Disentangling the effects of cannabis and cigarette smoking on impulsivity

Background: Cannabis smoking and cigarette smoking often co-occur, yet limited research has investigated the potentially different role impulsivity may play when these behaviours occur in isolation, compared with in combination. Aims: This study examined trait and behavioural impulsivity as a function of both cigarette and cannabis smoking. Methods: Trait impulsivity (BIS-11) was compared between 44 non-smokers, 76 cigarette only, 47 cannabis only and 58 cannabis plus cigarette smokers. The effects of cigarette and cannabis smoking on behavioural impulsivity (stop-signal and information sampling tasks) were then assessed in 87 of these participants during a laboratory session. Results: Trait impulsivity was significantly higher in cigarette smokers than non-smokers, irrespective of cannabis use, except for motor impulsivity, where cigarette smoking was only associated with elevated trait impulsivity in non-smokers of cannabis. Dimensions of trait impulsivity were significantly positively related to cigarette smoking frequency and nicotine dependence, but not to cannabis smoking frequency or dependence. Smoking cigarettes or cannabis was associated with significantly impaired reflection impulsivity relative to not smoking either substance. However, no additional increases in reflection impulsivity were observed in those who smoked both cigarettes and cannabis. No group differences in response inhibition were detected. Conclusions: Heightened trait impulsivity appears to be uniquely related to cigarette smoking, whilst the smoking of cigarettes or cannabis is associated with impairments in reflection impulsivity. Improved outcomes for treating cannabis dependence may result from encouraging concomitant cigarette smokers to cease using both drugs simultaneously in order to reduce heightened impulsivity and risk of relapse.


Introduction
Cannabis is the most widely used illicit substance globally (UNODC, 2018). In England, primary cannabis use accounts for the largest proportion of young people in treatment for substance misuse (Niblett, 2018). A major factor contributing to cannabis dependence may be co-dependence on nicotine, since many cannabis smokers also smoke cigarettes (Hindocha et al., 2016;Rabin and George, 2015), and are more likely to relapse to cannabis whilst continuing to smoke cigarettes (Haney et al., 2013). Furthermore, co-morbid cannabis-cigarette smoking is associated with heavier cigarette smoking relative to cigarette smoking only (Agrawal et al., 2011), while cigarette smoking mediates the relationship between cannabis use and cannabis dependence (Hindocha et al., 2015). Failure to consider the role of cigarette smoking when attempting to understand factors underlying problematic cannabis use may have hindered research progress. To date, limited research has attempted to disentangle the effects of cannabis smoking from cigarette smoking; in particular, the role impulsivity has in supporting dependence on these substances. Distinguishing between cannabis only smokers and those who smoke both cannabis and cigarettes may be a critical step towards this goal.
Impulsivity is a multifaceted construct encapsulating a range of trait and behavioural characteristics, where delayed rewards are discounted in favour of immediate gratification, actions are prematurely conceived without adequate forethought, and decision-making often lacks reflection despite resulting in negative consequences (Robbins et al., 2012). Moreover, impulsivity is thought to be involved in the transition to addiction for several drugs including cannabis (for reviews see de Wit, 2009;Wrege et al., 2014). Whilst heightened impulsiveness may predispose individuals to initiating substance use (e.g. Vergés et al., 2019), prolonged drug use may further reduce impulse control through drug-induced neuroadaptations, whereby dopaminergic prefrontal cortical processing becomes increasingly subverted by greater subcortical automatic conditioned responding, resulting in persistent drug-taking and making drug-related stimuli highly salient to users (Berridge and Robinson, 2016;Everitt et al., 2008;Goldstein and Volkow, 2002;Piazza and Deroche-Gamonet, 2013).
Trait impulsivity on the Barratt Impulsiveness Scale (BIS) is elevated in several dependent drug user groups relative to controls (e.g. Bozkurt et al., 2013;Ersche et al., 2008). Research has been less consistent in relation to whether trait impulsivity is elevated in cannabis smokers compared with non-smokers (NS). Some studies support this view (Dervaux et al., 2010;Moreno et al., 2012), while others indicate no trait differences between dependent (Johnson et al., 2010) or non-dependent cannabis smokers and NS (Clark et al., 2009;Field et al., 2006). A possible factor accounting for these discrepant findings is that studies have not always controlled for tobacco cigarette smoking in cannabis smokers sampled. In consideration of cigarette smoking, Beaton et al. (2014) found cigarette only smokers and cannabis smokers who smoke cigarettes both display elevated trait impulsivity on the BIS relative to NS and cannabis only smokerswho did not differ from each other -suggesting cigarette rather than cannabis smoking was associated with heightened impulsivity. In contrast, Gilman et al. (2015) found elevated trait motor and non-planning impulsivity in cannabis only smokers relative to NS. In terms of potential additive effects of cannabis and cigarette smoking on impulsivity, none were suggested in the work of Beaton et al. (2014), but these could not be assessed by Gilman et al. (2015) as no cigarette smokers were included in this study. Furthermore, behavioural impulsivity was not measured in either study.
Behavioural impulsivity is principally conceptualised as manifesting in two forms: response inhibition (the ability to inhibit a pre-potent response) and impulsive choice (a preference for smaller immediate over larger delayed reward), which have been the most widely studied aspects of impulsive behaviour to date (Weafer and de Wit, 2013). While evidence indicates that response inhibition, measured by the stop-signal task (SST), is impaired following acute cannabis administration (Ramaekers et al., 2006), it is unknown whether impulsive behaviour is elevated in cannabis smokers residually following long-term use. When not acutely intoxicated, existing evidence suggests that cannabis smokers do not differ from NS on response inhibition on the SST (Filbey and Yezhuvath, 2013;Grant et al., 2012), however neither of these studies controlled for cigarette smoking in the interpretation of task findings. Evidence also indicates that those who smoke both cannabis and cigarettes do not differ from NS on impulsive choice, measured using a delay discounting task (Johnson et al., 2010). Interestingly, in this study cigarettes smoked per day but not bouts of cannabis smoked per day predicted greater discounting of reward, suggesting impulsive behaviour more closely relates to cigarette smoking. However, no study has extended beyond this when comparing the role of cigarette smoking and cannabis use in impulsive behaviour. This may be problematic given (a) existing findings implicating chronic nicotine exposure in alteration of response inhibition (e.g. Luijten et al., 2011) and impulsive choice (e.g. Kolokotroni et al., 2014), and (b) increasing agreement that stimulant dependence may lead to more severe neurological impairment than cannabis dependence (Bonomo et al., 2019;Van Holst and Schilt, 2011).
A dimension of impulsive behaviour now recognised to be distinguishable from response inhibition is reflection impulsivity -the tendency to gather and evaluate information prior to decision-making (Kagan, 1966) -measured using the information sampling task (IST). On the IST respondents make correct or incorrect judgements under varying levels of uncertainty, with a lower probability of being correct (p(correct)) at point of decision-making being indicative of disrupted reflection impulsivity (Clark et al., 2006(Clark et al., , 2009. Using this paradigm, Clark et al. (2009) found those who smoke both cannabis and cigarettes display impaired reflection impulsivity, evidenced through sampling less information and lower p(correct), compared with non-users of either drug. Impaired reflection impulsivity on the IST has also been demonstrated in adolescent cannabis only smokers relative to controls (Solowij et al., 2012). At no point, however, have cigarette only smokers been compared on the IST relative to NS, cannabis only smokers and smokers of both cannabis and cigarettes. This is an important next step needed in order to understand the effects of cigarette smoking and cannabis smoking on this form of impulsivity.
In summary, given the high prevalence rates of co-morbid cannabis-cigarette smoking, it is critical to further understand the role of impulsivity when these behaviours are performed both in isolation and in combination. Past research that has attempted to establish the role of impulsivity in cannabis use has been limited by a failure to consistently control for co-morbid cigarette smoking. The present study builds on previous research by utilising a multivariate approach to understanding the role of impulsivity in cannabis use. Differences between cigarette and cannabis smokers on both trait and behavioural impulsivity were assessed using a sample that included NS, cigarette only, cannabis only and cannabis plus cigarette smokers. Importantly, both cannabis groups in this study used tobacco in their joints ensuring high ecological validity, accurately representing how cannabis is smoked in Europe. Critically, this design enabled the first step towards disentangling cannabis smoking from cigarette smoking with respect to the unique relationships different facets of impulsivity may have with each of these behaviours.

