Urine analysis of 28 designer benzodiazepines by liquid chromatography–high-resolution mass spectrometry

Highlights

• •An LC-HRMS method covering 28 designer benzodiazepines was developed.
• •The method included screening in FS (gradient run 3.5 min).
• •Positive samples were confirmed using MS/MS (gradient run 4.5 min).
• •The method was applied to patient samples from emergency wards and routine drug testing.
• •Around 30% of the patient samples contained an analyte covered in the LC-HRMS method.

Abstract

Hundreds of new psychoactive substances (NPS) covering most drugs-of-abuse classes have been introduced to the recreational drug market in recent years. One class of NPS drugs that has become more common recently is “designer” benzodiazepines. Due to a close structural resemblance with prescription benzodiazepines, some of these substances may elicit a positive response (i.e. cross react) in immunoassay screening. Consequently, it is increasingly important to include NPS benzodiazepines during method confirmation to ensure accurate identification of closely-related compounds as well as detection of the benzodiazepines themselves.

Here, we present our efforts to develop a screening and confirmation method for detection of 28 NPS benzodiazepines in urine using reversed-phase liquid chromatographic separation in combination with high-resolution mass spectrometry (LC–HRMS). MS was performed in positive electrospray mode on a Thermo Fischer Scientific Q Exactive Orbitrap instrument using either full scan (for screening) or parallel reaction monitoring (for confirmation).

We found the lower quantification limit of the method to range from 5 to 50 ng/mL. Analytical precision and accuracy were ≤15% for both screening and confirmation for all except one analyte. The method was used to analyze patient urine samples from routine drug testing and samples from acute intoxication cases presenting in emergency wards. Altogether, 16 of the 28 benzodiazepines (i.e., clobazam, clonazolam, deschloroetizolam, diclazepam, estazolam, etizolam, flubromazepam, flubromazolam, flunitrazolam, 3-hydroxyflubromazepam, 3-hydroxyphenazepam, ketazolam, meclonazepam, metizolam, nifoxipam, and pyrazolam) were detected in the urine samples.

The results from patient sample analysis indicate a high prevalence of NPS benzodiazepine use, emphasizing the importance of including novel drugs of abuse in drug testing menus.

1. Introduction

In recent years, hundreds of new psychoactive substances (NPS) have been introduced to the recreational drug market [1]. NPS are typically variants of classical drugs of abuse that have been structurally modified (i.e., “designed”) to circumvent legislative control and allow open sales online [2]. All common drug classes, including cannabinoids, stimulants, tryptamines, opioids and benzodiazepines, have experienced this phenomenon [2], [3]. Because NPS differ in chemical structure from their classical counterparts, intake may go undetected if standard drug testing by immunoassay screening is used. Additionally, NPS experience a high structural turnover on the market [3], requiring analytical methods be updated continuously to address the latest developments. Liquid chromatographic separation with mass spectrometric detection (LC–MS) provides an analytical methodology that allows the flexibility to reasonably maintain a relevant drug testing menu that will address the rapidly changing landscape of NPS [4].

In the Swedish STRIDA project on NPS [2], one drug class that has become more regularly used is benzodiazepines. Benzodiazepines are prescription drugs for treating anxiety, for example, but are also commonly abused due to their propensity to cause dependence, tolerance and addiction [5]. Of the benzodiazepines introduced as NPS on the EU drug market, some are already registered medicines outside Europe, whereas others are novel, designed variants [6]. The first NPS benzodiazepine of the latter category was pyrazolam, which appeared for sale online in 2012 [6]; it has since been followed by many more [7], [8], [9], [10], [11].

Due to structural similarity, and subsequent antibody cross-reactivity, NPS benzodiazepines are often detectable with immunoassays that target the standard set of prescription benzodiazepines [12], [13]. Accordingly, to avoid “false positive” screening results, confirmation methods should also include these NPS. Recently published methods for NPS benzodiazepines have used e.g. MS detection coupled to ultra-high performance LC, capillary electrophoretic, or gas chromatographic separation [14], [15], [16], [17]. However, an additional analytical problem with the NPS benzodiazepines is that some of these are precursors to, or metabolites of, prescription benzodiazepines [10], [18], which places high demands on the final interpretation of test results.

