential adverse health effects has been emerging only recently. Methylchloroisothiazolinone (MCI) and methylisothiazolinone (MI) are often used to enhance the overall antimicrobial effect and find application in preserving solutions in personal care products and industrial products. Human exposure is mainly the result of dermal applications. Both MCI and MI undergo rapid biotransformation in the human body with a mean half-life of 3.6 h. The main biomarker of MCI/MI exposure is N-methylmalonamic acid (NMMA), whereas acetylamino([3-(methylamino)-1-(methylthio)-3-oxopropyl]thio)acetic acid (M-12) is a minor metabolite with an excretion fraction of between 10 and 23% in humans (Schettgen et al. 2021b). MCI/MI are broad-spectrum antimicrobials with high efficiency at trace concentrations (Silva et al. 2020). Over recent years, concerns over their sensitization potential and allergic dermatitis and other adverse health outcomes such as cytotoxicity have increased, leading to an interest in monitoring these compounds (Castanedo-Tardana and Zug 2013; Park and Seong 2020). MCI has a higher potential for the mentioned health endpoints (Silva et al. 2020).

Occurrence of selected CECs in urine
Most of the publications included in this review verified their respective method by implementation in HBM studies of varying scales, thereby confirming the occurrence of CECs in human urine ( Table 1). Most of the studied CECs (fragrances, BTR and BTH, antimicrobials, pyrrolidones, and diisocyanates) are frequently detected in urine, whereas the UV filter MBC was rarely detected, and if detected at all, it was in trace amounts (Ao et al. 2018b;Frederiksen et al. 2017;Leng and Gries 2017). The included non-phthalate plasticizers, on the other hand, are frequently detected and in relatively high concentrations, with a maximum concentration of 1165 ng/mL of OH-MEHTP (Bastiaensen et al. 2020). Therefore, it is clear that the presumably nonexposed general population is frequently exposed to most of the included CECs and highlights the need to include these compounds in routine HBM studies.

Sample preparation and analytical methods for the determination of CECs in urine
Once in the human body, many compounds undergo phase I and II metabolism involving esterases, lipases, and enzymes of the cytochrome P450 family (phase I) (Iyanagi 2007). Phase II metabolism occurs mainly in the liver, although these enzymes are expressed to a lower degree also in other tissues such as kidneys and intestine. The main phase II reactions are glucuronidation, sulfation, and glutathione conjugation (James 2021). In order to measure total CECs (free and conjugated), enzymatic deconjugation is usually applied, using enzymes like β-glucuronidase/arylsulfatase or acid/base hydrolysis (Glauser et al. 2014).
Sample clean-up and extraction procedures are needed to remove unwanted matrix constituents, reduce matrix effects, and eliminate interferences. Furthermore, during extraction, the sample is usually concentrated to detect low levels of biomarkers of exposure at concentrations significant for HBM. Solid phase extraction (SPE) is a commonly used method for preparing biological samples (Bocato et al. 2019). Compared to liquid-liquid extraction (LLE), it offers higher selectivity due to variously functionalized sorbents for the retention of analytes of choice with a wide range of physicochemical properties. The availability of SPE in 96-well format offers an adaptation to high-throughput sample preparation workflows, which is beneficial in HBM since large cohorts are analysed. However, it requires extensive parameter optimization.
Along with common extraction methods, other more specific extraction methods have been utilized, such as solidsupported liquid-liquid extraction (SSLE) on microporous diatomite cartridges to extract fragrances (Liu et al. 2015), ultrasound-assisted emulsification microextraction (USAEME) with tetrachloromethane for the extraction of polycyclic and nitro musks (Chen et al. 2018), and solidphase microextraction (SPME) for extraction of BTRs and BTHs using polyacrylate fibre (Naccarato et al. 2014).
After extraction, which often includes a preconcentration step, the most commonly applied separation methods include liquid chromatography (LC) or gas chromatography (GC), but other approaches have been also used (see below). Due to the high degree of molecular functionalization and suitable polarity, most analysts use reversed-phase LC. High and ultra-performance LC (HPLC and UPLC) do not require high volatility and thermal stability of analytes and are suitable for analysing polar matrices, such as urine. For volatile analytes like fragrances, GC is the method of choice. GC can also separate semi-and non-volatile compounds and is often accompanied by an appropriate derivatization step, such as silylation.
