A validated reverse-phase LC-MS/MS method for the analysis of haloacetic acids in drinking water: supporting the transition from HAA5 to HAA9

Chlorination was first employed for disinfection and making water safer to drink in the 1890s, and since then its use has saved countless lives due to its ability to inactivate pathogenic organisms. However, chlorine also interacts with the organic matter present in water to form disinfection by-products (DBPs), with brominated as well as chlorinated DBPs formed when bromide is present in the water. DBPs were first identified in the 1970s when chloroform was detected in water supplies (Rook 1974), and since then many hundreds have been identified. Of these, the major classes by mass are trihalomethanes (THMs) and haloacetic acids (HAAs). DBPs are of concern as they have been identified as being potentially geno- and cyto-toxic (Richardson et al. 2007; National Toxicology Program (NTP) 2018) and have been associated with an increased incidence of bladder cancer in epidemiological studies (Evans et al. 2020). In response to this concern, DBPs are widely regulated, with levels for the four chlorinated and brominated THMs being limited to <100 μg/L in the United Kingdom.

More recently, attention has been turning to haloacetic acids (Table 1) and in particular to the brominated HAAs due to their enhanced toxicity relative to the chlorinated analogues (Richardson et al. 2007; Sawade et al. 2016). Currently, HAA levels are restricted to <60 μg/L for five HAAs (HAA5) in the United States, European Union (EU), and Scotland, but this excludes some of the brominated HAAs. Given the enhanced toxicity of the brominated HAAs, the monitoring and potential regulation of all nine chlorinated and brominated compounds (nine HAAs (HAA9)) has been discussed in both the EU and the United States (European Commission 2017; Samson & Seidel 2022) and is recommended by the World Health Organisation (WHO) (WHO 2022). Monitoring of all nine haloacetic acids is correspondingly likely to become more relevant as pressure on water resources results in the need to use lower quality source waters or to increase the use of desalinated water, resulting in higher levels of bromide and therefore increased formation of the more toxic brominated DBPs during chlorine disinfection (Chowdhury 2023).

Table 1

Haloacetic acids

Name .	Abbreviation .	CAS .	Structure .	HAA5 .
Monochloroacetic acid	MCAA	79-11-8		Y
Monobromoacetic acid	MBAA	79-08-3		Y
Dichloroacetic acid	DCAA	79-43-6		Y
Bromochloroacetic acid	BCAA	5589-96-8
Dibromoacetic acid	DBAA	631-64-1		Y
Trichloroacetic acid	TCAA	76-03-9		Y
Bromodichloroacetic acid	BDCAA	71133-14-7
Dibromochloroacetic acid	DBCAA	5278-95-5
Tribromoacetic acid	TBAA	75-96-7

Name .	Abbreviation .	CAS .	Structure .	HAA5 .
Monochloroacetic acid	MCAA	79-11-8		Y
Monobromoacetic acid	MBAA	79-08-3		Y
Dichloroacetic acid	DCAA	79-43-6		Y
Bromochloroacetic acid	BCAA	5589-96-8
Dibromoacetic acid	DBAA	631-64-1		Y
Trichloroacetic acid	TCAA	76-03-9		Y
Bromodichloroacetic acid	BDCAA	71133-14-7
Dibromochloroacetic acid	DBCAA	5278-95-5
Tribromoacetic acid	TBAA	75-96-7

View Large

Despite the evidence for the higher toxicity of the brominated HAAs, their current absence from many regulatory frameworks can be explained in part by the assumption that THMs and/or HAA5 can be used as a proxy for HAA9. A more detailed understanding of the formation of DBPs has identified that this assumption may not be valid, with evidence accumulating that although some correlation can be seen between THMs and HAA9 (Samson et al. 2017), this cannot be relied upon in all water types and treatment conditions (Malliarou et al. 2005; Kolb et al. 2017). In addition, their regulatory omission may be related to the analytical challenge they present, with many methods failing to detect all nine compounds adequately, and the non-availability of suitable analytical standards until relatively recently. Therefore, a substantial gap exists in the understanding of the prevalence of the unregulated brominated HAAs, with many studies excluding their analysis due to the difficulties of obtaining quality data. It is this absence of a simple and effective validated analytical method and the implications of that shortcoming which this article seeks to address.

