Ti

MU

a

ARRAA

KDACIDN

1

iihcttrasWt

UT

0h

Drug and Alcohol Dependence 125 (2012) 75– 80

Contents lists available at SciVerse ScienceDirect

Drug and Alcohol Dependence

jo u rn al hom epage: www.elsev ier .com/ locate /drugalcdep

he role of craving in AUDs: Dimensionality and Differential Functioningn the DSM-5

artina Casey ∗, Gary Adamson, Mark Shevlin, Adele McKinneyniversity of Ulster, N. Ireland, United Kingdom

r t i c l e i n f o

rticle history:eceived 16 November 2011eceived in revised form 5 March 2012ccepted 24 March 2012vailable online 17 April 2012

eywords:SM-5UDraving

RTifferential criterion functioningESARC

a b s t r a c t

Background: The dimensionality and the contribution of the proposed diagnostic criteria for the DSM-5model of alcohol-use disorders (AUDs) which will provide guidelines for future diagnoses have not beenexamined in depth.Method: Data from past year drinkers in the National Epidemiologic Survey on Alcohol and Related Con-ditions (NESARC), Wave 2 (n = 22 177) were analysed. Severity and discrimination of DSM-5 diagnosticcriteria was determined using a two-parameter logistic Item Response Theory model. Comparative anal-yses were conducted on the DSM-IV criteria. Differential functioning of the criteria across a number ofsocio-demographic variables was assessed.Results: The proposed criteria supported a unidimensional AUD model, with a factor loading range of0.625–0.914 (craving = 0.818). The model measured intermediate severity of AUDs with ‘reduced time onimportant/pleasurable activities’ and ‘failure to meet major role obligations’ criteria having the highestseverity and discrimination. Craving, endorsed by 4.2% of the general population, was in the mid-rangefor both severity (sixth) and discrimination (seventh). Significant measurement bias was found on fourcriteria across socio-demographic subgroups.

Conclusions: Application of the proposed DSM-5 changes yields an improved one-factor model of AUDover the existing DSM-IV model. Inclusion of a craving criterion improves the application of the diagnosticcriteria in a general population sample, covering a previously unrepresented problem area. Additionally,criteria measuring the milder end of the AUD continuum remain absent and some criteria exhibit mea-surement non-invariance. The AUD classification may require further refinement to enhance validity andreliability.

. Introduction

With earlier Diagnostic and Statistical Manuals (DSM) propos-ng alcoholism as a unitary construct, the DSM-III (APA, 1980)ntroduced the diagnostic labels of alcohol abuse (AA) and alco-ol dependence (AD). The DSM-III-R (APA, 1987) amended thisategorisation by creating a hierarchical relationship between thewo disorders and in the DSM-IV (APA, 1994) this hierarchical rela-ionship was retained. Diagnosis is based on a criterion count, AAequires 1 of 4 criteria and 3 of 7 criteria must be endorsed within

one-year period for an AD diagnosis. The hierarchical relation-

hip is reflected in that an AD diagnosis precludes an AA diagnosis.

ith the publication of each DSM edition it would be expectedhat some of the existent issues arising from previous editions

∗ Corresponding author at: Room MB205, School of Psychology, University oflster, Magee Campus, Co. Londonderry, BT48 7JL, N. Ireland, United Kingdom.el.: +44 2871 885129.

E-mail address: Casey-m4@email.ulster.ac.uk (M. Casey).

376-8716/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.drugalcdep.2012.03.019

© 2012 Elsevier Ireland Ltd. All rights reserved.

would be addressed and resolved. Two major issues that have beenquestioned are the current AA–AD dichotomy within Alcohol UseDisorders (AUDs) and their existent hierarchical relationship.

Research evidence is inconclusive on the dimensionality ofAUDs. A recent US population study (Hasin and Grant, 2004) foundthat of those who positively endorsed AD criteria, a third did notmeet an AA diagnosis, suggesting that the AA criteria may measure adifferent latent construct. Plus assessment in a cross-national studyshowed that when data were manipulated on the basis of responsesat either extremity, a two-factor model was found to be superior(Nelson et al., 1999). However, in an Australian general populationsample, Proudfoot et al. (2006) concluded that a one-factor modelwas the best fit to the data. Additionally, a study of data from theNESARC, Wave 1, found that with the exception of the AA legal cri-terion, both the AA and AD criteria formed a continuum of AUDseverity (Saha et al., 2006).

