Alcohol dependence (AD) is an international public health concern (1). Globally, in 2016, 2.6% of people over 15 years of age had AD (2). Although there are many treatments for AD, including pharmacological, behavioral, and psychosocial interventions, many AD patients still experience relapse and poor quality of life because of the chronic character of this disorder, highlighting a critical need to conduct additional research to expand treatment approaches (3).

The crucial question regarding AD treatment is why patients relapse after detoxification despite a strong intention to stay abstinent. Environmental cues are major factors thought to influence behavior and modulate the risk of relapse. Hence, one increasingly dominant view in the addiction literature conceptualizes the disease as one of aberrant learning and memory (4–6). Maladaptive motivational memory (MMM) associations link conditioned stimuli (CS; e.g., people, places, or bottles) with the intoxicating and rewarding effects of unconditional stimuli (US; e.g., alcohol) (7, 8). US and CSs contribute significantly to craving and subsequent alcohol consumption behaviors (9–11). Once formed, these MMMs are difficult to resolve because they are highly robust and resistant to destabilization.

Based on the classical conditioning theory, cue exposure therapy (CET) was designed to induce memory extinction through repeated exposure to CS without alcohol consumption to decrease conditioned responses (12). Several studies have reported positive effects of CET on reduced alcohol consumption (13), reduced craving (13), prolonged abstinence, and decreased cue reactivity (14). However, in a recent meta-analysis on CET targeting alcohol use disorder (AUD), CET showed no effect on drinking-related outcomes, and CET had a moderate effect on relapse (15). There are several possible reasons for such discrepant results. Notably, memory extinction is a largely passive procedure involving exposure to environmental cues, worsening negative responses by making the subjects recall painful experiences.

A single night of aversive olfactory conditioning during non-rapid eye movement sleep (NREM) sleep stage 2 (N2) reduced cigarette consumption in smokers (16). This suggests that sleep might interfere with substance-related MMMs. In seminal works, target memory reactivation (TMR) during sleep, similar to CET during wakefulness, reduced fear responses through memory extinction and augmented the efficiency of CET (17, 18). These findings support the hypothesis that new learning during N2 sleep or slow-wave sleep (SWS) can modify memory and alter later behaviors. The TMR paradigm poses a challenge to traditional CET, as sleep induces an anesthesia-like state during which a patient’s bad memory can be eliminated to optimize cognitive and behavioral therapy.

In a recent meta-analysis, the researchers found that TMR was only significant during the two NREM stages: N2 and SWS (19). Research about odor re-exposure during sleep found increased left-lateralized frontal slow spindle (11.0–13.0 Hz) and right-lateralized parietal fast spindle (13.0–15.0 Hz) activity, suggesting the possibility of a successful re-presentation of therapy-related memories during sleep (20). Specifically, during NREM sleep featuring cortical slow oscillations (SO) and thalamocortical spindles, covert memory reactivation can transform newly acquired, hippocampus-dependent learning such that neocortical representations become more stable and resistant to disruption (19). Furthermore, recent studies emphasized the importance of stage 2, spindles and the SO-spindle coupling in procedural memory consolidation (21). NREM sleep might be an excellent time window to enhance the effect of TMR. Notably, the patients with alcohol dependence have reported decreased SWS and increased latency of N2 sleep. A study indicated that N2 sleep during withdrawal accounts for 30% of total sleep time in alcohol-dependent patients, while SWS accounts for only 6.7% of total sleep time (22). Unlike healthy individuals, the reduction in sleep time might result in SWS sleep exposure alone not being able to provide stimulus exposure of sufficient duration.

Therefore, to elucidate whether repeated exposure to cues during sleep can be applied in a real clinical setting to reduce alcohol craving or consumption, we conducted a controlled study to determine the effects of re-exposure to an olfactory alcohol cue during NREM sleep in subjects with AD. Therefore, we chose to deliver odor cues at the N2 stabilization session and to continue the presentation when shifting to the N3 phase to ensure 20 min of cue exposure for extinction (23), in line with previous studies (20, 24). We hypothesized that repeated exposure to olfactory alcohol cues during NREM sleep would decrease alcohol craving and related physiological indexes in AD patients.

