Craving research: future directions

Addiction (2000) 95 (Supplement 2), S247–S255

SYNTHESIS AND FUTURE DIRECTIONS IN CRAVING RESEARCH

Craving research: future directions

D. COLIN DRUMMOND,1 RAYE Z. LITTEN,2

CHERRY LOWMAN2 & WALTER A. HUNT3

1Department of Addictive Behaviour and Psychological Medicine, St George’s HospitalMedical School, University of London, London, UK, 2Division of Clinical and PreventionResearch/Treatment Research Branch & 3Division of Basic Research, National Institute ofAlcohol Abuse and Alcoholism, National Institutes of Health, Bethesda, MD, USA

AbstractMany prospective clinical studies have concluded that craving does not reliably predict relapse and that theconcept is of little or no clinical utility. Contrary to earlier more simplistic clinical models of addiction, morerecent models do not require that craving be present for relapse to occur. New approaches to study humancraving may enhance its predictive validity and yield more knowledge of its nature, course, behaviouralsequelae and regulatory function in alcohol/drug consumption. These approaches include empirical researchthat focuses on: (1) the elucidation of the domains of craving (i.e. subjective experience, physiologicalresponses, behavioural sequelae and their inter-relationships); (2) the temporal dynamics of craving (i.e. itscourse over minutes or days, as well as its natural history over the course of a drinking career); (3) the factorsthat may mediate/moderate/determine the development and resolution of craving; (4) studies of the predictivevalidity of craving measures; and (5) the development of valid methods of measuring the different domainsof craving. The conclusions are that future craving research should: (1) incorporate more sophisticated generaltheories of behaviour (conditioning, cognitive social learning, neurobiological, and genetic); (2) apply moresophisticated and standardized measurement methods and experimental paradigms, including studies in whichalcohol is made available to human subjects; and (3) effective development of new pharmacological andbehavioural therapies for relapse prevention depend on greater understanding of the nature and measurementof craving.

IntroductionSince the National Institute of Drug Abuse con-ference on craving nearly a decade ago (Pickens& Johanson, 1992), there has been a growth inresearch activity in craving and the developmentof several new research paradigms and theoreti-cal models (Lowman et al., this issue). Someolder theories have been tested more rigorously

and some newer theories challenge hitherto� rmly held orthodoxy. However, the phenom-enon of craving remains an enduring puzzle, atthe same time � lled with both tantalizing prom-ise and frustrating disappointment. Continuedresearch interest in this area may be related tothe perceived clinical relevance of the cravingphenomenon to both patients and clinicians, as

Correspondence to: D. Colin Drummond, Department of Addictive Behaviour and Psychological Medicine, StGeorge’s Hospital Medical School, University of London, Cranmer Terrace, London SW17 0RE, UK. e-mail:[email protected]

Submitted 3rd November 1999; initial review completed 12th January 2000; � nal version accepted 3rd April 2000.

ISSN 0965–2140 print/ISSN 1360–0443 online/00/08S247–09 Ó US Government

Carfax Publishing, Taylor & Francis Ltd

S248 D. Colin Drummond et al.

well as the development of new pharmacologicaltreatments that aim to moderate alcohol cravingand relapse.

Nevertheless, craving remains elusive to com-plete explanation (Drummond, this issue). Opin-ion remains divided on the importance of craving,on one hand being seen as the key substrate ofaddiction, and on the other hand, a redundantepiphenomenon, neither necessary nor suf� cientto precipitate relapse. The term “craving” contin-ues to be used in different ways to describedifferent phenomena in spite of calls to tighten itsusage (Kozlowski & Wilkinson, 1987). Animalresearchers often use the term to denote human-like addictive behaviours in their subjects,whereas clinical researchers seek to elicit cravingin verbal reports; but are they studying the samephenomenon? Pharmacologists sometimes as-sume that craving is the principal target for thedevelopment of medications to attenuate drinkingbehaviour, whereas some psychologists argueagainst such an approach in advance of cravingparadigms and measures of satisfactory reliabilityand validity (particularly predictive validity).Neuroscientists, on the other hand, have appliedneuroimaging techniques to “look at” craving,avoiding, as some might argue, the necessity forself-report measures of unknown reliability andvalidity; but are they observing the same phenom-enon that patients experience and verbally de-scribe as “craving”?

