Validation of a microsampling-compatible liquid–liquid extraction method for cannabinoid quantitation in 50 µL of whole blood using liquid chromatography–mass spectrometry
Department of Emergency Medicine, University of British Columbia, VGH Research Pavilion, Vancouver, BC V5Z 1M9, Canada
Corresponding author. Department of Emergency Medicine, University of British Columbia, VGH Research Pavilion, 828 W 10th Avenue, Vancouver, BC V5Z 1M9, Canada. E-mail: aman.mohammed@ubc.caAbstract
Recently developed dried blood analysis methods for cannabinoid quantitation utilize small blood volumes, making them microsampling-compatible, but are limited by hematocrit-related bias for dried blood spots (DBSs) and higher consumable costs for volumetric absorptive microsampling (VAMS®). To address these issues, we developed a highly sensitive liquid chromatography-tandem mass spectrometry (LC-MS/MS) method capable of quantifying cannabinoids in 50 µL of liquid whole blood, providing a practical microsampling alternative to dried blood approaches. Using liquid–liquid extraction (LLE) with sodium hydroxide alkalinization and acetonitrile precipitation, followed by quantitative analysis on an Agilent 6495 liquid chromatography-triple quadrupole mass spectrometer, we achieved lower limits of quantitation (LLOQ) of 0.10 ng/mL for Δ9-tetrahydrocannabinol (THC), cannabinol (CBN), and cannabigerol (CBG), 0.20 ng/mL for cannabidiol (CBD), 0.50 ng/mL for 11-hydroxy-THC (11-OH-THC), and 1.0 ng/mL for 11-nor-9-carboxy-THC (THC-COOH). Calibration was linear from the LLOQ to 300 ng/mL for all analytes. To our knowledge, this is one of the first validated LLE approaches for cannabinoid quantitation in less than 100 µL of liquid whole blood. Further, it achieves sub-ng/mL sensitivity, exceeding the LLOQs of most published methods which require ≥100 µL of whole blood. We anticipate particular utility for our method in obtaining evidence from suspected impaired drivers at the roadside when paired with capillary microsampling, such as via finger prick. This approach enables measurement of THC levels at the time of driving, thereby overcoming current limitations, including the decrease in THC levels that occurs with delayed blood sampling, requirement for larger sample volumes (≥100 µL), and dependence on trained phlebotomists for venipuncture.
Introduction
Cannabis use has risen substantially in recent years, particularly following legalization in many jurisdictions, including Canada [1]. This trend has heightened concerns about cannabis-impaired driving, as Δ9-tetrahydrocannabinol (THC), the primary psychoactive compound in cannabis, is known to impair driving performance and increase collision risk [2]. In 2024, 16.3% of drivers involved in collisions tested positive for THC across Canada [3]. As such, many jurisdictions have established legal driving limits for THC in blood to support enforcement of impaired driving laws [4]. Accurate quantitation of THC in drivers is therefore critical for improving road safety, informing public policy, and enforcing traffic laws.
Despite existence of legal limits, interpreting blood THC concentrations remains challenging [5]. In current practice, blood samples from suspected impaired drivers are typically collected hours after driving. During this delay, THC undergoes distribution from blood into body tissue and metabolism to inactive metabolites such as 11-nor-9-carboxy-THC (THC-COOH); individual variability in cannabinoid pharmacokinetics further complicates back-extrapolation of the driver’s THC concentration at the time of driving [6], creating uncertainty when interpreting measured levels.
Capillary microsampling via fingerstick has emerged as a promising approach that could address these issues. Defined broadly as collection of ≤50 µL of blood [7], microsampling is minimally invasive, does not require trained phlebotomists, and allows for roadside collection of blood, capturing cannabinoid concentrations closer to the time of driving [8]. Dried blood spot (DBS) cards and Mitra® volumetric absorptive microsampling (VAMS®) tips are among the most popular tools, offering easy storage, reduced biosafety risk, and improved analyte stability due to enzymatic inactivation during drying [9]. These features make dried microsamples highly attractive for forensic toxicology applications.
