Opioid Craving Does Not Incubate Over Time in Inpatient or Outpatient Treatment Studies: Is the Preclinical Incubation of Craving Model Lost in Translation?

Abstract

Within addiction science, incubation of craving is an operational label used to describe time-dependent increases in drug seeking during periods of drug deprivation. The purpose of this systematic review was to describe the preclinical literature on incubation of craving and the clinical literature on craving measured over extended periods of abstinence to document this translational homology and factors impacting correspondence. Across the 44 preclinical studies that met inclusion criteria, 31 reported evidence of greater lever pressing, nose pokes, spout licks, or time spent in drug-paired compartments (i.e., drug seeking) relative to neutral compartments after longer periods of abstinence relative to shorter periods of abstinence, labelled as “incubation of craving.” In contrast, no clinical studies (n = 20) identified an increase in opioid craving during longer abstinence periods. The lack of clinical evidence for increases in craving in clinical populations weakens the translational utility of operationalizing the time-dependent increase in drug-seeking behavior observed in preclinical models as models of incubation of “craving”.

1. Introduction

Incubation of craving is a high-profile concept in the field of addiction at both preclinical and clinical levels of analysis which describes time-dependent increases in drug seeking or craving during periods of drug deprivation (Grimm et al., 2001; Shalev et al., 2001). This pattern of behavior was originally described by Gawin and Kleber based on interviews of 30 adults in treatment for cocaine use (Gawin and Kleber, 1986). The report suggested there is a protracted clinical phase that occurs after withdrawal symptoms have subsided, which is characterized by reports of “extreme craving” (Gawin and Kleber, 1986). In recent years, these time-dependent increases in drug seeking behaviors, referred to as incubation of craving, have been documented most often and studied most extensively in preclinical animal models (Lu et al., 2004; Pickens et al., 2011). As it stands, it is unclear whether self-reported craving in clinical research and the phenomenon observed in the preclinical incubation of craving model represent similar or distinct constructs.

The defining feature of incubation of craving applied in addiction science models is a time-dependent increase in drug-seeking behavior. Measurement of craving is operationally different in preclinical models, which measure solely drug seeking responses, and in clinical research, which largely measures subjective reports of craving based on interoceptive states. Preclinically, the occurrence of time-dependent increases in drug seeking was first identified with cocaine (Tran-Nguyen et al., 1998). Since these early studies were published, hundreds of preclinical studies have followed seeking to identify molecular and neurocircuit underpinnings of this putative phenomenon across different drugs of dependence, including opioids (Shalev et al., 2001). In the operant version of this model (although conditioned place preference paradigms may also be used (Sun et al., 2018)), non-human animals self-administer the drug under study for a period of time in an operant chamber in which responses on an operandum (e.g., lever , nose poke) result in the contingent delivery of the drug (usually intravenously) and the co-presentation of previously neutral cues (e.g., light, noise). Self-administration is typically followed by varied durations of abstinence (e.g., 1, 15, 30 days) in which the subject is no longer exposed to the operant context. Following this period of abstinence, subjects are again exposed to the operant context, where responses on the operandum (e.g., lever press, nose poke) produce the conditioned cues previously paired with the drug (e.g., light, noise), but not delivery of the drug itself. The operant behavior in this final phase – “cued extinction test” – is then quantified; increased lever pressing as a function of increasing periods of abstinence is considered evidence of incubated craving in these non-human animal subjects.

In clinical research, craving is typically conceptualized as an intense desire or urge for a drug. It is a complex and dynamic symptom experienced by individuals with substance use disorders and is broadly cited as a source of distress, a barrier to treatment retention, and a reason why some return to use (Bergeria et al., 2020; Biernacki et al., 2022; Kakko et al., 2019). The relevance of craving during periods of active drug use and during periods of drug deprivation is not fully understood in clinical addiction research and remains debated (Kleykamp et al., 2019b). Some have argued that craving is not a central feature of active drug addiction, nor does it compel use (Pickard, 2020; Smith, 2022) while clinical nosology has placed these symptoms within defining criteria for substance use disorder (e.g., DSM-5 criteria). Although craving can be predictive of future non-prescribed drug use (Vafaie and Kober, 2022), instances of self-reported craving do not deterministically lead to drug use or related behaviors (Epstein et al., 2009; Kowalczyk et al., 2018; Kowalczyk et al., 2015; Panlilio et al., 2019). Further complicating our understanding of how craving functions, craving is commonly characterized as a multidimensional construct (Kampman et al., 1998; Kleykamp et al., 2019a; Kleykamp et al., 2019b; Toll et al., 2006). It is possible that unique dimensions differentially predict future instances of drug use and related behaviors, however these unique relationships have not yet been fully realized (Saraiya et al., 2021).

Although craving can be assessed in many ways in clinical research (Kleykamp et al., 2019a), the method most concordant with the incubation of craving preclinical model are cue reactivity models (or cue-induced craving models). Cue reactivity models are used to investigate craving responses for varied drug classes and are most often used in human laboratory studies. This model involves exposing participants to common drug-related cues (e.g., syringes, white powder, lighters, spoons) and measuring self-report and physiological outcomes. Cue reactivity models are premised on principles of Pavlovian conditioning. That is, when neutral cues are reliably paired with drugs that produce an unconditioned response (e.g., euphoria, nausea), those initially neutral cues can become conditioned cues which signal to an individual or organism that drug effects are likely to onset. Conditioned cues elicit conditioned or preparatory responses, which can vary in nature and valence, can be specific to a person’s drug and treatment history, and can include drug craving, among other symptoms. Craving can occur after exposure to drug paraphernalia – drug-conditioned cues – in laboratory settings, but not always. Internal states can also serve as conditioned cues (e.g., withdrawal, stress, effects of another drug, emotion, arousal) that precede craving. Drug imagery cue reactivity models represent one mechanistically-unique way in which craving may be elicited and can be a valuable tool for understanding the interactions between individual-level traits (e.g., substance use history), internal states (e.g., stress), and environmental cues (e.g., images of drugs) on craving responses in controlled settings (Back et al., 2014; Back et al., 2015; Back et al., 2010; Kowalczyk et al., 2015; Sinha et al., 2007).

