Opioid-related overdose deaths in the United States have increased by 109% between 2015 and 2020, including a 38% increase in 2020 alone. Heroin use in particular increased by approximately 40% in response to the COVID-19 pandemic. Despite the magnitude of this drug use and overdose epidemic, however, and compared to other drugs of abuse, the neurobiological substrates of human opioid use disorder (OUD) have been largely understudied. In addiction to other drug classes (e.g., stimulants), drug cues have been shown to divert attention away from alternative reinforcers (e.g., social, food, sex), a core imbalance in motivational salience suggested to drive drug use and perpetuate the addiction cycle. Whether a similar hijacking occurs in opioid addiction, however, remains an open question. Here we were interested in the extent (i.e., its expression as a phenotype shared by groups of individuals) of such drug-biased neural processing in real-world contexts and its relationship to cue-induced craving in heroin addiction.

To explore the brain substrates of cue reactivity and craving in human drug addiction, neuroimaging studies commonly employ static images (e.g., pictures of drugs). Instead, here we used an engaging heroin-related movie, a dynamic, narrative-based, and context-rich natural stimulus, to better approximate actual real-world experiences in individuals with OUD (iOUD). Specifically, we used inter-subject synchronization to analyze fMRI BOLD responses to the movie, hypothesizing a shared brain signature in iOUD. Our primary region of interest was the Nucleus Accumbens (NAc), a major processing hub for motivational salience attribution, reward anticipation, and craving, where we expected to document unique synchronized tracking of drug-related content as predictive of cue-induced craving in the iOUD.

Functional MRI BOLD activity was recorded in 29 treatment-seeking medically stabilized iOUD (40.4±10.3 years, 23 Male, 19 White) and 16 age, sex, and race-matched healthy controls (HC) (43.8±10.3 years, 10 M, 11 W), while subjects watched the first 17 minutes of the movie “Trainspotting”, which contains scenes of explicit heroin use, as well as food and social scenes that are highly salient but not directly related to heroin use. Importantly, the overarching narrative structure of the movie is centered around drug-addiction, including complex contextual elements relevant to OUD (e.g., social, emotional, and economic challenges of addiction). To infer the movie features that led to synchronized NAc activity, we used an approach developed by Hasson et al., which was inspired by reverse correlation analyses of single unit recordings. We were interested in whether the iOUD would show differential synchrony in the NAc during the drug-related content at the expense of other typically motivating stimuli.

Following Hasson et al., we decomposed the BOLD signal in each voxel during movie watching into two components. The first component, which we call the global component, relies on the observation that engaging narratives evoke a wave of global synchronized activity across much of the brain, driven by surprising and emotionally engaging moments of a movie. Here, the global component was derived by averaging the BOLD signal across all grey matter voxels and then z-scoring the resulting time series within each subject. The second component, which we call a selective component, was derived by regressing the global component out of the time series at each voxel, yielding a residual signal that is not explained by the global activation wave throughout the brain and is specific to each voxel. Hasson et al. has shown that a reverse correlation analysis of this selective component can recover canonical response properties from traditional GLM-based fMRI tasks [e.g., fusiform face area responds to scenes with faces]. Here, for the first time, we applied this method for the purpose of tracking the neural signature of drug related motivation/incentive salience in subcortical regions in heroin-addicted individuals.

Applying this reverse correlation method to the mean selective component separately in the left and right NAc (LNAc and RNAc, respectively), we found a set of significant time points during the movie that elicited a synchronized NAc response in each subject group (see Identifying synchronized TR’s and region-specific reactivity in supplemental methods). Accounting for the hemodynamic lag, we then identified clips of the movie that contributed to this synchronization (5-10 sec before each significant time point). Stitching these clips together yielded separate NAc-specific “movies” for each group. Remarkably, the LNAc-specific movie contained almost exclusively clips of drug use in the OUD group and no scenes of drug use in the HC group (Fig. 1; see supplemental videos).