Another problem with craving research to dateis that craving can be conceptualized as both anentity and a process, but often researchers areconfused as to which they refer. As an entity,craving is something that patients experience anddescribe as a phenomenon of concern and at timesintrusive and perplexing. It is also a process,potentially interacting with a variety of otherphenomena to maintain drinking behaviour orleading to relapse (e.g. craving might undermineself-ef� cacy in the context of a high-risk situationin Marlatt’s conceptualization of craving andrelapse (Marlatt, 1978)). However, it would notbe appropriate simply to study craving as a processwithout at the same time trying to describe andunderstand craving as a clinical phenomenon.Nevertheless, it is important that researchers areclear about the way in which the term “craving”is being used.

The conclusion in many prospective clinicalstudies has been that craving does not reliablypredict relapse and that the concept is therefore

of little or no clinical utility. However, we take theposition that this may be because craving has beeninadequately conceptualized and studied in hu-mans (see also Li; Meyer; Shiffman; this issue).Contrary to earlier more simplistic clinical modelsof addiction, more recent models do not requirethat craving be present for relapse to occur (Low-man et al., 2000). Nevertheless, clinical observa-tions of craving in alcoholics and drug addicts(Drummond, 2000; Li, 2000; Meyer, 2000) sug-gest that there is considerably more to thephenomenon of craving in humans than can becaptured by the widely used unidimensional mea-sures of craving administered in most clinicaltrials. Such approaches do not, for example,represent or control for temporal variations in thecourse of craving and the independent variablesthat in� uence these (Drummond, 2000; Li, 2000;Sayette et al., 2000; Tiffany, Carter & Singleton,2000; Shiffman, 2000).

We intend to focus this commentary on ap-proaches to study human craving that may en-hance its predictive validity and yield moreknowledge of its nature, course, behavioural se-quelae and regulatory function in alcohol/drugconsumption. These approaches include empiri-cal research that focuses on (1) the elucidation ofthe domains of craving (i.e. subjective experience,physiological responses, behavioural sequelae,and their inter-relationships); (2) the temporaldynamics of craving (i.e. its course over minutesor days, as well as its natural history over thecourse of a drinking career); (3) the factors thatmay mediate/moderate/determine the develop-ment and resolution of craving; (4) studies of thepredictive validity of craving measures; and (5)the development of valid methods of measuringthe different domains of craving.

Our scan of the horizon of craving researchshould not be seen as an exhaustive list of researchquestions. Undoubtedly, our conclusions will beshaped by personal prejudices, and judgementshave had to be made about what to include andwhat to leave out. Some key issues for researchmay appear to be missing altogether. At this stagewe felt it important to identify some questions thatare potentially researchable, and that emerge fromthe totality of the papers in this supplement.

Future directions for craving researchWhat are the domains of craving?One of the striking observations that can be

Craving research: future directions S249

made of the current state of craving research isthe wide range of models and paradigms thathave been used to conceptualize craving. Thisleads one to ask if we are all studying differentaspects of the same phenomenon, or differentphenomena. What are the domains of craving,and what is the relationship between those do-mains? By domains, in this context, we mean theprincipal ways in which craving can be describedwithout recourse to explanation or metaphysicalassumptions. In other words, we seek to eluci-date the phenomenology of craving as proposed byMeyer (2000). (In this context we use the term“phenomenology” in the strict sense of“descriptive psychopathology”.) Without di-gression into the philosophy of science, in thiscontext we include the autonomic and be-havioural, as well as subjective dimensions ofcraving within the scope of “domains of craving”that need to be elucidated.

In terms of the subjective experience of crav-ing, it has been described variously as appetitiveand aversive, positively and negatively hedonic,liking and wanting or needing, an urge or adesire. It is not clear at present if these subjectiveelements are mutually exclusive, opposites, inter-related, or sometimes, or never, co-occurring.Are these different dimensions of the cravingexperience which, as some of the commentatorsin this supplement have suggested, are related todifferent underlying biological substrates of ad-diction? It seems that an important missing el-ement of craving research is the comprehensivestudy of subjective experience. Is craving cultur-ally or language bound, or is it a universal hu-man experience? Is craving different for differenttypes or classes of drug? How does the experi-ence of craving compare between different“behavioural addictions”, such as gambling dis-orders, obsessive compulsive disorder, eating dis-orders, sexual appetites (Marks, 1990), or is itunique to speci� c disorders? What is craving likein polydrug misusers, and do they recognizedifferent types of craving to different drugs? Howdoes craving relate to other conceptual elementsof the dependence syndrome? Edwards & Gross(1976) left open the possibility that not all ele-ments of the syndrome (including “subjectiveawareness of the compulsion to drink”) need bepresent simultaneously. How often and underwhat conditions (e.g. degree of dependence) iscraving present?