The current US Environmental Protection Agency (EPA) methods for HAAs which have been validated for use are EPA 552.3 (Domino et al. 2003), which entails derivatisation by methylation, liquid–liquid extraction, and analysis by gas chromatography electron capture detection (GC-ECD) and EPA 557 (Zaffiro et al. 2009), which is a direct aqueous injection and analysis by ion chromatography tandem mass spectrometry (IC-MS/MS).

The GC-ECD method (EPA 552.3) is widely used but is time consuming in terms of both extraction (2 h derivatisation plus multiple solvent exchange steps) and analysis (run times up to 1 h). Furthermore, there is dependence on phase separation and derivatisation, both of which can be highly variable, and the low selectivity of the ECD can result in potential false positives. For example, the mono-halogenated HAAs have been seen to extract more efficiently in high ionic strength samples, which can result in incorrectly high quantification for these compounds (Domino et al. 2004). Conversely, the derivatisation efficiency for the larger brominated tri-halogenated HAAs (Br-TXAAs) has been identified as being relatively low, with only 45% efficiency for the tribromoacetic acid (TBAA) observed (Domino et al. 2004). These Br-TXAAs have also been observed to be unstable in the extract solution if peroxide impurities are present, lowering the results obtained (Xie et al. 2002). Furthermore, the lack of a specific detector may result in reagent impurities or matrix interference producing false-positive results, with known interferants identified for BCAA (from dimethyl sulphide) and MCAA (TAME impurities) (Domino et al. 2003).

The IC-MS/MS method (EPA 557) has been less widely adopted but overcomes many of these issues by removing the need for derivatisation and utilising a highly specific tandem mass spectrometer. However, the method still suffers from some drawbacks. In particular, common background anions, including chlorite, are known to interfere and so have to be carefully separated chromatographically, resulting in a relatively long run time (∼1 h) or the risk of high matrix suppression effects. In addition, the HAAs are not stable in the alkaline conditions of the mobile phase, and so a cooled column compartment is required to improve analyte stability (Wu et al. 2017; Bruzzoniti et al. 2019).

Due to the shortcomings of the existing US EPA methods, a wide range of alternative methodologies have been developed (Ma et al. 2024). Proposed analytical strategies range from capillary electrophoresis with ultraviolet detection, to GC-mass spectrometry. A variety of sample preparation techniques have also been put forward, from simple direct injection to complex extractions utilising solid phase cartridges (see Supplementary Material Table S1). Reverse-phase liquid chromatography tandem mass spectrometry (RP-LC-MS/MS) has the potential to provide a robust and effective analytical method for HAA9 (Duan et al. 2011; Alexandrou et al. 2019); however, it has hitherto not been as widely adopted. One reason for this is the lack of regulatory imperative to analyse for all nine HAAs, or indeed any HAAs in many countries. This issue is compounded by the current absence of a reverse-phase LC-MS/MS method being approved and validated by the regulatory authorities. Another reason for this omission may be attributed to the fact that the LC-MS/MS instrumentation itself has been perceived to be relatively expensive to purchase and operate, with some less well-resourced laboratories lacking the funds for such equipment. However, these instruments are becoming more affordable and are now relatively commonplace in drinking water testing laboratories in wealthier countries. This is largely due to their high performance and high throughput, and their analytical versatility for contaminants such as pesticides and perfluorinated compounds. Laboratories may also have experienced difficulties during the implementation of the RP-LC-MS/MS method, which requires a degree of technical expertise and the use of modern highly retentive columns due to the analytical issues around detecting such small, polar molecules by this technology. However, once these method development challenges have been overcome and the laboratory demonstrates acceptable performance through a full validation, it should be considered acceptable to regulatory bodies. It is also prudent to investigate the impact of the change in methodology and to provide evidence for an improvement over the status quo. However, no such validation has been published for a reverse-phase LC-MS/MS method for HAA9 nor has a direct comparison to the widely used GC-ECD method previously been undertaken.