Here, we have undertaken to develop an LC–high-resolution MS (LC–HRMS) multi-component method for determination of designer benzodiazepines in urine, using direct dilution as the only sample preparation. This offers a sensitive, selective and flexible method and enables rapid inclusion of new analytes. The method is based on a published method for screening (using LC–HRMS) and confirmation (using LC–HRMS/MS) of several different NPS classes [19]. The benzodiazepines were selected for this study based on (i) frequency of mention in drug discussion forums, (ii) presence in the online marketplace, (iii) notification by the European Monitoring Centre for Drugs and Drug Addiction (EMCDDA), and (iv) not being prescribed in Sweden (i.e. NPS, deregistered pharmaceuticals, or pharmaceuticals approved in other countries) [6], [7], [8], [9], [10], [11], [20], [21], [22], [23], [24], [25], [26]. Method performance was evaluated on patient urine samples from routine drug testing and acute intoxication cases suspected to involve NPS benzodiazepines (STRIDA project cases).

2. Materials and methods

2.1. Chemicals

Adinazolam, bentazepam, clonazolam, cloniprazepam, diclazepam, deschloroetizolam, flubromazolam, flunitrazolam, 3-hydroxyflubromazepam, ketazolam, meclonazepam, metizolam, nifoxipam, nitrazolam, pivoxazepam and pyrazolam were obtained as 1 mg/mL standard solutions from Chiron (Trondheim, Norway). Bromazepam, clobazam, etizolam, estazolam, flurazepam, α-hydroxytriazolam-d4 (internal standard, IS), 3-hydroxyphenazepam, lorazepam-d4 (IS), N-desmethylflunitrazepam, nimetazepam, phenazepam and prazepam were obtained as 1 mg/mL solutions, and α-hydroxyalprazolam-d5, estazolam-d5, flunitrazepam-d7, nordiazepam-d5, oxazepam-d5 and temazepam-d5 (all IS) as 0.1 mg/mL solutions, from Cerilliant (Round Rock, TX, USA). Flubromazepam was obtained as a 1 mg/mL solution and tetrazepam as a 0.1 mg/mL solution from LGC (Teddington, UK). All solutions were obtained in acetonitrile or methanol.

Ammonium formate (LC–MS ultra grade), E. coli β-glucuronidase solution (80 U/mL at 25 °C), ammonia solution (25%), acetic acid (pro analysis grade) and formic acid (MS grade) were from Sigma–Aldrich (St. Louis, MO, USA), acetonitrile (HPLC gradient grade) from VWR International (Radnor, PA, USA), and ammonium acetate (≥98% purity) and methanol (LC–MS grade) from Thermo Fisher Scientific (Waltham, MA, USA). Ultra-pure water (>18 MΩ/cm) was prepared in house using an Integral 3 Milli-Q water system (Merck, Darmstadt, Germany).

2.2. Urine samples

Blank urines were obtained from laboratory staff members who self-reported that they had not taken benzodiazepines, and the samples were confirmed to be negative using the CEDIA benzodiazepine immunoassay. Patient urine samples were surplus aliquots collected from the routine drug testing service at the Department of Clinical Pharmacology, Karolinska University Laboratory (Stockholm, Sweden). These samples were collected consecutively from August 2015 until February 2017. The samples were stored for a maximum of 4 weeks at 4 °C and, thereafter, kept frozen at −20 °C. Patient samples were also obtained from the STRIDA project on NPS [2] (collected in 2012–2016) and these were stored at −20 °C upon arrival in the laboratory.

Ethical approval (No. 00-230 and No. 2013/116-31/2) was obtained from the regional ethical review board.

2.3. Preparation of solutions

2.3.1. Standard and quality control solutions

Four different stock solutions of benzodiazepines (0.1 mg/mL of each substance) were prepared in methanol. Solution A contained diclazepam, estazolam, etizolam, flubromazepam, meclonazepam, phenazepam and pyrazolam; solution B contained bentazepam, bromazepam, clonazolam, deschloroetizolam, flubromazolam, 3-hydroxyphenazepam, and nifoxipam; solution C contained adinazolam, cloniprazepam, fonazepam/N-desmethylflunitrazepam, 3-hydroxyflubromazepam, metizolam, nitrazolam, and pivoxazepam; and solution D contained clobazam, flunitrazolam, flurazepam, ketazolam, nimetazepam, and prazepam.