Regarding analysis, separation by LC and GC is most commonly combined with mass spectrometry (MS) analysers. MS enables increased sensitivity and selectivity, particularly when using tandem MS (MS2). Compared to MS, which is performed as a full scan or by following a single compound-specific ion (selected ion monitoring, SIM), MS2 follows multiple fragmentations (multiple reaction monitoring, MRM), which offers higher selectivity and specificity, which is especially important in the analysis of complex samples like urine. This specificity is reflected in the methods reported here, as CECs were detected almost exclusively in MRM mode. Besides MS, other approaches are suitable for detecting certain compounds. In one case, HPLC with fluorescence detection was used to detect biomarkers of   ). This required derivatization using 4-(1-pyrene)butanoyl chloride (PBC) ). Compared to the methods using negative chemical ionization (NCI) set-ups, such as GC NCI-SIM (Mirmohammadi et al. 2013), UPLC (Bhandari et al. 2016;Lépine et al. 2020Lépine et al. , 2019a, and nano-UPLC with MRM (Robbins et al. 2018) were described for the detection of MDA. This approach yields ten times lower LOQ (0.001 ng/mL) and does require any complex sample preparation procedures (offline-SPE). The suitability of the methods was tested following validation procedures, typically following validation guidelines, for example, Eurachem (Eurachem 2002) or the FDA Guidelines for Bioanalytical method validation (FDA 2001). As validation parameters, linearity, specificity, recovery, precision, and accuracy were tested. Precision and accuracy were always limited to at least 20%, but more often more stringent margins were used such as 15% (Asimakopoulos et al. 2013a;Stoeckelhuber et al. 2017) or even 10% (Bhandari et al. 2019;Höllerer et al. 2018b;Schettgen et al. 2017). Lépine et al. assessed the method's accuracy by comparing the results with the samples from an external quality control assessment, such as the German External Quality Assessment Scheme (G-EQUAS). In most studies, the method's LOD and LOQ were determined based on the signal-to-noise ratio, namely 3:1 and 10:1 signal-to-noise ratios for LOD and LOQ, respectively. An assessment based on 3-and 9-times standard deviation of the lowest limit of quantification (LLOQ) divided by the slope of the calibration curve was used in five studies (Asimakopoulos et al. 2013a;Höllerer et al. 2018aHöllerer et al. , 2018bLeón-González et al. 2013;Pinguet et al. 2019). The measurement uncertainty was assessed only once in a study published by Chen et al. (Chen et al. 2018).

Fragrances
For most fragrances, no deconjugation was performed before sample extraction and clean-up. The majority of nitro and polycyclic musks lack the functional groups for conjugation, whereas, for lysmerol, deconjugation was achieved enzymatically using β-glucuronidase. The reviewed methods for extracting fragrances from urine adopt an LLE approach using between 0.8 and 1 mL of sample volume. Due to the inherent volatility of fragrance compounds, they were mainly analysed using GC methods, except for lysmeral and 7-HCA, where both GC and LC-based methods have been applied. A derivatization step is required for these two compounds to improve their volatility, whereas polycyclic and nitro musks are sufficiently volatile to be analysed via GC without further adaptations. The two methods for detecting polycyclic musks in urine (Chen et al. 2018;Liu et al. 2015) yielded comparable results in terms of LOQ. However, regarding LOQs, the method of Liu et al. (2015) using an SSLE approach yields slightly better results (Table 2) compared to using USAEME, as suggested by Chen et al. (2018).
To determine lysmeral and 7-HC biomarkers of exposure, two UPLC-based methods using ESI-MRM detection are presented (Pluym et al. 2016;Stoeckelhuber et al. 2017). Separation was achieved on a 100-mm C18 analytical column with a binary-solvent mixture. To detect biomarkers of exposure of lysmeral, methanol (MeOH), and an aqueous solution of 5 mM ammonium acetate with 0.025% ammonium hydroxide at a pH of 9.2 were used as mobile phases (Pluym et al. 2016). The method for the determination of 7-HCA included 0.1% formic acid (FA) in acetonitrile (ACN) and aqueous 0.1% FA. Both methods report low LOQs for the studied compounds (≤ 0.5 ng/mL); however, internal standards have to be custom synthesized (Pluym et al. 2016;Stoeckelhuber et al. 2017), which limits the possibilities for many laboratories.