This article presents a simple and effective method for the analysis of HAA9 by LC-MS/MS, which is demonstrably fit for purpose for the regulatory analysis of haloacetic acids in drinking water. The implications of changing to such a method from the current widely used GC-ECD method are also assessed by a direct comparison using real drinking water samples, with the hypothesis that the LC-MS/MS method will offer worthwhile improvements. Such a comparison enables the impact of changing methods to be demonstrated, as well as an assessment of the validity of past research. It is intended that the improved analytical method will enable better assessment of the levels of HAAs in drinking water, including the more toxic brominated species, and thus improve our understanding of their prevalence and the risk they pose. This will inform the debate around whether regulation of HAAs, and HAA9 in particular, could or should be adopted more widely.

HAA mixed standard solution (HAA9, each at 1,000 mg/L in methyl tert-butyl ether (MTBE)) was purchased from Restek (Centre County, PA, USA). HAA mixed standard solution (HAA9, each at 2,000 mg/L in MTBE) and Internal Standards 2,3-dibromopropionic acid solution (2,3-DBPA, at 1,000 mg/L in MTBE) and 1,2,3-trichloropropane (TCP; at 200 μg/ml in methanol) were purchased from Merck (Darmstadt, Germany). Methanol (LCMS grade), formic acid (LCMS grade), MTBE (HPLC grade), sodium sulphate (extra pure), sulphuric acid (concentrated, reagent grade), and ammonium chloride (reagent grade) were purchased from Fisher Scientific (Pittsburgh, USA). The mineral water used was Harrogate Spring Water (Harrogate, UK). Ultra-pure water with resistance 18.2 MΩ cm was provided by an Elga Purelab Ultra Genetic system.

Validation was performed using non-chlorinated, treated water from representative UK locations including upland (high organic), lowland, and mixed sources, together with chlorinated tap water from a groundwater source. Method comparison was performed using treated and chlorinated tap water from locations, which included ground and surface water sources.

A method for the analysis of HAA9 by reverse-phase LC-MS/MS was developed and validated in house based on the procedure by Duan et al. (2011). Optimisation of the method was complicated by the multiple masses and adducts formed by HAAs when analysed by LC-MS/MS, along with the practical requirement to use a mixed analytical standard for HAA9. Each compound has the potential to ionise as the molecular ion [M-H]⁻ but also as the [M-COOH]⁻ decarboxylated ion. Combined with the potential isotopes of ³⁵Cl, ³⁷Cl, ⁷⁹Br, and ⁸¹Br results in a large number of possible precursor (parent) ions and even more product (daughter) ions, with many of these masses overlapping between the target compounds. Consideration of the 9 HAAs gives 150 possible mass transitions (precursor to product ion), and of these, 31 occur for more than one compound (see Supplementary Material Table S6 for all likely mass transitions). Therefore, a cautious approach to the method development is necessary and was adopted (detailed in the Supplementary Material Figure S1 and Table S3). This enabled the HAAs to be separated chromatographically before the selection of unique qualifier and quantifier transitions for each compound, which were then optimised for maximum sensitivity and selectivity. The optimised method, as detailed here and in the Supplementary Material, was submitted to full validation as utilised by UK drinking water laboratories when demonstrating method performance for regulatory purposes. A comparison to the existing GC-ECD method was also performed.

Intermediate standards of HAA9 and 2,3-DBPA were prepared at 1 mg/L in ultra-pure water and stored in the dark in the fridge for a maximum of 2 months (as detailed in EPA 557). Calibration solutions were prepared fresh with each batch at concentrations from 0.5 to 100 μg/L in the matrix-matched solution. The matrix-matched solution was non-chlorinated potable water (bottled mineral water) containing ammonium chloride preservative and adjusted to a pH of 2 ± 0.5 in the same manner as a sample.