Urine standards were prepared at 1000 ng/mL in two different sets by diluting either solutions A and B (standard C1) or solutions C, D and 0.1 mg/mL tetrazepam (standard C2) with blank urine. These solutions were then further serially diluted with blank urine to obtain C1 and C2 standards with concentrations of 700, 300, 100, 50, 10, 5 and 1 ng/mL. In addition, single analyte standard solutions were prepared to study selectivity. These single analyte solutions were prepared in blank urine to a final concentration of 500 ng/mL.

Quality control (QC) solutions at 900, 300, 75, 50, 10 and 5 ng/mL were prepared similarly from separate stock solutions. Solutions A, B, C and D were stored at −20 °C and urine standards and QC at −80 °C.

Solutions used for studying stability and matrix effects were prepared in 6 different blank urines as described above. Additional solutions were prepared in 10 mmol/L ammonium acetate (pH 6.0) and used for comparison in the matrix experiments.

2.3.2. Internal standard working solution

The IS working solution containing 100 ng/mL each of α-hydroxyalprazolam-d5, α-hydroxytriazolam-d4, estazolam-d5, flunitrazepam-d7, lorazepam-d4, nordiazepam-d5, oxazepam-d5 and temazepam-d5 was prepared in 10 mmol/L ammonium acetate (pH 6.0) and stored at 4 °C for a maximum of 2 months.

2.4. Sample preparation

Samples were prepared in 96-well polypropylene plates (Waters, Milford, MA, USA) by adding 170 µL IS working solution, 30 µL β-glucuronidase solution, and 50 µL urine sample to each well. The plate was then gently shaking for 30 s, placed at room temperature (20 °C) for 20 min, and, then, directly transferred to the autosampler followed by injection of samples into the LC–HRMS system.

2.5. LC–HRMS system

The LC–HRMS system was comprised of a Dionex Ultimate UHPLC System coupled to a Q Exactive MS equipped with a heated electrospray ionization (HESI)-II source (Thermo Fisher Scientific, Waltham, MA, USA). The analytical column was a YMC UltraHT Hydrosphere C18 column (2.0 µm, 100 mm × 2.0 mm) combined with a Hydrosphere C18 guard column, both from YMC Europe GmbH (Dinslaken, Germany). The autosampler temperature was 12 °C, the column temperature 60 °C, and the injection volume 2 µL. Mobile phase A was composed of 10 mmol/L aqueous ammonium formate with 0.005% formic acid (pH 4.8) and mobile phase B of 10 mmol/L ammonium formate in Milli-Q water and methanol (10:90, v:v) with 0.005% formic acid. The flow rate was 0.5 mL/min. A mixture of formic acid, acetonitrile, and Milli-Q water (5:900:95, v:v:v) was used for needle wash, and methanol and Milli-Q water (10:90, v:v) for seal wash.

The MS operated in positive electrospray ionization mode with the following settings: spray voltage 3 kV, capillary temperature 300 °C, sheath gas 60 AU, auxiliary gas 18 AU, heater temperature 450 °C, and S-Lens RF level 70.

The software used for data evaluation was Trace Finder Clinical Research 4.1 (Thermo Fisher Scientific).

2.5.1. Screening method

The screening method used the following gradient: 0–0.50 min, 4% B; 0.51–2.45 min, ramp from 50% to 95% B; 2.45–2.50 min, 95% B; 2.51–3.50 min, 4% B. The exact masses monitored are shown in Table 1. The MS settings were as follows: scanning m/z 100–650, 70,000 resolution, AGC target 1 × 10⁶, maximum IT 100 ms, and microscans 1. Identification criteria for the patient samples included a retention time of ±0.05 min from the standards and a mass of ±10 ppm from the exact mass of the analyte.

Table 1.

Chemical attributes (molecular formula, exact masses) and settings utilized in the screening and confirmation methods (MS/MS transitions, normalized collision energy (NCE) of the monitored MS/MS transitions, monitoring time in each parallel reaction monitoring method (PRM)).