Participants
Trait impulsivity participant sample. Trait impulsivity questionnaire data were gathered from a series of studies investigating the role of impulsivity in drug use with a total sample of 224. There were 44 NS (age M = 24.73, SD = 4.57; Male = 57%), 76 cigarette only smokers (age M = 23.55, SD = 5.75; Male = 55%), 47 cannabis only smokers (age M = 21.98, SD = 4.59; Male = 64%) and 58 cannabis plus cigarette smokers (age M = 22.62, SD = 3.37; Male = 74%). Individuals were recruited throughout the UK using opportunity and snowball sampling; all participants provided written informed consent. This study was approved by the Research Ethics Committees of Leeds Beckett University and the University of Leeds (approval reference: 15-0095) and was conducted according to the World Medical Association Declaration of Helsinki. Participants were excluded from participating if they were non-native English speakers; reported a history of psychiatric illness as defined in the Diagnostic and Statistical Manual of Mental Disorders (DSM-5;American Psychiatric Association, 2013), including disorders where impulsivity is a hallmark feature, e.g. attention deficit hyperactivity disorder; had a brain injury, or severe head trauma; were trying to abstain from smoking tobacco or cannabis; or were dependent on any drugs (including alcohol) other than nicotine or cannabis.
Participants were assigned to groups according to drug use frequency and duration of use, an approach that is common in the smoking literature (e.g. De Leon et al., 2003). In order to reach criteria for a cannabis smoker, individuals needed to have smoked cannabis once or more a week for the last 6 months and score ⩾13 on the Cannabis Use Disorders Identification Test -Revised (CUDIT-R), indicative of cannabis dependence (Adamson et al., 2010). Cannabis smokers who also smoked cigarettes daily and had done so for the last 6 months were classed as cannabis plus cigarette smokers; those who reported being non-smokers of cigarettes were classed as cannabis only smokers. Cigarette smokers had to have smoked tobacco cigarettes daily for the last 6 months and were selected in order to match cannabis plus cigarette smokers on frequency of cigarette smoking (this ranged between 3 and 20 cigarettes per day). NS reported no use of tobacco or any illicit substances and no current/past dependence on any substance.