BTRs and BTHs
To determine benzotriazoles (BTRs) and benzothiazoles (BTHs), they are first deconjugated using β-glucuronidase and then extracted using LLE (Asimakopoulos et al. 2013a;Li et al. 2017;Zhou et al. 2018) or SPE (Asimakopoulos et al. 2013a;Li et al. 2018a) and online-SPE (Gries et al. 2015). Online-SPE requires the lowest sample volume (0.5 mL), whereas the presented LLE methods require between 0.6 and 1 mL of sample and the SPE methods between 1.3 and 2 mL. The most commonly used sorbent for SPE is mixed-mode Oasis HLB (Waters), and the separation method of choice is LC. The SPME method by Naccarato et al. (2014) yields the highest LOQs (0.4-4.9 ng/mL) of all the presented methods; however, it is also the only GC approach among the reported methods. Therefore, the high LOQ cannot be attributed to the sample preparation procedure alone, but also to the instrument of choice. The analytes were separated using a 30-m TR-5MS column and detected by MS in MRM mode. Although detecting BTRs and BTHs using GC is possible, the more complex sample preparation procedure and the relatively high LOQs than LC-based methods suggest that GC separation is less suitable for these compounds. The online-SPE method included only one compound, 2-mercapto benzothiazole (2SH-BTH) and achieved comparable LOQ with increased efficiency (time) and less sample handling compared to offline approaches (Gries et al. 2015). Despite other studies commercially obtaining isotopically labelled internal standards, the authors of this study custom synthesized the standards. The results do not seem to be impacted by this.
For BTRs and BTHs, four out of five methods (see Table 2) used a LC-ESI-MS2, but with different columns.  Table 2). The similarity in the separation and detection methods means it is most likely that the slight differences in achieved LOQs are attributed to the differences in sample preparation.

Antimicrobials
Only two methods describe the determination of biomarkers of exposure of MI/MCI exposure in urine using both LLE (Schettgen et al. 2017) and SPE (Schettgen et al. 2021b). The LLE approach includes an additional derivatization step with pentafluorobenzyl bromide to increase lipophilicity. This step is omitted in the SPE procedure (Schettgen et al. 2017). However, the SPE method requires 0.5 mL of urine, whereas LLE yields better results (LOQ 0.5 ng/mL for NMMA) with 0.1 mL of sample, which is particularly important when sample volume is limited. The SPE method achieves an equally low LOQ (0.2 ng/mL); however, the analyte, in this case, is a different biomarker of exposure, a mercapturic acid metabolite of MI/MCI, named M12, which means that a direct comparison of the methods is not possible. However, both methods achieve low LOQs (Table 2), require a small sample volume and do not require complex procedures or installations in the laboratory. Important to note is that for NMMA isotopically labelled standards are commercially available, while for M-12, they have to be custom synthesized.  11, 2, 7, 4, 9, 6, 11, 4, 14 Li et al. (2017)             To determine MI/MCI biomarkers of exposure M-12, Schettgen et al. (2021a) used a LC-ESI-MRM with a 150mm C18 column using a tertiary solvent system: water (pH 2.5, adjusted with FA), MeOH, and ACN. Together with SPE, the authors obtained a LOQ of 0.2 ng/mL, which is suitable for HBM studies (Table 2). Schettgen et al. (2017) presented a GC-based method with MRM detection for MI/ MCI biomarkers of exposure NMMA based on a HP-5-MS 60m column. After LLE and derivatization with pentafluorobenzyl bromide (PFBBr) the method achieved a LOQ of 0.5 ng/mL (Table 2).

