A Framework for the Development of Targeted Gas Chromatography Mass Spectrometry (GC-MS) Methods: Synthetic Cannabinoids

Abstract

With the increased presence of novel psychoactive substances (NPSs) in casework, drug analysis has become more challenging. To address these challenges, new screening technologies with improved specificity are being implemented, allowing for the creation and adoption of targeted confirmatory analyses that produce more conclusive results. This paper outlines a six-step, data-driven, framework to develop and evaluate gas chromatography mass spectrometry (GC-MS) methods for targeted classes of drugs. The process emphasizes maximizing retention time differences (to minimize the potential for retention time acceptance windows to overlap) and understanding the tradeoffs between sensitivity and reproducibility using a test solution containing pairs of compounds that are difficult to distinguish. The method is then evaluated by expanding the panel of compounds analyzed, identifying limitations in compound discrimination, comparing to current methods, and analyzing representative casework to establish usability. To demonstrate this framework, a method for synthetic cannabinoids was created. The developed method utilizes a DB-200 column and an isothermal temperature program. It was found that sensitivity could be adjusted, without compromising reproducibility, by altering the split ratio and injection volume. The targeted method successfully differentiated 50 cannabinoids based on either retention time differences or mass spectral dissimilarity – determined using a newly developed spectral comparison test. Compared to the general method investigated for comparison, the targeted method was an order of magnitude more sensitive and a minute shorter while providing major increases in retention time differences. This framework can be implemented and adapted to develop targeted methods for other applications or compound classes.

Faced with increasing backlogs and more complex samples, forensic drug chemistry units need to identify new techniques or methods to assist in the analysis of cases(1,2). Implementation of new technologies can often be hindered by the initial upfront cost, time, and resources required to validate methods and train chemists. An interim or alternative approach to the adoption of new technologies is to develop new analytical methods on existing tools. This approach eliminates the training and cost components while simplifying validation efforts. When existing tools are adapted, such as through the development of new analytical methods, it is critical that the methods are created using a data-driven approach and that they are thoroughly evaluated to understand the strengths and limitations.

One of the challenges drug chemists often face, and where the adoption of new techniques or methods may be most beneficial, is the need to definitively detect and identify a large number of structurally similar compounds(2). This can be difficult using general purpose methods on platforms like gas chromatography mass spectrometry (GC-MS) since structurally similar compounds often have close or indistinguishable retention times and mass spectra. A number of alternative approaches to GC-MS, that can provide increased discrimination power, have been described in the literature(3–5) but these approaches usually require laboratories to adopt new instrumentation. Other efforts have focused on the use of existing techniques, like GC-MS, but instead leverage variations of the technique(6) which may require modification of existing instrumentation. The use of statistical approaches(7) to achieve compound discrimination has also been demonstrated, but only on small subsets of compounds not fully representative of the wide variety of compounds found in casework.

One class of compounds where isomeric differentiation is required is synthetic cannabinoids. Since the identification of Spice in 2008, there have been several studies focused on detecting, identifying, and differentiating isomers in this class. Methods for detection and differentiation of small (less than 10) suites of compounds by GC-MS have been developed(8,9). A number of other studies have focused on the use of GC-MS, and complimentary techniques, for the differentiation of isomeric synthetic cannabinoids(6,10–12). Additionally, GC-MS equipped with photoionization(13) as well as traditional GC-MS used in combination GC-PCI-MS (positive chemical ionization) and GC-NCI-MS (negative chemical ionization)(14) have also been proposed. A number of techniques beyond GC-MS, such as liquid chromatography mass spectrometry (LC-MS), liquid chromatography tandem mass spectrometry (LC-MS/MS), immunoassay, matrix assisted laser desorption/ionization MS (MALDI-MS) have also been demonstrated (15–17). DART-MS has been studied, but predominantly as a screening tool(18,19).

To compliment these efforts and provide another potential solution to the challenges faced by isomer differentiation a framework for the creation of targeted GC-MS methods was established. The framework is modular and can be modified or adapted to address specific needs of laboratories. It is intended to minimize the time required to develop targeted GC-MS methods while also ensuring objective, quantifiable results are obtained at each step, leading to data-driven decision-making. The framework is presented as a series of studies that can be used to establish and evaluate the performance of new targeted methods. To demonstrate the practical implementation and utility of the framework, a targeted method for analyzing synthetic cannabinoids was developed and evaluated.