Samples were preserved immediately at collection with 100 mg/L ammonium chloride and stored in the fridge in the dark for up to two weeks. Samples were then adjusted to pH 2 ± 0.5 with 10% sulphuric acid. Internal standard 2,3-DBPA was added to samples and standards at 10 μg/L and the relative response used for quantification.

The instrument used was a SCIEX (Framingham, MA, USA) ExionLC ultra high-pressure liquid chromatography (UHPLC) system with temperature-controlled autosampler and column compartment and Qtrap 6500+ mass spectrometer system with electrospray ionisation source. The chromatography column used was a Waters (Milford, MA, USA) Acquity HSS T3 Premier 100 × 2.1 mm 1.8 μm with the associated Vanguard column. The mobile phase comprised mobile phase A (0.2 mM formic acid in water) and mobile phase B (0.2 mM formic acid in methanol). Initial conditions were 1% B for 1 min, followed by a gradient to 70% B at 5 min then holding for 2 min before re-equilibration, with a total run time of 10 min. The flow was 0.2 ml/min, injection volume was 20 μL, column temperature was 50 °C, and autosampler temperature was 7 °C.

The mass spectrometer was operated in the negative ionisation multiple reaction monitoring (MRM) mode. Nitrogen curtain gas (CUR) was set to 25 psi, ion source temperature (TEM) 350 °C, compressed air ion source gas 1 (GS1) 50 psi, ion source gas 2 (GS2) 60 psi, nitrogen collisionally activated dissociation gas (CAD) was ‘high’, and ion spray voltage (IS) −4,500 V. The instrument was controlled with SCIEX Analyst version 1.6.3 software, and the quantification was performed with SCIEX MultiQuant version 3.0.2. The acquisition parameters and observed retention times for individual compounds are summarised in Table 2, and further details are provided in the Supplementary Material Table S4.

The GC-ECD method used was a modified version of EPA 552.3 (Domino et al. 2003; Tung et al. 2006).

An intermediate standard solution of HAA9 in ultra-pure water was prepared at 200 μg/L and used to prepare a set of calibration standards, all 30 ml in ultra-pure water, at concentrations from 1 to 100 μg/L. These were prepared fresh for each batch and extracted as per samples.

Samples were preserved immediately at collection with 100 mg/L ammonium chloride and stored in the fridge for up to 2 weeks. A sample of 30 ml was acidified using 1.5 ml sulphuric acid before addition of 3 ml MTBE containing 1,000 μg/L TCP internal standard and approximately 12 g sodium sulphate. Samples were shaken vigorously for 3 min before allowing phase separation. One millilitre of the upper MTBE layer was combined with 1 ml 10% sulphuric acid in methanol in a test tube and placed in a water bath at 50 °C for 2 h. Once cooled, 1 ml MTBE and 3 ml 10% sodium sulphate (aqueous) were added and mixed before removing the lower aqueous layer and discarding. A further 1 ml of 10% sodium sulphate was added and mixed before taking the upper MTBE layer for analysis.

The instrument used was a Hewlett-Packard 6890 GC-ECD (Agilent, Santa-Clara, CA, USA). The chromatography column was a Restek (Centre County, PA, USA) Rtx-1 30 m, 250 μm diameter, 1 μm thickness. The injection volume was 1 μL, and the injector was run in the split mode at 200 °C 12.7 psi with a split ratio 10:1. The initial oven temperature was 40 °C with a hold time of 5 min followed by ramp at 4 °C/min to 100 °C, then a hold time for 1 min before ramping at 8 °C/min to 200 °C, and then a hold time of 3 min before returning to 40 °C for post-run time of 5 min. The carrier gas was helium with an initial pressure of 12.7 psi and a flow of 1.1 ml/min. The ECD was operated at a temperature of 270 °C with nitrogen at 30 ml/min. The instrument was controlled, and quantitation was performed with OpenLab CDS Chemstation version C.01.07 SR4 software.

