All chemicals used were analytical reagent grade or better. Chloroform, ethyl acetate, isopropanol and methanol were purchased from Fisher Chemicals; eicosane was purchased from Acros Organics; diazepam, diazepam-d₅, flunitrazepam and temazepam were purchased from Sigma Chemical Company; bis(trimethyl)trifluoroacetamide (BSTFA) containing 1% trimethylchlorosilane (TMCS) derivatising agent was purchased from Supelco Analytical.

The beverages used were Bacardi Breezer, an alcopop, orange flavour, 4% by volume, Becks Beer, 5% by volume, Hardy’s Classic Selection Chardonnay 2010, 12.5% by volume, J2O apple and mango, Chekov Imperial Vodka, 37.5% by volume, and were all purchased from local supermarket stores in Cambridge, UK during the period Oct 10–July 2011. The five drinks tested in this research were selected to represent common drink types consumed by women in the 16–24 years age group – a group at considerable risk of being the victim of DFSA.

A mixed drug standard of diazepam, flunitrazepam and temazepam was prepared with each at a concentration of 100 µg/mL in methanol. When appropriate, the internal standard (IS), diazepam-d₅, was added to give a final IS concentration of 25 µg/mL.

To derivatise the drugs prior to analysis owing to thermal instability of temazepam, each solution was evaporated to dryness using a miVac sample concentrator (Genevac Ltd., UK) at 30°C. Derivatisation was performed by addition of BSTFA: TCMS (40 µL, 99∶1%) and IS eicosane (20 µL, 100 µg/mL) in ethyl acetate and heating to 70°C for 15 minutes. The samples were analysed immediately after derivatisation.

A Shimadzu QP 2010 GC-MS fitted with a ZB1 column (0.25 µm layer thickness, 0.25 mm i.d.) was utilised with split injection (14∶1) at 250°C. The injection volume was 1 µL. The column flow rate (He carrier gas) was 1.2 mL/min. The oven temperature programme was 150°C (2 min), then increasing at a rate of 20°C/min to 280°C and held at 280°C for 5 minutes. The quadrupole mass analyser was operated in electron impact (EI) mode. The ion source temperature was 230°C, interface temperature was 300°C and solvent delay time was 3 minutes. Samples were analysed in full scan and extracted ion modes which were used for data analysis. Ethyl acetate blanks were run between all samples, controls and standards to prevent carry over. Compounds were identified on the basis of retention index (compared to diazepam-d₅) and mass spectrum. An unspiked sample of each of the beverages was prepared using the extraction and derivatisation method and was analysed in an identical manner to all other samples. This was to demonstrate that there were no drugs or compounds giving identical data in the reagents and drink matrices. In addition, linearity on the drug response was established using pure drug standards as well as spiked beverages.

The Bacardi Breezer, Becks Beer, Hardys Chardonnay, Chekov Imperial Vodka and J2O (20 mL) were each spiked with individual drug solutions to give a final concentration of 10 µg/mL (diazepam and temazepam) and 20 µg/mL flunitrazepam. The difference in concentration is owing to the instrumental response. The beer and the Bacardi Breezer were degassed using an ultrasonic bath prior to addition of the drugs. Spiked beverages were divided into two aliquots (10 mL each) and stored at room temperature, which fluctuated diurnally, and in the refrigerator at 4°C, until liquid liquid extraction (LLE) and GC-MS analysis was carried out.

The IS (diazepam-d₅) was added to individual drinks (0.5 mL aliquots), to give a final concentration of 25 µg/mL, immediately prior to extraction. LLE was performed using chloroform: isopropanol (1∶1 v/v, 100 µL×3) directly after spiking and on unspiked beverages [18]. Once extracted, the organic phase (bottom) was collected and subjected to analysis.

Each of the beverages (10 mL) was spiked using individual drug solutions of 1 mg/mL diazepam and temazepam; and 2 mg/mL flunitrazepam to give the concentrations range of 2.5–50 µg/mL for diazepam and temazepam; 5–100 µg/mL for flunitrazepam. Analytes were extracted, derivatised and analysed as described above.

White wine samples were spiked with individual drugs to give final concentrations of 0.833 mg/mL each of diazepam and temazepam to simulate casework samples. This concentration was selected within the linear range. Each sample was analysed using the extraction and derivatisation method described above. Once the analysis was complete the identity of the drugs and the quantities detected were compared to those expected.