Materials and Methods

Framework for Developing Targeted GC-MS Methods

The framework consists of six discreet studies that should be conducted in series as summarized in Figure 1. Since a main driver behind using a targeted method would be to minimize the number of compound pairs with overlapping retention time acceptance windows the first two steps of the framework focus on arriving at parameters that maximize retention time differences between neighboring compounds in a test solution in a reasonable analysis time. In Step 1, a multi-component test solution is analyzed on different column types using identical method parameters to quantify the effect of different stationary phases. The results from this step provide information on retention time differences, defined here as the percent difference in retention times between neighboring peaks (%RTD) (Equation 1) since many laboratories use percentage-based acceptance windows. The sensitivity (peak area), peak width, peak purity, and mass spectral similarity between columns can also be compared. Once the desired column is chosen, the gas flow and temperature programs are varied (Step 2) to establish parameters that allow for maximum retention time differences attempting to minimize run time. For this step, a retention time acceptance window should be defined.

$$\mathrm{\%}RTD=\frac{(\left(RetentionTimeCompound2\right)-\left(RetentionTimeCompound1\right))}{\left(RetentionTimeCompound1\right)}\mathrm{*}100$$ Eqn.1

After assessing the column type, column flow rate, and temperature program, a design of experiments (DOE) approach (Step 3) is used to investigate parameters that may affect the sensitivity and reproducibility of the measurement but are likely to have minimal impact on separation. These parameters include injection volume, inlet temperature, split ratio, and source temperature. Use of a DOE approach provides information on multiple instrument parameters simultaneously, minimizing the number of required analyses and potential confounding effects that can occur with a one factor at a time approach. The results of this step provide parameters that maximize sensitivity and reproducibility while also identifying parameters that can be altered to adjust the sensitivity of the method without affecting reproducibility. Tune type can also be evaluated in this step to identify potential benefits or drawbacks from using either of the available tune options. Investigation of tune type can be completed independent of the other parameters or be incorporated into the DOE if desired. The results from Step 1 through Step 3 provide a targeted GC-MS method.

FIG. 1 –

Flowchart highlighting the steps of the targeted method development and evaluation framework.

Evaluation of the targeted method involves three additional steps that ensure the method is suitable for a larger set of compounds (Step 4), quantifies the potential gains of the targeted method compared to the current method (Step 5), and ensures the method is suitable for casework (Step 6). Measuring suitability of the method for a larger set of compounds (Step 4) is completed by analyzing all relevant standards available in the laboratory using the targeted method. A single analysis of each compound is first completed to determine whether modifications to the temperature program are required to achieve elution of all additional compounds. The method is then retention time locked to a locking compound. Retention time locking is incorporated to provide a mechanism for ensuring repeatable retention times with column aging or column maintenance. Replicate measurements of all compounds are then made to establish the locked retention times, measure the variability in retention times, and to obtain replicate mass spectra. This can be done by analyzing individual standards or by creating multi-component mixtures consisting of well resolved compounds. Additionally, an alkane ladder can be incorporated into this analysis to allow for the calculation of retention indices.

The resulting dataset from Step 4 provides the ability to identify instances where the method may fail to allow for unique compound identification based on overlapping retention time acceptance windows and/or mass spectral similarity. Sufficient retention time separation, in this study, is defined as greater than 2 %RTD but can be adjusted based on laboratory protocols. This value was chosen because it is one of the most conservative percentage-based acceptance windows used in the field and therefore provides a high number of potential indistinguishable pairs. For compounds that are not sufficiently separated by retention time, the mass spectral similarity between the compounds can be estimated using the replicate measurements and software tools or visual inspection. In this paper, we employ the newly developed min-max test, described in the Supplemental Information, for spectral differentiation. This automatable method compares the spectral similarity across replicates of each individual compound to the similarity computed between measurements of the pair of compounds. If the compounds are indistinguishable by their mass spectra, the scores obtained for the pair of comparisons are expected to be similar to those of the individual comparisons.

Following Step 4, a set of experiments can be completed to quantify the maximum possible gains provided by the targeted method when compared to the current method(s) (Step 5). Finally, the suitability of the method to analyze case samples should be examined (Step 6) by running a set of representative case extracts using the targeted method. During these experiments, the sensitivity of the method can be adjusted, using the results from Step 3, to arrive at a method with an appropriate level of sensitivity for the sample preparation method used. These experiments also provide information on whether carryover may be a concern or if excipients, other drugs, or complex matrices might cause issues when using the method. Steps 5 and 6 can be completed in reverse order, or simultaneously, if desired. The process used for analyzing the data generated from these studies is discussed in greater detail in the Supplemental Information.

