An article by Leung et al.¹ reviewed the history of efforts to measure how frequently cannabis users develop abuse and dependence. The first paragraph introduces some of the dilemmas perplexing this question.

'One in 10 cannabis users develop dependence' is the figure widely cited by clinicians, researchers, and policymakers in health and medicine. This estimate comes from the National Comorbidity Survey (NCS), a cross-sectional survey conducted in 1990–1992 in the United States, that estimated that the lifetime prevalence of cannabis dependence (CD; using DSM-III-R criteria) among adults who reported that they had ever used cannabis was 9.1 percent. These data were collected almost 30 years ago when cannabis products were less potent than today and before major changes in definitions of cannabis use disorders (CUD).²

The question of how prevalent cannabis dependence is among cannabis users is complicated. Including individuals who experimented with cannabis only once or twice among “cannabis users” leads to underestimating realistic dependence rates since they are hardly more likely to develop dependence than people who have never tried cannabis.

Should we only include people who have tried cannabis an arbitrary minimum number of times (e.g., five or more)? It is important to distinguish between those who have used it in the past year versus ever in their lifetime. For example, the past year's prevalence of CUD was 15.4 percent in 18–25-year-olds who used cannabis that year, against 9.8 percent among 18–25-year-olds who reported lifetime cannabis use.³ Should we differentiate between adolescent and adult-onset users? Are we mostly interested in weekly or daily users? Should we discard data from before THC potency increased to get a more accurate picture of today’s reality? It’s very complicated.

An additional curve ball thrown at researchers comes from changes in our diagnostic categories for cannabis abuse (CA), cannabis dependence (CD), and cannabis use disorder (CUD). The original 9.1 percent figure used the diagnostic framework in DSM-IIIR. DSM-4 was published in 1994, and this gave way to DSM-5 in 2013, which collapsed cannabis abuse and dependence into a continuum called cannabis use disorder. CUD requires any two of 11 criteria, including loss of control of use, continued use despite adverse consequences, use in risky situations, failed efforts to quit or fulfill responsibilities, tolerance, and withdrawal. Data gathered using DSM-5 confuses the comparison of the rate of cannabis dependence over time.

Leung et al. attacked the problem through a meta-analysis of articles published between 2009 and 2019 using data from the United States, Netherlands, Germany, Australia, New Zealand, Ireland, and France. Meta-analysis is a statistical process that pools data from studies with similar methodologies to find underlying commonalities. Leung et al.'s approach looked at different data sets to differentiate between individuals with a lifetime history of ever using cannabis, those who used it during the previous year, and those who used it regularly (weekly or more).

Among people who use cannabis, 22 percent (1 in 5) have CUD, with 13 percent meeting the diagnostic criteria for cannabis dependence. This figure of 13 percent is almost 50 percent higher than the generally accepted figure of 9.1 percent and reveals that closer to 1 in 8 people who have used cannabis will develop dependence. Meta-analysis of cross-sectional and longitudinal studies of people with regular cannabis use differ. Cross-section studies of daily or near-daily use found an 18 percent rate of dependence, while longitudinal studies found 33 percent of daily users were dependent, or 1 in 3.

Bottom line: People who use cannabis have a 1 in 5 risk of CUD (either abuse or dependence) and a 1 in 8 risk of cannabis dependence. The risk of developing dependence increases to 1 in 3 among people who use cannabis weekly or more often.

Accurately measuring the rate of cannabis dependence is important for both clinical and policy decisions. When clinicians understand one in every five patients who use cannabis likely satisfies the criteria for CUD, they will ask follow-up questions after positive screenings. This will provide the opportunity for relevant health education and the referral of those dependent to a specialist.

Second, policymakers need an accurate measurement of prevalence rates to make wise public health decisions. For example, do medical and recreational cannabis laws lead to increased rates of CUD? And, if they do, what portion of the population is most affected? Recent data show that, while such laws do increase CUD among Veteran Health Administration patients, the increase is relatively small compared to other factors and is most prominent in the 65-75 age group.⁴ In other words, rather than oppose legalization, public health should focus on correcting the public’s unrealistically low perception of risk and directing cannabis education toward older Americans.

Accurate data facilitate more effective clinical interventions and public health efforts.