Cue reactivity models can examine possible incubation of craving effects among human subjects (Bedi et al., 2011; Parvaz et al., 2016; Wang et al., 2013; Zhao et al., 2021). However, it has historically been difficult to establish translatability of the incubation of craving effects produced in preclinical research to clinical studies. This is in part because of the challenges associated with sustaining abstinence in humans for a long enough period of time to sufficiently evaluate cue reactivity craving over a protracted time period. To understand the extent to which the incubation of craving effect is preserved in clinical research we conducted a systematic review of cue reactivity findings with opioids. Specifically, we pooled results across clinical studies that investigated populations with varying lengths of opioid abstinence (between group comparisons) and who were assessed repeatedly across different temporal windows (within group comparisons). We present this systematic review of clinical research alongside a systematic review of preclinical studies that tested and characterized incubation of opioid craving. Doing so allows direct comparison of patterns of these two effects. We chose to focus the review on opioid cue reactivity craving as an examplar drug craving for two reasons. First, the persistent opioid crisis is defined, in part, by poor treatment retention outcomes and high rates of fatalities; scientifically and clinically addressing the opioid crisis can and should be informed by a better understanding of craving. Second, the greater likelihood of sustained recovery (vs. stimulant use which has no FDA-approved pharmacotherapy) more readily lends itself to the study of craving over time in people experiencing opioid use disorder (OUD).

2. Methods

The goal of the review is to characterize and compare craving patterns over time in (1) preclinical incubation of craving research and (2) clinical cue reactivity research. Reviewed papers were identified using PubMed between July 2023 and September 2023. For the preclinical literature search, PubMed was searched using the following terms: (incubation of craving OR cue-induced OR cue induced OR cue-reactivity OR cue reactivity OR cue-induced craving OR cue induced craving) AND (heroin OR opioids OR fentanyl OR oxycodone OR opiates) AND (rat OR rodent OR mouse). For the clinical literature search, PubMed was searched using the following terms: (cue-induced OR cue induced cue-reactivity OR cue reactivity OR cue-induced craving OR cue induced craving) AND (heroin OR opioids OR fentanyl OR oxycodone OR opiates).

Two independent reviewers (CB and JS) evaluated search results independently to determine if the studies met criteria for the systematic review. To be included in the preclinical systematic review, studies must have met the following criteria: (1) preclinical rodent research, (2) included a model of opioid reinforcement or reward, (3) included an opioid-paired cue-induced behavior tests without opioid delivery (i.e., cued extinction test) and (4) reported corresponding behavioral results following at least two periods of abstinence with varying lengths. Studies could be between- or within-subject. To be included in the clinical research review, studies must have met the following criteria: (1) human subjects research, (2) reported craving results following exposure to opioid-related cues across periods of abstinence either between or within subjects, (3) included individuals with OUD or a history of OUD, and (4) included results from individuals who were abstinent from non-prescribed opioid use. Finally, if the clincial study comprised an intervention, included studies had to describe a treatment as usual, placebo, or sham condition so that self-reported craving in a cue reactivity paradigm could be characterized without manipulation.

After PubMed entries were independently reviewed, lists of eligible studies compiled by each reviewer were compared and discrepancies were discussed until inclusion or exclusion of the studies were agreed upon. The preclinical review yielded 336 PubMed results, of which 34 met criteria for inclusion and were ultimately reviewed. The clinical research search yielded 554 PubMed results, of which 19 met criteria for inclusion and were ultimately reviewed.

3. Results

3.1. Preclinical Research Systematic Review

3.1.1. Study Parameters

Forty studies published between 2001 and 2023 met inclusion criteria for the systematic review and are detailed in Table 1, 2, and 3. Thirty-three studies examined incubation of opioid craving using drug self administration and three studies examined incubation of opioid craving using conditioned place preference (CPP). The majority of studies used heroin (N =20), followed by morphine (n = 9), oxycodone (n = 7), fentanyl (n = 1), and tramadol and tapentadol (n = 1 study with both drugs) to evaluate opioid craving. Forced abstinence was used as the only method for abstinence in most studies (n = 29). Seven studies evaluated a voluntary form of abstinence either utilizing an electric-barrier to reduce lever pressing (n = 4) or implementing a discrete choice task between lever pressing for drug and pressing a lever for a highly palatable food pellet (n = 3). Across all studies, shorter abstinence periods ranged from 1–14 days; longer abstinence periods ranged from 8–66 days.

Table 1.

Preclinical Research Studies Assessing Drug-cued Operant Behavior After Various Lengths of ‘Forced Abstinence’