Development of a targeted GC-MS method for synthetic cannabinoids – Chemicals & Instrumentation

To demonstrate this framework, a synthetic cannabinoid method was created. For this project, a custom test solution was created by Cayman Chemical (Ann Arbor, MI, USA). The solution contained FUB-AMB, MMB2201, MDMB-FUBINACA, EMB-FUBINACA, AB-FUBINACA, ADB-FUBINACA, 5F-ABICA, and 5F-ADBICA, each at a nominal concentration of 100 μg mL⁻¹ in methanol. The compounds were chosen as they provide a range of elution times and pairs of compounds that may co-elute. In addition to the test solution, all compounds were also analyzed as individual components, in methanol, at a concentration of approximately 100 μg mL⁻¹. Additional synthetic cannabinoids (listed in Table 4) were analyzed in Step 4 and were purchased from Cayman Chemical.

Table 4.

Expanded list of cannabinoids analyzed on the targeted method (Step 4). Retention times and indices are the averages of ten replicate injections. Uncertainties represent the standard deviation of the ten replicates. The %RTD, the retention time difference (in minutes) and the percent difference in retention index (ΔRI) between neighboring peaks are also presented. Note that an asterisk (*) indicates instances where the retention times for the neighboring compounds were not found to be statistically different based on a Student’s T-test (95 % confidence level).

Compound	RT (min)	%RTD	RTD (min)	RI	ΔRI (%)
Δ−8-Tetrahydrocannabinol	1.630 (±0.002)	3.03	0.049	2768 (±3)	1.95
Δ−9-Tetrahydrocannabinol	1.679 (±0.003)	7.52	0.126	2822 (±5)	4.04
Cannabinol	1.805 (±0.003)	3.58	0.065	2936 (±4)	2.28
CP 47,497	1.870 (±0.003)	6.79	0.127	3003(±5)	2.70
CP 47,497-C8-homolog	1.997 (±0.002)	1.77	0.035	3084 (±3)	0.75
UR-144	2.032 (±0.003)	12.62	0.256	3107 (±5)	3.70
2-Fluoro-ADB	2.289 (±0.002)	4.54	0.104	3222 (±3)	0.53
3-Fluoro-ADB	2.392 (±0.003)	4.46	0.107	3239 (±5)	0.62
XLR11	2.499 (±0.002)	1.19	0.030	3259 (±3)	0.21
4-Fluoro-ADB	2.529 (±0.003)	6.12	0.155	3266 (±4)	0.92
5-Fluoro-AMB	2.684 (±0.003)	13.20	0.354	3296 (±4)	2.06
ADB-PINACA	3.038 (±0.004)	0.84	0.025	3364 (±5)	0.06
FUB-AMB	3.063 (±0.004)	0.95	0.029	3366 (±4)	0.21
AB-PINACA	3.092 (±0.004)	1.66	0.051	3373 (±4)	0.33
MDMB-FUBINACA	3.143 (±0.012)	3.11	0.098	3384 (±12)	0.50
EMB-FUBINACA	3.241 (±0.005)	7.94	0.257	3401 (±5)	0.71
JWH-203	3.498 (±0.004)	5.29	0.185	3425 (±4)	0.50
JWH-250	3.683 (±0.005)	1.98	0.073	3442 (±5)	0.20
AKB48	3.756 (±0.005)	3.44	0.129	3449 (±5)	0.35
AB-FUBINACA 2’-indazole isomer	3.886 (±0.005)	0.39	0.015	3461 (±5)	0.03
JWH-302	3.901 (±0.005)	4.52	0.176	3462 (±4)	0.49
MMB2201	4.077 (±0.005)	2.55	0.104	3479 (±5)	0.34
5-Fluoro-MDMB-PICA	4.181 (±0.007)	1.24	0.052	3491 (±6)	0.09
JWH-201	4.233 (±0.006)	3.27	0.138	3494 (±5)	0.34
AM2201 benzimidazole analog	4.371 (±0.006)	1.70	0.074	3506 (±5)	0.31
ADB-CHMINACA	4.446 (±0.006)	2.47	0.110	3517 (±5)	0.23
JWH-073	4.555 (±0.007)	0.23	0.011	3526 (±6)	0.00
AB-CHMINACA	4.566 (±0.006)	0.78	0.035	3526 (±5)	0.09
THJ2201	4.601 (±0.008)	0.39	0.018	3529 (±6)	0.09
MDMB-CHMICA	4.619 (±0.006)	0.09*	0.004*	3532 (±5)	0.17
AB-FUBINACA isomer 2	4.623 (±0.009)	5.23	0.242	3538 (±7)	0.57
ADB-FUBINACA	4.865 (±0.003)	2.67	0.130	3558 (±2)	0.28
AB-FUBINACA	4.995 (±0.008)	2.15	0.107	3568 (±6)	0.03
JWH-018	5.102 (±0.002)	1.79	0.091	3569 (±1)	0.42
5-Fluoro-AKB48	5.194 (±0.008)	4.03	0.209	3584 (±6)	0.56
AB-FUBINACA isomer 1	5.403 (±0.008)	8.13	0.439	3604 (±6)	0.61
AB-7-FUBAICA	5.842 (±0.012)	2.90	0.169	3626 (±7)	0.19
4-Cyano-CUMYL-BUTINACA	6.011 (±0.009)	4.05	0.244	3633 (±5)	0.33
JWH-122 N-(4-pentenyl) analog	6.255 (±0.010)	0.39	0.025	3645 (±6)	0.00
JWH-122	6.280 (±0.011)	3.35	0.210	3645 (±6)	0.38
THJ	6.490 (±0.012)	0.13*	0.009*	3659 (±7)	0.14
5-Fluoro ADBICA	6.499 (±0.009)	3.56	0.231	3664 (±5)	0.25
5-Fluoro ABICA	6.730 (±0.000)	1.00	0.067	3673 (±0)	0.05
JWH-210	6.797 (±0.012)	1.46	0.099	3675 (±7)	0.05
5-Chloro-AKB48	6.896 (±0.009)	22.12	1.525	3677 (±5)	2.12
JWH-081	8.421 (±0.015)	6.95	0.585	3755 (±7)	0.64
NM2201	9.007 (±0.016)	0.29	0.026	3779 (±7)	0.42
MAM2201	9.032 (±0.017)	15.25	1.377	3795 (±7)	1.34
APP-CHMINACA	10.410 (±0.011)	0.69	0.072	3846 (±4)	0.34
FDU-PB-22	10.481 (±0.008)	N/A	N/A	3859 (±3)	N/A

