Controlled administration of cannabis to mitigate cannabis-attributable harm among recreational users: a quasi-experimental study in Germany

Study design, sample and recruitment

The aims of the study will be evaluated using a quasi-experimental two-group study design, as summarized in Figure 1. The intervention will constitute an individualized licence to purchase herbal cannabis in selected pharmacies for a duration of 12 months in Berlin, Germany. Control group participants will be recruited in Hamburg, a city comparable to Berlin in terms of sociodemographic indicators (population, deprivation) and user characteristics²⁹.

Figure 1. Study flow chart.

Prior to any data collection, study team members will take consent of each participant. A model consent form is available as Extended data⁶⁰. In both study groups, participants will be asked to complete a questionnaire and provide their usual consumption unit (e.g., joints, mix of tobacco and cannabis to smoke in water pipe, etc.) at both baseline and again after one year (T12). In predetermined intervals, additional questionnaires will be completed by all participants. Further, individual sales data will be collected for intervention group participants, which will be part of the primary outcome analyses.

All participants will be recruited via posters, or through consumer groups and local stakeholders in each city. To incentivize study participation and minimize loss to follow-up, participants will be reimbursed with 25€ upon completion of the questionnaires at each wave. Further, providing details regarding their usual consumption unit will be reimbursed with 15€ (at baseline and at one-year follow-up).

Every resident in the respective control or intervention municipality aged 18 years or above reporting to have used cannabis at least monthly in the past 6 months may participate in the study. Following international recommendations to minimize cannabis use risks³⁰, persons with a familial predisposition to psychotic or substance use disorders, pregnant women, persons in current psychiatric treatment or with a preceding psychotic disorder will be excluded from this study. Further, persons with a prescription to use cannabis for medical purposes will not be eligible to participate as the study focuses on recreational use only.

Details of the dispensary model

The term dispensary model describes how intervention group participants can acquire cannabis in pharmacies and all measures undertaken to achieve the postulated aim of mitigating use risks. Registered users will be eligible to purchase up to 5 grams of herbal cannabis per transaction and up to 15 grams of herbal cannabis per week. All herbal cannabis will be sold in 1-gram packages, which will be labelled with the concentration of THC and Cannabidiol (CBD,^31,32), and contain further information (e.g., on other constituents, producer, date of production, safety warnings).

Two core measures will deter intervention group participants from using high-potency cannabis varieties. First, we will restrict the sale to varieties containing at most 12% THC, which is just below the median value of 13% identified in herbal cannabis batches seized by the federal police in 2017³³. Second, the retail price will be bound to the THC level, with lower potencies being more affordable. Specifically, the variety lowest in THC will be sold at the current black-market price, as determined by the federal police. More potent varieties will be sold at higher prices, with the most potent variety to be sold at 50% above the current black-market price. In this way, purchasing varieties with high THC levels will be disincentivized.

To further mitigate the risks of THC, an upper threshold of the THC/CBD ratio will be established as CBD has been found to attenuate some of the adverse effects of THC (see e.g.^9,34). In the absence of any systematic evidence on the specific risk profile of THC/CBD ratios, an upper threshold of 50:1 was defined based on US data, indicating a worrying ratio increase from 23:1 to 104:1 between 2008 and 2017²⁵.

All herbal cannabis to be sold in the pharmacies will be acquired through the medical supply system³⁵. As of 2019, six out of the 20 varieties available for medical use in Germany³⁶ met the outlined criteria (cap of potency and cap of THC/CBD ratio) to be sold to the intervention group participants in this study. As prescribed for medical cannabis, each batch of herbal cannabis will undergo routine pharmaceutical tests. Further, producers of medical marijuana are legally required to perform systematic tests to detect variations in potency levels and contaminants. In accordance with the requirements of the narcotics law, cannabis stored in the pharmacies will not be accessible to anyone but trained staff. All cannabis products will be stored in a steel cabinet protected by a safety lock, in accordance with the standards prescribed by the German ‘Federal Institute for Drugs and Medical Devices’ (Bundesinstitut für Arzneimittel und Medizinprodukte, BfArM).