Analyte	Molecular formula	Exact mass (M)	Exact mass (M+H +)	Quantifier transition (m/z)	Qualifier transition (m/z)	NCE (V)	Time segment (min)
α-Hydroxyalprazolam-d5	C17H82H5ClN4O	329.1086	330.1165	–	–	–	–
α-Hydroxytriazolam-d4	C17H82H4Cl2N4O	362.0634	363.0712	–	–	–	–
Adinazolam	C19H18ClN5	351.1245	352.1324	352.1324 > 58.0658	352.1324 > 295.0745	50	2.77–3.17a
Bentazepam	C17H16N2OS	296.0978	297.1056	297.1056 > 166.0677	297.1056 > 269.1094	51	2.90–4.40c
Bromazepam	C14H10BrN3O	315.0002	316.0080	316.0080 > 182.0839	316.0080 > 209.0948	51	2.00–2.60b
Clobazam	C16H13ClN2O2	300.0660	301.0738	301.0738 > 259.0633	301.0738 > 224.0935	52	2.34–2.74a
Clonazolam	C17H12ClN5O2	353.0674	354.0752	354.0752 > 326.0563	354.0752 > 319.1064	40	2.00–2.60c
Cloniprazepam	C19H16ClN3O3	369.0875	370.0953	370.0953 > 316.0484	370.0953 > 302.0453	46	3.00–4.40a
Deschloroetizolam	C17H16N4S	308.1090	309.1168	309.1168 > 255.0951	309.1168 > 276.1367	41	2.50–3.10b
Diclazepam	C16H12Cl2N2O	318.0321	319.0399	319.0399 > 227.0496	319.0399 > 154.0418	55	2.70–4.40b
Estazolam	C16H11ClN4	294.0667	295.0745	295.0745 > 205.0760	295.0745 > 138.0106	60	2.20–2.80b
Estazolam-d5	C16H62H5ClN4	299.0981	300.1059	300.1059 > 272.0872		60	2.20–2.80b,c, 2.31–2.71a
Etizolam	C17H15ClN4S	342.0701	343.0779	343.0779 > 314.0388	343.0779 > 259.0216	46	2.50–3.10b
Flubromazepam	C15H10BrFN2O	331.9955	333.0033	333.0033 > 226.0889	333.0033 > 183.9751	55	2.50–3.10c
Flubromazolam	C17H12BrFN4	370.0224	371.0302	371.0302 > 343.0096	371.0302 > 292.1105	52	2.30–2.90c
Flunitrazepam-d7	C16H52H7FN3O3	320.1297	321.1375	321.1375 > 307.1344		50	0.70–4.40b, 2.10–2.70c, 2.23–2.63a
Flunitrazolam	C17H12FN5O2	337.0970	338.1048	338.1048 > 338.1048	338.1048 > 310.0861	40	1.95–2.35a
Flurazepam	C21H23ClFN3O	387.1508	388.1586	388.1586 > 315.0695	388.1586 > 317.0852	30	0.70–2.20a
3-Hydroxyflubromazepam	C15H10BrFN2O2	347.9904	348.9982	348.9982 > 302.9928	348.9982 > 330.9877	30	2.31–2.71a
3-Hydroxyphenazepam	C15H10BrClN2O2	363.9609	364.9687	364.9687 > 318.9632	364.9687 > 273.0022	51	2.35–2.95b
Ketazolam	C20H17ClN2O3	368.0922	369.1001	–	–	–	–
Ketazolam fragment	C16H13ClN2O	284.0711	285.0789	285.0780 > 193.0886	285.0780 > 257.0840	50	2.93–3.33a
Meclonazepam	C16H12ClN3O3	329.0562	330.0640	330.0640 > 316.0594	330.0640 > 285.0423	42	2.40–3.00c
Metizolam	C16H13ClN4S	328.0544	329.0622	329.0622 > 275.0386	329.0622 > 296.0823	40	2.38–2.78a
N-Desmethylflunitrazepam/Fonazepam	C15H10FN3O3	299.0701	300.0779	300.0779 > 198.0714	300.0779 > 225.0815	60	2.07–2.47a
Nifoxipam	C15H10FN3O4	315.0650	316.0728	316.0728 > 298.0623	316.0728 > 270.0673	26	0.70–2.30b
Nimetazepam	C16H13N3O3	295.0951	296.1030	296.1030 > 221.1073	296.1030 > 268.1081	48	2.33–2.73a
Nitrazolam	C17H13N5O2	319.1064	320.1142	320.1142 > 292.0955	320.1142 > 198.0900	48	1.98–2.38a
Nordiazepam-d5	C15H62H5ClN2O	275.0868	276.0947	276.0947 > 213.1309		50	2.71–3.11a
Oxazepam-d5	C15H62H5ClN2O2	291.0817	292.0896	292.0896 > 246.0841		40	2.38–2.78a
Phenazepam	C15H10BrClN2O	347.9660	348.9738	348.9738 > 183.9751	348.9738 > 242.0593	47	2.60–3.20c
Pivoxazepam	C20H19ClN2O3	370.1079	371.1157	371.1157 > 269.0476	371.1157 > 241.0524	15	3.00–4.40a
Prazepam	C19H17ClN2O	324.1024	325.1102	325.1102 > 271.0633	325.1102 > 208.0995	52	3.00–4.40a
Pyrazolam	C16H12BrN5	353.0271	354.0349	354.0349 > 167.0722	354.0349 > 206.0835	52	0.70–2.35c
Temazepam-d5	C16H82H5ClN2O2	305.0974	306.1052	306.1052 > 288.0947		40	2.52–2.92a, 2.40–3.00b,c
Tetrazepam	C16H17ClN2O	288.1024	289.1102	289.1102 > 169.0887	289.1102 > 117.0576	68	3.00–4.40a