Measures
Smoking behaviour and craving. Nicotine dependence was assessed using the Fagerström Test for Nicotine Dependence (FTND; Heatherton et al., 1991). Cannabis dependence was assessed using the CUDIT-R (Adamson et al., 2010). To assess smoking satiety on arrival at the behavioural session participants also self-reported recent drug use, drug craving and undertook expired carbon monoxide (CO) assessment with a Micro + Smokerlyzer (Bedfont, Maidstone, UK). Cigarette craving was measured using the Questionnaire of Smoking Urges -Brief (QSU-Brief; Cox et al., 2001), which consists of two sub-factors: positive (Cronbach's α = .95) and negative craving (α = .94). Cronbach's α for the QSU-Brief total score was .96. Cannabis craving was measured using the Marijuana Craving Questionnaire -Short Form (MCQ-SF; Heishman et al., 2009), which consists of four sub-factors: compulsivity, emotionality, expectancy and purposefulness craving. Cronbach's α for sub-factors ranged from .49 to .77 and the MCQ-SF total score α was .77.
Trait impulsivity. Trait impulsivity was assessed using the Barratt Impulsiveness Scale Version 11 (BIS-11;Patton and Stanford, 1995), which consists of three sub-factors: attentional, motor, and non-planning impulsiveness. Cronbach's α for BIS-11 sub-factors ranged from .70 to .75 and α for the BIS-11 total score was .84.

Behavioural impulsivity
Information sampling task (IST). The IST (Clark et al., 2006) was used to assess reflection impulsivity. The IST was composed of a single practice trial followed by 10 trials in each of two conditions: a fixed win (FW) condition and a decreasing win (DW) condition. On each trial a 5 × 5 matrix of grey boxes was presented, with a panel containing two boxes (one red and one blue) directly below. Selecting a grey box caused it to open immediately (for the duration of the trial) revealing a red or blue box. Respondents chose how many grey boxes to open, before deciding which of the two colours underlaid the majority in the matrix. Upon indicating a decision, any unopened boxes opened and a feedback message 'Correct! You have won (x) points' or 'Wrong! You have lost 100 points' was presented for 2 s. In the FW condition, the subject won 100 points for making a correct decision, regardless of the number of boxes opened. In the DW condition, the amount available to win started at 250 points decreasing by 10 points with each box opened. An incorrect decision resulted in a 100-point deduction from the total points scored. A variable delay (1s minimum) between trial onsets was used ensuring a minimum intertrial interval of 30 s. Mean probability of being correct at the point of decision, p(correct), is the primary index of reflection impulsivity; lower p(correct) is indicative of greater impulsivity. P(correct) is calculated as follows (equation (1); Clark et al., 2006): where Z = 25 -(number of boxes opened), and A = 13 -(number of boxes opened of the chosen colour). For example, if a decision is made towards red after opening 10 boxes (8 red, 2 blue), then Z = 25 -10 = 15, A = 13 -8 = 5, and p(correct) = [15!/(5! × 10!) + 15!/(6! × 9!) + . . . + 15!/(15! × 0!)]/2 15 = 0.94 (example taken from Clark et al., 2006: 518).
Secondary dependent variables include mean number of boxes opened, mean incorrect judgements, mean latency of box opening, and total points won.
Stop-signal task (SST). The SST ran through STOP-IT software and was used to evaluate response inhibition (Logan et al., 1997;Verbruggen et al., 2008). The present SST ran three blocks of trials (63 trials per block; 189 total), preceded by a practice block (15 trials). Trials began with a central fixation point presented for 500 ms, immediately followed by either a square or a circle for 1250 ms. Participants had to respond (on 75% of trials) by pressing left or right on the keyboard on presentation of a square or circle respectively. Randomly (on 25% of trials), participants were presented with an auditory stop-signal that occurred after varying delays following presentation of a square/circle. On receipt of this signal, participants had to withhold responding for that trial. Stop-signals were presented for 250 ms through external speakers and were randomly preceded by an equal number of squares and circles per block. The stop-signal delay (SSD) was initially fixed at 250 ms and subsequently titrated by 50 ms contingent on the previous stop-signal response using a staircase-tracking algorithm. This procedure adjusts the SSD until successful inhibition is achieved on 50% of stop-signal trials, allowing for estimation of the stop-signal reaction time (SSRT). SSRT was calculated by subtracting SSD from mean primary task reaction time. SSRT is the primary dependent variable; longer SSRT is indicative of poor response inhibition. Secondary dependent variables include: SSD, probability of responding on stop-signal trials, percentage of correct and missed responses, commission error reaction time, and correct response reaction time.