	Study Parameters					Change in Outcome Over Abstinence Periods
	General Study Aim	Animal Model	Study Drug	Incubation Test Days	Behavior Proxy for Craving
Shalev, Morales, Hope, Yap, and Shaham (2001)	Testing behavioral manipulation	Male Long-Evans	Heroin	1, 6, 12, 25, 66	Active lever press	Increase in lever pressing after 6, 12, and 25 days of abstinence compared to 1 day of abstinence. No difference in lever pressing after 66 days of abstinence compared to 1 day of abstinence.
Kuntz, Patel, Grigson, Freeman, and Vrana (2008)	Neurobiological characterization	Male Sprague-Dawley	Heroin	1, 14	Active spout licks	Increase
Zhou, Zhang, Liu, et al. (2009)	Testing behavioral manipulation	Male Sprague-Dawley	Heroin	1, 14	Active nose poke	Increase
Airavaara, Pickens, Stern, et al. (2011)	Neurobiological characterization	Male Long-Evans	Heroin	1, 11, 30	Active lever press	Increase
Theberge, Pickens, Goldart, et al. (2012)	Neurobiological characterization	Male Sprague-Dawley	Heroin	1, 11, 30	Active lever press	Increase
Fanous, Goldart, Theberge, Bossert, Shaham, and Hope (2012)	Neurobiological characterization	Male Sprague-Dawley	Heroin	1, 13–15	Active lever press	Increase
Theberge, Kambhampati, et al. (2013)	Testing pharmacological manipulation	Male Sprague-Dawley	Heroin	1, 13	Active lever press	Increase
Fanous, Guez-Barber, Goldart, et al. (2013)	Neurobiological characterization	Male Sprague-Dawley	Heroin	14, 30	Active lever press	No change from 14 to 30 days, data from day 1 not presented
Sun, Zhuang, Zhu, et al. (2015)	Neurobiological characterization	Male Sprague-Dawley	Heroin	1, 14	Active lever press	Increase
Venniro, Zhang, Shaham, and Caprioli (2017) *	Testing behavioral manipulation	Male and female Sprague-Dawley	Heroin	1, 21	Active lever press	Increase
Blackwood, Leary, Salisbury et a. (2019)	Neurobiological characterization	Male Sprague-Dawley	Oxycodone	5, 31	Active lever press	No change with short access self-administration Increase with long-access self-administration
Venniro, Russell, Zhang, and Shaham (2019)	Testing behavioral intervention	Male and female Sprague-Dawley	Heroin	1, 15	Active lever press	Increase
Fredriksson, Applebey, Minier-Toribio et al. (2020)*	Neurobiological characterization	Male and female Sprague-Dawley	Oxycodone	1, 15, 30	Active lever press	Increase
Gyawali, Martin, Sulima, Rice, and Calu (2020)	Neurobiological characterization	Male Sprague-Dawley	Fentanyl	1, 30	Active lever press	Decrease
Altshuler, Yang, Garcia, et al. (2021)	Neurobiological characterization	Male Sprague-Dawley	Oxycodone	1, 15	Active lever press	Increase
Barrera, Loughlin, Greenberger, Ewing, Hachimine, and Ranaldi (2021)	Testing behavioral manipulation	Male and female Long Evans	Heroin	3, 15	Active lever press	Increase with no environmental enrichment No change with environmental enrichment
Zhu, Zhuang, Lou, et al. (2021)	Neurobiological characterization	Male Sprague-Dawley	Heroin	1, 14	Active lever press	Increase
Salisbury, Blackwood and Cadet (2021)	Neurobiological characterization	Male Sprague-Dawley	Oxycodone	5, 30	Active lever press	Increase with long access self-administration No change with short access self-administration
Xu, Hong, Lin et al. (2021)	Neurobiological characterization	Male Sprague-Dawley	Heroin	1, 14	Active nose pokes	Increase
Zanda, Floris, and Sillivan (2021)	Testing pharmacological manipulation	Male Sprague-Dawley	Heroin	2, 21	Active lever press	Increase No change with low dose heroin during acquisition and among rats with high heroin intake (above median heroin intake during acquisition)
Mayberry, DeSalvo, Bavley, et al. (2022)	Testing behavioral manipulation	Male and female Long-Evans	Heroin	1, 30	Active lever press	Increase
Wong, Zimbelman, Milovanovic, Wolf, Stefanik. (2022)	Neurobiological characterization	Male Sprague-Dawley	Oxycodone	1, 15, 30	Active nose poke	Increase
Bossert, Townsend, Altidor, et al. (2022)	Testing pharmacological manipulation	Male and female Sprague-Dawley	Heroin	1, 8	Active lever press	Increase
Samson, Xu, Kortagere, España. (2022)	Neurobiological characterization	Male and female Long Evans	Oxycodone	1, 14	Active lever press	No change
Mayberry, Bavley, Karbalaei, et al. (2022)	Neurobiological characterization	Male and female Long- Evans	Morphine	1, 30	Active lever press	Increase
Gillespie, Mayberry, Wimmer, and Sillivan (2022)	Neurobiological characterization	Male Sprague-Dawley	Morphine	1, 30	Active lever press	Increase
Nett and LaLumiere (2023)	Testing behavioral manipulation	Male and female Sprague-Dawley	Heroin	1, 14	Active lever press	Increase
D’Ottavio, Reverte, Ragozzino, et al. (2023) *	Testing behavioral manipulation	Male and female Sprague-Dawley	Heroin (continuous and intermittent)	1, 21	Active lever press	No change in lever pressing among rats with intermittent heroin access during acquisition with forced abstinence Increased lever pressing among rats with continuous heroin access during acquisition and forced abstinence
Barrera, Timken, Lee, et al. (2023) *	Testing behavioral manipulation	Male and female Long Evans	Heroin	3, 21	Active lever press	No change in lever pressing

^*Asterisk demarks citations that included (an) experiment(s) with forced abstinence and voluntary abstinence.

Table 2.

Preclinical Research Studies Assessing Drug-cued Operant Behavior After Various Lengths of ‘Voluntary Abstinence’.

Study Parameters							Change in Outcome Over Abstinence Periods
	General Study Aim	Animal Model	Study Drug	Method for Abstinence	Incubation Test Days	Behavior Proxy for Craving
Venniro, Zhang, Shaham, and Caprioli (2017)	Testing behavioral manipulation	Male and female Sprague-Dawley	Heroin	Voluntary Abstinence-discreet choice procedure between drug and palatable food	1, 21	Active lever press	No change
Fredriksson, Applebey, Minier-Toribio et al. (2020)	Neurobiological characterization	Male and female Sprague-Dawley	Oxycodone	Electric barrier-induced abstinence	1, 15, 30	Active lever press	Increase
Reiner, Lofaro, Applebey et al. (2020)	Neurobiological characterization	Male and female Sprague-Dawley	Fentanyl	Voluntary Abstinence - discreet choice procedure between drug and food	1, 14 or 15	Active lever press	No change
Fredriksson I, Tsai PJ, Shekara A, et al. (2021)	Neurobiological characterization	Male Sprague-Dawley	Oxycodone	Electric barrier-induced voluntary abstinence	1, 15	Active lever press	Increase
D’Ottavio, Reverte, Ragozzino, et al. (2023)	Testing behavioral manipulation	Male and female Sprague-Dawley	Heroin (continuous and intermittent)	Voluntary Abstinence-discreet choice procedure between drug and palatable food	1, 21	Active lever press	No change
Barrera, Timken, Lee, et al. (2023)	Testing behavioral manipulation	Male and female Long Evans	Heroin	Environmental Enrichment Electric barrier-induced voluntary abstinence	3, 21	Active lever press Latency to active lever press in electric barrier abstinence model	No change in lever pressing Decrease in latency to lever press with electric barrier following no environmental enrichment, decrease in latency to lever press with electric barrier among female rats with environmental enrichment
Fredriksson, Tsai, Shekara, et al. (2023)	Neurobiological characterization	Male and female Sprague-Dawley	Oxycodone	Electric barrier-induced voluntary abstinence	1, 15 or 18	Active lever press	Increase

Table 3.