In all participating pharmacies, users will find printed information material on mitigating use risks and will be offered the option of renting devices for vaporizing herbal cannabis. The use of vaporizers can reduce exposure to some hazardous compounds and is hence considered a harm-reducing method³⁷. Further, the vending staff in all participating pharmacies will be trained in the following: legal framework, safer use guidelines (original:³⁰, for a German translation:³⁸), identification of vulnerable users, cooperation with drug counselling and prevention services. If required, study participants will be able to hold conversations with staff in private on the pharmacy premises.

Primary outcome

Two important aspects will determine the study’s success and will be combined in the primary outcome, measured at baseline and at 12-month follow-up: 1) dominance of legal over illegal consumption, and 2) reduction of THC exposure. The hypothesis for the primary outcome is:

“As compared to baseline, intervention group participants will reduce their use of illegally acquired THC within the past 30 days by at least 50% at 12-month follow-up. Taking into account baseline exposure levels, the total amount of THC consumed among intervention group participants within the past 30 days will fall below the total amount of THC consumed among control group participants within the past 30 days by at least 10% at 12-month follow-up.”

As first aspect, we expect that illicit cannabis use will be largely replaced by consumption of legally acquired products. A shift from an illegal to a legal consumption environment implies that exposure to contaminants and pollutants but also legal consequences (revocation of driver’s licence, imprisonment) will be largely diminished or even completely prevented. As an evaluation of changes in these specific outcomes would require very large sample sizes, and go beyond the scope of this study, we will restrict the first part of the primary outcome to a proxy, i.e., to the percentage of legal cannabis consumption. This will be operationalized as the percentage of an individual’s THC exposure within the past 30 days acquired through purchases in pharmacies.

For the second aspect, we expect to halt the increasing THC exposure among intervention group users. While THC levels in seized and legally available cannabis products are increasing, intervention group participants are expected to reduce their overall THC consumption as compared to control group participants. While THC exposure is the core determinant for cannabis-attributable harm, specific effects on adverse consequences (e.g. on incidence of psychotic disorders) require larger sample sizes which, again, are beyond the scope of this study.

While the first aspect will only be evaluated within the intervention group, the second aspect will require between-group comparisons across a 12-month period. Only if both aspects can be positively evaluated, can an overall success of the study be inferred.

Secondary outcomes

For secondary outcomes, we will adhere to the criteria proposed to evaluate the legalization of cannabis in Canada³⁹, encompassing all cannabis use behaviours linked to adverse consequences, for which substantial evidence exists³⁰. As intervention group participants will receive information on risky cannabis use³⁰ and enclosed within each cannabis package sold by the pharmacies, we hypothesize that risky cannabis use practices will not increase among these users – despite the presumably increased availability of cannabis products. Specifically, we hypothesize that intervention group participants will: A) not increase their use frequency, B) not increase frequency of THC-impaired driving, and C) will not smoke cannabis products more frequently (for operationalization, see Table 1). The secondary outcomes will be evaluated at one-year follow-up, which involves testing for differences between control and intervention group while taking baseline values into account.

As further secondary outcomes, we will examine acceptance and satisfaction of the dispensary model among intervention group participants, which may be important for explaining a possible dominance of illegal over legal consumption (see first aspect of primary outcome).

Assessment details

For both aspects of the primary outcome, THC exposure within the past 30 days needs to be determined for each respondent by combining the following data sources. First, THC levels per average use occasion will be determined through analyses of usual consumption samples provided at both baseline and after one year (T12). Second, questionnaire data will be used to determine the number of use occasions within the past 30 days. Third, and only for the intervention group, sales data will be used to determine THC exposure levels of legally acquired cannabis over a 12-month period.

In summary, for the control group, THC exposure levels within the past 30 days will be determined multiplying THC levels in a usual consumption unit with the number of use occasions reported in the questionnaire. For the intervention group, THC exposure levels will be determined analogously while correcting for the proportion of legally acquired and consumed THC using sales and questionnaire data.

At all waves — T0, T3, T6, and T12 — a set of questionnaires will be administered (see Table 2). For the primary and secondary outcomes, use characteristics will be assessed using items from the ‘Daily Sessions, Frequency, Age of Onset, and Quantity of Cannabis Use Inventory’ (DFAQ-CU), which will be translated and adapted for this study. All remaining questionnaire data will serve to explain unexpected findings, to control for potential confounders in the statistical analyses, or for the economic evaluation of this study.

At T18, we will conduct a post-intervention assessment of intervention group participants to examine how their cannabis use has developed after being denied further legal purchases of cannabis products. This assessment will include all instruments outlined in Table 2, in addition to several free-text items.

There will be three main types of study data: (1) chemical analyses of standard consumption units, (2) sales data, (3) survey responses of participants. For (1), we will adhere to the standard operating procedures issued by the BfArM⁴⁰. For (3), we will aim to carry out digital survey assessments to minimize human error in data entry. Further, consistency checks will be performed before data analyses.

Sample size and data analyses

The study outcomes will be analysed according to ‘Intention-to-treat’ (ITT) principles. Specifically, the sample to be analysed is defined as all users who have provided a usual consumption unit within 4 weeks after completing the baseline questionnaire. Over-recruitment will compensate for participants failing to provide a consumption unit within this period. According to ITT principles, only those participants who provided their baseline data, including their usual consumption unit, will be included in the analyses.

To examine between-group differences with a t-test, the required sample size of this study was calculated assuming a power of 80%, a 5% alpha error, and a THC standard deviation equal to one-third of its mean (approximated using Canadian data,⁴¹; adding 25% to account for uncertainty). The required sample size was estimated to sum up to n=698 participants (control: n=349; intervention: n=349) for detecting group differences in THC exposure.

Not relevant for sample size considerations, we expect that 20% (control: 30%; intervention: 10%) of all participants of the ITT sample will be lost to T12 follow-up. Subjects dropping out of the study will not be replaced and missing values will be imputed using the ‘Last Observation Carried Forward‘ (LOCF) technique. The critique regarding LOCF imputation methods⁴² does not apply to our study, as this method will bias the data towards the null hypothesis (assuming no change over time) and therefore representing a conservative imputation approach. The assumptions of LOCF will be examined in sensitivity analyses using advanced multiple imputation techniques.

To evaluate both aspects of the primary outcome, two analyses need to be conducted. First, the THC exposure levels in the past 30 days ascribed to illegal cannabis acquisition within the intervention group needs to be compared between baseline and intervention, with a 50% reduction denoting the minimum threshold for a positive evaluation. Second, a between-group comparison of THC exposure in the past 30 days at T12 adjusting for baseline data will be conducted using a t-test. The dependent variable will be calculated as follows:

The actual between-group comparison will be evaluated against the following condition:

Multilevel regression analyses will be performed additionally for evaluating the primary outcome, which serve to rule out the impact of possibly confounding variables. For evaluating the secondary outcomes, the proposed analyses (correction for baseline values, comparison via t-test, confirmation via multilevel regression analyses) will be performed analogously.

Adverse events and stoppage

If more than 60% of the intervention group participants drop out within the first three months, the study will be stopped immediately, because it will be taken as an indicator for subjects not accepting the administration model.

In addition, the Data Safety Monitoring Board (DSMB) will be formed by pharmacists and social workers working in the addiction and youth protection field. As stipulated in Good Clinical Practice guidelines, the DSMB, as an independent and multidisciplinary group will be established to review, at intervals, accumulating trial data, in order to monitor the progress of the trial and to make recommendations on whether to continue, modify or stop the trial for safety or ethical reasons.

