Water, methanol, formic acid (FA), and ammonium formiate were furnished by Biosolve. Drugs standard were purchased from Cerilliant or Lipomed, at either 1 mg/ml or 100 μg/ml. External quality controls (ECQ) were purchased from Medidrug, ACQ Science, or Clincheck. Blank lyophilized whole blood was acquired from ACQ Science.

Standard solutions containing tetrahydrocannabinol (THC), 11-Nor-9-carboxy-THC (THC-COOH), alprazolam, amphetamine, methamphetamine, 3,4-methylendioxymethamphetamine (MDMA), 3,4-methylene dioxy-amphetamine (MDA), methylphenidate, cocaine, cocaethylene, lorazepam, bromazepam, zolpidem, benzoylecgonine, morphine, codeine, methadone, tramadol, O-desmethyltramadol, diazepam, nordiazepam, oxazepam were prepared for calibration curve and internal quality control (IQC) preparation. In parallel, solutions containing THC-D3, THC-COOH-D9, cocaine-D3, benzoylecgonine-D3, amphetamine-D8, MDMA-D5, methylphenidate-D10, morphine-D3, codeine-D3, methadone-D3, tramadol¹³C-D3, O-desmethyltramadol-D6, nordiazepam-D5, alprazolam-D5, and zolpidem-D6 were prepared as internal standard (IS) solutions.

Calibration samples were prepared by spiking lyophilized whole blood at 5 concentration levels (Table 1). IS were added to reach a final concentration of 10 (THC-D3), 100, or 1,000 ng/ml depending on the specific calibration range.

IS solutions were spiked in Eppendorfs and evaporated to dryness before adding 100 μL of whole blood. The extraction was then performed by protein precipitation using 300 μL of methanol. After centrifugation for 10 min at 14,000 rpm, the upper methanolic phase was transferred into a new Eppendorf and evaporated to dryness under a nitrogen flow. Reconstitution was performed using 100 μL of 1:9 methanol: water and 10 μL were injected into the LC-HRMS system (Supplemental Figure 1).

A Thermo Scientific Ultimate 3000 LC system with a Phenomenex 2.6 μm C18 (10 cm × 2.1 mm) maintained at 45°C was used for chromatographic separation. Mobile phases were composed of A, ammonium formate 10 mM at pH 3.3, and B, methanol with 0.1% FA. Phase B was ramped linearly from 2 to 98% over 7.5 min. The column was then washed at 98% of B for 3.5 min, followed by a 6 min re-equilibration at 2% of B at 300 μL/min for a total analysis run of 17 min. The LC was coupled to a Q Exactive Plus system (Thermo Scientific, Bremen, Germany) via a heated electro spray ionization (ESI) source (H-ESI II probe, Thermo Scientific). The ionization spray voltage was set to 3 kV, sheath gas flowrate was set to 40, and auxiliary gas flowrate to 10 (both in arbitrary units). The method functioned in PRM, using an inclusion list containing the exact mass of the parent ion and the retention time windows for the different analytes. A polarity switch in negative was performed at 7.5 min for the specific detection of THC-COOH with a switch back in positive polarity at 8.7 min for the detection of THC. Resolution was set to 17,500 for the HCD fragmentation performed using an NCE at 50 eV with an AGC target of 1e5 and a maximum IT of 100 ms.

The validation criteria used to evaluate the analytical process was based on the directives of the “Société Française des Sciences et des Techniques Pharmaceutiques” (SFSTP) regarding bioanalytical methods and adapted to our specific requirements (Boulanger et al., 2003; Peters et al., 2007; Lynch, 2016). Two product ions (one quantifier and one qualifier) were used for data processing (Table 1) and full MS/MS spectra were compared with the online advanced mass spectral database m/z cloud. The validation was performed over 3 non-consecutive days (p = 3). The trueness and precision were evaluated using a variance analysis-based statistical treatment (ANOVA). Calibration (Cal) was performed in duplicate at 5 different concentration levels (k = 5) (Table 1) while quality controls (QCs) were prepared in quadruplicate at the two lowest and highest concentration levels (k = 4). Using the acquired data, trueness, precision, accuracy, linearity, limits of detection (LOD), and quantification (LOQ) were determined. Six different blank bloods were analyzed for selectivity assessment investigating for potential interferences. The approach developed by Matuszweski et al. was used for recovery (RE) and matrix effect (ME) evaluation (Matuszewski et al., 2003). In this optic, three sample sets were prepared, including all the substances of interest at two concentration levels (low being level 2 and high being level 4 described in Table 1). Sample set 1 represented neat standards spiked after the extraction, while sample set 2 represented blank blood spiked after extraction. Sample set 3 represented blank blood spiked before extraction. The absence of interfering peaks at the established retention times (RT) for the analytes and the IS was used to ensure the specificity. The chemical stability of all analytes was evaluated under sample handling and storage conditions at low and high concentrations in five replicates. Benchtop (6 h, room temperature), autosampler (24 h, 5°C), three cycles of freeze-thaw (−20°C), and short term (1 week, −20°C) conditions were used for stability determination.

In order to evaluate the method, 8 different EQCs were analyzed in duplicates using the exact same procedure.

