A limitation is that the sample size of the target sample was not large enough to carry out more advanced methods such as one-sample Mendelian Randomization to further explore causality between cannabis and other drug use. However, as far as we know this is the first study exploring genetic overlap between cannabis and ecstasy, stimulants and any other drugs and power was sufficient to detect these associations. In summary, PGS for cannabis use was significantly associated with use of ecstasy, stimulants, and any illicit drugs. An exploratory followup analyses indicated that this association was slightly stronger in cannabis users compared to non-users for ecstasy and stimulant use, but only in people born after 1968. The results of the MZ discordant twin analyses were in line with the suggestion that cannabis could be a causal factor for other drug use. Given the exploratory nature of this study, the present findings must be considered as preliminary rather than conclusive. Further unravelling the nature of the co-occurrence between substances will have implications for public health and intervention research. If there is no causal relationship then interventions that target reductions in one drug may not necessarily also lead to any change in use of another drug, and interventions that seek to target both drugs will need to incorporate active ingredients for each substance.Tobacco and cannabis use during adolescence, when the brain is still developing and undergoing considerable structural and function changes , is a major public health concern. The association between adolescent tobacco and cannabis use and subsequent cognitive functioning has received particular attention because certain cognitive functions  do not peak until early adulthood  in parallel with maturation of the prefrontal cortex . Due to the prolonged neurodevelopmental period and the potential for the endocannabinoid and nicotinic cholinergic signalling systems to be involved in altering development , it is plausible that tobacco and cannabis use during this potentially critical period could play a role in disrupting normal brain development . Nonetheless, there is still uncertainty regarding the nature of the association between tobacco and cannabis use and neurocognitive function.

A recent review of prospective studies of the association between cannabis use and cognition in young people  highlighted an association between cannabis use and neuropsychological decline . However, studies often fail to control for neurocognitive measures prior to cannabis use  and associations were largely found for the heaviest cannabis users and were often attenuated when potential confounders  were included . A recent study , using a co-twin design , assessed IQ prior to cannabis initiation and found insufficient evidence to suggest pot for growing cannabis use was associated with decline in general IQ. Findings from two recent longitudinal studies of adolescents  using a repeated measures design suggest that the association between cognitive functioning and cannabis use could be bidirectional. The direction of association between tobacco and cognitive functioning is also unclear as there is a lack of epidemiological studies that have prospectively examined this relationship. Evidence from animal studies suggests that nicotine exposure may have more deleterious developmental effects during adolescence, when the brain is thought to be more vulnerable . Furthermore, human studies suggest that nicotine has a more potent effect when consumed in late adolescence compared to in adulthood . One small prospective study  found that current smokers performed worse than non-smokers on a variety of cognitive assessments including language related IQ and working memory while controlling for earlier cognitive measures and other substance use . Finally, one large study  on Israeli male soldiers  found a dose-response relationship between number of cigarettes smoked and lower general cognitive ability compared to non-smokers. They also found diminished cognitive functioning in individuals who started starting smoking after 18 years of age. The literature is further complicated by the differential effects of acute, chronic, and withdrawal from chronic nicotine on cognitive functioning. Studies have reported beneficial effects of acute nicotine , negative effects of nicotine withdrawal on cognitive functioning , and the reduction of beneficial effects with nonacute nicotine consumption as tolerance develops . In an effort to strengthen the evidence, we used data from the Avon Longitudinal Study of Parents and Children , a large UK prospective birth cohort, to investigate whether patterns of adolescent tobacco and cannabis use were prospectively associated with cognitive functioning at 24 years of age. Separate measures of tobacco and cannabis use were assessed on six occasions across adolescence allowing distinct classes of tobacco and cannabis use to be established. As young people do not initiate tobacco or cannabis at the same time , we used longitudinal latent class analysis to identify heterogenous classes of individuals with different tobacco and cannabis use profiles across adolescence . As a next step we used genetic variants that are separately associated with smoking initiation and lifetime cannabis use to perform Mendelian randomization  to improve causal inference .

The aims were to investigate  whether separate patterns of tobacco smoking and cannabis use  were associated with working memory, response inhibition, and emotion recognition assessed at age 24, and  whether tobacco use and cannabis use were associated with these cognitive outcomes using MR. The Stop Signal Task  was used to assess response inhibition – the ability to prevent an ongoing motor response. The task consisted of 256 trials, which included a 4:1 ratio of trials without stop signals to trials with stop signals. Mean response times were calculated. An estimate of stop signal reaction time  was calculated using the median of the inhibition function approach . SSRT used as the primary outcome as it is a reliable measure of inhibitory control, with shorter reaction times indicating faster inhibition. SSRT data were available for 3201 participants. Individual Stop Signal indices  were examined as secondary outcomes. Emotion recognition was assessed using a six alternative forced choice  emotion recognition task  comprising of 96 trials  which measures the ability to identify emotions in facial expressions that vary in intensity. In each trial, participants were presented with a face displaying one of six emotions: anger, disgust, fear, happiness, sadness, or surprise. Participants were required to select the descriptor that best described the emotion that was present in the face, using the computer mouse. Emotion intensity varied across 8 levels within each emotion from the prototypical emotion to an almost neutral face. Each individual stimulus was presented twice, giving a total of 96 trials. An overall measure of emotion recognition  was used as the primary outcome. Emotion recognition data were available for n = 3368 participants. Each of the individual emotions were examined as secondary outcomes. Confounders comprised of established risk factors for cognitive functioning that could plausibly have a relationship with earlier substance use. Potential confounders included: income, maternal education, socioeconomic position, housing tenure, sex, and maternal smoking during first trimester in pregnancy. Working memory at approximately 11 years and experience of a head injury/unconsciousness up to 11 years were included to control for cognitive functioning prior to baseline measures of substance use. Finally, a measure of alcohol use asking whether they had ever had a whole drink of alcohol was collected at age 13 years . Further information is presented in Supplementary Material.This approach uses the weights derived from the latent classes to reflect measurement error in the latent class variable. Linear regression was used to examine the association between the cognitive outcomes and latent class membership controlling for the confounding variables. Results are reported as unstandardized beta coefficients with 95 % confidence intervals.