Several authors in this supplement raise the

issue of the importance of studying craving in thenatural drinking environment. Little is known ofthe “ecology of craving”: what is the experienceof craving like in the natural drinking environ-ment, and how does this sort of craving differfrom that experienced by someone who is under-going alcohol detoxi� cation, or in an experimen-tal laboratory undergoing cue exposure?

Drummond (2000) and Li (2000) have re-ferred to Isbell’s distinction between non-symbolic and symbolic craving (analogous to“withdrawal craving” and “cue-elicited craving”;Drummond, 2000). Are these phenomena em-pirically distinct, and how are they inter-related?In order to address some of these questions, avariety of research methods and paradigms willbe needed. It is apparent that the research oncraving is currently predominantly experimentaland/or psychometric. It seems that qualitative,anthropological and cross-cultural methodswould offer a new direction in investigating thephenomenology of craving. Further, since the1970s, researchers have not employed the poten-tially important method of alcohol-priming dosesand free-drinking paradigms to study craving inhumans (e.g. de Wit, 2000; Sinha et al., 1999).Even when these methods were in vogue, fewstudies examined subjective craving per se, withsome important exceptions (e.g. Rankin et al.,1982, 1983). We believe this is a missed researchopportunity, and one that could increase ourunderstanding of the phenomenology of craving,providing appropriate ethical safeguards are ap-plied (Dolinsky & Babor, 1997).

Several studies have examined autonomic re-sponses to alcohol- and drug-related cues (auto-nomic cue-reactivity), and their relationship tocue-elicited craving (see Drummond, 2000).Clearly it is possible that autonomic cue-reactivity and cue-elicited craving may be onlyloosely coupled (e.g. as is found in autonomicresponses to fear-inducing stimuli, as noted byRachman & Hodgson, 1974, and others). How-ever, in terms of cue reactivity, it appears thatautonomic reactivity and subjective craving toalcohol cues are inter-related (Glautier & Drum-mond, 1994), and similar across different classesof drug such as alcohol, opiates, cocaine, nic-otine (Carter & Tiffany, 1999). However, auto-nomic cue reactivity has smaller effect sizes thansubjective reactivity.

Further, there is considerable debate about thenature and signi� cance of the form of cue reac-


tivity (i.e. whether it conforms to a drug-like ora drug-opposite response). It is likely that thestudy of the effects of agonist and antagonistdrugs on various neurotransmitter systems (in-cluding opioidergic, dopaminergic, glutamater-gic, and GABAergic neurones) could help toelucidate the directional valence of cue reactivityeffects. Brain imaging techniques may also bevaluable in this area.

Behavioural cue reactivity has tended to bestudied more in the context of animal experi-ments where free access to alcohol is less contro-versial than in human studies. Opportunitiesexist to make a connection between animal andhuman research, particularly if free access andreward paradigms can be introduced into clinicalpopulations, with appropriate safeguards.

In summary, knowledge of the phenomenolog-ical domains of craving is an essential precursorto the experimental study of craving in bothhumans and animals. Such knowledge couldhelp to inform the development of measurementinstruments, and the design of parallel humanand animal experimental paradigms. Further,knowledge of the phenomenology and inter-relations of craving and other elements of depen-dence in the natural drinking environment aswell as in experimental settings, would allow thedevelopment of craving research with a greaterdegree of ecological validity.

What are the temporal dynamics of craving?As with craving domains, the study of temporaldynamics refers to a set of phenomenologicalresearch questions rather than seeking to explaincraving or its antecedents and consequences.The distinction between temporal dynamics anddomains of craving, in this context, is in thedynamic nature of craving rather than the di-mensionality. The main question here is, if crav-ing had to be designed, what would it look like intemporal space? Again this is a necessary de-scriptive precursor of developing measures andparadigms to study craving, that might allowunderstanding of the craving phenomenon.

Several authors in this supplement have notedthe dynamic nature of craving. It has been de-scribed variously as tonic, phasic and pulsatile;linear and curvilinear; continuous and discontin-uous (e.g. urges versus craving). The temporaldynamics of craving responses, particularly tostimuli (cue-elicited craving), will have implica-

tions for the design and timing of administrationof craving measures. For example, pulsatile orphasic craving responses may be missed by in-frequent measurements, or measures that aggre-gate craving over long periods. Similarly, thetemporal sequencing of craving may have anin� uence on the timing and type of measuresemployed (Sayette et al., 2000). For example,measures of craving taken before, during or afteran episode of drug use may yield different de-grees or types of craving experience.