The observed retention times for the individual compounds are provided in Supplementary Material Table S5. The relative response to the internal standard was used for quantification.

The LC-MS/MS method was validated in line with the recommendations of the UK Drinking Water Inspectorate (DWI) and ISO:17025 standards, and in compliance with NS30 and UK Drinking Water Testing Specification (DWTS) single laboratory validation with at least 10 degrees of freedom.

As such, 11 batches of duplicate samples were extracted and analysed separately, with each batch comprising neat, low spiked (1 μg/L) and high spiked (10 μg/L) samples from four water types as well as duplicate standards at 20 and 80% of the method range. Representative matrices were selected from a range of water types including lowland water, upland water, impounding reservoir water, and chlorinated groundwater.

The new LC-MS/MS method and the widely used existing GC-ECD method were directly compared using real world treated, chlorinated tap water selected from six different locations including ground and surface water sources. These samples were used directly and to prepare fortified samples by addition of 2 and 20 μg/L of each HAA, thus ensuring a range of levels and matrices were assessed. A total of 36 samples were prepared this way and then analysed by both methods simultaneously to enable direct method comparison.

Once validated, the ongoing quality of the results obtained was assured using appropriate quality control and assurance procedures. As such, a calibration set covering the reporting range of the method was prepared with each batch and analysed prior to any samples, with a mid-range check standard at least every 20 injections and at the end of the batch. Correlation was required to be r² > 0.995 and check standards within 15% of nominal. Samples outside of the calibrated range were not extrapolated but repeated with dilution. Blanks were prepared as if samples using ultra-pure water and determined to be <1/3 reporting limit. Spiked (fortified) samples containing 10 μg/L HAA9 were prepared for every matrix type and at least every 10 samples, with acceptable recoveries within 80–120%. Duplicate samples were prepared at least every 20 samples with results within 20% or 10 μg/L. Internal standard response was monitored and acceptance limited to within 60–140% of that obtained in calibration standards. The ratio between qualifier and quantifier transition for each analyte was monitored using the software, and failures were identified if the ratio was >20% different from that achieved in the calibration. Peak shape and resolution between 2,3-DBPA and BDCAA were monitored and used to identify deterioration in column performance and the need for corrective action (column clean or change).

The LC-MS/MS method was developed using the described protocol (see Supplementary Material) to ensure a method with good chromatography and spectrometry was achieved. This includes full resolution of BDCAA and 2,3-DBPA and two MRM transitions for quantitation and qualification of each analyte. Subsequent validation in drinking water matrices revealed that the performance exceeded standard requirements and thus can be deemed an appropriate method for HAA9 analysis in drinking water (Table 3).

Linearity was demonstrated across the range of 0.5–100 μg/L using matrix-matched calibration standards and duplicate analyses at 20 and 80% of the range. The correlation coefficient r² for the calibration curves was greater than the target of 0.995 and exceeded 0.999 in all cases. The precision at 20 and 80%, measured as 2× relative standard deviation, met the target of <25% in all cases, and the trueness, measured as bias, met the requirement of <± 25% and was <5% in all cases.

The LOD of the method was calculated as 3× standard deviation of non-chlorinated samples spiked at 1 μg/L. In all cases, the LOD was less than the target of 0.8 μg/L (<10% of the PCV).

The precision was calculated as 2× relative standard deviation of spiked samples at around the PCV and in all cases found to be significantly less than the target of 25%. The selectivity was demonstrated using two transitions for each compound, which gives a high degree of certainty due to the presence of separate precursor and product ions for each. In addition chromatographic resolution for overlapping transitions was required and in particular for 2,3-DBPA and BDCAA (see examples in Supplementary Material Figure S2). It was observed that no false positives were identified in known negative samples and that known positive samples demonstrated passing ion ratios in all cases.