Procedure
All participants initially completed a questionnaire pack including measures of drug use, drug dependence, and trait impulsivity. The behavioural impulsivity sample then attended a test session where behavioural impulsivity tasks were completed. Individuals were asked to refrain from consuming >2 units of alcohol in the 24 h before the session or any caffeine in the 2 h before the session, to avoid residual and acute effects of these substances on performance. Cannabis groups were permitted to smoke cannabis ad-libitum but were asked to avoid using cannabis 3 h before the session to avoid acute intoxication (Meyer and Quenzer, 2013). Cigarette smoking before the session was ad-libitum. Smoking satiety was verified through self-report of last drug use, drug craving and expired CO at arrival to the session.

Statistical analysis
Data were analysed using SPSS Version 24 (Armonk, NY: IBM Corp.). Prior to inferential analyses, normality was assessed using the Shapiro-Wilk test and homogeneity of variance was assessed using Levene's test. Between-subject t-tests were used to compare groups on smoking behaviour and craving scores, and to verify whether participants who completed the behavioural impulsivity measures differed significantly in age, smoking behaviour and trait impulsivity scores from those who completed only the trait impulsivity measures. Variables violating normality were assessed using the non-parametric Mann-Whitney U test. Pearson's chi-square test was used to compare sex ratio across groups and samples.
Two (cigarette smoking status: cigarette smokers vs non-cigarette smokers) × two (cannabis smoking status: cannabis smokers vs non-cannabis smokers) between-subject ANOVAs were used to compare groups on age, expired CO, trait and behavioural impulsivity scores. Simple effects analyses with Bonferroni adjustment for multiple comparisons were used to explore significant interaction effects. On the SST, SSRT can only be reliably estimated when the mean probability of responding on stop-signal trials is close to 50% (Verbruggen et al., 2019). Consistent with this recommendation and previous work (Congdon et al., 2012), all individuals who inhibited less than 40% or greater than 60% of stop-signal trials (reflecting values falling one standard deviation outside of the mean probability of responding on stop-signal trials), were therefore omitted from analysis. Ten participants (six NS, two cigarette only, one cannabis only, one cannabis plus cigarette smoker) were excluded from SST analysis. Pearson's correlations were used to explore relationships between trait impulsivity, drug use and drug dependence, independently by drug group. Alpha (two-tailed) was set at p < .05.

Trait impulsivity participant sample characteristics
As shown in Table 1, there was no significant main effect of cigarette smoking status or interaction with cannabis status on age. However, a significant main effect of cannabis smoking status revealed non-cannabis smokers were significantly older than cannabis smokers. Sex ratio did not differ significantly across groups.
Importantly the two cannabis groups reported comparable levels of cannabis smoking frequency and years smoking cannabis. Despite comparable cannabis use, the cannabis plus cigarette group reported significantly greater cannabis dependence than the cannabis only smokers. Cannabis plus cigarette smokers also reported using tobacco in joints significantly more than cannabis only smokers. Cigarette only smokers did not differ significantly from the cannabis plus cigarette group on daily cigarette smoking, years smoking tobacco, or levels of nicotine dependence.

Associations between trait impulsivity, cigarette use, and nicotine dependence
As shown in Table 2, in cigarette only smokers, higher total, motor and non-planning impulsivity were associated with increased daily smoking of cigarettes and greater levels of nicotine dependence. A negative association was revealed between total impulsivity and years smoking tobacco. In contrast, in those who smoked cannabis and cigarettes, greater total impulsivity was associated with increased daily cigarette smoking, while greater attentional impulsivity was associated with increased levels of nicotine dependence.

Associations between trait impulsivity, cannabis use, and cannabis dependence
No significant relationships were found between trait impulsivity and either cannabis dependence or frequency of cannabis use in either cannabis group, with the exception of greater motor impulsiveness, which was associated with fewer days per week smoking cannabis in the cannabis plus cigarette smoking group (see Table 3).

Behavioural impulsivity participant sample characteristics
There was no significant main effect of cigarette smoking, cannabis smoking or their interaction on age, and sex ratio was equal across groups (see Table 4). Cannabis groups did not differ significantly on weekly or monthly cannabis use, years smoking cannabis, cannabis dependence, use of tobacco in joints, or time (hours) since last cannabis use prior to the session (Table 4). The two cannabis groups also reported comparable cannabis craving on (total and subscales of) the MCQ-SF on arrival at the session.
Cigarette only and cannabis plus cigarette groups did not differ significantly on cigarettes smoked per day, years smoking tobacco, nicotine dependence, or time (min) since last cigarette prior to the session (Table 4). The two cigarette groups also reported comparable cigarette craving on (total and subscales of) the QSU-Brief on arrival at the session.
A significant main effect of cigarette smoking status for expired CO levels was found (Table 4), indicating that cigarette smokers had significantly greater CO levels compared with nonsmokers of cigarettes at arrival to the session. No significant main effect of cannabis smoking status or a cannabis × cigarette smoking status interaction were revealed.

