In this one-year, longitudinal investigation of adolescents and adults who use cannabis, we found that adolescents (aged 16–17) scored on average 3.7 points higher on the CUDIT-R than adults (aged 26–29) across all 5 assessment waves (3.68, 95% CIs 1.81, 5.56). This effect was only partially attenuated after adjustment for gender, COVID-19, and mean weekly standard THC units (3.66 95% CIs 1.99, 5.34). CUD symptoms decreased linearly over the year in both age groups (−0.47, 95%CIs −0.67, −0.27). Through the use of a longitudinal study with five assessment waves, and a comprehensive standardised assessment of cannabis exposure, these findings show that the increased number of CUD symptoms that have been observed in adolescents persists over 12 months and is robust after adjustment for variation in THC dose.

Evidence of the persistence of increased levels of CUD symptoms in adolescents compared to adults across the 12-month period builds on previous cross-sectional comparisons of the likelihood of CUD by current age [6–8, 29, 30]. To our knowledge, the current study is the first such longitudinal comparison of adult and adolescent symptoms. This study is important because longitudinal comparisons can provide higher quality of evidence than cross-sectional comparisons. Furthermore, they can provide insight into the time course of such associations. The findings indicate that this is a persistent effect over a year, highlighting the need for a comprehensive evaluation of the impact of cannabis use on adolescent health and wellbeing. Adolescents often endorsed items related to cannabis affecting their general functioning, including failing to meet obligations and dedicating a lot of time to cannabis use. This indicates that the use of cannabis at this age has the potential to disrupt adolescents’ personal or academic lives, which could result in difficulties with educational outcomes and transitions into adulthood [31]. Given these findings, it is crucial that appropriate healthcare resources are available for this age group; however, the transition from child (< 18) to adult (> 18) health services can be challenging, and there is a risk of young people falling through the cracks [32]. Our findings add weight to the idea of integrated young peoples’ services covering a wider age range (e.g., 12 to 25). Further avenues of support could include education and harm reduction advice tailored for young people, as well as public health/policy-related changes to reduce stigma and barriers related to treatment seeking for cannabis-related support and increasing accessibility of support for young people [33]. Additionally, there was a linear decrease in CUD symptoms over time in both groups. This could be an indication of ‘maturing out’ from CUD [34]. However, longer follow-up periods would be necessary to demonstrate robust changes in CUD symptoms such as long-term remission. Furthermore, group means on the CUDIT-R were still elevated in both groups at the 12-month follow-up. By including longer follow-ups, such studies could provide valuable insight into the course of adolescent risk of CUD.

Some previous investigations have used samples with matched or similar levels of cannabis frequency [7, 8]; however, most do not consider cannabis quantity. Cannabis use profile (including frequency, quantity, and potency of use) has been consistently linked to the risk of CUD [18–20], and adolescents may use higher quantities of cannabis than adults. Therefore, not accounting for this could have led to overinflation of estimates of risk in adolescents. The current analysis used a novel measurement of cannabis quantity, the THC unit (5mg THC). Here, we found that adolescents did report greater cannabis use using this measure that incorporates quantity, frequency, and potency. However, adjusting for this in models did not substantially alter the main effect of a greater number of symptoms in adolescents, indicating that increased cannabis use in adolescents was not primarily responsible for the increased CUD symptoms.

The current analysis investigated two important factors related to CUD: current age, and profile of cannabis use. However, there are several other factors that might influence the relationship between cannabis use during adolescence and CUD symptoms that were not accounted for in the main analysis model. For example, CUD often co-occurs with other mental health disorders and other substance use disorders [31, 35]. We chose not to adjust for this in the primary analysis due to concerns over the direction of causality, given that other mental health disorders could either act as a mediator of the relationship between adolescent cannabis use and CUD symptoms or as a common cause of both [36]. However, in an exploratory sensitivity analysis, we added mental health and other drug use to the model, and adolescents still scored on average 3.1 points higher on the CUDIT-R than adults. This provides more support for the role of adolescent vulnerability to CUD, as this will account for more relevant confounders. However, other factors could still confound the relationship between adolescent frequent cannabis use and CUD [37]. For example, we were unable to account for genetic risk factors that may have differed between the adolescent and adult groups. Further studies with larger sample sizes will be needed to provide adequate power to adjust for a more comprehensive set of potential confounds, to increase precision when estimating the risk of CUD in adolescents compared to adults.