Preclinical Research Studies Assessing Conditioned Place Preference after Various Lengths of Abstinence

Study Parameters							Change in Outcome Over Abstinence Periods
	General Study Aim	Animal Model	Study Drug	Method for Abstinence	Incubation Test	Days Behavior Proxy for Craving
Mucha & Iversen (1984)	Basic behavioral pharmacology	Sprague-Dawley, sex not specified	Morphine	Forced Abstinence	1 day, 1 month	Time spent in morphine-paired side minus time spent in saline-paired side	No change
Vezina & Stewart (1987)	Testing behavioral manipulation	Male Wistar	Morphine	Forced Abstinence	1, 16	Time spent on morphine-paired floor textures	No change
Tzschentke & Schmidt (1995)	Testing pharmacological manipulation	Male Sprague-Dawley	Morphine	Forced Abstinence	1, 9, 18, 20, 38	Time spent in drug-paired compartment	Decrease
Wang, Luo, Zhang and Ji-Sheng (2000)	Testing behavioral manipulation	Male Sprague-Dawley	Morphine	Forced Abstinence	1, 3, 5, 7	Time spent in drug-paired compartment	Decrease
Lu, Su, Yue, et al. (2001)	Testing pharmacological manipulation	Male Sprague-Dawley	Morphine	Not specified	1, 30	Time spent in drug-paired compartment	Decrease
Li, Li, Wang, et al. (2008)	Neurobiological characterization	Male Sprague-Dawley	Morphine	Forced Abstinence	1, 7, 14	Time spent in morphine- and saline- paired compartment	Increase
Xie, Zhang, Ma, et al. (2022)	Neurobiological characterization	C57BL/6N Mice	Morphine	Forced Abstinence	1, 14	Proportion of time spent in drug paired compartment / time spent in both compartments	Increase with larger heroin dose No change with smaller heroin dose
Barbosa, Leal, Pereira, Dinis-Oliveira, and Faria (2023)	Characterizing other opioids	Male Wistar	Tapentadol, Tramadol	Forced Abstinence	1, 7, 14	CPP difference scores (proportion of time spent in drug-paired compartments after drug conditioning - before drug conditioning) Number of entries into drug-paired compartments	Tapentadol: Decrease in CPP difference scores No change in number of compartment entries Tramadol: No change in CPP difference scores No change in number of entries at day 7 Decrease in number of entries at day 14

During the cued extinction tests following abstinence, cued-behaviors that were used to infer craving intensity included: (1) lever presses (n = 28), (2) time spent in a drug-paired compartment (n = 7), (3) nose pokes (n = 3), and (4) spout licks (n = 1).

3.1.2. Changes in Cue-Induced Behaviors as a Function of Abstinence Length

Across the 44 studies (40 citations), 31 reported evidence of greater lever pressing, nose pokes, spout licks, or time spent in drug-paired compartments relative to neutral compartments or decreases in latency to lever press after longer periods of abstinence relative to shorter periods of abstinence. The majority of these studies (n = 25) utilized forced abstinence and drug-cue induced operant behavior as the evidence of craving intensity. Nineteen studies showed evidence of no change (n = 16) or decreases (n = 6) in the frequency of cue-induced behavior across periods of abstinence under certain conditions.

Four additional studies which evidenced no change in lever pressing implemented “voluntary” models of abstinence (Table 2). Notably, though, four studies that also used voluntary models of abstinence found increases in lever pressing across periods of abstinence. Discreet choice voluntary abstinence models (n = 3/3) showed no change in responding over time.

Two of the nineteen studies that did not show increases in cued extinction test outcomes across abstinence periods used an intermittent schedule of drug delivery (i.e., 6 hour days with five minute access every 30 minutes for 10–12 days) during drug self-administration acquisition (D’Ottavio et al., 2023; Samson et al., 2022). In contrast, all other self-administration studies used protocols where periods with access to drug self-administration was proportionally larger than “timeout” periods. Two separate studies evaluated the impact of short versus long access self-administration acquisition periods on cued extinction tests; short access was associated with no changes, while long access – which more closely mirrored standard self-administration acquistion paradigms used in incubation of craving studies – was associated with increases in lever responding during cued extinction tests. One study reported no changes in cue-induced lever pressing as a function of abstinence length but only provided lever pressing data after 14 and 30 days of forced abstinence (Fanous et al., 2013).

Six of the eight studies which examined CPP across varying abstinence lengths showed no change (n = 4) or decreases (n = 4) in cued behavior over time (Table 3). One such study showed no changes in CPP of a tramadol-paired compartment and showed decreases in CPP of tapentadol-paired compartment over varying periods of abstinence. This study was unique not only because of its use of tramadol and tapentadol, which are atypical opioid agonists, but also in that CPP was tested in all rats at all timepoints (1, 7, and 14 days after pairing a compartment with tramadol) (Barbosa et al., 2023).