All analyses were completed on an Agilent 7890 / 5977-B GC-MS (Agilent Technologies, Santa Clara, CA, USA) system using ultra-pure helium as the carrier gas. A total of six different columns (DB-1UI, DB-5, DB-5UI, DB-35, DB-200, and VF-1701ms) were evaluated, all of which had dimensions of 30 m x 0.25 mm x 0.25 μm and were purchased from Agilent Technologies. All columns were conditioned according to the manufacturer’s suggested protocols. The MS was tuned daily using the standard spectra tune (stune), unless otherwise noted. Relevant MS parameters, that were kept constant throughout the method development, include an MS scan range of m/z 40 to m/z 550, a threshold of 150 counts, and a scan speed of N = 2. Additional method parameters are provided throughout the text and in the Supplemental Information.

Results and Discussion

Column Comparison (Step 1)

In Step 1 a unified method, the parameters of which are shown in Supplemental Table 1, was used for analysis of test solution on all six columns. The results from the six different stationary phases are shown in Figure 2 and Table 1. The unified method allowed for elution and detection of all compounds in the test solution on all columns, however co-elution of at least one pair of compounds was observed using the DB-1 UI, DB-5, and DB-5UI columns. Using the DB-35 column, AB-FUBINACA and ADB-FUBINACA were barely separated (0.02 %RTD), and two other pairs had %RTDs of less than 0.20 % (Figure 2A). Overall, the more polar stationary phases, DB-200 and VF-1701ms, led to greater %RTDs and greater retention time differences. The elution times between the two columns varied significantly, with later elution times and larger spans of elution times being observed on the VF-1701ms column. Representative chromatographs of the test solution analyzed on each of the columns can be found in Supplemental Figure 1.