Diisocyanates
In the case of diisocyanates, deconjugation is achieved using mineral acid hydrolysis. LLE is the most commonly applied method for extraction (Henriks-Eckerman et al. 2015;Lépine et al. 2019a;Mirmohammadi et al. 2013;Robbins et al. 2018) with SPE being used less often (Bhandari et al. 2016;Lépine et al. 2020;Sun et al., 2018) and cation exchange sorbents being the most used. Large differences in the required sample volume can be observed, as SPE methods require only a few hundred of microliters of the sample, while LLE requires larger volumes (1 and 2 mL). The exception is the method of Lépine et al. (2019a) where the authors use LLE with a volume of 250 μL. Only one method included a derivatization step with acetic anhydride (Bhandari et al. 2016;Lépine et al. 2019bLépine et al. , 2019a. By far, the lowest LOQ (0.001 ng/mL for MDA) was achieved by Sun et al. (2018) who applied SPE followed by derivatization with 4-(1-pyrene)butanoyl chloride. The data shows no difference in the LOQs obtained using LLE or SPE, albeit micro-SPE using a 96-well plate is more feasible for implementation in large-scale HBM schemes.
Most of the included LC methods involve UPLC, followed by HPLC, and nano-UPLC, with the C18 phase the most commonly used for separation. Three methods apply ESI-MRM detection (Lépine et al. 2020(Lépine et al. , 2019aRobbins et al. 2018), whereas Sun et al. (2018) applied fluorescence detection (FLD). Bhandari et al. (2016) used atmosphericpressure chemical ionization (APCI)(+)MRM detection. Three methods describe separation on a 100-mm or 150mm C18 column (Bhandari et al. 2016;Robbins et al. 2018;Sun et al. 2018), whereas two studies used a 50-mm TSS T3 column (Lépine et al. 2020(Lépine et al. , 2019a. Sun et al. (2018) achieved the lowest LOQ (0.001 ng/mL) for MDA, but did require a derivatization step, which is omitted by Bhandari et al. (2016), while achieving a higher but equally satisfactory low LOQ as well (0.03 ng/mL) for MDA ( Table 2). The method presented by Robbins et al. (2018) does not include MDA, but achieves a LOQ of 0.03 ng/mL for trisaminohexyl isocyanurate (TAHI). Internal standards for this compound are -in contrast to the other analytes presented here -not commercially available and have to be custom synthesized.
Henriks-Eckerman et al. (2015) and Mirmohammadi et al. (2013) used GC to determine MDA. Mirmohammadi et al. (2013) only briefly describe their method using negative chemical ionization (NCI) SIM detection and separation on a BP-5 column and do not report LOQ, whereas Henriks-Eckerman et al. (2015) achieved a LOQ of 0.1 ng/ mL using a 25 m HP-5 column for separation and NCI-MS-SIM detection. Despite being suitable for HBM, the latter method requires a higher sample volume (1 mL) and derivatization with HFBA, leading to higher LOQs (Table 2) than the LC-based methods.

Pyrrolidones
As a pre-treatment step, enzymatic deconjugation using β-glucuronidase is performed. The samples were either minimally prepared with the dilute-and-shoot approach using a 10-fold dilution with water (Bhandari et al. 2019;Haufroid et al. 2014;Suzuki et al. 2009) or extracted with SPE (Schindler et al. 2012. Based on the LOD/LOQ, required sample volume and overall sample preparation time, the dilute-andshoot approach seems to yield better results. Only Schindler et al. (2012) included a derivatization step with N-methyl-N-tert-butyldimethylsilyl trifluoroacetamide (MTBSTFA) in their sample preparation to adapt the compounds for GC analysis. Applying derivatization, however, did not lead to lower LOQ. Isotopically labelled standards are commercially available for these analytes.
Two methods are presented to determine NMP biomarkers of exposure using LC-ESI-MRM: Haufroid et al. (2014) and Suzuki et al. (2009) achieved compound separation on a 100-mm C18 column using aqueous 0.1% FA and ACN as mobile phase. The second method separated compounds on a 100-mm HSS-PFP column using 5 mM aqueous ammonium formate and MeOH as mobile phase (Bhandari et al. 2019). Schindler et al. (2012) present a method for determining NMP and NEP using cooled injection GC and separation on a 60-m DB-35 MS column. Their method achieves LOQs (5-20 ng/mL) significantly higher than LC-based methods (Bhandari et al. 2019;Haufroid et al. 2014;Suzuki et al. 2009). The LOQs obtained using this method are too high for HBM, where the analytes are often present in trace concentrations ( Table 1) and suggest that LC is a better approach for determining these compounds.