To take advantage of microsampling techniques, there has been a growing push to develop analytical methods for dried blood sampling in toxicology [10]. These approaches have demonstrated sensitivity comparable to liquid blood methods for cannabinoid quantitation, despite using lower sample volumes [11–13]. This is partly due to the use of dried matrices, including filter paper cards or VAMS® tips, which retain proteins and lipids, reducing the matrix effect that often suppresses analyte sensitivity. This feature allows analysts to bypass liquid–liquid or solid–phase extraction (SPE) steps and use a simple elution approach, minimizing analyte loss. However, dried blood sampling methods have limitations. Variability in hematocrit affects blood spreading and uniformity on cards, reducing quantitative accuracy [14]. The higher cost of VAMS® devices is also a barrier to widespread use [15]. As such, dried blood sampling methods have yet to be implemented in forensic casework or at the roadside. For now, liquid blood remains the gold standard for toxicology until these issues are addressed and more studies establish the feasibility of dried microsampling.
Currently, a limited number of published methods have demonstrated cannabinoid quantitation from liquid blood volumes in the microsampling range (≤50 µL), with the most sensitive reporting a lower limit of quantitation (LLOQ) of 0.50 ng/mL for THC [16, 17]. As summarized in Supplementary Table S1 (see online supplementary material for a color version of this table), several methods using ≥100 µL of liquid blood achieve LLOQs of 0.50 ng/mL, with very few reporting values as low as 0.10 ng/mL. However, unpublished data from our study of 13 765 injured drivers across Canadian emergency departments from 2018 to 2024 show that 22.9% of THC-positive cases have blood concentrations between 0.10 and 0.50 ng/mL. Considering that many jurisdictions have zero-tolerance driving laws particularly for young novice drivers [16], and are therefore limited by the LLOQs of analytical methods, existing approaches may fail to detect low but forensically relevant THC levels commonly observed several hours after a crash. This gap likely reflects the combined challenges of matrix effects in liquid blood, extraction-related sample loss, and the sensitivity limits of current gas and liquid chromatography-tandem mass spectrometry (LC-MS/MS) methods.
To address this, we developed a method that enables cannabinoid quantitation from 50 µL of liquid whole blood with an LLOQ of 0.10 ng/mL for THC. Additional analytes including cannabinol (CBN), cannabigerol (CBG), cannabidiol (CBD), 11-hydroxy-THC (11-OH-THC), and THC-COOH were also quantified with LLOQs suitable for forensic applications. This approach integrates microsampling with liquid blood, combining the practicality of low-volume sampling with the reliability of liquid blood analysis.
Materials and methods
Chemicals and reagents
THC, CBN, CBG, CBD, 11-OH-THC, and THC-COOH analytical standards (1.0 mg/mL in methanol) and their deuterated internal standards (THC-d3, CBN-d3, CBG-d3, CBD-d3, 11-OH-THC-d3, THC-COOH-d3; 1.0 mg/mL in methanol) were obtained from Cerilliant (Round Rock, TX, USA). HPLC-grade methanol, isopropanol, acetonitrile, hexane, and ethyl acetate were purchased from VWR (Radnor, PA, USA). Liquid chromatography-mass spectrometry (LC-MS) grade formic acid and sodium hydroxide (NaOH) pellets were obtained from Fisher Scientific (Fair Lawn, NJ, USA). Deionized water was produced in-house using a Millipore Direct-Q® 5UV system (Burlington, MA, USA). Ten sources of drug-free defibrinated horse blood were purchased from Dalynn Biologicals (Calgary, AB, Canada) and stored at 4°C until use. Three sources of drug-free human blood were obtained from volunteers and stored in BD Vacutainer™ K2 EDTA tubes. Medical-grade nitrogen gas (Linde, Vancouver, BC, Canada) was used with a Reacti-Vap™ system (Fisher Scientific, Fair Lawn, NJ, USA) for solvent evaporation.
Standard solutions preparation
A 10 000 ng/mL stock solution containing THC, CBN, CBG, CBD, 11-OH-THC, and THC-COOH was prepared in methanol and used to prepare the calibrators by serial dilution. Calibrators were prepared by fortifying drug-free horse blood to achieve final concentrations of 0.10–300 ng/mL for THC, CBN, and CBG, 0.20–300 ng/mL for CBD, 0.50–300 ng/mL for 11-OH-THC, and 1.0–300 ng/mL for THC-COOH.