FIG. 2 –

Results of the column comparison study (Step 1). Retention time differences (%RTDs) between subsequent compounds, sequentially labelled 1 through 8 because of changes in elution order depending on the column used, is shown in (A.), where points closer to the center of the web indicates a higher probability of having overlapping retention time acceptance windows. Note that (A.) is plotted on a log scale. The average peak areas (B.) and peak purities (C.) for the each of compounds analyzed on each column are also presented. Uncertainties indicate the standard deviation of triplicate measurements.

Table 1.

Summary results of key metrics for the different columns examined (Step 1). Uncertainties indicate the standard deviation of the averages obtained for each of the eight compounds in the test solution. The notation “CEP” denotes instances where there were overlapping co-eluting peaks. Weighted match factors were obtained from AMDIS using a subset of spectra from the SWGDRUG library (version 3.6). See Supplemental Information for more information.

Stationary Phase	Maximum RT (min)	Minimum %RTD (%)	Avg. Weighted Match Factor	Avg. Peak Width (min)
DB-1 UI	8.94	CEP	91.4 (±8.4)	0.07 (±0.04)
DB-5	9.53	CEP	90.3 (±5.2)	0.08 (±0.04)
DB-5 UI	16.76	CEP	84.6 (±15.8)	0.15 (±0.13)
DB-35	15.52	0.05 (±0.05)	89.1 (±8.6)	0.15 (±0.08)
DB-200	14.33	1.19 (±0.01)	92.5 (±7.1)	0.16 (±0.05)
VF-1701ms	29.79	1.44 (±0.03)	75.9 (±14.4)	0.25 (±0.05)

Differences in peak area (Figure 2B), peak purity (Figure 2C), mass spectral quality (Table 1), and peak width (Table 1) were also compared. Generally, the DB-200 column produced the largest peak areas while the VF-1701ms column produced the smallest. Peak purity measurements, which provide an estimate of the noise in the mass spectrum, were fairly consistent across the columns, though column bleed at the end of the VF-1701ms stationary phase runs was shown to be detrimental to the purity measurements of the late eluting compounds like 5F-ABICA and 5F-ADBICA. Mass spectral quality scores were consistent across the stationary phases with the exception of VF-1701ms which was also likely due to column bleed. Peak width increased with increasing polarity of the stationary phase, as expected, though overall peak shape remained acceptable in all instances. Given these results, the DB-200 stationary phase was chosen as the column to use for development of the targeted method.

Assessment of Temperature and Flow Parameters (Step 2)

Investigation of the temperature and flow parameters was completed using a series of ramped and isothermal runs combined with constant flow rates of 0.8 mL min⁻¹, 1.2 mL min⁻¹, or 2 mL min⁻¹ analyzed in triplicate. Table 2 presents the different programs that were examined as well as the minimum and median %RTDs and the maximum overall retention time for each run. Representative chromatograms for each of the conditions listed in Table 2 can be found in Supplemental Figure 2.

Table 2.

Summary results from the temperature and flow evaluation runs (Step 2). The notation “CEP” denotes instances where there were co-eluting peaks. Uncertainties represent the standard deviation for triplicate measurements.

Temperature Program	Flow Rate (mL min−1)	Minimum %RTD (%)	Median %RTD (%)	Maximum Retention Time (min)
250 °C Isothermal	2.0	2.75 (±0.07)	4.14 (±0.01)	6.59
290 °C Isothermal	0.8	2.80 (±0.00)	4.23 (±0.06)	9.62
290 °C Isothermal	1.2	2.70 (±0.00)	4.31 (±0.04)	7.92
290 °C Isothermal	2.0	2.73 (±0.13)	4.16 (±0.05)	6.54
200 °C – 290 °C at 2 °C min−1	1.8	1.45 (±0.01)	3.96 (±0.04)	35.55
240 °C – 290 °C at 2 °C min−1	1.2	2.10 (±0.04)	4.80 (±0.03)	21.85
240 °C – 290 °C at 2 °C min−1	2.0	2.25 (±0.02)	5.19 (±0.06)	19.25
240 °C – 290 °C at 5 °C min−1	2.0	CEP	3.74 (±0.05)	12.82

Isothermal runs provided better separation of nearly all compounds within the test solution compared to ramped temperature runs while also providing higher minimum %RTDs. Comparison of different flow rates using the same temperature program revealed nearly identical %RTDs between all pairs of compounds. Altering the isothermal temperature in the range of 250 °C to 290 °C did not affect the elution times, peak areas, or peak widths of compounds when equivalent flow rates were used. Because of this, a 290 °C isothermal run was chosen to minimize the potential of carryover and aid in elution of other compounds that may be encountered in case samples. A flow rate of 1.2 mL min⁻¹ was chosen to provide flexibility for variations in flow rates due to retention time locking.

