Galenic Preparations of Therapeutic Cannabis sativa Differ in Cannabinoids Concentration: A Quantitative Analysis of Variability and Possible Clinical Implications

Introduction: Magistral preparations of therapeutic cannabis are extracted from standardized products imported from Holland or from the Florence Military Pharmaceutical Chemical Works, but extraction protocols differ among galenic laboratories. This study assessed the inter-laboratory variability in concentrations of cannabidiol (CBD), cannabinol (CBN), tetrahydrocannabinol (THC), and tetrahydrocannabinolic acid (THCA) among different magistral oil preparations. Methods: 219 samples of Bediol, Bedrobinol, Bedrolite or FM-2 70 or 100 mg/ml in oil were collected from 3 laboratories. Concentrations of CBD, CBN, THC, and THCA were quantified by high-pressure liquid chromatography; inter-laboratories variability was assessed using the Kruskal–Wallis test. Results: A significant variability in CBD and THC concentrations was found for Bediol 70 mg/ml samples from 2 laboratories [for CBD: median 5.4 (range 4.8–6.6) vs. 6.1 (4.9–7.2) mg/ml, p = 0.033; for THC: 3.6 (3.1–3.9) vs. 4.0 (2.6–5.1) mg/ml, p = 0.020]. As for Bediol 100 mg/ml, a significant variability emerged in THC concentrations among the three considered laboratories [5.7 (-) vs. 4.2 (1.5–4.8) vs. 5.2 (4.2–6.9), p = 0.030]. No significant inter-laboratory variability emerged for Bedrocan and Bedrolite. Concentrations of CBD, CBN, and THC were <LOQ in all Bedrocan samples, and CBN and THCA were <LOQ in all Bedrolite samples. As for FM-2, a significant inter-laboratories variability was found for CBD concentrations. Conclusion: Quantitative variability of cannabinoids in magistral preparations might impact on the efficacy and safety of therapeutic cannabis. A standardized protocol is needed to guarantee a homogeneous product and patients’ therapeutic continuity.


INTRODUCTION
In Italy, use of Cannabis sativa for Therapeutic Purposes (CTP) was first authorized in 2006. Indications for its use include chronic pain, nausea and vomit associated to chemotherapy, appetite stimulation, hypotension effect in glaucoma, and reduction of uncontrolled body and facial movements (Ministero, 2016).
The main constituent of raw cannabis is tetrahydrocannabinolic acid (THCA) (Moreno-Sanz, 2016). Following heating, THCA is converted to 9-THC, a partial agonist on both cannabinoid receptors (CB1 and CB2), that mediates the well-known psychoactive effects of cannabis (Howlett and Abood, 2017). Another active principle, cannabidiol (CBD), has only a modest affinity for the cannabinoid receptors, thus has no psychoactive effects. However, both THC and CBD have a number of additional pharmacological targets, including the calcitonin gene-related peptide (GPR55) orphan receptor, the ligand-gated ion channels of human 5-HT3A receptors, and several additional ionic channels and enzymes (Pertwee, 2008;De Petrocellis et al., 2017).
Magistral preparations of CTP are prepared by extraction from standardized products, obtained from dried and minced cannabis inflorescences and containing standardized THC and CBD concentrations, that are imported in Italy from the Dutch Office of Medicinal cannabis (Ministero, 2016): Bedrocan R (mean amounts of 22% for THC and <1% for CBD), Bedrobinol R (13.5% for THC and <1% for CBD), Bediol R (6.5% for THC and 8% for CBD), and Bedrolite R (0.4% THC and 9% CBD) (Office of Medical Cannabis, 2018).
In addition, since 2016 a national production of the cannabis product FM-2 R has been started in the Military Pharmaceutical Chemical Works of Florence. In the FM-2 R product, mean amounts of THC and CBD range between 5-8% and 7.5-12%, respectively (Ministero, 2016).
According to the European Pharmacopeia, several magistral products can be prepared, including cannabis decoction filter bags, unit dose formulation for inhalation and cannabis extracts, mainly in olive oil (Council of Europe, 2017).
Although the preparation of cannabis oil is relatively simple (Romano and Hazekamp, 2013), the qualiquantitative composition of the final products is poorly characterized, particularly considering that variations in temperature and time of extraction may impact on the final concentration of cannabinoids (Citti et al., 2016;Pacifici et al., 2017).
Recently, a study conducted on over two hundred extracts highlighted a wide variability in THC and CBD concentrations among different preparations of Bedrocan R , Bediol R , and Bedrolite R 5 g/50 ml in olive oil, both inter-and intra-laboratory (Carcieri et al., 2018). In light of these recent findings, our study aimed to provide additional evidence on the variability in terms of concentrations of cannabinoids in different oil preparations from different Italian laboratories, focusing on oil preparations of Bedrocan R , Bediol R , Bedrolite R , and FM-2 R 70 and 100 mg/ml.