Separate methanolic stock solutions (10, 100, 1000, 10 000 ng/mL) were used to prepare the following blood quality controls (QCs): low (0.30 ng/mL for THC, CBN, CBG; 0.60 for CBD; 1.5 for 11-OH-THC; 3 for THC-COOH), mid (20 ng/mL for all analytes), and high (240 ng/mL for all analytes).
A mixed internal standard (ISTD) solution was prepared in methanol containing 10 ng/mL of THC-d3, CBN-d3, CBG-d3, CBD-d3, 11-OH-THC-d3, and 25 ng/mL of THC-COOH-d3.
Liquid–liquid extraction
Fifty microliters of calibrator or QC was fortified with 25 µL of ISTD solution in borosilicate glass tubes (13 × 100 mm). One hundred microliters of 0.1 M NaOH and 100 µL of 0.1% formic acid in water were added sequentially and mixed for 1 minute. Five hundred microliters of cold acetonitrile was added to precipitate proteins, followed by mixing for 5 minutes and centrifugation (4000 RPM, 10 minutes, 8°C). The supernatant was then transferred to clean glass tubes (13 × 100 mm) and mixed with 1.5 mL 7:3 (v/v) hexane: ethyl acetate for 10 minutes and centrifuged (4000 RPM, 10 minutes, 8°C). The upper organic layer was transferred to clean borosilicate glass tubes (16 × 100 mm) and evaporated at 25°C under nitrogen gas (flow rate of 1.5 mL/min for 5 minutes). Extracts were reconstituted in 80 µL of 75:25 (v/v) 0.1% formic acid in methanol: 0.1% formic acid in water and transferred to autosampler vials with glass inserts.
Instrumentation
Analyte separation was achieved using an Agilent 1290 Infinity II liquid chromatograph (Santa Clara, CA, USA) with an Agilent Poroshell 120 EC-C18 column (2.7 µm, 2.1 × 100 mm) held at 40°C. Mobile phase A was 0.1% formic acid in water and mobile phase B was 0.1% formic acid in methanol. The total run-time was 10.5 minutes with the following gradient profile applied: 75% B at 0.40 mL/min for 0.5 minutes, before increasing to 100% B over 4.5 minutes, hold at 100% B while increasing to 0.70 mL/min over 1 minute, hold at 0.70 mL/min for 1 minute, before decreasing back to 75% B at 0.40 mL/min by 7.2 minutes and re-equilibrating for 3.3 minutes. The injection volume used was 20 µL.
Quantitation was performed with an Agilent 6495 triple quadrupole mass spectrometer using positive electrospray ionization and dynamic multiple reaction monitoring (dMRM), acquiring one quantifier and qualifier transition per analyte and ISTD. The full list of monitored transitions is provided in Supplementary Table S2 (see online supplementary material for a color version of this table). The optimized MS source parameters are as follows: gas temperature (290°C); gas flow (11 L/min); nebulizer (43 psi); sheath gas temperature (370°C); sheath gas flow (12 L/min); capillary voltage (3100 V); and nozzle voltage (500 V). Agilent MassHunter Workstation (Version 12.0) was used for data acquisition and analysis.
Method validation
Method validation followed the Scientific Working Group for Forensic Toxicology (SWGTOX) guidelines [18]. Parameters evaluated included calibration model, bias, within- and between-run precision, limit of detection and quantitation, ion suppression/enhancement, interferences, carryover, dilution integrity, and processed sample stability. Drug-free horse blood was used as a surrogate matrix for all validation experiments unless otherwise noted.
Calibration was assessed using nine nonzero calibrators over five different days. Linearity was modeled by least-squares regression with a 1/x2 weighting and accepted when the residuals of individual calibrators were within ±3 standard deviations and the correlation coefficient (R2) exceeded 0.99. Limits of detection (LOD) were estimated using the regression method based on the y-intercept and mean slope of the five calibration curves. The LLOQ was defined as the lowest nonzero calibrator, which was extracted in triplicate from three unique blood sources across 5 days (n = 15). LLOQs were accepted when qualifying ion ratios were within ±20% of the calibrators, signal-to-noise ratios were ≥10, and retention times (RT) were within ±0.1 minute of the calibrators.