Comparison of trait-only participants and behavioural impulsivity sample
The behavioural impulsivity sample did not differ significantly from participants who had only completed the trait impulsivity measures in terms of age, sex ratio, drug use or trait impulsivity scores (see Table 5).
Reflection impulsivity P(correct). For FW trials, no significant main effects for cigarette or cannabis smoking status (both F < 1, p ⩾ .370, η p ⩽ .01), were revealed on p(correct); there was, however, a significant cannabis × cigarette smoking status interaction (F (1, 83) = 9.14, p = .003, η p = .10). The three smoking groups had similar p(correct) scores that were lower than NS of either substance; see Figure 2a. Simple effects analyses showed that cigarette only (p = .001) and cannabis only (p = .011) smokers had significantly lower p(correct) scores compared with NS. However, using both cigarettes and cannabis together did not lead to significantly further reductions in p(correct) scores relative to using either cigarettes (p = .101) or cannabis (p = .202) alone.
Boxes opened. For FW number of boxes opened, there was no significant main effect of cannabis smoking status (F < 1, p = .541, η p = .01), but a significant main effect of cigarette smoking status (F (1, 83) = 4.26, p = .042, η p = .05), and a significant cigarette smoking × cannabis smoking status interaction (F (1, 83) = 5.52, p = .021, η p = .06). Simple effects analyses revealed cigarette only (p < .001) and cannabis only (p = .043) smokers opened significantly fewer boxes compared with NS of either substance. However, using both cigarettes and cannabis was not associated with a further reduction in box opening, relative to using either cigarettes (p = .213) or cannabis (p = .863) alone, see Figure 3a.
For DW number of boxes opened there was a significant main effect of cigarette smoking status (F (1, 83) = 4.30, p = .041,   η p = .05), such that smokers of cigarettes opened significantly fewer boxes compared with individuals who did not smoke cigarettes (Figure 3b). There was no significant main effect for cannabis smoking status or cigarette × cannabis status interaction on the number of boxes opened (both F < 1, p ⩾ .513, η p ⩽ .01).

Incorrect judgements. For FW incorrect judgements there
were no significant main effects for cigarette or cannabis smoking status (both F < 1, p ⩾ .726, η p ⩽ .00); however there was a significant cigarette × cannabis smoking status interaction(F (1, 83) = 9.59, p = .003, η p = .10). Simple effects analyses showed cigarette only (p = .003) and cannabis only (p = .021) smokers made significantly more incorrect judgements compared with NS of either substance. In this case, cigarette only (p = .046) but not cannabis only smokers (p > .999) also made significantly more incorrect judgements relative to users of both drugs, see Figure 4a. For DW incorrect judgements there were no significant main or interaction effects (all F (1, 83) ⩽ 2.74, p ⩾ .102, η p ⩽ .03; see Figure 4b).
Latency of box opening. A significant main effect of cigarette smoking status on FW latency of box opening (F (1, 83) = 4.53, p = .036, η p = .05) revealed cigarette smokers took longer to open boxes than non-cigarette smokers There was also a significant main effect of cigarette smoking for DW latency of box opening (F (1, 83) = 4.08, p = .047, η p = .05), this time qualified by a significant cigarette smoking × cannabis smoking status interaction (F (1, 83) = 5.47, p = .022, η p = .06).
Simple effects analyses showed cigarette only (p = .778) and cannabis only (p = .300) smokers did not differ significantly from NS on time spent opening boxes. However, cannabis plus cigarette smokers spent significantly longer opening boxes than cigarette only (p = .024) and cannabis only smokers (p = .010), see Figure  5b. There was no significant main effect of cannabis status (F < 1, p = .408, η p = .01) Total points won. For FW total points won there were no significant main effects for cigarette or cannabis smoking status (both F < 1, p ⩾ .726, ηp ⩽ .00); however, there was a significant cigarette × cannabis smoking status interaction (F(1, 83) = 9.59, p = .003, ηp = .10). Simple effects analyses showed cigarette only (p = .003) and cannabis only (p = .021) smokers won significantly fewer total points than NS. Cigarette only (p = .046) but not cannabis only smokers (p > .999), also won significantly fewer points than cannabis plus cigarette smokers, see Figure 6a.  In contrast, for DW trials there was no significant main or interaction effects on total points won (all F(1, 83) ⩽ 2.26, p ⩾ .136, ηp ⩽ .03; see Figure 6b).

Response inhibition
There were no significant main or interaction effects on SSRT or any secondary dependent variables of the SST (all F (1, 73) ⩽ 2.86, p ⩾ .095, η p ⩽ .04); see Table 6 for SST dependent variable descriptive statistics.