^aPRM method 1.

^bPRM method 2.

^cPRM method 3.

2.5.2. Confirmation method

The confirmation method used the following gradient: 0–0.50 min, 4% B; 0.51–3.45 min, ramp from 50% to 95% B; 3.45–3.50 min, 95% B; 3.51–4.50 min, 4% B. The MS was operated in parallel reaction monitoring (PRM) mode with 17,500 resolution, AGC target 2 × 10⁵, maximum IT 130 ms, microscans 1, and isolation window 1.5 m/z. Three different PRM methods were used (same general settings, but different PRM inclusion lists). The analytes found in each method, their normalized collision energy (NCE), and the monitored transitions and transition times are shown in Table 1. Identification criteria for the patient samples included a retention time of ±0.05 min from the standards, a mass value of ±10 ppm from the expected mass (Table 1), and a correct peak area ion ratio according to Maralikova and Weinmann [27].

2.6. Method validation

2.6.1. Precision, accuracy and limit of detection

The precision and accuracy were determined over five different days by analyzing the QC samples in triplicate. The lower limit of quantification (LLOQ) was defined as having an accuracy of 80–120% and a coefficient of variation (CV) ≤20%. The limit of detection (LOD) was determined from quintuplicate analysis of samples over five days. The LOD of the screening method was defined as the spiked concentration at which ≥75% of the investigated samples had an acceptable peak shape. The LOD of the confirmation method was defined as the concentration at which ≥75% of the investigated samples had a correct ion ratio.

2.6.2. Matrix effects

Matrix effects were determined at the LLOQ and at 900 ng/mL by comparing the mean area of duplicates of six different blank urines to the mean area of duplicates in 10 mmol/L ammonium acetate (pH 6.0), according to Matuszewski et al [28]:

$$\mathit{Matrix}effect=\left(\left(\frac{\mathit{mean}totalareaofanalyteinurine}{\mathit{mean}totalareaofanalyteinbuffer}\right)\times 100\right)-100$$

2.6.3. Selectivity

Selectivity was studied using six different blank urines and 500 ng/mL urine solutions of each analyte. Possible interference from >100 other NPS [19] was investigated by analyzing urine solutions spiked with a range of analytes included in a substance abuse monitoring program at the Karolinska University Laboratory. Furthermore, the ability to accurately detect each analyte in the presence of a number of interferents was investigated by spiking urine with 100,000 ng/mL ethyl glucuronide, 30,000 ng/mL each of morphine-3-glucuronide, morphine, codeine-6-glucuronide, codeine, and ethylmorphine-6-glucuronide, 20,000 ng/mL ethyl sulfate, 10,000 ng/mL of α-hydroxyalprazolam, α-hydroxymidazolam, α-hydroxytriazolam, 7-aminoclonazepam, 7-aminoflunitrazepam, 7-aminonitrazepam, desmethyldiazepam, diazepam, lorazepam, oxazepam, and temazepam), 8000 ng/mL of amphetamine, methamphetamine, MDMA, and MDA, 1000 ng/mL buprenorphine, 750 ng/mL 6-acetylmorphine, and 450 ng/mL benzoylecgonine.