Carryover was investigated over five runs by injecting an extracted blank immediately after a 500 ng/mL fortified sample (n = 5) and considered absent if the response was below the signal of the lowest calibrator, and no peaks were observed at the expected RT. Bias and precision were assessed at three quality control levels across 5 days. Bias (%) was considered acceptable when within ±20% and precision (%CV) was required to be <20%. Within-run precision was calculated separately for each QC level per run, while between-run precision was determined from the mean values across all 5 days.
Ion suppression/enhancement was evaluated using post-extraction addition at the low and high QC levels. Percent matrix effect was calculated by dividing the mean peak areas of the analyte and ISTD fortified in 10 unique sources of blood extracted in duplicate (n = 20) by the mean peak areas of eight neat samples of analyte and ISTD. Acceptability was defined as <25% matrix effects and %CV <20%.
Matrix interferences were investigated by extracting 10 sources of blank blood without analyte or ISTD, which allowed assessment of endogenous compounds that could contribute to background signals. Responses in these blanks were considered negligible if found to be below the LLOQ and no peaks were observed at the expected RT. Interference was further assessed by analyzing samples containing only the ISTDs (n = 3) and evaluating the signal of the analytes. Interference was determined to be insignificant if the signal fell below the LLOQs. To examine potential ISTD crosstalk from high concentrations of cannabinoids, blank blood samples (n = 3) were fortified with 500 ng/mL of cannabinoids in the absence of ISTDs, and the deuterated transitions were monitored.
Exogenous interferences were evaluated by fortifying low QC blood samples (n = 5) with a 98-compound mixture of common impairing drugs (>5000 ng/mL) and determining if acceptable bias and precision criteria were still met (±20%). The interference mix included opioids, stimulants, benzodiazepines, antidepressants, and other drug classes. The complete list is provided in Supplementary Table S3 (see online supplementary material for a color version of this table).
Additional validation parameters
Dilution integrity was evaluated by 1:1 dilution of a fortified blood sample with 4 ng/mL of THC, CBN, CBG, CBD, and 11-OH-THC, and 100 ng/mL of THC-COOH, analyzed in triplicate over 5 days. Acceptability criteria included bias and precision within ±20%.
Processed sample stability was assessed over 72 hours in the autosampler (8°C). Samples fortified at the low and high QC levels were separately extracted, pooled, and aliquoted into multiple autosampler vials. The first vial from each set was injected in triplicate to establish the mean baseline response (t0), with subsequent vials being analyzed in triplicate at 8-hour intervals. Stability was considered acceptable if peak areas remained within ±20% of t0.
To assess the suitability of horse blood as a surrogate matrix, linearity was evaluated in 3 unique sources of drug-free human blood over 3 days. Bias and precision at the LLOQ, low, mid, and high QC levels, prepared in triplicate in each of the three human blood sources, were assessed against a horse blood calibration curve over 3 days. Matrix interferences and ion suppression/enhancement were also examined in the 3 human blood sources. To investigate potential differences in analyte binding to matrix components between human and horse blood, mid-QC samples were prepared in three sources of each matrix type. Spiked samples were either extracted immediately or stored at room temperature for 3 hours before extraction, with ISTD added after the storage period. At each timepoint, samples were extracted in duplicate and the measured concentrations of extracts from each matrix type were compared to evaluate differences in analyte recovery.
Results
Method development
The objective of this study was to develop a highly sensitive LC-MS/MS method capable of quantifying cannabinoids in 50 µL of whole blood with an LLOQ of 0.10 ng/mL for THC, providing a microsampling alternative to dried blood approaches while addressing the sensitivity limitations of existing techniques, which often cannot quantify THC below 0.50 ng/mL.
Ionization and fragmentation parameters were optimized by injecting 100 ng/mL of each analyte and ISTD. Initial methods using 3 minutes of postrun column equilibration showed progressive signal loss, with analyte and ISTD responses falling below 30% of the first injected standard after 20 repeated injections. Extending the equilibration to 5.5 minutes stabilized responses, maintaining consistent responses across 100 consecutive injections (±10%). The maximum instrumental injection volume of 20 µL provided the highest signal intensity at the LLOQs while preserving acceptable peak shapes as shown in Fig. 1.