Discussion
It is unclear from existing research how the role of impulsivity may differ across cannabis only, cigarette only and cigarette plus cannabis users, and to what extent cigarette smoking may account for the heightened impulsivity previously found in cannabis users. To explore this, the present study assessed the effects of cigarette and cannabis smoking on trait and behavioural impulsivity across a sample of NS, cigarette only, cannabis only, and cigarette plus cannabis smokers. Findings demonstrated clear differences   between these groups on elements of trait and behavioural impulsivity, and how these relate to key cigarette and cannabis use outcomes, suggesting that unique aspects of impulsivity may be differentially related to cannabis and cigarette smoking. Each of these findings will be discussed in more detail below.
Total, attention and non-planning trait impulsivity were significantly higher in cigarette smokers (cannabis and non-cannabis users) compared with individuals who did not smoke cigarettes. Elevated trait motor impulsiveness was also observed in cigarette smokers, however, on this component of impulsivity the effect of cigarette smoking was specific to those who did not also use cannabis. In contrast, trait impulsivity appeared to be unrelated to cannabis use. The present findings stand in opposition to previous research demonstrating greater trait motor and non-planning impulsivity in cannabis only smokers compared with NS (Gilman et al., 2015), instead suggesting that impulsive tendencies may be especially prevalent in cigarette smokers compared with NS. The present findings are however consistent with previous work demonstrating elevated trait impulsivity on the BIS-11 in both cigarette only and cannabis plus cigarette smokers compared with NS (Dougherty et al., 2013;Mitchell, 1999;Moreno et al., 2012), and also with those reporting comparable levels of trait impulsiveness between cigarette only and cannabis plus cigarette smokers (Beaton et al., 2014).
Exploratory correlations between self-reported drug use and trait impulsivity in each drug user group revealed significant positive relationships between the number of cigarettes smoked per day and total trait impulsivity in both the cigarette only and cannabis plus cigarette groups. Number of cigarettes smoked per day was also significantly positively associated with greater motor and non-planning impulsiveness in cigarette only smokers. In addition, greater nicotine dependence was significantly positively associated with trait total, motor and non-planning impulsiveness in cigarette only smokers, whilst in cannabis plus cigarette smokers only attentional impulsiveness was related to level of nicotine dependence. Therefore, whilst relationships between trait impulsivity and indicators of nicotine dependence were detected in both cigarette smoking groups, these associations differed between cigarette only and cannabis plus cigarette groups across subscales of trait impulsivity. These findings suggest that in cannabis plus cigarette smokers, greater tendencies towards distractibility and inattention may play critical roles in the transition to nicotine dependence. By contrast, in cigarette only smokers, greater tendencies towards rapid and unpremeditated behaviour may instead be integral traits that increase vulnerability for transitioning to nicotine dependence. In direct opposition, no significant positive associations were found between trait impulsivity and cannabis dependence or frequency of cannabis smoking in either cannabis group. Interestingly, motor impulsivity was found to be significantly negatively associated with daily frequency of cannabis smoking, providing further evidence of the limited role heightened trait impulsivity may play in cannabis use relative to cigarette smoking.  Data presented as mean (SD). NS: non-smoker of cigarettes and cannabis; CAN NON-CIG: cannabis only smoker; CIG: cigarette only smoker; CAN CIG: cannabis plus cigarette smoker; SSRT: stop-signal reaction time; SSD: stop-signal delay; P(respond|signal): probability of responding on stop-signal trials; ms: milliseconds; RT: reaction time. a Ten participants (six NS, two cigarette only, one cannabis only, one cannabis plus cigarette smokers) were excluded from the stop-signal task analysis due to inability to accurately estimate SSRT.
Taken together, these differences between cigarette and cannabis users in elements of trait impulsivity and in how they relate to key outcomes, suggest that a failure to assess and/or control for cigarette smoking may be responsible for inconsistencies in past research exploring trait impulsivity in cannabis smokers. Given the robust associations shown between variation in mesocorticolimbic dopamine functioning, integrity of frontostriatal circuitry and levels of trait impulsivity (e.g. Buckholtz and Meyer-Lindenberg, 2012;Volkow et al., 2011), differences in trait impulsivity observed in cigarette smokers compared with non-cigarette smokers may indicate that heightened trait impulsiveness increases the risk of tobacco cigarette smoking but may not be uniquely related to cannabis smoking. Alternatively, it is possible that trait impulsivity is influenced by prolonged cigarette smoking, suggesting trait propensities are not entirely stable constructs. It may be that regular exposure to greater levels of nicotine with or without co-morbid cannabis smoking leads to distinct changes in some dimensions of trait impulsivity as compared to cannabis smoking alone. Supporting this argument, preclinical studies indicate that chronic nicotine administration increases behavioural disinhibition (Kolokotroni et al., 2012) and impulsive choice (Kayir et al., 2014) in rodents. Furthermore, longitudinal assessment of other psychostimulants has found increased cocaine intake to be associated with elevated trait impulsivity on the BIS-11 while decreased intake was associated with reduced impulsivity at follow-up (Hulka et al., 2015). Against this view however, no significant positive correlations between trait impulsivity and years smoking tobacco were revealed in the present study, which would be expected if heightened trait impulsivity was a consequence of chronic nicotineinduced neural adaptations. It is imperative that future longitudinal research is conducted to determine whether trait impulsivity predates and/or is affected by extended drug use, to further understand its role in addiction and clarify its potential as an intervention target.
No significant effects of cigarette or cannabis smoking status were revealed on response inhibition. The present findings extend existing research (Filbey and Yezhuvath, 2013;Grant et al., 2012) demonstrating that capacity for response inhibition on the SST is comparable in cannabis smokers compared with non-cannabis smokers. Whilst impairments in response inhibition on the SST have been reported in cannabis smokers previously (Ramaekers et al., 2006), such observations have been limited to performance whilst under the acute influence of δ-9tetrahydrocannibinol. Cannabis smokers in the present study on average smoked approximately 13 h before the session, suggesting that previously reported cannabis-induced impairments in response inhibition may be transient. Findings are also consistent with previous work comparing cigarette smokers and NS reporting comparable response inhibition on the SST (Bekker et al., 2005), a Stroop task (Powell et al., 2002) and an antisaccadic task (Domier et al., 2007), but not with others indicating reduced response inhibition in cigarette smokers on the go/no-go task (Luijten et al., 2011). Discrepancies in findings may be due to differences between tasks in the aspects of the behavioural inhibition they measure and the partially distinct neuropharmacological mechanisms they rely on (Eagle et al., 2008). The SST assesses action cancellation (i.e. stopping a pre-potent motoric response after it is initiated) and task performance is improved by administration of atomoxetine, a selective norepinephrine reuptake inhibitor (Broos et al., 2012). In contrast, the go/no-go task assesses action restraint (i.e. inhibiting of a pre-potent motoric response before it is initiated) and task performance is impaired following serotonin depletion (Harrison et al., 1999).
During FW trials on the IST, impaired reflection impulsivity was demonstrated in individuals who smoked cigarettes alone or cannabis alone relative to NS of either substance. Impaired reflection was demonstrated in both these drug groups compared with NS, evidenced through lower p(correct), reflective of decision-making under high levels of uncertainty, with limited information (opening fewer boxes) resulting in more incorrect judgements and fewer total points won. This heightened impulsive responding was restricted to FW trials where there were no penalties for sampling information, consistent with previous research in other drug users (Clark et al., 2006). However, there was some indication of reduced information sampling in DW trials in cigarette smokers (irrespective of cannabis use) compared with non-cigarette smokers, suggesting cigarette smoking may increase impulsive decision-making during situations involving potential risk.
Importantly, on FW trials, neither p(correct) -which is the main index of reflection impulsivity -nor number of boxes opened were altered further in those who smoked cigarettes and cannabis, suggesting co-morbid cigarette-cannabis smoking does not have additive effects on reflection impulsivity. On the contrary, those who smoked cigarettes alone made more incorrect judgements and won fewer total points than those who smoked both cigarettes and cannabis. However, the relevance of these findings is diminished given that the accuracy of judgements and total amount of points won do not directly quantify the amount of information sampled prior to decision-making (Clark et al., 2006), making these variables less valid measures of reflection impulsivity than p(correct).
Group differences were also observed in latency of box opening on the IST. During FW trials, cigarette smoking was associated with slower responses, whilst it was combined cigarette and cannabis smoking that was associated with slower responses in the DW condition. Speed of box opening has been argued to offer an insight into task-related arousal (Clark et al., 2006), suggesting that these findings may reflect differences in motivation towards the task as a function of cannabis and cigarette smoking. Reassuringly however, the group differences found in latency of box opening did not reflect those found for p(correct), suggesting any differences in motivation that may be indicated by the latency findings are unlikely to explain the differences in reflection impulsivity that were found. Furthermore, all groups sampled less information during DW than FW trials, suggesting they were sensitive to the differing reward contingencies of the task and motivated to win points.
These findings are the first to demonstrate that cigarette smoking or cannabis smoking alone significantly impair reflection impulsivity relative to no drug use. Data suggest that comorbid cigarette-cannabis smoking does not have additive impairing effects on reflective decision-making. Findings extend previous work reporting poor reflection impulsivity in cannabis smokers where the additive effect of these drugs was not tested (Clark et al., 2009;Solowij et al., 2012). Less thoughtful decision-making may therefore be one mechanism by which cannabis smoking and cigarette smoking are independently maintained, through users being prepared to accept greater risks regarding their decisions, such as the negative health and financial consequences of their decision to smoke. Importantly, these findings have implications for drug rehabilitation. Continuing to smoke either cigarettes or cannabis whilst attempting to quit the other may sustain drug-induced impulsive behaviour leading to poorer judgements and increased likelihood of drug relapse during attempted abstinence. Indeed, previous research strongly supports this view, demonstrating cigarette smoking during cannabis abstinence significantly increases risk of relapse compared with those who do not smoke cigarettes (Haney et al., 2013). Taken together with the present study's findings suggesting that cigarette smoking or cannabis smoking each impair reflective decision-making, it is imperative that interventions aim to stop use of both drugs in parallel to reduce risk of relapse, in line with previous suggestions (Rabin and George, 2015).
In the present study, different findings were observed between distinct dimensions of behavioural impulsivity. Differential performance on the IST and SST have been previously reported in other participant groups (e.g. Huddy et al., 2013;Jepsen et al., 2018), and is likely to reflect variations in task nature that rely on the integrity of partially distinct neurobiological systems (Caswell et al., 2015). Specifically, inhibitory control tasks involve withholding motoric responding to pre-potent stimuli, responding is highly time-sensitive, and the activity of the right inferior frontal gyrus and anterior insula is integral for inhibitory control (Aron et al., 2014;Eagle et al., 2008;Hampshire and Sharp, 2015). In contrast, the IST involves decision-making under uncertainty and responding is not time-sensitive. Although the neurobiological mechanisms underlying reflection impulsivity are currently not well defined, recent evidence suggests that inhibitory control and reflection impulsivity may be mediated by distinct physiological mechanisms (Herman et al., 2019). An important further step for future research is to explore differences between NS, cannabis only, cigarette only and cigarette plus cannabis smokers on measures of impulsive choice (e.g. delay discounting) given that this form of impulsivity is also behaviourally and neurobiologically distinct from the two dimensions studied (de Wit, 2009;Caswell et al., 2015). Behavioural impulsivity may also be greater when performance is measured in the context of drug-related rewards. Indeed, heightened response disinhibition and impulsive choice has been previously found in various drug users when drug-related stimuli are presented relative to non-drug stimuli (e.g. Pike et al., 2013;Weafer and Fillmore, 2012;Zeeb et al., 2010).