2.6.4. Stability

The stability of the analytes in methanol stock solutions A, B, C and D at −20 °C was determined by diluting both stored and freshly prepared methanol solutions with 10 mmol/L ammonium acetate (pH 6.0) to concentrations of 900 ng/mL, as described in Section 2.3.1. The solutions were then analyzed in triplicate, and the relative difference in analyte/IS area ratio of the stored stock solutions was compared to freshly prepared solutions. Analytes were considered to have retained stability during storage if the stored methanol stock solutions differed by ≤20% from the freshly prepared solutions.

The stability of the analytes in a urine matrix was investigated in triplicate at the LLOQ and at 700 ng/mL using six different blank urines, by comparison against a standard curve of the corresponding blank urine. Stability was studied at room temperature (∼20 °C), 4 °C, and −80 °C. Stability was also examined during three freeze/thaw cycles at −80 °C, and for prepared samples stored at 12 °C in the autosampler. Analytes were considered to have retained stability if the mean concentrations of the six spiked blank urines were within ±20% of the spiked value at both the LLOQ and 700 ng/mL.

2.6.5. Carry-over

Carry-over was examined for the screening and the confirmation methods by injection of a blank urine immediately after injecting the highest standards (1000 ng/mL).

2.7. Application of the method

The method was applied to the analysis of 126 routine patient urine samples that tested positively in the CEDIA benzodiazepine immunoassay, but were confirmed negative for the standard set of prescription benzodiazepines and metabolites evaluated by the in-house LC–MS/MS confirmation method. The analytes included in the in-house method were α-hydroxyalprazolam, α-hydroxymidazolam, α-hydroxytriazolam, 7-aminoclonazepam, 7-aminoflunitrazepam, 7-aminonitrazepam, desmethyldiazepam, lorazepam, oxazepam, and temazepam. Additionally, 329 patient urine samples from intoxication cases were obtained from the STRIDA project [2] and analyzed.

3. Results and discussion

3.1. Method validation

3.1.1. Precision, accuracy and limit of detection

The LOD (1–50 ng/mL) and measuring ranges (5–1000 ng/mL) of the screening and confirmation methods are shown in Supplementary material 1. A screening method chromatogram displaying all analytes included in the study is presented in Fig. 1. The measuring range of the confirmation method was similar to that of the screening method, but three analytes (i.e., adinazolam, meclonazepam and nimetazepam) showed a higher LLOQ in the confirmation method. The lower sensitivity of these analytes in the confirmation method was due to the signal intensity of the qualifier ion. The measuring ranges obtained for both the screening and confirmation methods corresponded to the substance concentrations previously observed in patient urine samples [29].

Fig. 1.

Chromatograms showing the retention time (RT) of all NPS benzodiazepines and internal standards included in the method in a screening run with the highest standard (i.e., 1000 ng/mL). The standards were prepared in two sets (left and right). Ketazolam is shown as both the M+H ⁺ ion (RT 2.67) and its in-source fragment at m/z 285.0789 (RT 2.66).

The screening method was less sensitive for ketazolam (i.e., LOD > 50 ng/mL) which was assumed to be due to in-source fragmentation into an ion with m/z 285.0789 (i.e., loss of C₄H₄O₂). This supposition was supported by difficulties encountered while optimizing the NCE for this analyte in the confirmation method. Consequently, the m/z 285.0789 ion (herein named “ketazolam fragment”) was monitored together with ketazolam in the screening method and used as the parent ion for the transitions monitoring ketazolam intake in the confirmation method. However, 285.0789 is also the exact mass of the M+H⁺ ion of diazepam, which has been demonstrated to be a degradation product of ketazolam when analyzed by LC–MS with thermospray ionization [30], and on storage of ketazolam in methanol [31]. Therefore, a patient sample was not considered positive for ketazolam, unless both ketazolam and the fragment ion were detected in the screening method and the fragment in the confirmation method.

The parent M+H⁺ ion of flunitrazolam was chosen as the quantifier ion in the confirmation method together with the fragment of m/z 310.0861 (qualifier), since using other fragments of flunitrazolam resulted in an ion ratio dependent on the sample concentration. This was eliminated by using the parent mass as quantifier.