Liquid–liquid extraction (LLE) was selected for sample preparation due to its low cost and ability to effectively remove cellular components. Parameters such as sample pH and solvent composition were systematically adjusted to maximize THC response. Given that THC binds plasma proteins in blood [19], raising the pH was hypothesized to increase protein negative charge and weaken hydrophobic interactions with THC. Alkalinization with NaOH increased the extraction pH from 5.0 to 8.0 and enhanced responses for THC (+29%), CBN (+22%), CBG (+40%), CBD (+57%), and 11-OH-THC (+36%), while the THC-COOH response decreased (−6%) relative to nonalkalinized samples prepared at the LLOQ (n = 6). Protein precipitation with acetonitrile was hypothesized to further reduce THC binding by destabilizing protein structure and increased THC (+96%), CBN (+38%), and CBD (+61%) responses, but lowered CBG (−6%), 11-OH-THC (-46%), and THC-COOH (−76%) responses relative to non-precipitated samples prepared at the LLOQ (n = 6). The reduced THC-COOH signal under both conditions likely reflects deprotonation to THC-COO- at basic pH, which may lower its solubility in acetonitrile and 7:3 hexane: ethyl acetate.
Organic solvent ratios of 9:1, 7:3, and 1:1 (v/v) hexane: ethyl acetate were compared, with 7:3 providing the most favorable partitioning across all 6 cannabinoids. Reconstitution conditions were optimized by testing 100% methanol, 90:10 methanol: water, 75:25 methanol: water, and 75:25 mobile phase B: mobile phase A, with and without mild heating at 30°C during vortexing. Sequential reconstitution was also assessed, where half the volume was added, vortexed, and transferred to autosampler vials before adding the second aliquot. Neither mild heating nor sequential addition improved peak areas at the lowest calibrator levels. The selected reconstitution condition was 80 µL of 75:25 mobile phase B: mobile phase A.
Method validation
The calibration ranges achieved for the target cannabinoids are summarized in Table 1. All analytes demonstrated linear responses up to 300 ng/mL, with LLOQs corresponding to the lowest calibrators: 0.10 (THC, CBN, CBG), 0.20 (CBD), 0.50 (11-OH-THC), and 1.0 ng/mL (THC-COOH). Residual plots confirmed homoscedasticity, and correlation coefficients were consistently ≥0.998, supporting the use of a linear regression model. At the LLOQs, chromatographic peaks were well defined, with signal-to-noise ratios ≥10, consistent RTs, and acceptable qualifier ion ratios. Chromatograms at the LLOQ are provided in Fig. 1. Using the regression method, LODs were 0.023 (THC), 0.028 (CBN), 0.062 (CBG), 0.061 (CBD), 0.27 (11-OH-THC), and 0.35 ng/mL (THC-COOH). No carryover was observed in extracted blanks following injection of 500 ng/mL samples.
| Analyte | Internal standard | LOD (ng/mL) | LLOQ (ng/mL) | Slope ± SD (n = 5) | y-int ± SD (n = 5) | R 2 ± SD (n = 5) | Linear range (ng/mL) |
|---|---|---|---|---|---|---|---|
| THC | THC-d3 | 0.023 | 0.10 | 0.052 ± 0.018 | −0.001 ± 0.000 | 1.000 ± 0.000 | 0.10–300 |
| CBN | CBN-d3 | 0.028 | 0.10 | 0.321 ± 0.060 | 0.008 ± 0.003 | 0.999 ± 0.000 | 0.10–300 |
| CBG | CBG-d3 | 0.062 | 0.10 | 0.113 ± 0.015 | −0.002 ± 0.002 | 0.999 ± 0.000 | 0.10–300 |
| CBD | CBD-d3 | 0.061 | 0.20 | 0.136 ± 0.008 | 0.003 ± 0.002 | 1.000 ± 0.000 | 0.20–300 |
| 11-OH-THC | 11-OH-THC-d3 | 0.27 | 0.50 | 0.216 ± 0.061 | 0.030 ± 0.018 | 0.999 ± 0.000 | 0.50–300 |
| THC-COOH | THC-COOH-d3 | 0.35 | 1.0 | 0.004 ± 0.000 | −0.001 ± 0.000 | 0.998 ± 0.001 | 1.0–300 |
Bias and precision were evaluated by fortifying horse blood with each cannabinoid at low, mid, and high concentrations. The low QC level for each cannabinoid was established as three times the LLOQ, with the high QC level being 80% of upper limit of quantitation. The mid QC level was set at 20 ng/mL. Across all cannabinoids and QC concentrations, bias was −14.0% to 14.4%. Within-run precision was 0.9%–14.0%, and between-run precision was 1.4%–8.0%. All values met the 20% acceptance criteria. Dilution integrity was also assessed with a two-fold dilution of a blank fortified with 4 ng/mL of THC, CBN, CBG, CBD, and 11-OH-THC, and 100 ng/mL of THC-COOH. Acceptable bias and precision criteria were met. Table 2 details bias and precision results for each QC concentration, the dilution integrity study, and the LLOQs.