To return to the ecological validity of cravingmeasures, little is known about the way in whichcraving varies during a typical drinking day foran alcoholic in his or her natural drinking en-vironment (e.g. on waking, following the � rstdrink, after the second and subsequent drinks,walking into the pub, when highly intoxicated,when access to alcohol is denied). Again, thismay require more qualitative and/or observa-tional methods of study, rather than quantitativeor experimental designs. Drummond (2000), Li(2000), Meyer (2000) and Koob (2000) drawattention to the potential in� uence of withdrawaland protracted withdrawal on the craving experi-ence. Earlier studies of alcohol withdrawalplaced limited emphasis on the measurement ofsubjective craving, partly because it was believedto be either an indication of personality disorder,or an epiphenomenon, rather than being a mea-sure of interest in its own right (Kolb & Him-melsbach, 1938; Mendelson & Mello, 1966).The study of the in� uence of alcohol withdrawalon cue-elicited craving is also a potential area tounderstand the temporal dynamics of craving(Drummond, 2000).

What are the determinants/mediators/moderators ofcraving?One of the potential problems in the study ofcraving is in assuming that our subject popula-tions, both animal and human, are homoge-neous. There is extensive evidence noted inseveral papers in this supplement that there ex-ists considerable heterogeneity with respect tothe factors that might in� uence craving. Animalresearch has provided clear differences in alcoholpreference and degree of conditionability withinspecies (Samson, 2000). It would be worthwhileexploring the existence of genetic differences inhumans in relation to susceptibility to craving (aswell as the development of other aspects of de-


pendence). It would be sensible to explorewhether or not there are genetically determineddifferences in humans in the activity of neuralsystems believed to be involved in craving experi-ences. Psychopharmacology and neuroimagingcould help to elucidate such biological diversity.

Several papers refer to the need to study thein� uence of contextual cues, or occasion setters,in studying craving. These could include intero-ceptive cues such as cognitions, emotions andpriming doses of alcohol (see Beck et al., 1993;Drummond, 2000). There are some pointers tothe likely effects, but this has never been studiedsystematically in human cue-reactivity. There issome evidence that personality factors canin� uence cue-elicited craving. What aboutspeci� c psychiatric co-morbidities such as pho-bic anxiety or depression? Is craving different inthese contexts? Brain damage, related to alcoholor otherwise, can interfere with emotionality,learning and memory, all of which are likely tobe important in craving and cue-reactivity. Braindamage may alter the neural “hard wiring” of thebrain. Does this interfere with acquisition andextinction of cue-elicited craving, and could lossof tolerance to alcohol sometimes observed inthe later stages of alcoholism (Edwards et al.,1997, p. 46) be related to such lesioning throughdamage to a conditioning mechanism? Again,neuroimaging may help in the study of thesequestions.

The way in which experiments are designedhas been recognized as an important determi-nant of craving experience. Alcoholics in treat-ment who are attempting to abstain, may befaced with very different demand characteristicsand expectancies than social drinking volunteers.How far can experiments with social drinkersinform an understanding of craving in addictedindividuals?

Overall, there needs to be a greater awarenessof the potential of biological and psychologicaldiversity to in� uence the craving experience.Systematic study of these factors could help toinform research design, and potentially � ltersome of the “noise” out of craving measurement,as well as help to understand the nature anddeterminants of craving.

What is the predictive validity of craving?One of the key reasons for the continuing studyof craving is its potential role in relapse to drink-

ing behaviour following a period of abstinence inalcoholics, which is arguably the most importantissue to address in treatment. If craving has nopredictive power it has little heuristic or practicalclinical importance. Thus, a key issue debated inseveral of the papers here has been about thepredictive validity of craving. There appear to befour main possibilities concerning the predictivevalue of craving and relapse. First, craving andrelapse may be completely unconnected randomevents. On the basis of existing evidence, how-ever, this does not appear to be the case. Sec-ondly, craving is predictive of relapse, butexisting measures are inadequate, i.e. the signal-to-noise ratio is low. In this case, greater invest-ment in developing more reliable and validmeasures of craving is needed. Thirdly, cravingis predictive of relapse, but only under certainconditions. Thus, study of the conditions underwhich craving leads to relapse is needed.Fourthly, the subjective experience of craving isnot predictive of relapse, but autonomic and/orbehavioural correlates of craving or the determi-nants/mediators/moderators of craving are pre-dictive. This will require the systematic study ofthe relationship between correlates of cravingand relapse.