For statistical confirmation purposes, and based on previous methods, we quantified the content of the LNAc-specific movies by first breaking the entire movie into 24 distinct scenes. Each scene was a priori classified as drug or non-drug, based on whether drugs or drug use were explicit in the scene (see supplemental methods section Labeling of region-specific movies for detailed criteria). Each TR from the LNAc-specific movie could then be matched to a scene (multiple TRs could be matched to the same scene) and thereby labeled as drug or non-drug, yielding a count of scene types to be used for statistical comparisons. Using this count as the dependent measure, chi-square tests revealed a significantly different scene type distribution between the iOUD and HC for the LNAc-specific movie, such that for iOUD it fell almost exclusively within the drug scenes, while for HC it was close to what could be expected if scenes were selected at random (iOUD: 40/8, HC: 25/40 drug/non-drug; χ 2 (1)=20.95, p=0.0000047). Indeed, nonparametric tests also indicated that the number of drug scenes was significantly different (𝛼=0.05) from random phase shuffled signals for iOUD, but not HC (Fig. S1). We further compared the magnitude of the LNAc-selective component directly between the two groups and found that it was significantly different during each of the LNAc-specific movies (Fig. S2).

Given the skewed responsiveness of the LNAc to drug stimuli in iOUD and this region’s role in craving, we inspected whether the magnitude of the LNAc-selective component predicts cue-induced craving for drugs. In-scanner subjective heroin craving ratings were collected immediately before and after subjects watched the movie (for calculation of movie-induced craving). After leaving the scanner (within 45 minutes after watching the movie), subjects also completed a custom questionnaire intended to probe scene-specific cravings. The LNAc-specific component averaged over significant time points for each subject predicted both sceneand movie-induced cravings (Fig. 2A,B). Interestingly, the correlation with baseline craving (measured with the heroin craving questionnaire, HCQ) (12) was not significant (Fig. 2C), suggesting that the LNAc synchronous reactivity scales with the dynamic experience of craving induced by watching the movie.

We next tested whether the LNAc reactivity was related to other potential explanatory variables including the demographic measures that differed between the groups and, in iOUD, measures of recent and lifetime drug use, withdrawal symptoms and severity of dependence (Table S2). The LNAc reactivity signal was not significantly correlated (𝛼=0.05) with any of these measures, further suggesting that it was not driven by the demographic factors that commonly differ between individuals with and without addiction, and that it may specifically be predictive of craving induced by movie-watching. To better examine the specificity of the above results to the LNAc, we repeated the above analyses (including statistics and correlations) for the global component, several other regions of interest involved in drug cue reactivity [ventromedial prefrontal cortex (13), dorsolateral prefrontal cortex, anterior cingulate cortex (15), insula (16), putamen], and a control region [fusiform cortex (18)]. Other than the LNAc, only the left anterior insula showed significant drug-biased reactivity in OUD (Table S3). However, only the LNAc showed a complete reversal of scene-type preferences between the groups (i.e., drug scenes for iOUD and non-drug scenes for HC) and a correlation with the cue-induced craving measures, suggesting specificity of these results to the LNAc.

Supporting the impaired response inhibition and salience attribution model, and other prominent neurobiological theories of addiction, which posit an aberrant incentive salience attribution to drug-related stimuli at the expense of other typically motivating stimuli, we revealed a bias toward naturalistic depictions of drug-use in the LNAc of iOUD. Documenting this imbalance during exposure to a movie, a narrative-based and context-rich dynamic stimulus, suggests that such neural bias may dominate during real-life daily ongoing experience. Here we documented such a shared narrowing of responses in addiction towards drug-related stimuli even in currently abstinent, treatment-seeking and medication-maintained individuals. Predicting cue-induced craving, this interpretable brain signal may also be a powerful predictor of later drug use and relapse, as remains to be ascertained with longitudinal follow-up studies in these subjects.