Maximizing Sensitivity and Reproducibility (Step 3)

A two-level, 2⁴⁻¹ partial factorial design of experiments (DOE) approach was used to study the effect of MS source temperature, split ratio, injection volume, and inlet temperature on the sensitivity and reproducibility of the method. Peak area and peak height were used as measures for sensitivity while the percent relative standard deviation (%RSD) of the peak area, peak height, retention time, and %RTD across triplicate measurements were used as measures of reproducibility. The levels chosen each of the parameters in the DOE were 230 °C and 280 °C for the source temperature, 10:1 and 30:1 for the split ratio, 0.5 μL and 2.0 μL for the injection volume, and 200 °C and 300 °C for the inlet temperature. An outline of the DOE setup can be found in Supplemental Table 2.

The results for the DOE are shown in Figure 3. A Student’s T-test was completed on outputs disaggregated by each pair of parameters (e.g., the peak areas for source temperatures of 230 °C and 280 °C) to determine which, if any, were statistically different at a 95 % confidence level. A statistically significant difference was found, as expected, when comparing injection volumes for peak area (p = 0.045) and peak height (p = 0.029), due to the higher amount of material injected. No statistical differences, however, were found for the reproducibility measures. For inlet temperature, no significant difference was found in peak area or peak height, however a significant difference was observed for reproducibility (%RSD) of peak area (p = 0.005), peak height (p = 0.020), and retention time (p = 0.049). All metrics showed better reproducibility was obtained using an inlet temperature of 300 °C, which was chosen for the targeted method. An injection volume of 2 μL was initially chosen, given that no statistically significant difference in reproducibility was observed, since it could be lowered if the method was deemed too sensitive. Given there was no statistical difference in source temperature or split ratio, a source temperature of 280 °C was chosen to minimize potential build up in the source and a split ratio of 10:1 was chosen to increase sensitivity.

FIG. 3 –

Results of the DOE study for source temperature (A. and E.), split ratio (B. and F.), injection volume (C. and G.), and inlet temperature (D. and H.) (Step 3). Peak area (blue) and peak height (orange) are shown in A. through D. while the average %RSD of peak area (blue), peak height (orange), retention time (green), and %RTD (yellow) across triplicate injections are shown in E. through H. Boxes with an asterisk (*) indicate settings where a statistically significant difference was observed at the 95 % confidence level.

Prior work has demonstrated that tune type can affect the resulting mass spectra and spectral reproducibility, though frequency of tuning is more important than the type of tune used(20). The test solution was analyzed in triplicate using both autotune (atune) and standard spectra tune (stune) after which the peak areas and mass spectral search scores were compared. Between the two tune types, mass spectral search scores were above 90 a.u. and within 1 a.u. of each other for all compounds except for MDMB-FUBINACA. Peak areas were approximately equal for both tune types. The stune was chosen since it is the tune currently used at MSP-FSD and no obvious advantage was gained from atune. The final settings for the targeted method are presented Table 3.

Table 3.

Settings for the targeted method. Settings in parenthesis represent those chosen after the analysis of representative case samples (Step 5b), which required a decrease in sensitivity.

Column	DB-200 30 m × 0.25 mm × 0.25 μm
Temperature Program	Isothermal at 290 °C
Flow Rate	1.2 mL min−1
Injection Volume	2.0 μL (1.0 μL)
Inlet Temperature	300 °C
Split Ratio	10:1 (30:1)
Transfer Line	300 °C
Quad Temperature	150 °C
Source Temperature	280 °C
Tune Mode	stune
Solvent Delay	1.4 min
Mass Scan Range	m/z 40 – m/z 550
Threshold	150 counts
Scan Speed	N = 2 [≈4 scan s−1]
Total Run Time	12.0 min

Analysis of Additional Cannabinoids (Step 4)

Evaluation of the method began with the analysis of an expanded compound set consisting of a total of 50 cannabinoids, listed in Table 4, which comprised all available standards at the laboratory. During the initial analysis of the compounds it was found that the method needed to be extended by one minute to allow for elution of all compounds, after which the method was locked to the retention time of AB-FUBINACA (4.995 min). Locked retention times for the expanded panel of cannabinoids were then established through ten replicate analyses of multi-component mixtures containing the compounds. An even-numbered alkane ladder was incorporated into the sequence so retention indices could be computed.