In the present study, 22 analytes (15 IS) included in the main classes of drugs of abuse, as well as the major benzodiazepines, were analyzed using a single simultaneous multi-analyte quantitative approach. This list of substances was established based on the FEDRO recommendations and on the knowledge of the prevalence of psychoactive compounds among suspected impaired drivers. A nationwide study performed on 4,668 samples collected on suspected drivers in 2010 in Switzerland proved cannabinoids (48%), alcohol (35%), cocaine (25%), opiates (15%), amphetamines (7%), and benzodiazepines (6%) to be the most detected substances (Senna et al., 2010). The use of such multi-analyte approaches is challenging due to the various physico-chemical properties of the substances of interest and requires specific care during method development. To ensure a proper quantification, retention time windows were set for the acquisition ensuring the acquisition of a sufficient number of acquisition points (Figure 1). For good-quality integration and reproducible quantification, a minimum of 10–15 points is necessary to define exactly the peak start, peak apex, and peak end. The method was designed to resolve the wide majority of DUID cases using a single procedure and a limited amount of biological sample (Supplemental Figure 1) (Senna et al., 2010). The method allows the successful PRM-based quantification of cannabinoids, amphetamines, cocaine and its metabolites, opiates and opioids, and the major benzodiazepines at the sensitivity necessary for legal thresholds and therapeutic ranges (OOCCR-OFROU, 2008; Schulz et al., 2012).

Independent QC samples at 4 different calibration levels were injected in 4 replicates over 3 non-consecutive days for the determination of trueness and precision. Accuracy represents the total error and is divided into trueness (representing the “bias” or the systematic error) and precision (referring as the standard deviation or random errors) (Gonzalez et al., 2010). The trueness can be evaluated by calculating the percentage difference between the experimental and the expected theoretical values. In the present study, the systematic error varied from −4.7 to 7.6% (Table 2). Precision was divided into two parameters: the relative standard deviation (repeatability or R_R.S.D.) and the inter-day variability (intermediate precision or IP_R.S.D.). R_R.S.D. represents the variability under similar conditions, meaning that the analyses are performed by the same operator using the same reagents and samples. On the other hand, IP_R.S.D. represented the variability associated with the use of the same samples on different days with different reagents. Precision parameters were evaluated to be between 1.1 and 11.6% (Table 2). Accuracy profiles are visual representations combining both the trueness and the precision to represent the uncertainty measurement (Figure 2). Precision is represented by the calculated confidence limit at 95% at each concentration level. Accuracy profiles also include the representation of acceptance limits of ±20% at the LLOQ suggested for method validation (±15% at the other calibration levels). All analyzed QCs were within the acceptance limits.

The definition of linearity stands as the method capacity to provide a result proportional to the actual sample concentration. To determine this parameter, a linear regression model based on the least square method was applied on the fit of the obtained concentration as a function of the theorical concentration. Slopes values were comprised between 0.9688 and 1.0623 with coefficients of determination above 0.9921 for all the compounds confirming the method linearity within the concentration ranges of interest (Table 2). LOQs were fixed according to the lowest point of the calibration curve (Table 1).

Selectivity is defined as the ability to differentiate the analyte of interest from potential interferences. To assess the good selectivity of the method, six different blank blood samples were analyzed using the complete extraction procedure. No compounds impairing the detection and quantification of the analytes of interest were observed. HRMS technology offers a high selectivity due to its resolving power, therefore reducing the number of potential interferences (Chindarkar et al., 2014). However, ME, including ion suppression or enhancement, are often associated with the use of ESI as ion source challenging the method selectivity. The determination of such ME is therefore crucial to ensure a proper detection and quantification of the substances of interest. ME ranged from −51% (15% CV) of ion suppression for THC at low concentration and 3% (13% CV) of ion enhancement for lorazepam, being consistent with the existing literature (Table 3) (Simonsen et al., 2010; Fernandez Mdel et al., 2013; Montenarh et al., 2014; Steuer et al., 2014; Vaiano et al., 2016; De Boeck et al., 2017). All values concerning RE and ME are summarized in Table 3. To compensate for those undesirable ME, isotopically-labeled internal standards was used for normalization.

Results regarding analytes' stability are listed in Table 4. Overall, stability ranged between 86 and 115%, assuring that the samples were stables within the tested conditions (auto-sampler, bench-top, 3 cycles of freeze-thaw and short-term stability.

Eight commercial EQCs were analyzed in duplicates to ensure the robustness of the developed method and procedure. In total, the quantification was performed on 28 samples of amphetamines, 24 samples of cocaine and its metabolites, 16 samples of cannabinoids, 36 samples of opioids and opiates, and 28 samples of benzodiazepines and z-drugs. Results comparison between the described method and the expected EQCs values are represented in Figure 3 for the different classes of molecules involved in DUID cases. A good correlation was observed between the expected and the obtained values. The relative standard deviation was lower than 20% for all tested substances, confirming the efficiency of the PRM quantitative acquisition mode for toxicological analyses. This method confirms the potential of PRM as a solid alternative to classical MRM approaches (Li et al., 2016; Lv et al., 2018; Joye et al., 2020).