Whether the dynamic bias of the NAc to a drug-related narrative is generalizable to other substance use disorders, or to individuals at risk for developing drug addiction, remains an open question. Future efforts may also benefit from more sex-balanced samples to interrogate potential sex-differences related to mapping drug cue reactivity and craving in a naturalistic context. Such efforts may also warrant larger samples, although the inter-subject correlation based statistics yield stable results with comparable samples in traditional block designs, and even smaller samples in naturalistic designs. Furthermore, future efforts could employ a whole-brain search for inter-brain synchronization during movie watching, complementing our regional focus (that accounted for the global component and corrected for multiple comparisons). A real-time measure of engagement during movie watching (e.g., with eye-tracking or online scene-specific ratings) may have provided more information about participants’ experiences during the movie, but we decided in this first effort not to disrupt the continuous narrative structure and engagement during movie viewing, instead obtaining a comprehensive post-movie survey.

To the best of our knowledge, this is the first use of dynamic naturalistic stimuli for the demonstration of drug-biased processing as a predictor of cue-induced craving in the addicted human brain. The interbrain synchronization and scene classification methods introduced here for the first time in the context of heroin addiction allow us to highlight specific scenes that drive the shared striatal reactivity; as compared to typical block- or event-related picture-based studies, these methods can better decode the actual features/identity of the craving-inducing stimuli even in this treatment-seeking group. Crucially, this approach effectively operationalizes daily-life drug-related experiences, opening a window into the neurobiology underlying drug-biased processing and cue-induced craving as they unfold in the real world. As one iOUD referred to the movie stimulus “it gives me that feeling like even though I'm not getting high, I just, I feel like I'm doing it with them.”

Fig. 1. LNAc responds almost exclusively to drug use in OUD and non-drug scenes in HC

The mean LNAc-specific component for each group is displayed with filled circles indicating significantly synchronized time points for each group (left: OUD; right: HC). Movie clips 5-10 s prior to each significant time point were labeled by independent raters as drug (red) or non-drug (blue). Representative still images are displayed for the four most significant peaks (top) and valleys (bottom) of the LNAc-specific component. While OUD responded almost exclusively to drug use, HC peaks corresponded to other typically motivating stimuli (e.g., food or social cues) and HC valleys to typically aversive stimuli (e.g., pain, embarrassment, feces). A 3 dimensional rendering of the LNAc mask (Harvard-Oxford Atlas) is displayed in MNI space.

Fig. 2. LNAc reactivity correlates with cue-induced heroin craving in OUD

A. LNAc reactivity correlated with mean scene-induced craving ratings. These craving ratings were collected outside the scanner in response to 3 sec video clips from the movie (averaged over all clips). B. LNAc reactivity correlated with immediately post-minus-pre movie subjective ratings for heroin craving. C. LNac reactivity did not correlate with baseline craving as measured by the heroin craving questionnaire (HCQ). For all plots, outliers with a ±3 SD from the mean threshold criteria were removed. Line and shaded area represent least squares linear fit with 95% CI. All reported r and p values are based on Pearson correlation.

Summary

Opioid-related overdose deaths in the United States have escalated significantly, including a substantial increase in 2020. Despite this crisis, the specific brain mechanisms underlying human opioid use disorder (OUD) remain largely understudied compared to other substance use disorders. Prior research on other drug classes indicates that drug-related cues can divert attention from alternative motivators, suggesting an imbalance in motivational processing that drives addiction. The extent to which such a drug-biased neural process occurs in opioid addiction, particularly in real-world contexts and its relationship to craving, has been an open question.