To our knowledge, this was the first study to explore the effect of re-exposure to olfactory alcohol cues during NREM sleep on alcohol craving and physiological responses in subjects with AD. We found that after sleep, repeated exposure to an alcohol stimulus during NREM sleep was ineffective in reducing subjective craving or physical responses, including changes in the SCR and blood pressure, in the CS group, contrary to our a priori hypothesis.

Unexpectedly, studies on memory manipulation during sleep have reported contrasting results. Because of the paucity of literature on TMR in drug users, we referred to related studies on both fear and drug memories, focusing on the effect of TMR similarly to CET. One study promoted fear extinction, in which participants underwent re-exposure to the odorant stimulus during SWS; this was more efficient than repeated exposure to cues during wakefulness (17). Another study reported re-exposure to cues during SWS reinstated fear responses in humans (24), and some animal studies also found enhanced fear responses (40, 41). Given the replication crisis in the fear memory field, recent papers have cautioned about the lack of standardization of procedures, and it is possible that the significant findings within the field of psychology are exaggerated (42). In previous human fear memory studies, fear conditioning training was conducted in healthy subjects (23), whereas our study took advantage of pre-existing addiction memory. We recruited alcohol-dependent patients in an attempt to explore the effects of TMR on maladaptive memories. However, unlike a learned fear memory, maladaptive drug-related memories are learned over hundreds of thousands of trials in multiple contexts and therefore likely to be highly resistant to destabilization and extinction (43). Moreover, memory extinction just creates a novel association between only special CS and US to compete with maladaptive alcohol memories. As a largely passive procedure, extinction training lacks reinforcement. Counterconditioning, which re-pairs reward cues (e.g., pictures of beer) with negative outcomes (e.g., disgust-inducing bitter liquids and odors), may provide a more powerful corrective learning experience than extinction (44). In seminal work, researchers found that olfactory aversive conditioning during N2 sleep resulted in a greater reduction in smoking behavior that lasted for a longer time than the reduction following similar conditioning in REM sleep (16). A recent study highlighted greater long-term reductions in alcohol consumption when counterconditioning was conducted following retrieval (4).

In addition, it is worth considering whether some aspect of the design might reduce the efficacy of TMR. The length of exposure is likely to influence the efficacy of extinction training. During this NREM sleep period, we delivered odor cues to participants for 20 min, resulting in each cue being presented multiple times (20 presentations). Previous research using tone cues suggested that the number of sound replays is not important for benefits of TMR to emerge (36). Furthermore, TMR benefits have been observed following the presentation of each cue only once during the sleep period (45). Some researchers have also pointed out that it is possible that amassed presentation of the cues within 3–40 min leads to extinction, whereas presenting single cues over a longer time period of 2–4 h strengthens the underlying memory trace (23).

Another possibility is the lack of sensitivity of acute laboratory assessment of the reinforcing effects of alcohol since self-reported craving might not truly reflect the urge to drink. Alcohol-related olfactory cues may be powerful appetitive cues, eliciting a craving response (34). In test 1, alcohol cues elicited strong physical reactivity but not a subjective craving increase in both groups (Supplementary Table 1). Participants who were re-exposed to alcohol cues or water during sleep all showed a trend toward a stronger craving for alcohol in test 2, but the difference was not significant in either group. It has been difficult to assess craving in clinical settings, and there have been inconsistent conclusions regarding the association of craving with relapse (46). We used the VAS to assess the desire for alcohol because of its ease of use in facilitating multiple evaluations of craving. VASs are easy-to-use measurement tools, but they are not as accurate as multi-item questionnaires. In comparison to multi-item questionnaires, subjects report less nuanced ratings on a unidimensional scale. Furthermore, given that craving on VAS was rated during hospitalization, when subjects were undergoing psychosocial and pharmacological treatment that may have interfered with the assessment of craving, lower VAS scores seem possible. However, to exclude the limitations of evaluating subjective craving via the VAS, we assessed the SCR and cardiovascular measures. Throughout NREM sleep, the SCR and blood pressure responses to cues both decreased over time in each group. The SCR and blood pressure showed strong generic habituation over repeated stimulus presentation. It is plausible that we found no effect of memory manipulation due to the lack of accurate measurement indexes.