The GC-MS method reported here was followed after derivatisation that allowed identification of temazepam, whereas diazepam and flunitrazepam could have been analysed without derivatisation. Sample injection without derivatisation of temazepam resulted in its decomposition. Temazepam is known to be thermally labile [24]. The data obtained for the standard drugs and drugs extracted from the drinks were identical with respect to retention index (Table 1) and mass spectrometry (Table 2). This allowed identification and quantification of the drugs. Solvent blanks and sample matrices were analysed to ensure there was no carryover between sample injections. Results from the matrix blank and negative controls demonstrated the absence of drugs in these samples. This proves that the chromatographic peaks and mass spectroscopic data could only have come from the drugs themselves and that there was nothing in any of the drinks tested which gave identical data indicating selectivity of the method.

Extracted ions (diazepam m/z = 256; diazepam-d₅ m/z = 261; flunitrazepam m/z = 343 and temazepam m/z = 312) were utilised to construct calibration graphs for the pure drugs – the regression equations and r² values are provided in Table 3. A linear detector response range was established for each drug in the ranges 0.025–0.25 mg/mL for diazepam and temazepam; and 0.05–1 mg/mL for flunitrazepam. The limit of detection (LOD) and limit of quantification (LOQ) were calculated using IUPAC (International Union of Pure and Applied Chemistry), 2009 [25] and are also given in Table 3. LOD ranged from 0.50–0.71 ng and LOQ ranged from 0.64–1.29 ng on column.

Calibration curves were also constructed for the drugs extracted from the drinks in a fashion analogous to those for pure compounds. The LOD and LOQ data are given in Table 4. The recovery of each of the drugs from each of the drinks was measured (Table 4) with values of diazepam 84–99%, flunitrazepam 56–101% and temazepam 81–105% depending upon the drink (Table 4). Recovery was least good from the white wine and flunitrazepam was particularly poor from the white wine (56%) and the Bacardi Breezer (77%). The reasons for this are more complicated than recovery merely being a function of the pH values of these drinks which lay in the range between pH 3.2 to pH 3.4 and the pK_a value of flunitrazepam (pK_a = 1.8) otherwise temazepam (pK_a = 1.6) should have been extracted equally poorly. The factors affecting recovery were not investigated further. Recovery of all drugs from both the beer and the vodka were good with mean values of 84–101%.

All of the spiked beverages stored at room temperature and at 4°C were subjected to LLE at the given time intervals. The results show that the drugs were detected under both storage conditions throughout the study period, although some drugs were more stable than others. Exemplar results for analytes extracted from drinks stored at room temperature for example, diazepam (Figure 1) flunitrazepam (Figure 2) and silylated temazepam (Figure 3) show a clear difference between the stability of the individual analytes in different beverages. In order to determine whether or not there is a significant difference in these results the Kruskal-Wallis U test was used. The chi-squared (χ2) statistic and p value were calculated and compared with the critical χ2 value at 95% confidence interval for 3 degrees of freedom and the summary of results are given in Table 4.

The Kruskal-Wallis U test showed that for diazepam there was no significant difference between the results obtained after storage at room temperature and at 4°C in any beverage except J2O (Figure 1, Table 4). In J2O, at room temperature, there was evidence of diazepam degradation. The cause of this was not investigated further. For flunitrazepam significant differences were observed after storage under both conditions in wine and J2O (Figure 2, Table 4). The instability is not linked to the sample pH and alcoholic content of the beverage because both of them have (i) same pH of 3.2 (ii) the wine is 12.5% v/v ethanol, the J2O is alcohol free. There was significant loss of temazepam in all the beverages analysed after storage in both conditions (Figure 3, Table 4) except beer at 4°C. In general the stability of benzodiazepines depended on the type of beverage in which they were dissolved. J2O in particular showed significant differences for all the analytes under both storage conditions.

Factors which might potentially impact upon the stability of benzodiazepines in a beverage are pH, alcohol concentration, extraction and analysis. Our results do not show any pattern related to them in the Kruskal-Wallis U significance tests.

Blind trial samples were extracted, derivatised and analysed following the method developed and presented in this paper. The drugs were each correctly identified. The concentrations of the drugs in the drinks lay between 95–100% of the true values. Against a true value of 0.083 mg/mL, the concentration of diazepam was detected at 0.079 mg/mL (95% recovery), 0.083 mg/mL (100% recovery) for temazepam from white wine.