There is an important distinction to be madebetween, on one hand, the causal role of cravingfor relapse, and the prognostic value of cravingon the other hand. A causal relationship is localin time and space: if a certain instance of relapseis causally “explained” by craving, then cravingmust have been present immediately before therelapse occurred. Within this de� nition it maywell be the case that craving of one sort oranother is an important factor in most or allcases of relapse.

On the other hand, the predictive value ofcraving may be more limited. We have suggestedthat craving may vary considerably over timeand, hence, the temporal dynamics of cravingmay limit its predictive value depending on howand when it is measured. If the auto-correlationof craving is weak, it may be impossible to pre-dict future craving (and therefore future relapse)on the basis of past craving, even if one coulddevelop a “perfect” measure of craving. Thus,craving may well explain many cases of relapse,but be unable to predict relapse.

The timing of craving measurement may proveto be a critical determinant of its predictivevalidity. Typically, little attention has been paid


to this issue in previous research. It would beunsurprising if craving measured at a time pointdistant from a relapse event had low predictivevalue (e.g. during a residential treatment pro-gramme several weeks or months before an absti-nent alcoholic returned to their usual drinkingenvironment). Further, if craving varies widelyover the course of a day or even one hour, it mayhave variable predictive power depending onwhen it is measured. Shiffman’s Ecological Mo-mentary Assessment (EMA) which carries outfrequent measures throughout a day using ahand-held computer, may have a greater abilityto capture craving occurrences close to the re-lapse event than an infrequent or retrospectivemeasure (Shiffman, 2000). However, this mayprove more dif� cult to measure in alcoholicscompared with smokers, where cognitive effectsof alcohol may interfere with assessment compli-ance. Similarly, cue-elicited craving may havegreater predictive power than measures of gener-alised or “background” craving level becausecue-elicited craving may be more closely relatedto an individual’s craving reaction in a futurehigh risk situation (Rohsenow & Monti, 1999;Drummond, 2000).

It may also be the case that the extent to whichan individual’s craving � uctuates widely acrosshours, days or weeks may be an important pre-dictive factor in itself. If, for example, an individ-ual’s average craving is generally low, but theyexperience short, extreme “bursts” of cravingonce a day, this may place them at greater risk ofrelapse than another individual who has a gener-ally high but stable (i.e. non-� uctuating) level ofcraving. This could potentially be tested usingEMA.

Another possibility is that correlates of cravingmay predict relapse, whereas craving itself doesnot. Thus, the determinants/mediators/modera-tors of craving may have greater prognostic valuethan craving measures. Several studies have al-ready examined the relationship between cravingand subsequent drinking behaviour. In animalexperiments one can only make inferences aboutthe animal’s craving experience. However, it isclear that, in animals, contextual cues associatedwith alcohol consumption are predictive ofdrinking behaviour. The inference could be thatif contextual cues elicited animal “craving”, andanimal craving could be measured, cravingwould be predictive of drinking behaviour. If thiswas true, however, craving would be a somewhat

redundant concept unless it added predictivevalue to that of contextual cues alone, i.e. cuesmay be predictive of relapse, but only whencraving is present. Clearly this can only bestudied in humans.

We have suggested earlier that more researchis needed into the determinants/mediators/moderators of craving. This will be particularlyimportant in relation to establishing their predic-tive power in relapse. It may be, for example,that while a craving and depression are corre-lated, relapse only occurs when depression andcraving are present simultaneously. It might bethe case that self-ef� cacy is an intervening vari-able in the craving–relapse relationship (as sug-gested by Niaura, 2000). Hence, possibly, bothcraving and low self-ef� cacy need to be presentfor relapse to occur. Clearly one could generatea whole series of plausible causal hypotheses ofthis sort, based on existing theories of relapsethat could be empirically tested.

It is also important to study further the condi-tions under which craving and alcohol consump-tion co-vary, in order to establish the predictivevalidity of craving. It may be the case that differ-ent types of craving (e.g. withdrawal-related,cue-elicited, and priming dose-related) may havedifferent predictive power.

Finally, different domains of craving (cogni-tive, autonomic and behavioural) may have dif-ferent predictive power. There is some evidencethat attentional bias and autonomic cue-reactivity have a higher predictive value thansubjective craving, but this needs to be studiedmore systematically and using experimental de-signs with greater statistical power (Tiffany et al.,2000). Tiffany’s cognitive processing theory ofcraving (Tiffany & Conklin, 2000) needs to beresearched in human alcoholics, including bothexperimental and observational methods.