To investigate the brain mechanisms of cue reactivity and craving in OUD, this study employed a novel approach, utilizing a dynamic, narrative-based movie stimulus rather than static images, to more accurately represent real-world experiences. Functional magnetic resonance imaging (fMRI) BOLD responses were analyzed using inter-subject synchronization to identify shared brain signatures in individuals with OUD (iOUD). The primary region of interest was the Nucleus Accumbens (NAc), a key area involved in motivational salience, reward anticipation, and craving. The investigation hypothesized that iOUD would exhibit a shared, synchronized brain signature in the NAc predictive of cue-induced craving. Twenty-nine treatment-seeking, medically stabilized iOUD and sixteen healthy control participants watched scenes from the movie "Trainspotting" that included explicit heroin use, as well as non-drug related social and food scenes. The BOLD signal within each voxel during movie watching was decomposed into a global component, reflecting broad synchronized activity, and a selective component, representing voxel-specific responses. This selective component was then analyzed to track neural signatures related to drug-related motivation.

Applying this analytical method to the left and right NAc selective components, specific time points within the movie elicited synchronized NAc responses in each group. When clips preceding these synchronized moments were identified, the Left Nucleus Accumbens (LNAc)-specific "movie" for the OUD group contained almost exclusively scenes of drug use. Conversely, the LNAc-specific "movie" for the healthy control group contained no drug-use scenes. Statistical analysis confirmed a significantly different scene type distribution between iOUD and healthy controls for the LNAc, with iOUD responses overwhelmingly linked to drug scenes, while control responses were more randomly distributed or linked to non-drug stimuli.

Given the observed bias of the LNAc towards drug stimuli in iOUD and its established role in craving, the study further examined whether the magnitude of this LNAc-selective component predicted cue-induced craving for drugs. The LNAc-specific component, averaged over significant time points for each participant, was found to predict both scene-specific and overall movie-induced heroin craving. However, no correlation was observed with baseline craving levels, suggesting that the LNAc's synchronous reactivity dynamically scales with the craving experienced during movie viewing. Further analysis revealed that the LNAc reactivity signal was not significantly correlated with demographic factors or measures of drug use history or dependence severity, indicating its independence from these common confounding variables. While the left anterior insula also showed some drug-biased reactivity in OUD, only the LNAc demonstrated a complete reversal of scene-type preferences between groups and a correlation with cue-induced craving, underscoring the specificity of these findings to the LNAc.

These findings support prominent neurobiological theories of addiction, which posit an aberrant attribution of incentive salience to drug-related stimuli at the expense of other typically motivating stimuli. The documented bias towards naturalistic depictions of drug use in the LNAc of iOUD suggests that such neural bias may dominate during daily real-life experiences, even in currently abstinent, treatment-seeking individuals. This interpretable brain signal, predictive of cue-induced craving, may also serve as a powerful predictor of future drug use and relapse, a hypothesis requiring longitudinal follow-up studies. This study represents a novel application of dynamic naturalistic stimuli to demonstrate drug-biased processing as a predictor of cue-induced craving in the addicted human brain, offering insights into the neurobiology of drug-biased processing and cue-induced craving as they unfold in a real-world context.

Summary

Opioid overdose deaths in the United States increased significantly between 2015 and 2020, with a notable rise in heroin use during the COVID-19 pandemic. Despite the seriousness of this drug use and overdose crisis, the specific brain processes involved in human opioid use disorder (OUD) have received less study compared to other addictions. In other types of addiction, drug-related cues are known to draw attention away from normal rewards like social interaction or food. This study investigated whether a similar shift in brain processing occurs in heroin addiction, particularly how drug-related brain activity in real-world situations connects to craving.

To better understand brain activity related to drug cues and craving, this research departed from typical neuroimaging studies that use static images. Instead, researchers used a dynamic, story-based movie about heroin use to more closely mimic real-life experiences for individuals with OUD. The study analyzed how brain activity, measured with fMRI, synchronized across participants while they watched the movie. The primary area of interest was the Nucleus Accumbens (NAc), a brain region crucial for processing motivation, reward, and craving. The expectation was to observe unique synchronized NAc activity linked to drug content, predicting cue-induced craving in those with OUD. Participants included 29 individuals undergoing treatment for OUD who were medically stable, and 16 healthy individuals matched for age, sex, and race. All participants watched a 17-minute segment of the movie "Trainspotting," which depicted both explicit heroin use and other compelling non-drug scenes, like those involving food and social interactions. The movie's overall story focused on the challenges of drug addiction.