The existent hierarchical relationship proposes that in termsof severity, AD ranks higher than AA, thus, by default imply-ing that those criteria measuring AD are higher in severity thanthose comprising an AA diagnosis. However, empirical evidence has

7 hol D

c(lcur

seiot2

m(eIpao

ad2ateThmivc((

bDoEi(iO2aio

tcireatsttata

ah

6 M. Casey et al. / Drug and Alco

onsistently refuted this proposal. For example, Kahler and Strong2006) found that in severity terms, AA symptoms were not alwaysower than AD symptoms whilst Saha et al. (2006) found that allriteria constituted a severe AUD range on a dimensional contin-um, with hazardous use, an AA criterion, at the higher end of thisange.

The DSM criteria are assumed to be equally effective, that is,howing measurement invariance across different subgroups, how-ver the DSM-IV criteria have been shown to exhibit differentialtem functioning (DIF) of some criteria in subgroup comparisonsn a range of variables including gender, age, race and assessmentime frames (Kahler and Strong, 2006; Saha et al., 2006; Keyes et al.,010).

When the DSM classification of AUDs is compared to the otherajor taxonomy in use, the International Classification of Diseases

ICD), currently the ICD-10 (WHO, 1992), one major differencemerges, the ICD assess a craving item. Compared to the DSM-V, the ICD-10 has shown improved identification of AD in generalopulation studies (Rounsaville et al., 1993) and slightly higher reli-bility (Rounsaville, 2002) thought in part to be due to the inclusionf craving.

Craving, viewed as a multidimensional concept, has beenpproached from different theoretical posits and how it is bestefined and measured remain controversial topics (Rosenberg,009). Many theories of craving are not specific to alcohol butlso pertain to drugs, most falling under the umbrella of addictionheories (see Addolorato et al., 2005; Drummond, 2001; Lowmant al., 2000; Skinner and Aubin, 2010; Singleton and Gorelick, 1998;iffany, 1999 for reviews of major theories). The role of craving inuman addictive behaviour has been studied for some time, dueainly to the reported experiences of both addicts and the clin-

cians treating them. However researchers have questioned thealidity of a craving construct (Lowman et al., 2000), disagree on theonditions under which it arises and how it relates to consumptionDrummond, 2001) even suggesting it may be an epiphenomenonTiffany and Conklin, 2000).

Despite this, clinical and laboratory studies show craving toe correlated with severity of alcohol dependence (Anton androbes, 1998; Yoon et al., 2006), in some instances to be predictivef relapse subsequent to treatment (Bottlender and Soyka, 2004;vren et al., 2010; Flannery et al., 2003) whilst some argue that crav-ng reduction should be an immediate goal of addiction treatmentO’Brien, 2005). The development of pharmacological ‘anti-craving’nterventions, most commonly naltrexone (Volpicelli et al., 1992;’Malley et al., 1992; Anton et al., 1999) and acamprosate (Kranzler,000; Mason and Ownby, 2000; Mason, 2001) often used as andjunct to verbal therapies, have also been shown to be effectiven reducing craving in patients, resulting in improved treatmentutcomes.

Whilst little work has been conducted assessing craving in rela-ion to DSM-5 criteria, two studies assessed an adapted DSM-IVraving model, showing that craving had medium to good discrim-nation, was among the medium to high severity items and overallesults for a model including craving showed good fit (Cherpitelt al., 2010; Keyes et al., 2010). One study that has specificallyssessed proposed DSM-5 criteria, using data from the 1997 Aus-ralian National Survey of Mental Health and Well-Being, foundupport for a one factor AUD model. Assessment of the craving cri-erion showed that it was the fifth most prevalent criterion, hadhe third highest discrimination value and fell in seventh place on

severity spectrum (Mewton et al., 2011). This study also foundhat there was differential item functioning (DIF) present across

ge and gender on some of the criteria.