Limitations
Despite clear distinctions in trait and behavioural impulsivity across drug user groups, some limitations of the study need to be addressed. Although cannabis groups did not differ in frequency and duration of cannabis use, cannabis plus cigarette smokers indicated significantly higher levels of cannabis dependence than cannabis only smokers on the CUDIT-R. This distinction was only present in groups being compared on trait impulsivity. A possible reason for heavier levels of cannabis dependence in cannabis plus cigarette smokers may be related to their co-morbid smoking of cigarettes, consistent with previous work indicating cigarette smoking promotes greater levels of cannabis dependence (Hindocha et al., 2015). However, the impact of this on data in the present study is likely to have been minimal given that (a) no significant correlations between trait impulsivity and cannabis dependence were found, and (b) heavier cannabis dependence reported in the cannabis plus cigarette group would bias results in favour of an additive effect of co-morbid cannabis-cigarette smoking on impulsivity, which was not found. Furthermore, previous research has indicated that level of dependence may be less accurate at predicting motivation for cannabis than weekly use (Cousijn et al., 2011), a measure that was consistent across groups in the current study.
The present findings also need to be considered in view of the fact that cannabis only smokers were not truly naïve to nicotine, as the majority of individuals in both cannabis smoking groups acknowledged using tobacco in their joints. However, this is highly consistent with European norms of cannabis smoking, where the majority of users smoke cannabis joints with tobacco (Hindocha et al., 2016;Hindocha et al., 2015). Furthermore, cannabis plus cigarette smokers reported using significantly more tobacco in their joints than cannabis only smokers, and selfreported smoking status of cannabis only smokers was further confirmed through biochemical (CO) assessment, demonstrating CO levels equivalent to a NS (Cox and Whichelow, 1985), thereby providing strong support for the legitimate classification of groups. Importantly, impulsivity differences were still observed between cannabis groups, strongly supporting the idea that despite both groups being exposed to nicotine via their joints, these individuals likely face different challenges relating to their drug use, which may in turn have implications for distinct treatment needs.
Although groups did not differ in time of last cigarette or cannabis use prior to the session, and current smoking status of individuals was supported by CO data, the amount of each drug used before the session was not restricted. Although consistent with protocols of previous investigations in these drug users (e.g. Lawn et al., 2017;Smith et al., 2011), stricter control of quantity of drug use prior to the session is an important consideration for future research. The level of alcohol use was also not controlled for in the present study. Whilst participants were excluded if they reported alcohol dependency, levels of recreational alcohol may have differed across groups and could have potentially impacted on levels of impulsive behaviour (e.g. Reed et al., 2012).