To understand which movie features caused synchronized NAc activity, a specific analytical approach was used, which involved separating the overall brain activity signal into two parts. The first, a "global component," represented general brain synchronization driven by surprising or emotionally engaging moments in the movie. The second, a "selective component," isolated brain activity specific to individual regions, after removing the general brain response. This allowed researchers to pinpoint how the Nucleus Accumbens specifically responded to particular movie content. The goal was to see if the NAc in individuals with OUD would show more synchronized activity for drug-related content, even when other motivating scenes were present.

When this analysis was applied to the left Nucleus Accumbens (LNAc), specific moments in the movie were found to cause synchronized LNAc responses in each group. By identifying the movie clips that triggered this synchronization, researchers created LNAc-specific "movies" for both groups. Notably, the LNAc-specific "movie" for the OUD group almost exclusively contained scenes of drug use, whereas the control group's LNAc activity was not primarily driven by drug scenes. Statistical tests confirmed that the distribution of scene types driving LNAc activity differed significantly between the OUD and healthy control groups, with the OUD group showing a strong preference for drug scenes. Furthermore, the magnitude of LNAc activity also differed significantly between the groups during these specific movie moments. The study also explored whether the LNAc's responsiveness to drug stimuli predicted craving. It was found that the level of LNAc activity did predict both craving experienced during the movie and immediately after, but it did not correlate with a person's general baseline craving levels. This indicates that LNAc activity reflects the dynamic, moment-to-moment experience of craving triggered by the movie.

These findings support theories of addiction that suggest drug-related stimuli gain excessive importance in the brains of individuals with OUD, overshadowing other typical motivations. Observing this brain bias in response to a realistic movie suggests that such a neural preference may be active during daily life. Importantly, this brain response was observed even in individuals who were currently abstinent, seeking treatment, and on medication. This interpretable brain signal, which predicts craving, could potentially also predict future drug use and relapse, though this requires further long-term studies. Future research should consider if these findings apply to other substance use disorders or individuals at risk, and might benefit from more diverse participant groups and broader brain analyses. This study represents a novel use of dynamic, real-world stimuli to demonstrate how the brain processes drug-related information and how this processing predicts craving in individuals with addiction, offering a clearer view into the neurobiology of craving as it unfolds in daily life. As one participant with OUD commented about the movie, it provided "that feeling like even though I'm not getting high, I just, I feel like I'm doing it with them."

Summary

Deaths from opioid overdoses in the United States rose sharply, increasing by 109% between 2015 and 2020. This included a 38% rise in 2020 alone. Heroin use also increased by about 40% during the COVID-19 pandemic. Despite this serious drug use and overdose crisis, the brain changes linked to human opioid use disorder (OUD) have not been studied as much as those related to other drug addictions. For other drug classes, such as stimulants, drug cues are known to shift a person's attention away from other important things like social activities, food, or sex. This imbalance in what a person finds rewarding is thought to drive drug use and keep the cycle of addiction going. This study aimed to explore if a similar shift in brain processing happens in heroin addiction and how it relates to drug cravings triggered by cues in real-life settings.

To study how the brain reacts to drug cues and craving, most brain imaging studies use still pictures. However, this research used an engaging movie, "Trainspotting," which is dynamic and rich in context, to better reflect real-world experiences for individuals with OUD (iOUD). Participants watched the first 17 minutes of the movie, which included explicit scenes of heroin use, as well as scenes involving food and social interactions not directly related to heroin. Researchers specifically focused on the Nucleus Accumbens (NAc), a brain area crucial for motivation, reward, and craving. They expected to find a unique, synchronized brain response in the NAc of iOUD when watching drug-related content, which would predict craving. The study included 29 iOUD patients who were seeking treatment and medically stable, along with 16 healthy individuals matched for age, sex, and race.