3.2. Clinical Research Systematic Review

3.2.1. Between-subjects Comparisons of Cue Reactivity Craving Between Groups with Varying Lengths of Abstinence

3.2.1.1. Study parameters.

Nine studies (eight citations) were reviewed that assessed cue reactivity craving between groups with varying lengths of abstinence (Table 4). All studies were observational and published between 2007 and 2021. There were two ways cue-induced craving outcomes were analyzed: (1) craving change score (post-cue craving score – pre-cue craving score) and (2) craving score post-cue exposure. Four studies reported both craving outcomes, four studies reported craving change score only, and one study reported post-cue craving score only.

Table 4.

Clinical Research Studies with Comparisons of Cue-reactivity Craving Between Groups with Varying Lengths of Abstinence

Citation	Groups	Medication for Opioid Use Disorder	Treatment Setting	Sample Size(s)	Craving Outcome	Difference in Craving Between Groups with Different Abstinence Lengths
Shi, Zhao, Epstein, et al. (2007)	(1) Receiving methadone for 15–45 days (2) Receiving methadone for 5–6 months Methadone	Outpatient	(1) n = 17 (2) n = 17	Craving change score	No difference
	(1) 15–45 days post methadone detoxification (2) 5–6 months post methadone detoxification	None	Outpatient	(1) n = 18 (2) n = 18	Craving change score	No difference
Zhao, Fan, Du, et al. (2012)	(1) Recently Abstinent - less than 1 month (2) Long-term Abstinent - at least 12 months	None	Residential/Inpatient Outpatient for follow-up	(1) n = 27 (2) n = 29	Craving change score	No difference
Wang, Zhang, Zhao, et al. (2012)	(1) 1 month abstinent (2) 3 months abstinent (3) 12 months abstinent (4) 24 months abstinent	None	Residential/Inpatient	(1) n = 18 (2) n = 18 (3) n = 12 (4) n = 19	Craving change score Craving scores post-cue	No difference
Preller, Wagner, Sulzbach, et al. (2013)	(1) Recently Abstinent - 1 day (2) Long-term Abstinent - At least 1 year	None	Residential/Inpatient	(1) n = 15 (2) n = 14	Craving change score Craving scores post-cue	No difference in change score Lower craving score post-cue in long-term abstinent participants compared to recently abstinent participants
Li, Wang, Zhang, et al. (2013)	(1) Short-term Abstinent - 7–72 days (2) Long-term Abstinent - 150–300 days	None	Residential/Inpatient	(1) n = 19 (2) n = 18	Craving change score Craving scores post-cue	No difference
Wang, Wang, Li, et al. (2014)	(1) Methadone less than 1 year (2) Methadone at least 2 years	Methadone	Outpatient	(1) n = 15 (2) n = 15	Craving scores post-cue	No differences
Su, Yang, Wang, et al. (2017)	(1) Short-term Abstinent - 2–6 months (2) Long-term Abstinent - 19–24 months	None	Residential/Inpatient	(1) n = 26 (2) n = 26	Craving change score	No difference
Chen, Wang, Zhu, et al. (2021)	(1) Short-term Abstinent - 7–15 days (2) Long-term Abstinent - 6–12 months	None	Not specified	(1) n = 19 (2) n = 35	Craving change score Craving scores postcue	No difference in change score. Significantly lower craving score post-cue among individuals with long-term abstinent compared to individuals with short-term abstinent.

3.2.1.2. Participant Characteristics.

Studies included samples of individuals who were maintained on methadone (n = 2) and individuals not receiving medication for OUD (no MOUD; n = 7). All studies reported biochemical and/or circumstantial (e.g., inpatient treatment facility) proof of abstinence throughout the study. Participants were categorized as having comparatively shorter or longer abstinence lengths based on criteria that varied widely across studies: short term abstinence ranged from 1 day to less than 1 year and long term abstinence ranged from 5–6 months to at least 2 years. All studies compared cue reactivity craving between groups, assessed at a single timepoint.

3.2.1.3. Differences in Cue Reactivity Craving Between Groups with Varying Abstinence Lengths.

None of the reviewed studies reported greater cue reactivity craving among cohorts with long-term abstinence relative to cohorts with short-term abstinence. All studies reported no differences in cue reactivity craving across groups with different abstinence lengths. Two studies that analyzed craving change scores and post-cue craving scores reported significantly lower post-cue craving among individuals with relatively longer abstinence than individuals with shorter abstinence (Chen et al., 2021; Preller et al., 2013). However, those studies reported no significant differences in craving change scores across groups.

The two studies which examined cue reactivity craving among individuals receiving methadone for OUD reported no differences in craving between groups with shorter v. longer lengths of abstinence.

3.2.2. Within-subject Comparisons of Cue Reactivity Craving Over Time

3.3.2.1. Study parameters.

Eleven clinical studies published between 1993 and 2022 assessed cue reactivity craving across study timepoints (Table 5) and met inclusion criteria for the systematic review. Nine of 11 reviewed studies were randomized controlled trials (RCTs) with cue reactivity craving assessments across at least two timepoints (range: 2–4 timepoints). Timepoints were scheduled over the course of a minimum of 1 day and a maximum of 12 months. Of the nine RCTs reviewed, five tested a behavioral intervention or psychotherapy, three tested brain stimulation, and one tested a pharmacotherapy. All RCTs had a treatment as usual or control condition. Two non-RCT studies were observational reports of cue reactivity craving assessments measured within a single group of individuals at two timepoints: baseline, and +6 weeks or +6 months.

Table 5.