| Concentration (ng/mL) | Bias (%) (n = 25) | Within-run precision (% CV, n = 25) | Between-run precision (% CV, n = 25) | |
|---|---|---|---|---|
| THC | ||||
| LLOQ | 0.10 | 1.2 to 9.6 | 4.6 to 9.2 | 2.3 |
| Low | 0.30 | −6.7 to 9.3 | 2.2 to 9.0 | 5.7 |
| Mid | 20 | −4.9 to 11.4 | 1.4 to 3.7 | 5.5 |
| High | 240 | −2.5 to 2.1 | 1.8 to 4.5 | 1.8 |
| 2-fold dilution | 4.0 | −17.3 to 12.4 | 2.0 to 7.1 | 13.5 |
| CBN | ||||
| LLOQ | 0.10 | −1.0 to 14.1 | 1.5 to 7.1 | 4.3 |
| Low | 0.30 | −7.1 to 14.4 | 3.6 to 6.8 | 8.0 |
| Mid | 20 | −6.8 to 9.4 | 2.5 to 4.8 | 6.4 |
| High | 240 | −14.0 to 6.3 | 2.3 to 4.0 | 7.0 |
| 2-fold dilution | 4.0 | −14.4 to 8.1 | 1.3 to 6.6 | 8.2 |
| CBG | ||||
| LLOQ | 0.10 | −7.8 to 8.3 | 3.9 to 9.7 | 6.1 |
| Low | 0.30 | −7.1 to 12.0 | 3.1 to 8.3 | 7.7 |
| Mid | 20 | −9.7 to 3.0 | 2.0 to 5.7 | 4.8 |
| High | 240 | −7.8 to 4.2 | 1.3 to 3.2 | 4.4 |
| 2-fold dilution | 4.0 | −8.7 to 5.2 | 2.1 to 7.9 | 5.2 |
| CBD | ||||
| LLOQ | 0.20 | −5.6 to 8.8 | 1.2 to 7.1 | 4.7 |
| Low | 0.60 | −2.3 to 8.1 | 1.9 to 4.8 | 3.4 |
| Mid | 20 | −0.8 to 7.3 | 1.6 to 5.9 | 4.0 |
| High | 240 | −3.6 to 0.2 | 0.9 to 3.3 | 1.4 |
| 2-fold dilution | 4.0 | −10.0 to 5.7 | 2.7 to 9.4 | 6.4 |
| 11-OH-THC | ||||
| LLOQ | 0.50 | −2.7 to 6.5 | 2.9 to 4.2 | 3.2 |
| Low | 1.5 | −5.3 to 11.1 | 2.1 to 7.9 | 5.5 |
| Mid | 20 | −7.4 to 5.7 | 3.7 to 6.8 | 3.3 |
| High | 240 | −0.9 to 4.7 | 0.9 to 5.8 | 5.0 |
| 2-fold dilution | 4 | −11.4 to 9.1 | 1.2 to 7.8 | 8.8 |
| THC-COOH | ||||
| LLOQ | 1.0 | 7.4 to 12.6 | 1.9 to 7.5 | 1.6 |
| Low | 3.0 | −6.5 to 7.5 | 4.3 to 14.0 | 5.3 |
| Mid | 20 | −7.2 to 9.1 | 3.7 to 8.3 | 7.0 |
| High | 240 | −6.7 to 2.5 | 4.2 to 7.2 | 3.2 |
| 2-fold dilution | 100 | −14.2 to 16.1 | 1.6 to 8.7 | 12.2 |
Matrix effects were examined by post-extraction addition at the low and high QC levels. Ion suppression/enhancement across 10 unique blood matrices was ≤±25% with %CV <20%, indicating effective removal of interferents during extraction. 10 extracted sources of horse blood were free of interferents. No crosstalk between deuterated and non-deuterated cannabinoids was found when evaluating samples fortified with either ISTDs or analytes. Extracted samples at the low QC level fortified with 98 common impairing drugs showed accuracy and precision within 20% and were concluded to be acceptable.