Overall, the important point being made hereis that more sophisticated models of the relation-ship between craving and relapse need to begenerated and tested. Many of the papers in thissupplement provide such models. It is clear fromexisting research presented in this supplementthat the relationship between craving and relapseis likely to be complex. Thus, the observationthat craving and relapse are not closely coupledempirically does not negate the potential causalexplanatory power of craving both as a phenom-enon and a process. Rather, it suggests thatgreater attention needs to be focused on the


temporal dynamics and correlates of cravingin future research on the predictive validity ofcraving.

How should craving be measured?Several papers (particularly Sayette et al., 2000)point to a host of ways in which the measure-ment of craving could be improved. While sev-eral subjective craving measures already exist,consensus is required regarding the optimal mea-sure for use in different experimental contexts(e.g. cue-reactivity, medications development,EMA). A compromise will need to be madebetween questionnaires that provide adequatecoverage to protect internal validity (number ofitems/dimensions), and those that are practicallyuseful and have external validity in experimentalsettings (ease of use, ecological validity). It ispossible that questionnaires or indeed verbalquestions about craving may themselves induce(or cue) craving (Sayette et al., 2000). Thus,some further thought needs to be given to theoptimal design of craving experiments to providean adequate degree of ecological validity. Fur-ther, de Wit’s (2000) proposal that craving ef-fects of priming may be dose-dependent, needsto be tested in human alcoholic subjects.

The relationship between different existingmeasures of craving (e.g. Obsessive CompulsiveDrinking Scale (Anton, 2000), the Desires forAlcohol Questionnaire (Love et al., 1998), andsingle-item craving measures) needs to be exam-ined. Shorter variants of existing questionnairesneed to be studied for concurrent validity withfull versions as comparisons, in different exper-imental paradigms and populations.

In relation to the measurement of cue-reactivity (including cue-elicited craving) severaladvances in experimental design and measure-ment have taken place over the past decade.However, different research groups tend to usedifferent measures and designs (measurementparameters, measurement techniques, cues usedto elicit craving), which leads to problems inmeta-analysis of cue reactivity (Carter & Tiffany,1999). Further, Sayette et al. (2000) point toemerging cue-reactivity measures that may proveto be of greater predictive value than typicalsubjective or autonomic measures (e.g. startlere� exes, neuroimaging, cognitive processing,Stroop, expressive behaviour, reinforcementparadigms, drug self-administration paradigms).

Overall, there is now a much wider range ofmethods of measurement of craving than wasavailable 10 years ago. In addition to a continu-ing search for further novel approaches, there isa need to reach a consensus on the most usefuland practicable existing measures and methodsof craving measurement. What should go intothe standard craving assessment “toolkit” thatcan be taken into the laboratory or the clinicalsetting? Many other areas of addiction researchhave achieved greater consensus on standardmeasures. For example, clinical trials now com-monly use measures such as the Time-LineFollow-Back method of measuring alcohol con-sumption (Sobell & Sobell, 1992), or Form-90from Project MATCH (Miller, 1996). Can simi-lar standardized approaches be found for cravingmeasurement in clinical and experimental re-search?

ConclusionsThe original aim of the “craving workshop”which has given rise to the papers in this sup-plement was to extend craving research in thecontext of alcoholism treatment. Looking at the� nal product of this initiative it would be easy toform the conclusion that we are some way fromreaching a consensus on the clinical relevance ofcraving research. However, can some conclu-sions be reached regarding the potential of crav-ing research in the clinical arena?

It must be concluded that we are not yet at astage where research-derived measures of craving(be they subjective, physiological or behavioural)can impact directly on patient care of the al-coholic. However, several promising leads havebeen identi� ed. The workshop, and this sup-plement which resulted from it, have helped informulating some preliminary conclusions on thecurrent state of craving research, and allow thedevelopment of new research directions.

First, there is a need to move away fromclinically based and simplistic causal de� nitionsand models of the relationship between cravingand relapse. The phenomena are clearly muchmore complex than would be implied by suchmodels. There is a need to move towards moresophisticated models that incorporate advancesin general theories of behaviour (including con-ditioning, cognitive, social learning, neurobiolog-ical and genetic theories). It is likely that nosingle theoretical model will be able fully to


explain the craving phenomenon and its relation-ship to drinking behaviour, although several pa-pers in this supplement provide valuableattempts to integrate these models (e.g. Abrams,2000; Koob, 2000; Meyer, 2000; Monti et al.,2000; Niaura, 2000).

