Defining what constitutes a meaningful ‘reduction’ in CUD is central to advancing treatment research yet remains one of the least resolved challenges in the field. Clinically, changes in use often occur along a trajectory rather than as a binary shift from use to abstinence (Figure 2). Frequency can be operationalized across multiple levels, including days of use, number of use occasions per day and time spent using within a day. For example, clinically meaningful reduction may include a ≥50% decrease in the number of days of use, a shift from daily to intermittent use (e.g. ≥25 to ≤10 days/month) or reductions in within-day use (e.g. from 10+ to 3–5 use episodes), particularly when accompanied by improvements in functioning. Although many trials collect detailed longitudinal data on cannabis use, there is no shared framework for how reduction should be measured, interpreted or compared across studies. As a result, potentially important treatment effects, particularly in pharmacotherapy trials, are often difficult to detect or contextualize. Reviews of outcome assessment emphasize wide variability in end-points and analytic approaches, limiting synthesis and comparisons across studies [34].

A critical but underdeveloped aspect of defining reduction in cannabis use is its relationship with downstream medical risk, which provides a necessary anchor for determining whether reductions are clinically meaningful. While reductions in frequency or quantity are often reported, their clinical significance depends on their relationship with health outcomes. Emerging evidence suggests that cannabis use may affect multiple physiological systems, including cardiovascular, respiratory, gastrointestinal and reproductive domains. For example, cannabis exposure has been associated with acute cardiovascular effects (e.g. tachycardia, blood pressure changes) and potential longer-term vascular dysfunction, while chronic inhalation is linked to bronchitic symptoms, and cannabis hyperemesis syndrome represents a well-characterized complication of heavy use [46]. In the reproductive domain, cannabinoids may adversely affect both male and female fertility, including spermatogenesis, ovulation and implantation [47]. However, most studies do not systematically quantify dose, frequency or severity of use in relation to these outcomes, limiting the ability to determine what degree of reduction confers meaningful benefit. Establishing these dose–response relationships will be essential for defining clinically meaningful CUD reduction thresholds [48].

Other substance use fields demonstrate how progress can follow from clearer agreement about outcomes. In alcohol use disorder (AUD), large epidemiologic and clinical trial data sets have demonstrated that reductions in heavy drinking, clearly and consistently defined by agreed upon definitions of what quantifies a drink as well as defining a heavy drinking day, were associated with improvements in blood pressure, liver enzymes, cardiovascular risk, sleep, mood and overall mortality [49]. Treatment development then accelerated once trials routinely incorporated agreed-upon non-abstinent end-points that captured meaningful improvement, even when abstinence was not achieved. The 2015 FDA guidance for AUD drug development explicitly recognizes outcomes beyond abstinence, including reductions in heavy drinking [50]. More recently, the FDA qualified reductions in World Health Organization (WHO) risk drinking levels as an acceptable drug development tool, reflecting a broader shift toward end-points tied to improvements in health and functioning rather than cessation alone [51]. In AUD, reduction-based end-points gained traction when heavy drinking and WHO risk drinking levels were shown to map onto improvements in health and functioning. This shift did not occur because alcohol reduction is inherently easier to measure than cannabis reduction, but because the field converged on standardized definitions for both quantity and frequency, empirically supported thresholds, and outcomes that are meaningful to clinicians and patients.

Nicotine research—leading to the development of FDA-approved treatments including nicotine-replacement therapy, bupropion and varenicline—offers a similar lesson. Early smoking cessation trials used inconsistent definitions of abstinence, leading the field to develop consensus standards for outcome measurement and reporting [52, 53]. Longitudinal cohort studies demonstrated that reductions in cigarettes per day were associated with improvements in cardiovascular risk markers, respiratory symptoms and lung function, with these benefits following a dose–response relationship [54–56]. Later studies further showed that smokers who reduced consumption substantially had lower risks of cardiovascular events, chronic obstructive pulmonary disease progression and all-cause mortality, compared with those who maintained heavy smoking, even when abstinence was not achieved [57–59]. Although nicotine is more amenable to biochemical measurement than cannabis, the broader lesson still applies: a field progresses when investigators agree on a defined set of outcomes that are practical to measure, consistently reported and meaningful in clinical settings.

Cannabis poses distinct measurement challenges that make it difficult to apply alcohol or nicotine outcome frameworks directly. As noted above, the prolonged detection of THC metabolites among frequent users limits the value of urine testing for assessing recent behavior and complicates the interpretation of quantitative results [34]. Self-reported measures, particularly TLFB assessments, are better suited to capture meaningful changes in use over time, with biological measures used selectively to support interpretation rather than define treatment success [40]. This strategy is consistent with alcohol research, where carefully collected self-report data form the basis of validated, regulator-accepted end-points [50].

Defining the ‘quantity’ of cannabis use presents a second challenge. Unlike standard drinks or cigarettes, cannabis consumption varies widely by product type, route of administration, potency and session structure. Many patients cannot reliably estimate grams or milligrams of THC, particularly in contemporary markets dominated by high-potency products and diverse delivery systems in the absence of federal standardization of cannabis products [34, 60]. Together, these limitations make it difficult to rely on a single quantitative dose metric across trials. One response to these limitations has been to combine quantity and frequency rather than privileging either alone. In population-level measurement work, Asbridge and colleagues integrated typical quantity with recent frequency to create a quantity–frequency typology modeled on alcohol volume measures, which showed clearer gradients of cannabis-related problems and Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) scores than frequency- or quantity-only approaches [61]. Rather than abandoning quantity altogether, a more workable approach is to prioritize measures that most participants can report reliably, while supplementing them with additional detail when feasible. Emphasizing within-subject change in quantity over time allows each participant to serve as their own control and supports analyses based on mean change scores, even when between-subject comparisons of absolute dose are unreliable [34, 43]. In addition, incorporating in vivo assessments, such as ecological momentary assessment, may further improve the measurement of use patterns and reduce the reliance on retrospective quantity estimates [62–64].