Table 4 shows the average retention times and retention indices obtained using the locked targeted method. Retention times ranged from 1.630 min to 10.481 min, with retention indices ranging from 2768 to 3859. All compounds were separable, though not baseline separable, with a minimum %RTD between neighboring compounds of 0.09 % (0.004 min) and a maximum of 22.12 % (1.512 min). The average %RTD between neighboring compounds was 4.22 % (0.192 min) with a median of 3.15 % (0.107 min). Reproducibility of retention times was high, with an average and median %RSD of 0.14 %, and a max %RSD of 0.20 % respectively. Comparison of retention times of neighboring compounds using a Student’s T-test showed all but two pairs were statistically different at the 95 % confidence level, MDMB-CHMICA | AB-FUBINACA isomer 2 (p-value = 0.22) and THJ | 5-Fluoro-ADBICA (p-value = 0.08).

Next, identification of compounds pairs that could not be differentiated by retention time was established. While recent work has suggested a retention time acceptance window as small as 0.2 % can be used to distinguish compounds(21), a 2 % window (a 2 %, %RTD value) was chosen as it is one of the most conservative windows used in the field and therefore provides the most exhaustive list of instances where discrimination based on mass spectra may be required. Using the 2 % window a total of 27 pairs of compounds, listed in Supplemental Table 3, were found to have acceptance windows that would overlap. The min-max test, which is explained in detail in Supplemental Information, was applied to these 27 pairs to determine which could be differentiated based off their replicate mass spectra. The min-max index (Supplemental Equation 1) was calculated for each pair to provide an aggregate measure of how different the spectra between compounds are. Higher values (on the scale of −999 to 999) indicate higher dissimilarity and negative values are definite indications that the spectra between the two compounds are not numerically discernable. None of the 27 min-max indices were negative and a minimum index of 116 was obtained when comparing JWH-122 N-(4-pentenyl) analog to JWH-122, which was sufficiently high for compound differentiation. Generally, min-max indices were greater than 200 indicating that all 27 pairs could be sufficiently differentiated based on their mass spectra. Therefore, all 50 cannabinoids could be differentiated based on either retention time or mass spectral similarity.

Since some laboratories use a fixed retention time difference for their acceptance window, the same study was completed on all compound pairs where the retention time difference was less than or equal to ±0.1 min. This approach produced four additional compound pairs that needed to be investigated, none of which had negative min-max indexes.

Limits of Detection and Comparison to the General Screening Method (Step 5)

The next evaluation step compared the targeted method to the general confirmatory method currently used at the laboratory, the parameters of which can be found in Supplemental Table 4. Using the most sensitive settings (a 2 μL injection volume and 10:1 split ratio) analysis of the test solution produced an overall average peak area increase of 6518 %. Peak area increases ranged from 14-fold to 93-fold, depending on the compound, as shown in Supplemental Table 5. More importantly, the targeted method led to major improvements in %RTD values of neighboring peaks. Unlike the general method, which had multiple instances of co-eluting peaks as shown in Figure 4, the targeted method produced peaks retention time differences of at least 2.5 % for all compounds in the test solution. The increased separation of compounds was accomplished using a method that was 1.33 min shorter than the general confirmatory method studied. Approximate LODs were also measured (Supplemental Table 5) and were between a factor of two and ten lower for the targeted method using the most sensitive settings. The approximate LODs ranged from 1 μg mL⁻¹ to 10 μg mL⁻¹ for the targeted method and from 10 μg mL⁻¹ to 25 μg mL⁻¹ for the general method. Details on how the approximate LODs were obtained can be found in the Supplemental Information.

FIG. 4 –

Comparison of the test solution analyzed using the targeted method (red) and the general confirmatory method (blue) (Step 5a). Note that the general method chromatogram (blue) is plotted on a secondary axis.