Analysis of spiked samples across the four matrices demonstrated recoveries within the range 90–110%, which is well within the target of 75–125%.

Expanded method uncertainty was calculated based on a 95% confidence limit and found to be <15% for all analytes. This compares with the published US EPA methods, where the expanded uncertainty (calculated from the published validation data) was between 2 and 26% for EPA 552.3 and 4 and 20% for EPA 557 (data from Domino et al. 2003; Zaffiro et al. 2009).

Thirty-six samples were analysed in parallel by the validated LC-MS/MS method and the existing EPA 552.3 GC-ECD method, covering six water sources and including extracts of fortified samples from each water source. Observation of the time required for both extraction and analysis by each method, as well as the requirement for solvents and other reagent chemicals, demonstrated that the LC-MS/MS method is substantially quicker and involved lesser amounts of potentially hazardous and environmentally damaging chemicals. To enable better comparison of the practical considerations for each analytical option for HAA9, this assessment was quantified for a hypothetical testing laboratory performing 20 samples per week and extended to encompass the other alternative validated method, EPA 557 (Supplementary Material Table S2). This review confirmed the potential of the LC-MS/MS method to reduce laboratory sample preparation time and reagent use compared to GC-ECD, as well as making optimal use of mass spectrometry equipment when compared to EPA 557. However, it highlighted that capital and operational costs would be significantly higher for both LC-MS/MS and EPA 557 than for the GC-ECD method.

Assessment of the analytical results obtained from the real-world sample comparison shows that for unfortified waters where HAA levels are low (less than 20 μg/L), the two techniques demonstrate agreement across the different water types to within 3 μg/L in all bar one case (Table 4).

The difference in recovery was most pronounced for BDCAA, DBCAA, and TBAA, where observed recoveries by the LC-MS/MS method were 105.9, 107.0, and 107.2%, respectively, compared to 68.2, 65.3, and 67.8% by GC-ECD. Importantly, in the case of the GC-ECD method, for these compounds, the recovery values are less than the target of ±25% trueness (recovery 75–125%) set for validation, and therefore, from this dataset, they would be excluded from regulatory reporting. The previous research has identified that the derivatisation efficiency of brominated tri-halogenated acetic acids is lower than for the smaller HAAs (Domino et al. 2004), and these results suggest that the derivatisation efficiency is further adversely affected by the matrix, resulting in lower values when compared with the calibration standards (which are derivatised in ultra-pure water and therefore have a minimal matrix affect).

In addition, differences were observed with respect to MCAA, which reported a lower mean recovery of 76% and a large variance between samples when using the GC-ECD method compared to the LC-MS/MS method. Significantly, Sample 5 reported a concentration of 21.7 μg/L with the GC-ECD method, which was identified as a false positive caused by a co-eluting impurity. Issues with measuring MCAA have been previously reported whereby it has not been possible to report MCAA due to analytical issues (Malliarou et al. 2005) or where unexpectedly large values have been seen for MCAA in real water samples (Simpson & Hayes 1998; Avşar & Kılıç 2023). Given that the expectation is for a low formation potential of MCAA in most circumstances (thought to be due to the compounding effect of promotion of electrophilic substitution and low preference for hydrolysis from singly halogenated intermediates (Liang & Singer 2003; Weng et al. 2022)), it raises the likelihood that the GC-ECD method is sometimes leading to overestimation of MCAA levels, which requires further investigation.

In addition to the comparisons to the approved US EPA methods, our proposed method can be assessed against some of the alternative available methods. Efforts to overcome the obstacles presented by the EPA methods have resulted in a wide range of methods being published, many of which include the benefits of speed and accuracy such as offered by reverse-phase LC-MS/MS, but also have their own inherent downsides (Table S1).