Conclusions
Previous research attempting to understand factors underlying problematic cannabis use may have been limited by several studies overlooking the effects of co-morbid cigarette smoking. The present study was the first step towards disentangling the effects of these substances by exploring the effects of both cannabis and cigarette smoking on dimensions of trait and behavioural impulsivity in a sample of NS, cigarette only, cannabis only and cannabis plus cigarette smokers. The main findings of this study demonstrate that significantly higher trait impulsivity was found in cigarette smokers, with elevated motor impulsiveness being confined to those who did not also use cannabis. Elevated trait impulsiveness in cigarette only and cannabis plus cigarette groups may be a risk factor for the initiation of cigarette smoking and greater levels of nicotine dependence. In contrast, trait impulsivity does not appear to be related to increased cannabis smoking or dependence. Cigarette smoking or cannabis smoking appear to independently significantly impair capacity to reflect on information prior to decision-making, but interestingly, when used co-morbidly did not have an additive effect on the degree of uncertainty tolerated at the point of decision-making. In support of recent recommendations, that tobacco use should be controlled in research on health outcomes in cannabis use (Lovell et al., 2018), the present findings highlight the importance of cigarette smoking within the context of understanding the relationship between cannabis smoking and impulsivity. Treatment of cannabis dependence may benefit from tailoring interventions depending on whether individuals are co-morbid cigarette smokers; programmes targeting those who smoke both substances may benefit from ceasing use of both drugs in parallel to reduce the risk of impaired self-reflection thereby increasing likelihood of abstinence.

Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by Leeds Beckett University through a PhD studentship awarded to JR.