SUDs can be conceptualized as pathologies of decision-making (Bickel et al. 2014; Ekhtiari et al. 2017; Redish et al. 2008). We used an experimental medicine approach to study the neural valuation of drug (cannabis) and non-drug (snack) rewards in PWUCF, following neutral and active (cannabis/snack) cues. We found positive cannabis SV signals within pre-specified ROIs (vmPFC, dPCC and VS), but the expected positive snack SV signals could not be detected. In the vmPFC and VS, cannabis SV signals were larger than snack SV signals. Additionally, within the dPCC, active cues had differential effects on cannabis and snack SV signals.

Numerous studies (Chib et al. 2009; Hare et al. 2009; Plassmann et al. 2010) and meta-analyses (Bartra et al. 2013; Clithero and Rangel 2014) have demonstrated that SV signals exist in a valuation network including the vmPFC, striatum, and dPCC. We observed cannabis SV signals in these regions, revealing that the brain’s computation of value for cannabis functions as expected in PWUCF (comparable to that for non-drug rewards in other populations). This is contrary to theories that in addiction the valuation of drugs is impaired, with decision-making habitual (Everitt and Robbins 2005). The only prior study to directly investigate neural SV signals in relation to substance use (Lawn et al. 2020) also observed expected cigarette SV signals in cigarette smokers. Moreover, three other studies – while not modelling SV encoding – found that people who use cannabis, alcohol, and tobacco recruited valuation circuitry during drug purchase decisions (Bedi et al. 2015; Gray et al. 2017; MacKillop et al. 2014). This converges with current findings to indicate that the brain’s computation of drug value is not “impaired” in people who frequently use drugs.

Conversely, the expected snack SV signals were not observed. Several studies have investigated value computation guiding choices for food rewards (e.g.,13,32,36); these have overwhelmingly reported SV signals for palatable food in the regions investigated. We therefore expected to observe positive snack SV signals. That these could not be detected with the fMRI measure employed – either in ROI-based or exploratory whole-brain analyses – may be indicative of dysregulation of the neural valuation of non-drug rewards in our sample. Similarly, weaker SV signals for shop vouchers (but not cigarettes) were found in the PCC of dependent compared to occasional cigarette smokers in a prior study (Lawn et al. 2020). Together, these results support theories that non-drug reward value processing may be compromised in people who frequently use cannabis or tobacco (Goldstein and Volkow 2011). In other words, the valuation system’s functionality may be moderated by reward type, such that the neural processing of drug value is intact while that of non-drug reward value is altered. Speculatively, it is possible that progressive changes to non-drug reward SV signaling plays a causal role in the pathophysiology of CUD, such that relative SV encoding for drug vs. non-drug rewards shifts during development of addiction (Hogarth and Field 2020). Future research must compare people who use cannabis on a spectrum of no, mild, moderate, and severe CUD to formally examine this.

At a behavioral level, participants interacted with snack rewards in a typical way, as they did with cannabis rewards. They were motivated to receive snacks, willing to spend money on snacks (although significantly less money than on cannabis, in the behavioral choice task), and they responded to snack cues with increased craving. Hence, the apparent lack of a typical SV signal for non-drug rewards warrants further investigation as a potential early marker of choice dysregulation in CUD. It is important to note that we designed the behavioral choice task to produce indifference points that equalize the number of choices for snacks/cannabis vs. money in the fMRI choice task so we could clearly probe the neural value signals, thus we did not expect, nor observe, behavioral differences in the fMRI task.

How might near-daily cannabis use itself affect valuation processes? The endocannabinoid system is thought to modulate reward circuitry (Gardner 2005; Parsons and Hurd 2015) and CB1 receptors are expressed densely in the vmPFC and VS (Curran et al., 2016). Frequent recent cannabis use could, via downregulation of the CB1 receptor (D’Souza et al. 2016), alter neural reward processing, as has been shown in some (Skumlien et al. 2021), but not all (Skumlien et al. 2022), studies of people who regularly use cannabis. However, it remains unclear why such effects would preferentially affect SV encoding for non-drug, but not drug, rewards.

Our results are concordant with literature demonstrating that people who frequently use drugs are oftentimes economically rational, and not compulsive, when they use drugs (Heyman 2013; Higgins 1997; Hogarth 2020). Although the value assigned to drugs is higher in people with SUDs (Hogarth 2020) and the SV signals potentially stronger (Lawn et al. 2020), value-based decision-making about drugs remains intact.

Findings did not support our hypothesis that active cues would preferentially augment cannabis SV signals. In the dPCC, there was an interaction between reward and cue; the cannabis SV signals were differentially affected by the presence of active cues relative to snack SV signals. The PCC is thought to be particularly sensitive to cues and craving (Brewer et al. 2013; Li et al. 2013). A similar pattern was also observed in whole-brain analyses in the angular gyrus, another region considered important in the valuation network (Clithero and Rangel 2014). This may indicate that the presence of cues weakens the relationship between encoding in some regions and subjective valuation of cannabis rewards. However, the current results do not present a clear picture regarding cue effects; further investigation into the impact of cues on drug SV encoding is therefore warranted.

As the first investigation of SV encoding in relation to cannabis use, this study had limitations. We did not include a light or non-cannabis-using control group with which to compare the non-drug SV signals of our PWUCF group. However, a wealth of evidence indicates that healthy participants show positive non-drug-reward SV signals in these ROIs (Chib et al. 2009; Clithero and Rangel 2014; Plassmann et al. 2010), including snack SV signals observed during a similar task (Hare et al. 2009). A second limitation was that cue exposure occurred outside of the scanner, with some small delays between exposure and scan initiation. This may have contributed to the relatively minor effects of the cues observed. To counter this, cues were designed to be highly potent (e.g., multisensory including engagement with personalized cannabis paraphernalia). However, future studies could valuably explore the possible impacts of value-encoding during choices for drugs in the presence of drug-paired cues, rather than after exposure. A third limitation is that we did not have sufficient power to assess the impact of factors such as gender, or variables known to be clinically-relevant markers in cannabis use, such as tobacco smoking. A fourth limitation is that we did not assess cannabis use disorder (CUD) or CUD severity. However, our participants smoked cannabis near daily, at a mean rate of 6.5 days/week, smoking an average of 7.7 joints/day. This type of high frequency, high quantity use is indicative of high risk for cannabis use disorder (Callaghan et al. 2020). Ongoing work by our team will investigate relationships between CUD severity and SV signals for cannabis and non-drug rewards.

Limitations notwithstanding, the current results suggest that this novel, controlled, multidisciplinary methodology represents a promising approach to further elucidate the neurobehavioral choice processes subserving the development and maintenance of frequent substance use. There are numerous potential avenues for future research. It remains unclear whether similar patterns of preferentially disrupted SV encoding would be observed in people using drugs other than cannabis and cigarettes (Lawn et al. 2020). Similarly, the effects of state factors like withdrawal on SV encoding for drug and non-drug reinforcers in different populations remains unknown. A key question for future investigation will be the extent to which these neural SVs change following cessation of drug use. Links between neural SV signals, behavior, and clinical outcomes should be investigated.

This was the first study to investigate neural SV signals for a real, tangible, illicit drug: other studies of drug-related choice used hypothetical rewards (Hogarth 2020) or drug images (Moeller et al. 2013) or studied legal drugs (Lawn et al. 2020; MacKillop et al. 2014). In PWUCF, cannabis SV signals were intact, while snack SV signals could not be detected, tentatively suggesting dysregulation of these neurobiological choice processes.