Stability studies of extracted samples at the low and high QC levels stored in the autosampler (8°C) showed that all analytes remained stable (±20% of t0 peak area) after 24 hours. At 48 hours, CBN, CBG, 11-OH-THC, and THC-COOH fell below 80% of the t0 peak area at one QC level. By 72 hours, THC, CBN, 11-OH-THC, and THC-COOH were within acceptance criteria, while CBG and CBD at the low QC level showed >20% loss relative to t0. Results expressed as the percent difference from t0 peak areas are shown in Supplementary Table S4 (see online supplementary material for a color version of this table).
Horse blood was shown to be a suitable surrogate matrix for the quantitative analysis of cannabinoids in human blood. Calibrations prepared in 3 unique sources of human blood were linear, with R2 ≥ 0.995, as summarized in Supplementary Table S5 (see online supplementary material for a color version of this table). Human blood samples spiked at the LLOQ, low, mid, and high QCs, quantified against a horse blood calibration curve, showed bias of −16.5% to 12.0%, within-run precision of 0.5%–17.7%, and between-run precision of 0.5%–10.4% across all cannabinoids and concentrations (Supplementary Table S6, see online supplementary material for a color version of this table). Matrix interferences were negligible and ion suppression/enhancement across the three human blood sources were ≤±25% with %CV <20%. Additionally, no differences were observed between the measured concentrations of human and horse blood samples spiked with cannabinoids at 20 ng/mL, even after storage for 3 hours at room temperature (Supplementary Fig. S1, see online supplementary material for a color version of this figure), suggesting any differences in analyte binding between the two matrices are normalized following ISTD addition.
Discussion
To our knowledge, this study presents the first validated method capable of quantifying THC at 0.10 ng/mL using <100 µL of whole blood. Previously published methods have required larger sample volumes, relied on the cleaner matrix plasma to increase sensitivity, employed SPE, or a combination of these strategies to reach comparable sensitivity [20]. Some approaches also use separate acidic and basic extractions to optimize recovery of both THC and THC-COOH [21]. Our method incorporates both NaOH and formic acid within a single extraction, allowing pH to be adjusted depending on the analytical application. Increasing basicity enhances extraction efficiency for THC, which is relevant when assessing recent cannabis use or impairment, whereas more acidic conditions can improve recovery and LLOQs for THC-COOH, which is more informative for evaluating cannabis use in general. It should be emphasized that our results were achieved using an Agilent 6495 liquid chromatography-triple quadrupole mass spectrometer which has an instrument detection limit of 0.30 fg of reserpine when injected onto the column [22]. Comparable results may depend on access to instrumentation of similar sensitivity. In this study, the method was optimized for THC, reflecting its envisioned application in roadside microsampling, where timely collection provides accurate estimates of THC levels at the time of driving.
While SPE can improve sensitivity through reduced analyte loss and enhanced matrix clean-up [20], our LLE method achieves superior LLOQs for THC. Although less amenable to automation, LLE remains practical for high-throughput workflows, as it avoids the substantial cost of solid-phase cartridges. Our LLOQs for other cannabinoids are also comparable to the most sensitive SPE methods [23]. Inclusion of 11-OH-THC is particularly important, as it is psychoactive and often reaches higher blood concentrations after oral cannabis consumption than inhalation [24]. Our method also quantifies CBN, CBG, and CBD, which are frequently excluded but may provide valuable interpretive context. CBG, a biosynthetic precursor in cannabis, is rapidly metabolized in the body and may indicate recent use [25]. CBN forms through oxidative degradation of THC and may potentiate the sedative effects of THC [26]. CBD measurement is increasingly relevant due to widespread use of CBD-rich products [27] and its known interactions with CYP enzymes that can alter the effects of other drugs [28]. Including these cannabinoids therefore extends interpretive power beyond THC alone and strengthens forensic assessments of cannabis use.