MBC
Following enzymatic deconjugation, the most commonly applied sample preparation method for methylbenzylidene camphor (MBC) is offline SPE (Ao et al., 2018b;León-González et al. 2013), followed by online 1 3 SPE Krause et al. 2017;Leng and Gries 2017). Janjua et al. (2008) used urine lyophilisation with reconstitution before analysis. Among those methods, online SPE requires the lowest sample volume (500 μL) for analysis, whereas other methods require a minimum of 2 mL. Additionally, the method presented by Leng and Gries (2017) yields the lowest LOQ (0.15 ng/ mL for CBC), whereas a LOQ of 0.9 ng/mL Krause et al. 2017) is the second-lowest achieved showing that online-SPE is the most suitable sample preparation method for determining these compounds. Only one study (Ao et al. 2018b) included a derivatization step with N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) for the instrumental analysis (GC), which did not increase the LOQ.
Four methods describe the detection of MBC and biomarkers of exposure of MBC using LC and, most commonly, a C18 phase. While most studies achieve separation on a 50-250-mm C18 column (Janjua et al. 2008;Leng and Gries 2017;León-González et al. 2013 and Krause et al. (2017) used a 50-mm Tur-boFlow Cyclone P and a 50-mm hypersil Gold aQ column. Classical ESI-MRM detection is used by León-González et al. (2013) and Leng and Gries (2017), whereas Janjua et al. (2008) applied UV detection, specifically SPD-6 UVD, while Frederiksen et al. (2017) and Krause et al. (2017) chose APCI-MRM. Different mobile phase compositions are presented in the included studies. A binary mixture of aqueous 0.1% FA and 0.1% FA in ACN/MeOH (1:1) is used by León-González et al., while Frederiksen et al. (2017) and Krause et al. (2017) use a tertiary system without specifying the solvents. Isocratic elution with MeOH/water (88:12) is used in one study (Janjua et al., 2008), whereas Leng and Gries (2017) do not specify their mobile phases. Krause et al. (2017) achieved by far the lowest LOQ for MBC (0.87 ng/mL), whereas the LOQs of the other studies range between 2.9 and 6 ng/mL for the same compound. Leng and Gries (2017) also achieved low LOQs, but for different biomarkers of exposure: CBC (0.15 ng/mL) and CBC-OH (0.3 ng/mL), which makes a comparison with other approaches difficult. Since Leng and Gries's (2017) method involves time-saving online extraction with minimal sample handling and no derivatization, this method appears the most efficient one for the determination of MBC. Ao et al.'s (2018a) method for detecting MBC involves derivatisation with BSTFA+ trimethylchlorosilane (TMCS) followed by GC (30 m with TSQ Quantum XLS Column), with MS (MRM) detection. The obtained LOQ (3.5 ng/mL) is higher than that (0.9 ng/mL) obtained by Frederiksen et al. (2017) and Krause et al. (2017), making GC slightly less advantageous.

Non-phthalate plasticizers
Following enzymatic deconjugation, extraction methods for biomarkers of exposure of TEHTM all involve SPE. The methods of Höllerer et al. (2018b) and Pinguet et al. (2019) include online-SPE, whereas Bastiaensen et al. (2020) and Been et al. (2019) describe an offline version. Only Pinguet et al. (2019) use 50 μL of the sample, whereas the other methods require 1 mL. In terms of LOQs, Pinguet et al.'s (2019) method gave the best results for mono-esters. In their method, Bastiaensen et al. (2020) and Been et al. (2019) determined di-esters at trace concentrations (LOQ 0.1 ng/ mL). Although Höllerer et al.'s (2018b) method had the highest LOQ, they also analysed the highest number of analytes (n = 11) of interest in HBM studies. From this work, it is clear that online-SPE achieves the best results for determining TEHTM metabolites, although the inclusion of too many compounds can increase the LOQ by decreasing sensitivity, and researchers have to compromise between the advantages and disadvantages of each method based on their research question. Labelled standards are commercially available for TEHTM biomarkers of exposure.