Clinical Research Studies with Comparisons of Cue-reactivity Craving Within Subjects Across a Period of Abstinence

	Study parameters			Participant Characteristics			Craving Results
Citation	RCT or Observational Study	Study Comparisons	Timepoints	MOUD	Treatment Setting	Sample Size(s)	Craving Outcome	Change in Craving Across Abstinent Timepoints (Decrease, No Change, Increase)
Dawe, Powell, Richards, et al. (1993)	RCT	(1) Methadone taper (1) Methadone taper + Cue Exposure Therapy (CET) (3) Clonidine-assisted detoxification (4) Clonidine-assisted detoxification + CET	Baseline, + 3 weeks	None	Residential/Inpatient	(1) n = 13 (2) n = 13 (3) n = 9 (4) n = 5	Average craving score from five ratings taken throughout a five-minute video	(1) Methadone taper - Decrease (2) Methadone taper + CET - Decrease (3) Clonidine - Decrease (4) Clonidine + CET - Decrease
Franken, de Haan,van der Meer, Haffmans, and Hendriks (1999)	Observational	(1) After therapy (2) Follow-up	Baseline, + 6 weeks	None	Residential/Inpatient	n = 16	Average craving score from five ratings taken throughout a five-minute video	Decrease
Marissen, Franken, Blanken, et al. (2007)	RCT	(1) Control Therapy (2) Cue Exposure Therapy	Baseline, +3 month	None	Residential/Inpatient	(1) n = 55 (2) n = 59	Craving scores post cue	(1) Cue Exposure Therapy - Decrease (2) Control Therapy - Decrease
Xue, Luo, Wu, et al. (2012)	RCT	(1) No memory retrieval-extinction procedure (2) Memory retrieval-extinction, 10 minute delay (3) Memory retrieval-extinction, 6 hour delay	Baseline, +3, +33, +183 days	None	Not specified	(1) n = 22 (2) n = 22 (3) n = 22	Craving change score	Statistics not reported for effect of time on change scores. Visual analysis shows decreases in all groups
Du, Fan, Jiang, et al. (2014)	RCT	(1) Usual Care (2) Usual Care + Cue-Exposure Therapy	Baseline and +2 months	None	Residential/Inpatient	(1) n = 23 (2) n = 22	Craving change score Craving scores post-cue	Usual Care-No change Usual Care + Cue-Exposure Therapy - Decrease
Navidian, Kermansaravi, and Tabas, et a. (2016)	RCT	(1) Treatment as Usual (2) Motivational Interviewing	Baseline, +2, +6, +12 months	Methadone	Outpatient	(1) n = 50 (2) n = 50	Craving scores post cue	Treatment as Usual: Decrease Motivational Interviewing: Decrease
Hurd, Spriggs, Alishayev, et al. (2019)	RCT	(1) Placebo (2) 400 mg cannabidiol (3) 800 mg cannabidiol	Baseline, +1 and +7 days	None	Outpatient	(1) n = 14 (2) n = 13 (3) n = 15	Craving change score	Placebo: Decrease 400 mg cannabidiol: Decrease 800 mg cannabidiol: Decrease
Liu, Zhao, Liu, et al. (2020)	RCT	(1) Sham (2) 1 Hz rTMS (3) 10 Hz rTMS	Baseline, +30, +60, and +90 days	None	Residential/Inpatient	(1) n = 27 (2) n = 34 (3) n = 31	Craving scores post cue	Sham: Decrease 1 Hz rTMS: Decrease 10 Hz rTMS: Decrease
Liu, Wang, Zhang, et al. (2021)	Observational	(1) Baseline (2) 6-month follow-up	Baseline, +6 months	None	Residential/Inpatient	(1) n = 31 (2) n = 15 (n = 16 lost to follow up)	Craving scores post cue	Decrease
Yue, Yuan, Bao, et al. (2022)	RCT	(1) No Treatment (2) Memory Reconsolidation Updating Procedure (MRUP)10 minutes post methadone (3) Memory Reconsolidation Updating Procedure 6 hours post methadone	Baseline, +1, +2, +3, +4, and +5 months	Methadone	Outpatient	(1) n = 30 (2) n = 26 (3) n = 27	Craving change score	No Treatment: Decrease MRUP 10 minutes: Decrease MRUP 6 hours: Decrease
Kang, Ding, Zhao, et al. (2022)	RCT	(1) Sham (2) iTBS	Baseline and +11 days	None	Residential/Inpatient	(1) n = 22 (2) n = 20	Craving scores post cue	Sham: No change iTBS: Decrease

Note. RCT = randomized controlled trial, MOUD = medications for opioid use disorder

There were three methods used to characterize cue reactivity craving across timepoints: (1) craving change score, (2) post-cue craving score, and (3) an average of multiple craving scores recorded throughout cue exposure. Six studies reported craving as a post-cue score, three studies reported craving as a change score, one study reported craving as an average score based on multiple scores throughout cue exposure, and one study reported both craving change scores and post-cue scores.

3.3.2.2. Participant Characteristics.

Studies included samples maintained on methadone (n = 2) and samples not receiving MOUD (n = 9). All studies reported biochemical and/or circumstantial (e.g., inpatient treatment facility) proof of abstinence throughout the study. Two studies included participant data at timepoints while they remained abstinent but did not include participants if they subsequently tested positive for non-prescribed opioid use (Liu et al., 2021; Navidian et al., 2016). As such, sample size gradually decreased across timepoints.

3.3.2.3. Differences in Cue Reactivity Craving Within Groups Across Timepoints

Of the 11 studies making within group comparisons of cue reactivity craving across timepoints, none reported increases in craving across timepoints. Among the nine RCTs, all reported decreases in craving across timepoints in the experimental conditions. Six out of nine studies reported decreases in cue reactivity craving across timepoints in the control conditions and the remaining three reported no changes in cue reactivity among the control conditions.

The two observational studies reported decreases in cue reactivity craving over timepoints (i.e., baseline and six week or month follow-up).

Among the two studies which enrolled individuals receiving methadone for OUD, one reported decreases in cue reactivity craving among the treatment as usual condition (Navidian et al., 2016). The second reported no change in cue reactivity craving over time in the sham condition (Jin et al., 2022). Both experimental conditions tested among among individuals receiving methadone for OUD showed decreases in cue reactivity craving over time.