The present method offers an attractive alternative to existing liquid blood assays. It enables quantitation of THC between 0.10 and 0.50 ng/mL, a range we commonly observed in injured drivers, while providing a practical counterpart to dried blood sampling approaches. By using 50 µL of whole blood, it preserves the advantages of microsampling while avoiding limitations associated with DBS and VAMS®, including hematocrit-related bias and higher consumable costs. Liquid microsamples collected using devices such as the Sarstedt Minivette® POCT can be directly transferred into borosilicate glass tubes for extraction. Utilizing <50 µL of sample is possible but will sacrifice some sensitivity. However, a 1:1 dilution can be performed without compromising accuracy, supporting the use of as little as 25 µL of blood.
A major advantage of dried blood sampling is the ability to store samples at room temperature, since water loss during drying halts most enzymatic degradation. THC, 11-OH-THC, and THC-COOH have been shown to remain stable on DBS cards for up to 3 months at room temperature [11], and on VAMS® devices for at least 30 days [12]. In contrast, cannabinoid stability in liquid blood is highly variable and depends on storage material, blood volume, and preservatives used [29]. Future studies should therefore evaluate optimal storage conditions for liquid microsamples in borosilicate glass tubes. Additionally, the comparability of capillary and venous blood warrants further investigation. Studies examining CBD show similar concentrations between capillary finger prick and venous blood [30], whereas another study using capillary blood drawn from the shoulder reported up to 40% lower THC, with THC-OH and THC-COOH being largely comparable [31]. Other studies in toxicology suggest that such differences may depend on the sampling site [32].
A critical distinction between our method and other microsampling approaches lies in sample preparation. DBS, VAMS®, and current 50 µL liquid blood methods rely on simple protein precipitation with methanol or acetonitrile [11–13, 16, 17], bypassing the organic extraction phase used in LLE and SPE. While this simplifies workflows, it provides less effective removal of endogenous materials, including proteins and, importantly, phospholipids. This is relevant considering phospholipids such as lysophosphatidylcholines and lysophosphatidylethanolamines can co-elute with THC in LC-MS workflows and have been shown to cause substantial ion suppression of cannabinoids in acetonitrile extracts of whole blood [33]. The long-term impact of residual phospholipid co-extraction on instrument performance for protein precipitation-based methods remains unclear. Future studies should therefore evaluate instrument robustness, including sensitivity and column backpressure, over extended injection series (>1000 injections) between protein precipitation and organic extraction workflows.
The use of horse blood as a surrogate matrix was chosen for both practical and analytical reasons. Being more widely available, horse blood is more economical, and is simpler to source ethically than human blood, making our method more accessible to laboratories without easy access to human blood banks. Additionally, many sources of “blank” human blood may contain trace amounts of cannabinoids. Given the high sensitivity of our LC-MS/MS, such contamination would render these matrices unsuitable for use as calibrators and controls. This risk is supported by the variability in donor eligibility guidelines, as many blood collection organizations permit donations from individuals who use cannabis provided they are not intoxicated, and a substantial proportion of donors report cannabis use within 72 hours [34]. Combined with the long detection window for THC-COOH in frequent users [35], trace cannabinoid presence in “blank” matrices is plausible. We demonstrated that horse blood performs similarly to human blood for cannabinoid quantitation using our method, with comparable LLOQs and acceptable accuracy, precision, and matrix effects, supporting its use as a surrogate matrix.
Conclusion
The present LC-MS/MS method enables quantitation of 6 cannabinoids of forensic interest: THC, CBN, CBG, CBD, 11-OH-THC, and THC-COOH. Using an LLE approach with 50 µL of whole blood, we achieved an LLOQ of 0.10 ng/mL for THC, allowing reliable measurement in the low-concentration range commonly observed in injured drivers. This represents an advance in liquid blood analysis of cannabinoids and provides a feasible alternative to dried and liquid blood microsampling approaches that could be paired with roadside collection of blood for measurement of THC levels in suspected impaired drivers.
Supplementary Material
Supplementary material
Supplementary material is available at Journal of Analytical Toxicology online.
Conflicts of interest
None declared.
Funding
Funding for this research was provided by the Canadian Institutes of Health Research (375264).
Data Availability
The data underlying this article are available in the article and in its online supplementary material.