Three online-SPE methods and one offline SPE method are published for determining DEHA biomarkers of exposure. None of the presented methods require a derivatization step for LC analysis. The offline SPE procedure (Bastiaensen et al. 2020;Been et al. 2019) requires a much higher sample volume compared to the online-SPE methods (1 mL vs 50-300 μL); however, it yields the highest LOQ (0.15 ng/ mL for MEHA and OH-MEHA). The online-SPE methods by Nehring et al. (2019) Ringbeck et al. (2020) describe a method for the determination of DnBA biomarkers of exposure (MnBA, 3OH-MnBA, and cx-MnPrA) using online-SPE and report low LOQs (0.05-0.5 ng/mL). From this, it is evident that all of the presented methods are suitable for implementation in HBM studies, although Nehring et al.'s (2019) method has advantages in terms of preparation procedure (online), LOQ (comparable with others) and the number of analytes included (n = 3). While isotopically labelled internal standards are commercially available for DEHA transformation products, internal standards for DnBA biomarkers of exposure have to be custom synthesized.
For determining terephthalates, namely, DEHTP biomarkers of exposure, an online-SPE-based method (Pinguet et al. 2019) and an offline-SPE-based method (Bastiaensen et al. 2020;Been et al. 2019) were reported. The offline method requires a much larger sample (1 mL) than the online method (50 μL) and none of the procedures require 1 3 derivatization. Although there are significant differences in sample preparation, the online method requires minimal sample handling compared to offline SPE. The benefits of this method are visible in terms of achieving lower LOQs; the online-SPE-based method achieved LOQ for MEHTP of 0.018 ng/mL, whereas the offline method achieved a LOQ of 0.1 ng/mL for the same analyte.
Two studies determined DEHTP biomarkers of exposure. Pinguet et al. (2019) used a phenyl hexyl column and UFLC-ESI-MRM, while Bastiaensen et al. (2020) and Been et al. (2019) used a biphenyl column and HPLC-ESI-dMRM. Both studies use the same binary solvent system (aqueous 0.1% AA and 0.1% AA in ACN). Pinguet et al. (2019) achieved a lower LOQ for MEHTP (0.018 ng/mL), whereas 0.1 ng/mL was achieved by Bastiaensen et al. (2020) and Been et al. (2019). The latter also includes OH-MEHTP. A benefit to HBM of the UFLC approach is that it requires minimal sample (50 μL) and minimal sample handling (Pinguet et al. 2019).

Conclusions
In this review, we reviewed six groups of compounds that have been overlooked in existing HBM schemes but are good candidates from the recent increase in their detection in urine and reports concerning their potential toxicity. This review summarizes the emerging techniques and current state-ofthe-art analytical methods for inclusion in on-going HBM schemes. The methods were assessed in terms of their LOQs.
Looking at the analytical procedures, the most common first step in sample handling is enzymatic hydrolysis, which hydrolyses phase II metabolites to the parent compound or phase I metabolites. Accordingly, deconjugation is generally applied before analysis, except in the case of diisocyanates and nitro musks. Deconjugation was achieved using mineral acid hydrolysis for diisocyanates, while deconjugation was not applied for nitro musks.
Sample preparation varied widely within compound groups, but online-SPE stands out regarding the low sample volume required, limited sample handling, and achieved LOQs. However, the main drawback is the specific setup, which might not be feasible for every laboratory and offline SPE approaches were, therefore, the most commonly used. Despite being more time-consuming than online SPE and costlier than LLE, SPE offers high flexibility due to the variety of available SPE sorbents with broad polarity mixed-mode sorbents used most frequently, followed by ionexchange sorbents for diisocyanates and certain phthalate alternatives. New techniques, such as solid-supported liquid-liquid extraction and ultrasound-assisted emulsification microextraction, represent new directions to achieve even lower detection limits and low sample volumes needed for HBM studies, as sample amount is limited, particularly for vulnerable populations, such as infants and children.