MATERIALS AND METHODS
Samples of Bedrocan R , Bediol R , Bedrolite R , and FM-2 70 and 100 mg/ml prepared by three different community pharmacies (named Lab1-Lab4) of the Metropolitan Area of Florence, Italy, were collected.
Samples were analyzed at the Department of Clinical Toxicology and Antidoping of the Local Health Authority of Florence, Italy.
For buffer solutions, MilliQ water was obtained from the purification system Millipore R (Billerica, MA, United States), whereas K2HPO4 e HCl were purchased from Carlo Erba (Milan, Italy).

Chromatographic Analysis
High pressure liquid chromatographic analysis with Diode-Array Detection (HPLC-DAD) was performed on a Thermo-Fisher Surveyor Plus (Thermo Fisher Scientific, Waltham, MA, United States).
Briefly, 4 preparations of KH2PO4 5 mM (Solution A) at pH of 3.45 were prepared. The calibration lines for analytes were prepared from stock solution of 0.4 mg/ml THC in methanol and of 1 mg/ml CBD in methanol, that were further diluted in methanol to the final concentration of 0.1 mg/ml. Standard solutions were further diluted to obtain six different samples, with concentrations ranging between 0.001 and 0.040 mg/ml.
Conditioning was performed according to the procedure reported in the data sheet of the column, i.e., with the flow of 10-20 volumes at the speed of 0.4 ml/min over 26 min. The adopted work conditions were: mobile phases A, ACN/5 mM phosphate butter pH 3.45 at the rate of 75/25 v/v; flow, 0.38 ml/min, temperature of 53 • C; injection volume 10 µl full loop; pressure ∼ 210 bar; detector UV channel 222 nm; time of 8 min (isocratic).

Sample Preparation
Samples were stored at room temperature, until time of analysis.
Forty microliter of the sample were diluted in 960 µl THF. After Vortex mixing, 50 µl of such solution were added to 950 µl of ACN, and the solution was mixed using the Vortex. 10 µl of this solution were injected in the chromatographic system.

Software
Obtained chromatographs were analyzed using the software ChromQuest TM , version 4.2.34 (Thermo Fisher Scientific).

Statistical Analysis
Concentrations of different active principles in analyzed samples were expressed both as mean value and related standard deviation (SD) and as median value and range (min-max). For the statistical analysis, the value 0.9 mg/ml was arbitrarily attributed to concentrations below the limit of quantification (<LOQ; i.e., <1 mg/ml).
Differences in mean and median concentrations among different pharmacies were tested using the One-way ANOVA test and the Kruskal-Wallis test, respectively.
Statistical analysis was performed using the Software STATA version 14. Statistical significance was considered for p-value < 0.05.
Concentrations of CBD, CBN, THC, and THA in analyzed cannabis olive oil preparations of Bediol, Bedrocan, Bedrolite, and FM-2 are summarized in Table 2 and expressed as mean value and SD, and as median value and related range. Median value and related interquartile range and range are further illustrated in the Supplementary Figures S1, S2.
As for Bediol 70 mg/ml, median CBD concentration in the 23 analyzed samples was of 5.6 (4.8-7.2) mg/ml, with a statistically significant variability between the two pharmacies from whom samples were derived (p-value from Kruskal-Wallis test = 0.033). As for THC, median concentration was of 3.7 (2.6-5.1) mg/ml, with significant variability between the two pharmacies (p = 0.020). Median CBN and THCA concentrations were <LOQ, with no variability between the two pharmacies.
No statistically significant difference emerged in the concentrations of the active principles CBD, CBN, THC, and THCA among Bedrocan 70 or 100 mg/ml prepared in the three different pharmacies. Of particular note, median levels of CBD, CBN, and THC in Bedrocan samples were <LOQ in all three pharmacies. Also considering Bedrolite 70 or 100 mg/ml, no difference in the median concentrations of CBD, CBN, THC, and THCA emerged among the three pharmacies. Of note, concentrations of CBN, THC and THCA were <LOQ in all 14 samples of Bedrolite 70 mg/ml, and concentrations of CBN and THCA were <LOQ also in all 5 samples of Bedrolite 100 mg/ml.
Considering FM-2 preparations, we found that median CBD concentrations significantly differed between the pharmacies, both considering FM-2 70 or 100 mg/ml.