Analyses were carried out using Mplus 8.4 . Missing data was dealt with in three steps. First, full information maximum likelihood  was used to derive trajectories tobacco  and cannabis  based on individuals who had information on at least one timepoint between 13 and 18 years. For a detailed description of missingness at each timepoint see Tables S2a and S2b. Next, multiple imputation was based on 3232 participants  who had information on at least one of the cognitive outcomes. The imputation model  contained performance on all of the cognitive tasks, all measures of tobacco and cannabis use, and potential confounding variables, as well as a number of auxiliary variables known to be related to missingness . Finally, inverse probability weighting was used where estimates of prevalence and associations were weighted to account for probabilities of non-response to attending the clinic. See Table S3 for a detailed description of attrition for completing the cognitive assessments at age 24 years. See Tables S4a and S4b for a detailed description of confounding factors associated with tobacco and cannabis use class membership. See Table S5 for a detailed description of sample characteristics. Our aim was to triangulate the findings from the observational analyses with one- and two-sample MR analyses. However, due to insufficient power in the two-sample MR analyses, we will primarily focus on the one-sample MR results. Two-sample MR are still included as a set of sensitivity analyses as they allow us to conduct some of the pleiotropy robust methods , but must be interpreted with caution. Information on genotyping and quality control are presented in the Supplementary Material. This observational study provided evidence to suggest an association between tobacco and cannabis use across adolescence and subsequent cognitive functioning. Early- and late-onset regular tobacco smokers demonstrated poorer working memory and poorer ability to recognise emotions; while, early-onset regular tobacco smokers had slower ability to inhibit responses compared to non-tobacco smokers. Early-onset regular cannabis users had poorer working memory performance and slower ability to inhibit responses compared to non-cannabis users. Our results remained largely consistent when controlling for prior measures of substance use and cognition allowing for clear temporality between exposure and outcomes. Genetic analyses were imprecise and did not provide sufficient evidence for a possible causal association between smoking initiation and lifetime cannabis use and cognitive functioning in the ALSPAC sample. It is likely that these analyses were underpowered. To our knowledge, this is the first study to assess the relationship between separate tobacco and cannabis use in adolescents, and subsequent cognitive functioning using a combination of observational and genetic epidemiological approaches. Overall, we found an adverse association between tobacco/cannabis use and working memory, response inhibition, and emotion recognition in ALSPAC. Those who initiated regular use at earlier and later ages demonstrated poorer performance on the cognitive tasks. There was some evidence to suggest cannabis use with associated with emotion-specific impairments in emotion recognition. This is in line with previous research suggesting container for growing weed cannabis users may have poorer recognition of negative emotions . Our results also tentatively suggest that recognition deficits may be related to specific patterns of cannabis use, with different patterns in early- and late-onset use. The observational findings contribute to a literature of mixed findings regarding the direction of association between tobacco and cannabis exposure and subsequent cognition by suggesting that adolescent tobacco and cannabis use precede observed reductions in cognitive function. These findings support studies that have demonstrated effects may depend on the frequency, duration, and age at onset of use .

Our study extends previous findings in a number of ways. First, the observational study was better powered than most of the previous studies as it used data from over 3200 participants providing information spanning birth to 24 years of age. Second, identifying heterogeneous patterns of tobacco and cannabis use across this crucial period allows individuals who follow markedly different developmental trajectories to be captured . Third, the cognitive measures were assessed at a time when they are expected to have reached maturity in some individuals , in comparison to previous studies which have examined cognitive functioning at earlier ages while they are still maturing. Examining mature levels of cognitive functioning reduces the possibility that cognitive functioning is influencing earlier tobacco and cannabis use, effects that cannot be disentangled in purely cross-sectional studies. Further, our ability to control for earlier measures of cognitive functioning and substance use, prior to the baseline measures of tobacco and cannabis use helps to rule out the possibility of reverse causation. Fourth, our study sought to examine specificity in cognitive functioning, by using well-validated tests to probe different domains of cognitive functioning instead of focusing on general intelligence. Finally, we sought to triangulate our results by using one- and two-sample MR approaches to assess tobacco and cannabis use as causal risk factors for cognitive functioning.