4. Discussion

The primary finding of this review is that the patterns observed in opioid cue reactivity craving across varying periods of abstinence in clinical research do not match the patterns of behavior operationalized as incubation of craving observed in preclinical models. Our review of opioid cue reactivity clinical studies evidenced either no changes or decreases in craving as a function of longer periods of opioid abstinence. No studies reported increases in craving as a function of longer time abstinent. In contrast, in preclinical studies that modeled incubation of craving for opioids, the majority of studies showed outcomes that suggested proxies labeled as opioid craving (e.g., lever presses, nose pokes associated with opioid delivery) were greater after longer periods of abstinence. These findings emphasize recent arguments that preclinical models of incubation of craving likely measure a phenomenon that is mechanistically distinct from cue reactivity opioid craving experienced by individuals with OUD. These findings also highlight the importance of precision and parsimony in language to avoid conflation of operational labelling in preclinical models with behavioral and/or subjective outcomes observed in clincial settings.

4.1. Findings Related to the Measurement of Craving Over Time

Several themes emerged in the systematic review worth noting. First, an important finding in the preclinical systematic review is that two of the eight studies that showed circumstances where incubation of opioid craving did not occur used intermittent schedules of drug self-administration. These data are especially interesting given that intermittent access self-administration is associated with unique behavioral profiles of relapse and conformational neural changes (Ahmed and Koob, 1998; Algallal et al., 2020; Calipari et al., 2013). Intermittent access schedules of drug self-administration arguably represent a more ecologically valid model of opioid exposure and opioid use, so this may explain the more mixed findings for incubation of craving in these models and the closer homology to the clinical data. Additional preclinical studies should evaluate the differential effects of intermittent versus continuous access for drug self-administration acquisition on incubation of opioid craving to bolster findings from the two published studies. These findings stand in contrast to patterns observed in preclinical incubation of cocaine craving research which shows intermitent schedules of self administraiton is associated with greater increases in responding relative to continuous schedules of self administraiton (Nicolas et al., 2019; Nicolas et al., 2021). This systematic review highlights differences in preclinical opioid incubation cracing effects as a function of drug acquisition patterns and how subtle design differences may impact conclusions made about the lasting effects of reinforcing drugs. Not all drug exposure equally impacts subsequent behavior and it is important to model various clinically-relevant drug taking patterns in the laborartory when possible. More broadly, these data highlight the role of individual differences and variations in drug use history in the unique and heterogenous expression of OUD symptoms.

Another trend we observed in the clinical research systematic review is that changes in cue reactivity craving did not differ among individuals receiving methadone compared to individuals not receiving medications for OUD. Of the four studies that reported on changes in cue reactivity craving among individuals receiving methadone, two reported no change in craving and two reported decreases in craving. The majority of reviewed clinical studies included individuals who were not receiving MOUD. Among those 16 studies which enrolled individuals not receiving MOUD, five reported no changes in cue-induced craving as a function of time abstinent, seven reported decreases, and four reported mixed results (one craving outcome did not change, another craving outcome decreased). Although the sample of studies is small, these results preliminarily suggest that incubation of craving does not occur among individuals with or without MOUD (i.e., that results do not differ based on the provision of FDA-approved pharmacothearpies).

Finally, in the clinical research systematic review there was a greater likelihood to observe decreases in cue reactivity craving in within-group comparisons across time than in between-group comparisons. Eight of 11 within-group studies (73%) reported decreases in cue reactivity craving, while only two of nine between-group studies (22%) reported decreases in cue reactivity craving. It is possible that within group studies are impacted by practice effects or placebo/sham effects that result in systematically reduced reporting on craving measures over time or habituation to the images used in cue reactivity procedures. Regardless, neither between- nor within-group studies evidenced increases in cue reactivity craving across time.

4.2. What can preclinical research tell us about subjective, self-reported treatment endpoints used in clincal research?

In clinical research, craving is a complex and dynamic subjective state that can be induced by exposure to external cues or the experience of internal cues. Despite variations in methods to assess craving, craving is ultimately operationalized by an internal state (e.g., somatic, psychological, or cognitive state). Therefore, although craving may be the predictor of future substance use or a symptom that is troubling for a patient in its own right, craving is not a directly observed and causal drug-seeking or drug-taking behavior. This is a nontrival point for several reasons. First, craving uniquely functions depending on the context in which it is experienced and may be conceptualized differently between drug classes and individuals, and within-person at different timepoints during abstinence. Second because craving in humans is primarily measured by the self-reporting of subjective response – though physiolgical or neural indicators might also be assessed they should ultimately must be validated as indicators of the internal subjective experience of craving – in all practical manner, clinical practice relies on patient reported experiences. As such, one question we might ask is if craving, in the clinically relevant ways needed for predictive validity or mechanistic homology, can be modeled in nonhuman animals in a manner that clinically translates to humans? Any such translation is limited by: (1) the absence of higher-order language functions among nonhuman animals in laboratory settings, (2) the difficulty situating experiences within and beyond the temporal horizon of short-term abstinence in nonhuman animals, especially when considering distal or delayed goals with differing probability of likelihood, competing choices, proximal context, and individual traits that all influence how craving is experienced in the moment for patients in treatment, and (3) the lack of understanding how craving is relevant to and changes during remission and recovery goals over time for clinical populations relates to the behaviors observed in nonhuman animals. All of these factors call into question the ecological validity of incubation of craving preclinical models and largely undermine the translatability of preclinical work on craving (de Wit, Epstein, & Preston, 2018; Field & Kersbergen, 2020; Venniro et al., 2020).

4.3. What is the translational value of incubation of craving models?

Although the translational homology of preclinical research on the incubation of opioid craving to the clinical understanding of opioid craving is disconnected, there has been substantial work to evaluate the neurobiological mechanisms involved in the behavior observed in these models which provides basic science information about the brain circuits that are relevant to drug intake and behavior during drug abstinence. Studies have evaluted these mechanisms in tissues following the targeted behavior, with some having incorporated manipulations in an attempt to reveal causal relationships between a neurobiological finding and the target behavior. In these preclinical studies, defining the involvement of key nodes within the mesocorticolimbic reward pathway in the incubation of opioid seeking phenomenon is a primary focus. For example, one study found that orbitofrontal cortex (OFC) neuronal ensembles were activated following 14 days of abstinence from heroin self-administration and a cued extinction test. Selective inactivation of these cells reduced lever pressing during the test, and thus the results from this study indicate that a small subset of OFC neurons are involved in this specific behavior (Fanous et al., 2012b). In another study, cAMP response element binding protein (CREB) signaling within the nucleus accumbens was found to be involved in incubation of heroin seeking (Sun et al., 2015). In a more recent study, the incubation model has been applied to fentanyl dependence and found evidence that corticotropin releasing factor (CRF) in the bed nucleus of the stria terminalis (BNST) played a siginficant role in incubation of fentanyl craving (Gyawali et al., 2020). These identified regions and neurobiological underpinnings may inform novel therapeutics to address OUD or a potential treatment endpoint (although likely not craving based on the data reviewed above). To date, no pharmacotherapeutics have come from these neuroscience efforts (Fredriksson et al., 2021b; Strickland et al., 2022), which may be attributable to the the specificity of the implicated regions and the lack of exisiting medications or technology to target these mechanisms.