Secondly, there is a need for greater sophisti-cation and standardization in methods ofmeasurement of craving, including further devel-opment of research paradigms used to investi-gate craving. Again, several papers in thissupplement point to directions in which suchdevelopment can progress. Of critical import-ance is the need for development of, and consen-sus on, the most appropriate subjective measuresof craving, which possess a suf� cient degree ofinternal and external validity. Only when re-search studies begin to share common measureswill comparisons between different populationsand experiments become possible. Further, thedevelopment of measures and researchparadigms are necessary to allow some of thequestions identi� ed in this commentary aboutthe domains, temporal dynamics and correlatesof craving to be studied. We believe that studiesin which alcohol is available to human subjectswould help to advance knowledge in a way thatprevious research has been unable to achieve.Clearly, careful consideration for the safeguardof human subjects needs to be balanced with thepotential for such research to improve the devel-opment of treatments aimed at alleviating thesuffering related to alcoholism.

Finally, we believe that advances in our under-standing of the nature of craving will allow fur-ther development of promising new treatments.The identi� cation of individuals in alcohol treat-ment programmes who are at high risk of relapsecould help to target treatment interventions atthose most in need. With the arrival of new,promising medications aimed at preventing re-lapse, there is potential for better cravingmeasurement (including subjective, autonomicand behavioural measures) to allow theidenti� cation of which individuals are most likelyto bene� t from which drugs, and so improvetreatment cost-effectiveness (Anton, 1999; Swift,1999). Such research is likely to include the useof neuroimaging studies to investigate the effectsof medications on craving and brain function(Hommer, 1999). Furthermore, improved crav-ing measurement could help to identify morepromising drugs before progressing to costly

clinical trials. A better understanding of the na-ture of craving and the factors that mediatecraving and relapse may also help in the develop-ment of psychological interventions aimed atpreventing relapse. Overall, the key challengefacing craving research is to prove its relevanceto the real-world problems facing patients andclinicians every day. Any advances that can im-pact on the high rate of relapse following alcoholtreatment would be welcome. We feel that thepapers in this supplement advance the research� eld further in that direction.

ReferencesABRAMS, D. B. (2000) Transdisciplinary concepts and

measures of craving: commentary and future direc-tions, Addiction, 95 (suppl. 2), S237–S246.

ANTON, R. F. (1999) What is craving? Models andimplications for treatment, Alcohol Research andHealth, 23, 165–173.

ANTON, R. F. (2000) Obsessive–compulsive aspects ofcraving: development of the Obsessive CompulsiveDrinking Scale, Addiction, 95 (suppl. 2), S211–S217.

BECK, A. T., WRIGHT, F. D., NEWMAN, C. F. & LIESE,B. S. (1993) Cognitive Therapy of Substance Abuse(New York, Guilford Press).

CARTER, B. & TIFFANY, S. T. (1999) Meta-analysis ofcue reactivity in addiction research, Addiction, 94,327–340.

DE WIT, H. (2000) Laboratory-based assessment ofalcohol craving in social drinkers, Addiction, 95(suppl. 2), S165–S169.

DOLINSKY, Z. S. & BABOR, T. F. (1997) Ethical, sci-enti� c and clinical issues in ethanol administrationresearch involving alcoholics as human subjects,Addiction, 92, 1087–1097.

DRUMMOND, D. C. (2000) What does cue-reactivityhave to offer clinical research? Addiction, 95 (suppl.2), S129–S144.

EDWARDS, G. & GROSS, M. M. (1976) Alcohol depen-dence: provisional description of a clinicalsyndrome, British Medical Journal, 1,1058–1061.

EDWARDS, G., MARSHALL, E. J. & COOK, C. C. H.(1997) The Treatment of Drinking Problems: a guide forthe helping professions (Cambridge, CambridgeUniversity Press).

HOMMER, D. W. (1999) Functional imaging of craving,Alcohol Research and Health, 23, 187–196.

KOLB, L. & HIMMELSBACH, C. K. (1938) Clinical stud-ies of drug addiction. III. A critical review of thewithdrawal treatments with a method of evaluatingthe abstinence syndrome, American Journal ofPsychiatry, 94, 759–772.

KOOB, G. F. (2000) Animal models of craving forethanol, Addiction, 95, (suppl. 2).

KOZLOWSKI, L. T. & WILKINSON, D. A. (1997) Use andmisuse of the concept of craving by alcohol, tobacco,and drug researchers, British Journal of Addiction, 82,31–36.