Compounds were almost exclusively analysed using separation techniques (GC and LC) hyphenated to mass spectrometry. Although both GC and LC can separate many compounds, LC-based approaches achieved better results in terms of detection limits. Due to the broader range of compound polarities, most studies chose LC. UPLC methods, achieving higher chromatographic resolutions than conventional HPLC and reduced analytical times, are in rapid development to fill the analytical gap, for example, in the case of lysmeral, 7-HC, BTRs, BTHs, and MBC. Furthermore, an even faster method is being developed, for example, UFLC, which was used to separate non-phthalate pesticides. It is also apparent that mass spectrometric detection is the most favoured in trace-level analyses in HBM, with a triple quadrupole analyser dominating the scene, followed by quadrupole-linear ion-trap (QTrap) analysers. To date, HRMS is scarcely used in the targeted analysis of these compounds. The most common ionization method was ESI with compound-dependent polarity followed by APCI, with the MS being operated almost exclusively in MRM mode ensuring high specificity and selectivity, which is important when dealing with complex samples like urine.
Various columns were used for GC separation depending on the compound type. Most studies used low polarity 5% diphenyl/95% dimethylsiloxane-based columns or equivalent for fragrances, MI and MCI, diisocyanates, and MBC, achieving the best separation of these compounds. Other column phases, such as non-polar 5% phenyl polysilphenylene-siloxane and mid polarity 35% phenyl/methylpolysiloxane were used for BTRs and BTHs and pyrrolidones, respectively. Like in the case of LC-based methods, the compounds were commonly detected using tandem MS in MRM mode, using GC-specific EI ionization and, less often, negative chemical ionization. Also, in the case of GC separation, derivatization was often applied to achieve sufficient volatility. Silylation forming trimethylsilyl ethers was most commonly used for reviewed compounds, followed by the formation of pentafluorobenzoyl and heptafluorobutyrate ethers and esters.
Most of the methods achieved LOQs in a ng/mL and sub ng/mL range; however, as data on exposure of broader populations are still scarce, actual levels might be lower and, thus, quantification is hindered. To improve the existing methodology, there remains a need for even more sensitive, specific, and accurate analytical methods that aid in comprehensive exposure assessment and inform health risk assessment. A necessary step is the lowering of LOQs and thus improving the sensitivities of existing methods to assess the exposure even at pg/mL levels.
The methods must also be fit-for-purpose and should comply with strict QA/QC procedures and report method validation procedures, such as accuracy, reproducibility, linearity, and matrix effects. Due to the complex nature of urine as a matrix it is essential to prepare matrixmatched calibration. A calibration matrix would be, in an ideal case, prepared by pooling different urine matrices to assure an average matrix effect. As urine samples vary by water content so does the dilution of matrix and, hence, the matrix effects which must be considered. This can be compensated by using isotopically labelled internal standards, although these are not yet available commercially for some compounds (namely selected biomarkers of exposure for: lysmeral, 7-HC, MI and MCI, and HDI). The suitability of the methods should be rigorously tested, not only by internal quality measures but also by establishing robust quality control systems and participating in external quality control protocols, such as EQUAS. The overview of the methods here shows that only one method reported the measurement uncertainty, indicating a general lack of comprehensive assessment and reporting of measurement uncertainty with the associated results in the exposure studies. Every sample processing and analytical protocol introduces a certain amount of measurement uncertainty through several laboratory steps and as every method differs in specific validation parameters, such as achieved LOD/LOQ and accuracy, so does the measurement uncertainty. Therefore, it is of paramount importance that it is reported along with the study results. Future development of certified reference materials for here presented compounds and production of certified calibration standards would enable a basis for traceability of results, which is fundamental for their comparability, particularly across different populations, as studied by HBM.
With advances in HRMS, the specificity and selectivity of targeted analysis could be improved even further as HRMS enables global-scale non-targeted analysis which informs on known and yet unknown chemicals present in urine. Identifying possible new metabolites and biomarkers of effect could further aid in understanding toxicity and the health risks associated with exposure to emerging contaminants. This review highlights several analytical techniques available to analyze these six groups of chemicals and should be applied to various populations to thoroughly assess the distribution and occurrence of the emerging contaminants discussed in this study in a large population. In addition, to improve cost-and timeefficiency, which is of immense importance in large cohort HBM studies, one comprehensive multi-residue method would be applied in an ideal case. However, with analysis of such a heterogeneous group of chemicals, this is impossible and compound prioritization in compliance with the research question should be performed and the most appropriate method applied. Nevertheless, to improve the existing knowledge of potential health risks associated with the exposure to the chemicals presented here, more studies into their biological effects are needed, especially when present in a real-life scenario, simultaneously, as a mixture.