GC-based methods have been developed, which improve selectivity by utilising mass spectrometry (Franco et al. 2019) or increase speed by using microextraction techniques (Wu Gabryelski & Froese 2002). However, although these do offer improvements, they do not overcome the inherent problem that haloacetic acids require derivatisation to be suitable for the GC analysis. Not only does this introduce a time and cost factor by requiring a sample preparation step but it also increases the risk of introducing a negative bias towards the brominated tri-halogenated HAAs.

LC-based methods remove the need for derivatisation, but in many cases, the sensitivity required means that sample preparation is needed to increase concentrations and remove interfering matrices. This increases the time for sample preparation and potential for losses, but can allow for more complex matrices to be analysed (Duan et al. 2013; Alsharaa et al. 2016). In contrast, the direct injection reverse-phase chromatography presented here would be unlikely to adapt to wastewater analysis due to the high susceptibility of LC-MS/MS to matrix suppression.

Direct analysis methods, such as direct injection liquid or ion chromatography, can reduce sample extraction errors and delays by minimising the sample preparation, but generally require expensive detection systems. Despite using high-end detectors, both LC and IC-MS/MS are prone to matrix interference by co-eluting organic matter and anions, respectively. This can be overcome by selective, fully optimised chromatography. For ion chromatography, this often entails using an alkali mobile phase and long-run times, which decreases the stability of the HAAs during analysis (Bruzzoniti et al. 2019). For the reverse phase, some methods have utilised ion pair agents to improve retention (Takino Daishima & Yamaguchi 2000). This enables better separation, but there are downsides, which include lower sensitivity due to suppression in the mass spectrometer and the potential for a long-term impact on mass spectrometer performance due to buildup of ion pair agents. Other alternatives include two-dimensional chromatography, which improves separation from background contamination but adds an extra layer of cost and complexity to the analysis (Wagner et al. 2017).

Therefore, reverse-phase chromatography using specialised polar columns, such as this method presents, remains a useful option in low organic matrices such as drinking water. Comparisons to other published direct injection reverse-phase LC-MS/MS methods confirm that it has the potential to efficiently provide accurate results, with Duan et al. (2011) demonstrating comparable method performance albeit over a much smaller sample size. However, it is observed that if chromatography is not properly optimised, this has implications for data quality (for example, an expanded uncertainty of up to 116% was obtained for MCAA under less optimal LC-MS/MS conditions (Planas et al. 2019)). Other useful applications of reverse-phase methods include utilising high-resolution mass spectrometry (HRMS), which has a distinct advantage over MS/MS in the ability to identify non-target DBPs in the same analytical run. However, they are equally if not more expensive and generally have lower sensitivity than the tandem mass spectrometers (Planas et al. 2019).

Further considerations when comparing methods include the ability to combine with other analyses if required. The method presented here could be expanded to include other, small, negatively charged target analytes potentially encompassing emerging contaminants such as the smaller perfluorinated acids, pesticide metabolites, or iodinated HAAs. However, because the analytical run is in negative ionisation mode and includes a relatively slow, aqueous gradient, it will not be readily combined with other drinking water LC-MS/MS testing suites (most pesticides and other contaminants will run faster and in positive mode). Also, unlike most HRMS, the data are purely targeted and cannot be re-visited retrospectively to investigate as yet unknown contaminants.

The results of the current study reveal that adopting the reverse-phase LC-MS/MS method outlined in this article improves measurement accuracy and reduces the risk of false positives for all the HAAs, not just MCAA. This is a result of the additional selectivity of the MS/MS detector when compared to the ECD due to the multiple mass transitions used and a further degree of certainty due to the good spike recoveries in real matrices. In addition, the reduction in time associated with sample preparation and analysis affords the potential for the inclusion of enhanced quality checks, such as spiked or duplicate samples within each analytical batch, imparting further confidence in the data obtained.