LI, T. K. (2000) Clinical perspectives for the study of


craving and relapse in animal models, Addiction, 95(suppl. 2), S55–S60.

LOVE, A., JAMES, D. & WILLNER, P. (1998) A compari-son of two alcohol craving questionnaires, Addiction,93, 1091–1102.

LOWMAN, C., HUNT, W. A., LITTEN, R. Z. &DRUMMOND, D. C. (2000) Research perspectives onalcohol craving: an overview, Addiction, 95 (suppl.2), S45–S54.

MARLATT, G. A. (1978) Relapse prevention: theoreticalrationale and overview of the model, in: MARLATT,G. A. & GORDON, J. (Eds) Relapse Prevention: mainte-nance strategies for addictive behaviors, pp. 3–70 (NewYork, Guilford).

MARKS, I. (1990) Behavioural (non-chemical) addic-tions, British Journal of Addiction, 85, 1389–1394.

MENDELSON, J. H. & MELLO, N. K. (1966) Experimen-tal analysis of drinking behavior of chronic alco-holics, Annals of the New York Academy of Sciences,133, 28–45.

MEYER, R. E. (2000) Craving: what can be done tobring the insights of neuroscience, behavioral scienceand clinical science into synchrony, Addiction, 95(suppl. 2), S219–S227.

MILLER, W. R. (1996) Form 90: a structured assessmentinterview for drinking and related behaviors: test manual(Bethesda, MD, National Institute on AlcoholAbuse and Alcoholism).

MONTI, P. M., ROHSENOW, D. J. & HUTCHISON, K.(2000) Toward bridging the gap between biological,psychobiological, and psychosocial models of alcoholcraving, Addiction, 95 (suppl. 2), S229–S236.

NIAURA, R. (2000) Cognitive social learning and re-lated perspectives on drug craving, Addiction, 95(suppl. 2), S155–S163.

PICKENS, R. W. & JOHANSON, C. E. (1992) Craving:consensus of status and agenda for future research,Drug and Alcohol Dependence, 30, 127–131.

RACHMAN, S. & HODGSON, R. I. (1974) Synchrony anddesynchrony in fear and avoidance, BehaviourResearch and Therapy, 12, 311–318.

RANKIN, H., STOCKWELL, T. & HODGSON, R. (1982)Cues for drinking and degrees of alcohol depen-dence, British Journal of Addiction, 77, 287–296.

RANKIN, H., HODGSON, R. & STOCKWELL, T. (1983)Cue exposure and response prevention with alco-holics: a controlled trial, Behaviour Research andTherapy, 21, 435–446.

ROHSENOW, D. J. & MONTI, P. M. (1999) Does urgeto drink predict relapse after treatment? AlcoholResearch and Health, 23, 225–232.

SAMSON, H. H. (2000) The microstructure of ethanoldrinking: genetic and behavioral factors in the con-trol of drinking patterns, Addiction, 95 (suppl. 2),S61–S72.

SAYETTE, M. A., SHIFFMAN, S., TIFFANY, S. T.,NIAURA, R. S., MARTIN, C. S. & SHADEL, W. G.(2000) The measurement of drug craving, Addiction,95 (suppl. 2), S189–S210.

SHIFFMAN, S. (2000) Comments on craving, Addiction,95 (suppl. 2), S000–S000.

SINHA, R., KRISHNAN-SARIN, S., FARREN, C. &O’MALLEY, S. (1999) Naturalistic follow-up ofdrinking behaviour following participation in an al-cohol administration study, Journal of SubstanceAbuse Treatment, 17, 15–162.

SOBELL, M. B. & SOBELL, L. C. (1992) Time-LineFollow-back: a technique for assessing self-reportedalcohol consumption, in: LITTEN, R. Z. & ALLEN, J.(eds) Measuring Alcohol Consumption: psychosocial andbiological methods, pp. 41–72 (New Jersey, HumanaPress).

SWIFT, R. M. (1999) Medications and alcohol craving,Alcohol Research and Health, 23, 207–213.

TIFFANY, S. T. & CONKLIN, C. A. (2000) Cognitiveprocessing model of alcohol craving and compulsivealcohol use, Addiction, 95 (suppl. 2), S145–S153.

TIFFANY, S. T., CARTER, B. L. & SINGLETON, E.G.(2000) Challenges in the manipulation, assessmentand interpretation of craving relevant variables,Addiction, 95 (suppl. 2), S177–S187.

Documents

Craving research: future directions