DISCUSSION
Increasing literature evidence supports the use of therapeutic cannabis in the treatment of different painful or degenerative clinical conditions (Balash et al., 2017;Neale, 2017;Habib and Artul, 2018;Landa et al., 2018;Lattanzi et al., 2018;Poli et al., 2018), although dosing standards and quality assurance of cannabis-derived medications is still a matter of debate.
Our study highlighted a significant inter-laboratory variability of cannabinoids concentrations among magistral oil preparations. The highest variability in the extraction yields was observed for Bediol R -based preparations, both considering CBD and THC concentrations. Significant variability was observed also for CBD concentrations in FM-2 preparations.
Bedrocan R -based preparations had the highest extraction yields of THC, whereas CBD, CBN, and THCA were undetectable in these samples; on the other hand, Bedrolite R -based    (1.000)

1.000
Bedrolite 100 mg/ml These results are in line with a recent study by Carcieri et al. (2018), highlighting a wide inter-and intra-laboratory variability in THC and CBD concentrations in different preparations of Bedrocan, Bediol, and Bedrolite 100 mg/ml. Specifically, authors reported that Bediol R -based preparations had significantly higher extraction yields both of THC as compared to Bedrocan R and of CBD as compared to Bedrolite R .
Our study confirms such variability in concentrations of active principles in Bedrocan, Bediol, and Bedrolite magistral preparations, while adding new evidence on the variability of extraction yields in FM-2 preparation. As extensively described in Carcieri et al. (2018), variability in active pharmaceutical ingredients can be explained by the lack of a unique standardized extraction protocol (Williamson and Evans, 2000;Ware and Tawfik, 2005) as well as by the (although declared) variability in the concentrations of cannabinoids in the imported products (Office of Medical Cannabis, 2018).
Taken together, our results remark the need of introducing a standardized protocol and of providing concentrations data for each preparation; furthermore, these findings open a question on the possible clinical implications of such variability.
A recent study evaluating the effects of different doses of CBD on anxiety in healthy individuals showed that the anxiolytic effect of CBD follows a U-shaped dose-response curve (Linares et al., 2018): personal anxiety was reduced with CBD 300 mg, but not with CBD doses of 100 and 900 mg.
Besides implications on efficacy, dose variability may be associated with safety concerns.
In a phase 1 randomized controlled trial (Ahmed et al., 2014), high THC doses (6.5 mg vs. 3 mg) were associated with significantly higher rates of adverse events. Furthermore, high THC dose has been correlated with deficits in fine motor control and motor timing (Boggs et al., 2018).
Notably, clinical implications of the described heterogeneity in active principles might be particularly relevant in specific population subsets, such as elderly. The elderly is one of most represented population of cannabis users, with an estimated prevalence of use ranging between 6.5 and 37% (Ware and Tawfik, 2005;Hazekamp and Heerdink, 2013;Ahmed et al., 2014). Considering the well-known pharmacokinetic alternation occurring in such population, the high variability in concentrations of cannabinoids might suggest relevant safety as well as efficacy concerns, highlighting the need of a close phytovigilance monitoring (Lucenteforte et al., 2017).
The major limitation of this study is the small number of analyzed samples and of galenic laboratories from which they were derived. The evaluation of a broader spectrum of samples prepared in different laboratories throughout Italy would be of help to better estimate the real entity of the described variability in cannabinoids extraction yields.
Our results claim the need of a unique defined protocol, to be adopted by all galenic laboratories in order to guarantee reproducible and controlled extraction yields, with the final aim of limiting the possible clinical effects related to over-or undertherapeutic doses. Further studies are needed to evaluated the real correlation between cannabinoids doses and clinical outcomes in the real clinical practice.

DATA AVAILABILITY
Datasets are available on request.

AUTHOR CONTRIBUTIONS
RB, AB, and AV contributed to study design, assisted by NL, GC, VM, EG, AM, and FF. RB and AB took the lead in data analysis, assisted by NL, GC, VM, EG, AM, FF, and AV. GC, FF, NL, AM, and AV performed the data interpretation, assisted by AB, VM, EG, and RB. AB wrote the manuscript, assisted by NL, GC, VM, EG, AM, FF, RB, and AV. EG, VM, and AV revised the manuscript. All authors approved the final version of the manuscript.