The adoption of this LC-MS/MS method will therefore have particular significance for lowland waters, which have traditionally been of lesser concern for DBP formation due to the relatively low THM levels observed (Whitaker et al. 2003). However, many such lowland sources have the possibility of containing significant levels of bromide, and therefore, the potential of such waters to form enhanced levels of the more toxic brominated HAAs exists. Moreover, HAA precursor organic matter molecules are often different from those associated with THM formation and can be more hydrophilic and thus recalcitrant to removal by coagulation and subsequent water treatment processes (Li et al. 2014). This further increases the likelihood that the DBP-associated risks of lowland waters may have been historically under-estimated when brominated HAAs were not considered (Lau et al. 2023). Similarly, the choice of analytical method will be significant if more complex water matrices are required to be used for water supply as climate change and environmental considerations impact water resource availability. For example, during indirect water re-use, treated chlorinated effluent may be discharged to the environment to replenish drinking water sources, and this can contain significant levels of HAAs (Munné et al. 2023). The impact of the higher levels of the matrix in such samples would increase the possibility of underestimation of the risk due to larger, brominated HAAs if the derivatisation-based GC-ECD method was used. Utilising LC-MS/MS would allow the ready use of spiked samples, affording additional confidence in the results and a better appreciation of the risks associated with such emerging practices.

As the water sector moves beyond THMs and embraces routine consideration of HAAs, it is posited that the inclusion of all nine HAAs is prudent to ensure as complete a picture as possible is considered when establishing management strategies to minimise the risk associated with DBPs. This is most apparent when we extend beyond mass concentration to also incorporate relative toxicity through the lens of hazard indices (Goslan et al. 2020 ). This reveals the relatively high importance of low concentrations of specific compounds, most notably the brominated haloacetic acids. Consequently, inclusion of all nine HAA species has the potential to modify our understanding of the most appropriate ways to minimise the risk associated with DBPs. Thus, it is posited that the routine measurement of all nine HAAs is the most appropriate course of action going forward. While equivalent arguments could be made for other emerging DBPs (e.g., nitrogenous DBPs, iodinated DBPs), it is contended that HAA measurement is starting to become routine already and that the required LC-MS/MS is becoming more commonplace. This reflects the utility of such an instrument not only for HAA analysis but also when measuring other micropollutants such as specific pesticides and per and polyfluorinated compounds. It is further proposed that additional utilisation can be obtained by expanding the method presented here to include polar, acidic compounds such as dalapon (2,2-dichloropropionic acid) or herbicide metabolites (Postigo et al. 2020).

In support of the advisability of controlling all nine haloacetic acids, this article proposed a reverse-phase LC-MS/MS method for their measurement and compared its efficacy to the currently used US EPA GC-ECD method. The LC-MS/MS method was fully validated across four representative drinking water matrices, and good sensitivity (<0.8 μg/L), precision (<7%), and bias (<10%) were observed for all nine HAAs. The comparison to GC-ECD demonstrated improved performance by the new method, particularly with regards to the brominated tri-halogenated HAAs, which appear to be under-reported by GC-ECD with low recoveries. Consequently, it is posited that adoption of the presented method is critical in establishing, with appropriate accuracy, the current prevalence, and then future approaches to minimising the risk posed by haloacetic acids.

The work was supported by UK Water Industry Research Ltd (UKWIR) and Engineering and Physical Sciences Research Council (EPSRC) through their funding of the Water Infrastructure and Resilience (WIRe) Centre for Doctoral Training [EP/5023666/1].

Polly Grundy: visualisation, investigation, methodology, formal analysis, data curation, validation, writing – original draft, and writing – review and editing. Emma Goslan: conceptualisation, supervision, funding acquisition, and writing – review and editing. Peter Jarvis: conceptualisation, supervision, funding acquisition, and writing – review and editing. Bruce Jefferson: conceptualisation, supervision, funding acquisition, and writing – review and editing. John Fawell: writing – review and editing. John Haley: supervision and writing – review and editing.

There are no conflicts to declare.

All relevant data are included in the paper or its Supplementary Information.