The standard curves used in the accuracy and precision experiments comprised 5–7 concentration levels and all coefficients of determination (R²-values) were >0.98. The total CV of the screening (Supplementary material 2) and confirmation (Supplementary material 3) method was ≤15% at all investigated concentrations for all substances. Furthermore, the accuracy was within ±15% at concentrations above the LLOQ and within ±20% at the LLOQ for all analytes, except pivoxazepam. However, in the screening method, the accuracy for pivoxazepam was within ±20% and the total CV was also <15%.

3.1.2. Selectivity

Background peak areas from the six different blank urines were <20% of the peak areas for all analytes at the LLOQ levels, except for flurazepam and tetrazepam where the responses were ≤44%. The background interference for flurazepam was <20% in the confirmation method, whereas the interference for tetrazepam remained at a high level. However, because the response was always less than the LLOQ area, the interference was considered a minor problem.

Metizolam gave a response at the MS trace used for meclonazepam, but they were separated chromatographically (Fig. 1). Flurazepam gave a response for nifoxipam and these were not well separated in the chromatographic systems. The final identity, therefore, had to be based on the ion ratio. Because of this, metizolam and flurazepam were separated from meclonazepam and nifoxipam in the urine standards.

Unfortunately, diazepam was not separated from ketazolam and its in-source fragment, but as the identification criterion for a ketazolam positive sample included finding both ketazolam and its fragment in full scan mode, samples containing only diazepam would not be incorrectly identified as ketazolam. Lorazepam gave a response of 61% of the LLOQ area of 3-hydroxyphenazepam in the screening method, but as no response was noted in the confirmation method this was not considered an issue. Otherwise, the peak response area for classical drugs of abuse and the >100 NPS examined were similar to the results for blank urines and they were, consequently, considered not to cause selectivity problems.

3.1.3. Matrix effect

The analytes showed matrix effects ≤±20% at both the LLOQ and 900 ng/mL in the screening method, except for N-desmetylflunitrazepam, flunitrazolam, ketazolam, nifoxipam, pivoxazepam and tetrazepam where they ranged between −38% and 63%. However, since the CV of the mean calculated concentration was within 20% for all of these compounds, it appears that the IS was able to compensate for any matrix effects. In the confirmation method, all analytes, except nifoxipam, showed a matrix effect of ≤±20% at both the LLOQ and 900 ng/mL. As in the screening method, the CV of the mean calculated nifoxipam concentration was within 20%, indicating that the IS was also able to compensate for matrix effects in the confirmation method.

3.1.4. Stability

The stability of analytes in urine, methanol, and as prepared extracts was studied under various conditions and the results are presented in Supplementary material 4. All analytes, except clonazolam and nifoxipam, were stable at all conditions, as confirmed by the measured concentrations being within ±20% of the expected values (Supplementary material 4). For clonazolam, the mean stability at −80 °C over 4 months (i.e., −21%) was just outside the acceptance limit (i.e., ±20%). The concentration of nifoxipam decreased slightly below the acceptance limit after storage in methanol for 15 months, at 4 °C for 1 week, and at −80 °C for 4 months. It was, therefore, not recommended to store patient urine samples longer than 1 week at 4 °C. However, previous studies investigating the stability of NPS benzodiazepines in urine at −20 °C suggested that they are stable for at least 2 months under these storage conditions [17]. Urine standards and QC samples were stable at −80 °C over a 4-month period and could be thawed at least three times without significantly altering the results. The methanol solutions were stable for 15 months at −20 °C, and prepared extracts were stable for 4 days at 12 °C in the autosampler.

3.1.5. Carry-over

The peak area for blank urines injected immediately after the highest standard was <20% of the LLOQ peak area in both the screening and the confirmation method for all analytes, except pyrazolam and tetrazepam. For pyrazolam, the peak areas were always lower than the LLOQ, whereas for tetrazepam a response similar to that for the LLOQ was obtained. Due to this, it is recommended that tetrazepam positive samples be re-analyzed if showing a concentration near the LLOQ and if the sample was analyzed directly after a high concentration sample.