Importantly, some researchers have argued that the role of individual differences has been neglected in behavior studies (10). In the context of individual differences, whether each subject can perceive cues, beyond strengthening existing associated memory, and forming new memory (i.e., memory extinction) require to expand studies. On a neural basis, hippocampal reactivation is suggested as the underlying mechanism behind consolidation during sleep. Nevertheless, little is known about the precise time course and boundary conditions of fading of hippocampal representations (47). Accordingly, the lag between learning and the posttest is also a factor that may explain the inconsistent results. To date, parametric studies on memory reactivation are lacking, and it is unclear which parameters are optimal.

There were several limitations to our study. First, the sample was relatively small, and patients were enrolled during hospitalization. Large sample size and randomized controlled studies to investigate the effect of paradigm during sleep on AD are needed in the future. The study lacked the power to detect potential effects of re-exposure to alcohol cues during sleep. Second, in our paradigm, we did not probe the role of other sleep stages in memory (47). REM sleep is known to contribute to memory consolidation (48). In both rodent and human studies, it has been reported that REM sleep deprivation was associated with fear memory recall (49–51). Even so, a recent meta-analysis of TMR provided substantial support for the notion that TMR was not effective during REM sleep or wakefulness (19). Also, we did not explore the respective effects of N2 and SWS sleep periods on MMMs. Stimulation with learning-associated cues, mainly during SWS, has been shown to enhance declarative as well as procedural and emotional types of memory (17, 52, 53). Although the Meta-analysis reported TMR during SWS benefits on memory (19), recent work also suggested no effect of TMR on the recognition of emotionally negative or neutral memories (54). Due to the importance of stage 2, spindles and the SO-spindle coupling in memory consolidation (21), the relationship and different roles of NREM and SWS in the TMR are still unclear and need to be explored in future studies. Cue exposure might activate the addiction memory trace, leading to reduced cue response test result in the same context when tested immediately after sleep. Moreover, sleep itself might restabilize the trace during the short term. Additionally, we did not perform a similar procedure during the daytime to investigate the effect of CS exposure during wakefulness, which would determine whether the paradigm we used could prevent craving in patients with AD. Third, alcohol odor delivery was hand controlled, but the use of a respirator induced a delay in the experimenter’s reaction time. Although many studies on odor cueing used a computer-controlled olfactory meter to deliver odor, it was challenging to persuade clinicians to use the machine in real clinical settings. These limitations could be overcome in the future. Finally, our sample comprised only men. Therefore, the generalization of our conclusions to female patients with AD is limited. As previous studies have shown that cue-induced alcohol craving and physiological cue reactivity are associated with alcohol relapse risk and that well-tolerated re-exposure to cues during sleep could facilitate extinction memory, the hypothesis in the current study requires further testing to evaluate effects on alcohol craving and improve treatment outcomes.

To our knowledge, the present study is the first to examine the effect of CS re-exposure during NREM sleep on addiction memory in individuals with AD. We observed no impact of re-exposure to alcohol cues during NREM sleep on cue-induced craving and physiological response. The replication crisis in psychology highlights the need for further examination of factors that may moderate memory extinction manipulation effects and the lack of standardized procedures.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of Peking University Sixth Hospital, Beijing, China. The patients/participants provided their written informed consent to participate in this study.

RZ, ZN, RT, YX, YM, WS, JD, and HS contributed to the study conception and design. RZ, SZ, JC, and LP evaluated all the patients, performed the data collection, and conducted the study. YZ and WS conducted the sleep data scoring. RZ and JD performed the data analyses. RZ drafted the manuscript. LL and HS critically revised the important intellectual content in the manuscript. JD and HS were in charge of supervision. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.