One way to better translate findings from preclinical incubation of craving models would be to create a behavioral task homologue for clinical research that could be tested over periods of abstinence. Such a model would be distinct from drug self-administration and would evaluate the effects of drug-paired cues on decision making (e.g., Strickland, Lile, Stoops, 2019). Importantly, conclusions about whether such a human homologue is related to craving still remains an empirical question.

4.4. If patterns observed in the incubation of craving model do not correspond with changes in self-reported craving, what is the incubation of craving model measuring?

Although none of the clinical studies reviewed here evidenced increases in craving as a function of time abstinent, it may be that some other clinically relevant psychological construct or behavioral outcome increases in intensity over time and therefore better corresponds to the patterns observed in preclinical incubation of craving models. Ultimately, it remains unclear whether this clinically-relevant behavioral or self-report endpoint exists and corresponds to the phenomenon captured in incubation of craving preclinical models. Identifying such an endpoint could ultimately allow for a more thorough mechanistic understanding of symptoms that may require ongoing monitoring and treatment in clinical care. Until those such preclinical endpoints are identified, refined, and incorporated into incubation of craving models, craving in whatever manifestation it takes among individuals with OUD, should continue to be studied among clinical populations where this subjective state is most amenable to study. Relatedly, clinical investigations should consider novel ways for scrutinizing both cue-induced and organic/endogenous craving experiences among persons who cannot maintain abstinence as this may be a more urgent area of craving-related research relative to the possibility of an incubation of craving phenomenon in humans.

4.5. Limitations

This review focused on opioid craving; therefore, we can only make conclusions about the divergent patterns observed in the context of opioid use. Although craving is a common feature across all substance use disorders, it is possible that it may function differently temporally as a function of the specific use disorder, perhaps because of unique withdrawal profiles or because of the unique drug mechanisms of action that may differentially impact conformational neurobiological changes after chronic use. Regardless, this systematic review demonstrates that incubation of craving, at the very least, is not a uniform phenomenon observed across drug classes and that, in fact, the opposite is observed with respect to opioid craving in the clinical setting.

A second limitation is that there was heterogeneity in the methods used in the clinical studies included in this systematic review. However, craving time course do not appear to differ as a function of population enrolled nor the method by which craving is reported. Importantly, despite the variability in study design, craving did not increase as a function of time abstinent in any clinical study reviewed.

Next, this review focused on cue reactivity craving because it most closely resembles cued extinction tests used in the incubation of craving model. As noted, craving is multi-dimensional and efforts are ongoing to determine the best ways to measure, monitor, and treat this endpoint. Therefore, it is possible that some other dimension of craving – not cue reactivity-could follow similar patterns as the incubation of craving model. At this point in time, it would not be possible to evaluate this in a systematic review because there is great variability in (1) opioid craving assessments and (2) the dimensions that are captured in the various assessments. A standard, comprehensive assessment of opioid craving would lend itself to determining which – if any-craving dimensions are relevant to what is observed in the incubation of craving preclinical model.

Voluntary abstinence models (Table 2) may more closely approximate certain treatment conditions. The results from these new models are varied, but voluntary ‘choice-induced’ abstinence is more reliably associated with suppressions in time-dependent increases in drug-cued behavior. However, in clinical research, reported craving did not increase across time in any of reviewed treatment conditions (e.g., compulsory, residential, outpatient) suggesting forced abstinence versus voluntary abstinence does not necessarily influence the time course of craving. One preclinical study on the incubation of methamphetamine craving suggests that antecedents to drug taking (e.g., drug-paired cues, drug-paired contexts, and drug effects [i.e., reinstatement]) differentially influences incubation of craving responses (Adhikary et al., 2017). While this effect has not been replicated in clinical research and is not addressed by our systematic review, such associative relationships could be further explored in ecological momentary assessment studies.

Finally, clinical research which evaluates changes in cue reactivity craving over time is often restricted to individuals who can abstain over a given period of time. Therefore, it is not possible to understand how craving may differentially function during abstinence among individuals who cannot abstain. It is possible that such individuals represent a truly unique subgroup with their own “craving time course” phenotype that was not captured by this systematic review. Relatedly, in the reviewed clinical research studies the effects of time were observational in nature and individuals were never randomly assigned to a specified length of time abstinent. Importantly, such a design factor has been difficult to execute in prospective clinical studies on incubation of craving due to variable (and unpredictable) attrition rates. The data we characterize in this review reflect patterns in cue reactivity craving among individuals presenting to treatment or research settings and therefore are likely more clinically relevant to our understanding of how craving functions.

4.5. Conclusions

We found no evidence that measurable incubation of opioid craving reliably occurs in clinical research in OUD. The lack of clinical evidence for consistent increases in craving in clinical populations weaken the translational utility of operationalizations of the time-dependent increase in drug-seeking behavior observed in preclinical models as models of incubation of craving. The use of craving to describe these time-dependent increases likely undermines the translatability of these models given the lack of mechanistic homology observed across species.

Highlights.

• Incubation of opioid craving is not observed in clinical research

• Opioid craving decreases in humans over time during periods of abstinence

• The translational utility of the incubation of craving model needs refinement