Strengths of this study include its longitudinal design with five assessment waves, the use of a validated outcome variable (CUDIT-R score), and a comprehensive standardised assessment of THC exposure validated in this sample [21]. The current findings should be considered in the light of several limitations. Firstly, the study was limited to only a 12-month follow-up duration, restricting the conclusions that can be drawn about longer-term CUD across adolescence and into adulthood. Furthermore, the measurement of CUD symptoms was the CUDIT-R, rather than the diagnostic DSM-5 clinical interview. However, at the baseline assessment, mean CUDIT-R scores increased across DSM-5 severity classifications (see Online Resource, Fig. 1). The CUDIT-R has not been validated for use over periods shorter than 6 months and, therefore, this may have induced unintended consequences. For example, it could be that assessment over a shorter period of time influences cannabis use in some way. However, there was no indication that this had a different effect on either age group given the lack of time by age interaction on CUDIT-R scores. Our finding of reduced CUDIT-R symptoms across groups could be in part due to the influence of being part of this longitudinal study. Repeated assessment of the CUDIT-R as well as administration of the TLFB to assess drug use may have in some way acted as an intervention (e.g., due to increased self-monitoring of drug use), bringing participants’ attention to their cannabis use and encouraging reduction of use. Another consideration is whether instruments assessing CUD symptoms are appropriate for comparison across age groups. The CUDIT-R has been implemented in adolescent samples previously; however, little research has investigated measurement invariance of CUD assessments, a key assumption underlying comparison of age groups, and therefore this should be considered a necessary avenue for future research into adolescent/adult comparisons.

Additionally, whilst standard THC units can estimate the dose for all cannabis products and methods of administration, they may be influenced by participant error in reporting (e.g., estimation of grams). However, estimated standard THC units using these methods were associated with objectively verified THC exposure (THC:COOH/creatinine) with a large effect size (r = 0.52), with a stronger correlation than any other measure of cannabis use from the CannTeen dataset [21]). Our estimates of cannabis potency were based on available UK seizure data [29]. Cannabinoid potency testing can potentially be biased due to degradation in sample quality prior to testing for cannabinoid testing, which could lead to underestimates of the recorded THC concentration [38]. However, in the investigation of UK seizure data researchers found that CBN concentrations were low in their samples and this was not related to the length of time the samples were in storage for, with the authors indicating that these samples were a fair representation of the original seized materials [29].

Additionally, the sample size was modest and not sampled to be representative of the general population of people who use cannabis due to inclusion criteria around use. This enabled purposeful sampling of matched groups of adolescents and adults who use cannabis at the same high, mean frequency to increase the meaningfulness of comparisons. This approach can be considered advantageous to population cohort studies, as the prevalence of regular adolescent cannabis use in the general population is rare resulting in small sample sizes. Additionally, greater levels of cannabis use in adolescence than in adulthood could lead to overestimates of adolescent risk. Therefore, purposively sampling matched adult and adolescent groups can overcome these limitations. However, because of this sampling approach, these findings may not be representative of CUD risk in people who use cannabis less than weekly.

Criteria for the adult group to have had minimal cannabis use under age 18 means that they are likely not representative of the average adult who uses cannabis and are likely to differ from the adolescent group on other variables related to CUD. This design was chosen to isolate the effects of adolescent cannabis use compared to those from adult use, to investigate whether cannabis is associated with more harm when used frequently in adolescence. Given our inclusion criteria for adult participants to have no regular use of cannabis before the age of 18, our adult and adolescent groups varied based on their reported age of first cannabis use. As a sensitivity analysis, we included the age of first use as a covariate in the models, which did not substantially alter the pattern of results (see Online Resource, Table 5). Furthermore, the CannTeen sample was limited to those in the London area, and those willing to take part in a study with relatively frequent assessments and therefore high levels of engagement. The findings from this study should be viewed in the light of this.

These findings add to a wider literature on adolescent vulnerability to CUD, predominantly comprised of large-scale surveys. These research designs tend to have good statistical power to adjust for important confounding factors, but they typically lack detailed data on participants’ cannabis use and are mostly cross sectional. The current study therefore adds to the literature by examining the one-year course of CUD symptoms, adjusting for a detailed assessment of cannabis use and other relevant covariates. However, the clinical meaning of the current observed difference in CUDIT-R scores is yet to be determined, as we are not aware of work that has assessed the clinical meaning of CUDIT-R scores. Lived experience feedback from people who use and support those who use cannabis, including from adolescents themselves, is needed to further establish the implications of these findings.

In conclusion, the current study provides the first evidence of longitudinal persistence of increased severity of CUD in adolescents compared to adults, with adolescents on average scoring 3.7 points higher on a measure of CUD symptoms, over one year. This pattern of results remained after adjustment for a comprehensive measure of cannabis quantity. This study indicates the increased risk of CUD symptoms in adolescents and provides evidence to support the importance of delaying or minimising the use of cannabis during this developmental period.