Evaluation of the Method with Case Samples (Step 6)

The final portion of this work looked to evaluate the suitability of the targeted method against a set of eight representative, adjudicated case samples to identify if any changes to the method were required prior to completing a full validation. Given the sample preparation method employed, details of which are provided in Supplemental Information, the targeted method was found to be too sensitive as peak saturation for some extracts was readily observed. This was not unexpected given the low concentration of the test mixture relative to the case samples. The method was desensitized, based on the results of Step 3, by lowering the injection volume to 1 μL and increasing the split ratio to 30:1, as noted in Table 3. Using these settings, the synthetic cannabinoids, previously identified using the general confirmatory method, were readily identified in all case samples and co-elution of compounds of interest with cutting agents was not observed. Carryover was also not observed. Chromatographs for each of the samples can be found in Supplemental Figure 3. Peaks heights were in the range of 10⁵ cps to 10⁷ cps and retention times between the case samples and the standards were within the ±2 % acceptance window even though the retention times for the standards were collected two months earlier than the case samples using the more sensitive settings. Current efforts are focused on completing a full validation study of the method so that it can be incorporated into the workflow at MSP-FSD.

Continued Expansion of the Method and Data Availability

Method parameters, raw mass spectra of standards, and other relevant information are freely available to download(22). Current efforts are focused on adding additional compounds to the method. As they are added they will be updated here(22).

Conclusion

This paper presents a framework that can be leveraged and adapted for the development of targeted GC-MS methods for seized drug analysis. As a test case, a targeted method for synthetic cannabinoids was established. The method, which differed significantly from the general confirmation method used at the laboratory, utilizes a different stationary phase, MS source temperature, and inlet temperature as well as an isothermal temperature program. Studies completed to evaluate split ratio and injection volume – the two factors that provide the largest change in sensitivity – showed that the sensitivity of the method can be adjusted without compromising reproducibility. A total of 50 cannabinoids were analyzed using the method, and differentiation based on retention time or mass spectral similarity was achieved for all. Analysis of case samples showed that the method is likely suitable for casework, though this is being definitively established in ongoing validation studies. As with any analytical method, this approach is not without its limitations. Confident compound differentiation can only be achieved for the panel of compounds investigated, though new compounds can be readily added to the panel. To do so, replicate measurements of the new compound would have to be made after which the retention time differences and mass spectral similarities of that compound to all other compounds in the method panel could be compared. Since the method was designed for synthetic cannabinoids and uses a high temperature isothermal oven program it may preclude the detection of lower volatility drugs such as cathinones. This is why targeted methods should be used either in conjunction with information-rich screening techniques like DART-MS or to supplement generic GC-MS methods. Additionally, since the method uses a high-polarity column, retention time drift should be closely monitored, and is currently being studied, as column degradation may occur more rapidly than with less polar columns. Development of targeted GC-MS methods is part of a larger project to evaluate the benefits and drawbacks of implementing different analytical procedures in the screening and confirmatory steps of seized drug analysis. Ongoing work is focused on the development, and validation, of additional targeted methods and on the comparison of workflows utilizing targeted methods versus general confirmatory methods. Strategies for incorporating internal standards and utilizing locked retention times or retention indices are also being developed.

Table 5.

Identities and results from the analysis of representative case samples (Step 5b). Only results for the synthetic cannabinoids are listed. Samples originated from powders unless otherwise noted.

Case #	Case Contents	Cannabinoid Peak Height (cps)	Cannabinoid RT (min)	Standard RT (min)	MS Match Score (Weighted)
1	FUB-AMB	1.8×106	3.018	3.063	97
2	FUB-AMB (Plant Material)	7.8×105	3.030	3.063	97
3	5-Fluoro-AKB48	1.5×107	5.160	5.194	97
4	5-Fluoro-AKB48, Mannitol, α-PBP	4.9×105	5.077	5.194	94
5	JWH-018	1.4×106	5.002	5.102	93
6	JWH-250, JWH-073, JWH-018 (Plant Material)	9.7×105 7.8×105 2.2×106	3.645 4.490 5.015	3.683 4.555 5.102	92 96 93
7	JWH-250	3.4×106	3.644	3.683	96
8	XLR11	2.1×106	2.487	2.499	92

Highlights.

• A framework for the development and evaluation of targeted GC-MS methods is presented.

• Results for a synthetic cannabinoid method are presented with a total of 50 compounds analyzed.

• An automated approach was created to determine if spectra from two compounds are differentiable.

• The method was quicker than the general method and able to differentiate all compounds studied.