3.2. Application of the method

Both methods were employed for analysis of 126 patient urine samples collected between 2015 and 2017 that had tested positive in the CEDIA benzodiazepine immunoassay, but were confirmed negative for α-hydroxyalprazolam, α-hydroxymidazolam, α-hydroxytriazolam, 7-aminoclonazepam, 7-aminoflunitrazepam, 7-aminonitrazepam, desmethyldiazepam, lorazepam, oxazepam, and temazepam by the routine LC–MS/MS confirmation method. An NPS benzodiazepine was detected in 40 out of the 126 samples (i.e., 32%). The 9 detected compounds were clobazam, estazolam, flubromazepam, flubromazolam, flunitrazolam, 3-hydroxyflubromazepam, 3-hydroxyphenazepam, metizolam, and nifoxipam, and the number of cases of each are shown in Fig. 2. More than half of the positive cases contained flunitrazolam, which became classified as a narcotic in Sweden in January 2017 [32]. Example chromatograms for a flunitrazolam positive case are shown in Fig. 3A, Fig. 3B.

Fig. 2.

Number of urine samples containing each analyte among 126 patient samples collected from routine drug testing.

Fig. 3A.

Chromatograms from the screening method showing a patient sample containing 14 ng/mL flunitrazolam (left) and a standard spiked with 10 ng/mL flunitrazolam (right).

Fig. 3B.

Chromatograms from the confirmation method showing a patient sample containing 14 ng/mL flunitrazolam (left) and a standard spiked with 10 ng/mL flunitrazolam (right). Quantification transitions are shown above and qualification transitions below.

Furthermore, 329 urine samples from the STRIDA project on NPS [2], constituting acute intoxication cases presented to emergency wards in Sweden, were investigated. The following 14 NPS benzodiazepines were detected in 95 (i.e., 29%) of the STRIDA samples: clonazolam, deschloroetizolam, diclazepam, estazolam, etizolam, flubromazepam, flubromazolam, 3-hydroxyflubromazepam, 3-hydroxyphenazepam, ketazolam, meclonazepam, metizolam, nifoxipam, and pyrazolam. Many of these 95 samples were multi-intoxication cases (e.g., co-administrations of flubromazepam and pyrazolam).

The concentration ranges found in the routine and STRIDA samples are presented in Table 2. As expected, the substance concentrations were typically higher in the intoxication cases. Many patient samples showed concentrations for NPS benzodiazepines <50 ng/mL, which is the recommended analytical cut-off for prescription benzodiazepines in Sweden. Hence, based on the present and previous [17] results, a lower cut-off value of 10 ng/mL is suggested for urinary analysis of clonazolam, estazolam, flubromazepam, flunitrazolam, meclonazepam and metizolam, whereas 50 ng/mL may be sufficient for the other analytes.

Table 2.

Concentration ranges of NPS benzodiazepines found in urine samples from routine drug testing and from intoxication cases in emergency wards (STRIDA project samples).

Substance	Concentration, ng/mL (Routine samples)	Number of positive samples (Routine samples)	Concentration, ng/mL (STRIDA samples)	Number of positive samples (STRIDA samples)
Clobazam	86	1	–	–
Clonazolam	–	–	10–132	8
Deschloroetizolam	–	–	5	1
Diclazepam	–	–	8	1
Estazolam	30	1	8	1
Etizolam	–	–	40–1000	5
Flubromazepam	5	1	5–89	11
Flubromazolam	15–61	6	5–1082*	52
Flunitrazolam	5–121	24	–	–
3-Hydroxyflubromazepam	191–410	2	90–11,580*	10
3-Hydroxyphenazepam	84–16,549*	5	258–12,090*	5
Ketazolam	–	–	–	1
Meclonazepam	–	–	5–126	10
Metizolam	5–14	3	8	1
Nifoxipam	149	1	973–3635*	2
Pyrazolam	–	–	6–1995*	11

^*Samples with a substance concentration >1000 ng/mL were diluted 1:10 or 1:100 with Milli-Q water prior to re-analysis.

4. Conclusion

The multi-component LC–HRMS(/MS) method offered comprehensive and sensitive detection of a large number of benzodiazepines sold as NPS. Since the screening method is based on LC–HRMS analysis in full scan mode, it enables easy inclusion of new analytes as they appear on the market for recreational drugs. Analysis of NPS benzodiazepines in urine samples from the routine drug testing or acute drug intoxication cases revealed a high prevalence (∼30%) of NPS benzodiazepine use in Sweden in the last years.

Declaration of interest

None to declare.

Appendix A. Supplementary data

The following are the Supplementary data to this article: