Developing a new national MDMA policy: Results of a multi-decision multi-criterion decision analysis

Ecstasy (MDMA) is commonly used by urban, higher-educated young adults at dance events and house parties and typically only a few times per year; last-year prevalence in the Netherlands has stabilised at around 3%. MDMA was placed on List I of the Dutch Opium Act in 1988, a decision that the authors characterise as having an unclear evidence base at the time and one that was driven more by concerns about large-scale trade than by demonstrated health harm. Since that scheduling, evidence has accumulated that MDMA has relatively low dependence liability compared with other List I substances, that it may have therapeutic utility (notably as an adjunct in PTSD psychotherapy), and that illegal MDMA production in the Netherlands has contributed to organised crime, environmental damage and other social harms. Against this background, Van Amsterdam and colleagues convened an 18-member multidisciplinary expert panel to develop a science-based, politically feasible national MDMA policy. Using a multi-decision multicriteria decision analysis (MD-MCDA) framework, the group generated 95 policy instrument options and assessed their effects on 25 outcome criteria to identify an "optimal" policy model that maximises net benefits (public health, crime reduction, fiscal outcomes and related factors) and a slightly modified, more politically feasible variant (the "X-shop" model). The exercise compared the optimal model with four reference models (coffeeshop, adapted coffeeshop, free market, repressive) and with the current legal situation.

The investigators applied a consensus MD-MCDA procedure in iterative sessions. A steering group selected 18 experts chosen for domain-specific expertise and independence from political parties or ministries; represented fields included pharmacology, toxicology, pharmacy, ethics, anthropology, drug enforcement, epidemiology, neurobiology, medicine, philosophy of law, criminology, law, national and international drug policy, prevention and behavioural sciences. The extracted text does not list the individual experts by name. In Stage 1 the panel defined the decision space: 25 outcome criteria were selected to capture health, social, economic and criminal impacts, and 22 policy instruments were specified, each with 2–7 mutually exclusive options, yielding 95 policy instrument options in total. A policy model was defined as a particular combination of choices across the 22 instruments. Besides generating an "optimal" model by selecting the highest-rated option for each instrument, the team also specified four comparator policy models (coffeeshop, adapted coffeeshop, free market, repressive) as well as a politically adjusted variant called the X-shop model. Scoring of effects was conducted under a moderated consensus process. For each of the 95 options the experts rated the relative impact on each of the 25 outcomes, producing 2,375 scores (95 x 25). Prior to scoring, anchor values were set for each outcome so that the current legal situation equalled zero, a maximal negative impact equalled -100 and a maximal positive impact equalled +100; anchors were adjusted for 12 outcomes where the status quo approximated a bound. Scoring took place over three days in two parallel expert groups; when consensus could not be reached the arithmetic mean of individual scores was used. After each session, experts rated their confidence in the scores (0–100) and could revisit ratings in plenary. To combine heterogeneous outcomes, the panel applied a two-level weighting procedure. Within each of six outcome clusters (A–F; cluster names are not clearly reported in the extracted text) experts designated the single most important outcome and assigned it a weight of 100, then assigned relative weights to the remaining outcomes within the cluster (W2). Separately, experts assigned relative weights to the six clusters (W1); mean expert weights were then normalised (the highest cluster mean set to 100). Final model scores were computed by multiplying option-outcome scores by the relevant weights and summing across all outcomes. The optimal model was assembled by taking the top-scoring option for each instrument; the worst model used the bottom-scoring options. The optimal model was subsequently adjusted to create the X-shop model to improve political feasibility and resolve incompatible options. Sensitivity analyses assessed robustness to score uncertainty and to weight variation. For scores with low confidence the authors replaced values with either the maximum, zero or the minimum possible score and repeated this across confidence thresholds from 0.0 to 1.0 in 0.1 increments. Robustness to weighting was assessed by recomputing model scores using each expert's individual weight set instead of the averaged weights.

Under the current legal framework (MDMA on List I) production, import, export, possession, advertising, trade and sales of MDMA are prohibited in the Netherlands, although consumption per se is not illegal; packaging, age limits, price and quality regulation are not prescribed in law. The expert panel completed 2,375 option–outcome ratings and assigned individual weights to 25 outcomes and six clusters; the mean weighting factors are reported in the paper (table not reproduced in the extracted text). Numerically, the derived optimal policy model achieved an overall score 13,270 points higher than the current situation (status quo set to zero), whereas the worst possible model scored 7,252 points lower than the current situation. The authors highlight that the principal gains of the optimal model are improvements in health and social outcomes, stronger prevention of MDMA-related organised crime, increased state revenues, and enhancements in product quality and user information. These gains were attributed principally to policy elements such as legal requirements for sales, monitoring, and quality control. Key characteristics of the optimal model included regulated legal sales to users (only licensed sellers), legalised and licensed production under pharmaceutical-type rules, tolerance of a user quantity for possession, allowance of advertising, fixed minimum retail pricing, plain packaging with a prevention message, strong quality-assurance requirements with severe penalties for violations, prioritisation of investigation and confiscation of serious MDMA-related crime, legalised export with intent to seek an "inter se" treaty position, and a government-led prevention and subsidised health-education programme focused on harm reduction with extensive adverse-effect monitoring. The X-shop model is a politically tailored variant that differs from the optimal model on six points: (1) user quantity is legal and larger quantities are tolerated, (2) advertising is prohibited, (3) sales to users are regulated under pharmaceutical legislation, (4) an age limit of 18 years is applied, (5) export is illegal, and (6) responsibility for prevention policy is allocated to all governmental bodies rather than solely national/regional government. Changing the possession option to legal (and allowing larger quantities) and changing advertising from allowed to prohibited produced the largest single negative impacts on the overall score relative to the optimal model, decreasing the score by 148 and 203 points respectively; the other four adjustments had smaller effects. At the level of outcome clusters the optimal model outscored comparators on nearly all clusters, with the X-shop model producing cluster scores that were virtually identical to the optimal model except for some international-status measures. Sensitivity analyses produced two stable clusters of models. Replacing low-confidence scores with extreme values and varying confidence thresholds sometimes altered rank order within clusters but never caused a model from the high-performing cluster (optimal, X-shop, coffeeshop, adapted coffeeshop) to fall below any model in the low-performing cluster (free market, repressive). Using each expert's individual weighting values instead of the averaged weights yielded the same high/low clustering, and inspection of individual weights showed broad agreement in model rankings across experts.

Van Amsterdam and colleagues interpret their MD-MCDA exercise as having generated a coherent, evidence-based optimal MDMA policy model that would, relative to the current situation, reduce health harms, lessen MDMA-related organised crime and environmental damage, and increase state revenues while improving product quality and user information. They further argue that the X-shop variant represents a modestly adjusted, more politically feasible option that preserves most benefits of the optimal model but may result in a small increase in prevalence of ecstasy use due to greater availability and decriminalisation of user quantities. The authors attribute the anticipated health gains chiefly to legal obligations for sales standards, systematic monitoring and quality control of MDMA products, combined with a prevention strategy focused on harm reduction. A central public-order rationale for the model is reduction of organised crime through legalisation and licensing of production and sales plus strengthened confiscation and international cooperation. To address the international legal context, the authors propose pursuing an "inter se" treaty modification (a mechanism permitting like-minded states to agree changes among themselves) to enable regulated production and trade while limiting impacts on non-participating states. Strengths discussed include the multidisciplinary composition of the expert panel, provision of an extensive state-of-the-art literature review to inform ratings, a transparent structured decision-making process, and public availability of the data and model to permit reweighting and reanalysis. The authors emphasise robustness: sensitivity analyses indicated that the principal conclusions were stable against large changes in individual scores and weighting assumptions. Limitations they acknowledge centre on subjectivity inherent in expert selection and judgement. To mitigate this, the panel intentionally included law-enforcement and more conservative experts and moderators repeatedly challenged experts to justify ratings with scientific arguments. The authors also note that selection of instruments and outcomes was informed by prior studies and that sensitivity testing reduced concern about undue influence from individual experts. The authors conclude that, given the robustness and transparency of the MD-MCDA outputs, the optimal/X-shop models provide credible, evidence-based starting points for revising Dutch MDMA policy and may have applicability to other countries contemplating similar regulatory approaches. They acknowledge that operational details—especially the specifics of law-enforcement implementation and investigative practice—fall outside the scope of the present analysis and would need further work.

Using MD-MCDA, the authors derived an optimal MDMA policy model and a slightly adjusted X-shop model which they present as proposals to inform revision of the Dutch MDMA legal framework. Both models predict substantial public-health benefits while taking into account the current criminal burden, and the authors argue that the robustness of their results suggests relevance for other countries considering MDMA policy reform.