Researchers analyzed the brain activity patterns recorded during the movie by breaking down the signals into two parts. One part represented general brain activity that occurs when watching any engaging story. The other part, called the selective component, showed brain activity specific to a certain area, not explained by the general activity. By using a special analysis method, the research team identified specific moments in the movie that caused synchronized activity in the NAc for each group. They then looked at movie clips from immediately before these moments to see what kind of content triggered this activity.

A striking finding was that the left NAc (LNAc) in the OUD group responded almost entirely to scenes of drug use, but not to drug use scenes in the healthy control group. For the healthy group, their LNAc activity was more related to other motivating things like food or social cues. Statistical tests confirmed that the types of scenes that activated the LNAc were significantly different between the OUD and healthy groups, with OUD showing a strong bias towards drug scenes. Furthermore, the strength of the LNAc's response to these specific scenes in individuals with OUD predicted how much craving they reported both during and after watching the movie. The LNAc's response was specific to drug cues and was not linked to other factors like demographics, recent drug use, or withdrawal symptoms.

These findings support theories that suggest addiction involves an abnormal focus on drug-related cues, ignoring other typically rewarding experiences. By observing this imbalance while people watched a movie, the study showed that such a brain bias might be present in daily life. This focused brain response was seen even in individuals with OUD who were currently abstinent, seeking treatment, and taking medication. Since this brain signal predicted craving, it could potentially help predict future drug use and relapse. Future research will explore if these findings apply to other substance use disorders, individuals at risk, and whether there are sex differences. This study represents a significant step forward as the first to use dynamic, realistic stimuli to show how drug-biased brain processing predicts cue-induced craving in humans with addiction.

Summary

Deaths from opioid drugs have gone up a lot in the United States, with a big jump in 2020. Heroin use also increased around the time of the COVID-19 pandemic. Even though this is a serious problem, scientists have not studied how opioid addiction changes the brain as much as they have studied other drug problems. For other drugs, like stimulants, seeing drug-related things can make the brain focus only on drugs, ignoring other important things like friends, food, or sex. This study wanted to see if the same thing happens in the brains of people with heroin addiction, causing them to crave drugs.

To study the brain's reaction to drug-related things and the craving it causes, most brain studies use still pictures of drugs. But this study used a movie about heroin use called "Trainspotting." A movie is more like real life because it has a story and lots of details. Scientists used a special brain scan called fMRI to look at how the brains of people with opioid addiction reacted to the movie. They especially looked at a part of the brain called the Nucleus Accumbens (NAc), which is known for controlling what the brain finds important and for causing cravings.

The study looked at 29 people with opioid addiction who were getting treatment and 16 healthy people. While watching the movie, a part of the brain called the left NAc reacted very strongly in people with opioid addiction. For these individuals, this brain area mostly reacted when drug use was shown in the movie. But for the healthy people, the same brain area reacted to other things in the movie, like food or social scenes. This showed a clear difference in what each group's brain paid attention to.

This brain activity in the left NAc was also linked to how much people with opioid addiction craved heroin after watching the movie. This was true for cravings reported right after the movie and for specific cravings related to different scenes. This suggests that the brain's reaction to drug-related sights directly relates to a person's craving. The study found this brain difference even in people who were not using drugs at the time and were getting medical help for their addiction.

This study shows that the brains of people with opioid addiction strongly focus on drug-related things, even when they are trying to get better. This brain signal could help doctors understand who might be at higher risk for using drugs again. More studies are needed to see if these findings apply to other drug problems or to different groups of people. This new way of using movies to study the brain helps scientists learn more about how drug cravings happen in real-life situations.