Adverse events will be closely monitored and documented. Criteria have been pre-specified to define a preterm stop to the study. Further, the study team will be in close contact with the participating pharmacies in order to capture all events not foreseen at study inception. In regular meetings, the DSMB will evaluate the progress of the study and may decide to stop the study.

Economic evaluation

In a ‘cost-benefit-analysis’ (CBA), the proposed dispensary model will undergo an additional economic evaluation, in which the economic benefits from attenuated THC exposure will be contrasted to the programme costs⁵⁰, with estimates applied to Germany as a whole. Building on the approach of a previous CBAs for Australia⁵¹ and using the extended framework of generalized cost-effectiveness analyses⁵², we will compare the following scenarios:

For the CBA, the so-called net social benefit will be calculated as the difference between projected costs and benefits discounted over the study period. On the cost side, we will consider all economic costs that can be ascribed to cannabis-related law enforcement, treatment of cannabis use disorders and cannabis-attributable diseases (e.g. psychoses), loss of productivity, and all costs pertaining to establishing and maintaining the dispensary model. On the benefit side, we will consider all economic benefits arising in the following domains: reductions in law enforcement and health-care costs, as well as increases in productivity. Based on the CBA results, a cost-utility-analyses will additionally be conducted by estimating the costs required to avoid one ‘disability adjusted life year’ (DALY⁵³) for the scenario of a nationwide implementation of the dispensary model.

Data collection and dissemination

Data collection will be completed using electronic means and all required measures will be implemented to protect the data and anonymity of all study participants at all times. In particular, the study will adhere to the European ‘General Data Protection Regulation’ (GDPR) according to which, health care data is particularly sensitive and should be handled with the greatest care. Further, cannabis possession will remain a federal crime, further emphasizing the sensitive nature of the study data. Thus, all data will be stored on hard drives encrypted using state-of-the-art encryption techniques (AES-256). Upon completion of the study, the study data will be kept on encrypted hard drives in physically locked cabinets.

Conditioned on the approval from the public study sponsor, all study data shall be made available to other researchers. All efforts will be undertaken to anonymize the study data (in the sense of GDPR) in order to publish the data in public repositories making the data findable, openly accessible, interoperable, and re-usable (FAIR principles issued by the European Union). Upon publication of the primary outcome analyses, study data shall be published alongside the respective code of the statistical programme. Through these means, we hope that our analyses will be reproducible and that the data will be used for other purposes than those described, increasing the merit of this study. Lastly, study results will be published under an open access licence to allow for a widespread recognition of our findings.

Ethical considerations

This study has not undergone ethical review yet. A full-length study proposal is currently being evaluated by the BfArM. According to the German narcotics law, studies on cannabis may, by an exception, be allowed for scientific purposes and only if approved by the BfArM. Only once a positive decision is received from the BfArM will a complete study outline be reviewed by the responsible ethics board. The study has been pre-registered with the German Clinical Trials Register (a primary registry within the WHO registry network) and will be formally registered upon ethical approval (registry number: DRKS00020829). If any amendments of the outlined study design or protocol will be requested by the BfArM or ethics board, they will be reflected in this publication, as well.

Limitations

There are several limitations of this study. First, given a lack of randomization, causal inferences cannot be drawn from the study findings without controlling for relevant confounders. We have sought to include a broad variety of questionnaires to capture information on all possible determinants; however, we cannot rule out that some important confounding variables will not be assessed or be biased through self-report. Further, resin will not be part of the dispensary model for it is not available from medical suppliers. As far as we know, resin cannot be prescribed for medical purposes despite favourable THC/CBD ratios (for Dutch data, see 54). For cannabis users preferring resin over herbal cannabis, study participation may therefore be unattractive, and this may bias the study sample towards predominant herbal cannabis users.

Underlying data

No data are associated with this article.

Extended data

Figshare: Model consent form, https://doi.org/10.6084/m9.figshare.11903301.v1⁶⁰

Reporting guidelines

Figshare: SPIRIT checklist, https://doi.org/10.6084/m9.figshare.11903322.v1⁶¹

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).