As a result of cumulative research in these areas and in anttempt to address such issues, the DSM-5 work group (APA, 2010)as recently made public provisional changes for inclusion in said

ependence 125 (2012) 75– 80

edition, to include the merging of the AA and AD categories, removalof the legal criterion and the addition of a craving criterion. Thisstudy, using IRT analyses, will examine the newly proposed DSM-5(2010) criteria for AUD in terms of each criterion’s discriminationand severity estimates and supplementary analyses will comparethe newly proposed criteria with the current criteria. Secondly thecriteria will be examined for differential functioning across a rangeof socio-demographic variables including gender, age, socioeco-nomic status (SES), education plus a family history of alcoholproblems.

2. Methods

2.1. Sample

Data for this study was drawn from the National Epidemiologic Survey on Alco-hol and Related Conditions (NESARC), Wave 2, conducted between 2004 and 2005by the National Institute on Alcohol Abuse and Alcoholism (Grant and Kaplan, 2005).This was a large US general population study (n = 34,653) of non-institutionalisedcivilians. The Wave 2 sample included data from 14,564 (42%) males and 20,089(58%) females, with an age range of 20–90 plus years (M = 45.9, SD = 15.9 years).The NESARC had a multistage design and oversampled Blacks, Hispanics and youngadults (18–24 years) with data then weighted to account for various strands of thestudy design, further detail available elsewhere (Grant et al., 2006). All weightingfactors for Waves 1 and 2 were included in our analyses. The participants in this studywere restricted to those who were past year (PY) drinkers N = 22,177 (male = 10,395,female = 11,782).

2.2. Measures

2.2.1. Diagnostic assessment. The NESARC used the Alcohol Use Disorder and Asso-ciated Disabilities Interview Schedule DSM-IV version (AUDADIS-IV; Grant et al.,2001). Reliability and validity studies of the AUDADIS-IV have been extensively doc-umented (Grant et al., 1995, 2003, 2008) for AUD diagnosis (good to excellent). Theproposed criteria were constructed using 34 symptom items (SI) from the AUDADIS,subsequently coded as: (C1) Roles (major roles obligations not met; (C2) Hazard(use in hazardous situations; (C3) Interpersonal (continued use despite friend/familyproblems; (C4) Tolerance (less effect from usual amount; (C5) Withdrawal (includ-ing relief/avoidance of; (C6) More/Longer (drank more/or for longer than intended;(C7) Control (tried to cut down or control amounts; (C8) Time (substantial timespent procuring, using or recovering from drinking; (C9) Activities (reduced timeon important/pleasurable activities; (C10) Health (continued to drink despite recur-rent physical/psychological problems). A positive endorsement of either of two SIquestions ‘Want a drink so badly that you couldn’t think of anything else?’ or ‘Feel avery strong desire or urge to drink?’ were used to measure craving, coded as (C11)Craving (strong desire to drink).

2.2.2. Statistical analysis. Mplus version 6.11 (Muthen and Muthen, 2011) was usedfor the analyses. Specifically in the IRT analyses, a 2 parameter logistic Item Responsemodel (2-PL IRM), allowing both discrimination and severity parameters to be esti-mated for each item (criterion) was used. IRT contains two somewhat interrelatedassumptions, one, that the underlying latent construct which the items measureis unidimensional, and secondly, that all the item indicators are locally indepen-dent. Local independence is assumed present if unidimensionality is established.To ensure unidimensionality, confirmatory factor analysis (CFA) was conductedusing the weighted least squares means and variance adjusted (WLSMV) estima-tor, best suited in the presence of binary indicators (0 = not endorsed, 1 = endorsed).Differential criterion functioning (DCF) was assessed using the DIFFTEST option inMplus, entering modification indices for each subgroup category and comparing thisto the more constrained model without modifications. For gender, females werethe referent (R) category, age was operationalized as a five level variable (20–29,30–39 [R], 40–49, 50–59, ≥60) and race also as five (White, non-Hispanic [R], Blacknon-Hispanic, American Indian/Alaska Native non-Hispanic, Asian/Hawaiian non-Hispanic, Hispanic any race). Education had six levels (no formal education, gradeschool, some/all high school, graduate equivalency degree (GED), bachelor’s degree(Grad), >bachelor’s degree [R], whilst SES had five categories (<$12,999, $13–24,999,$25–39,999, $40–79,999, >$80,000 [R]). Family history included parents, brothers,sisters and both maternal/paternal aunts, uncles and grandparents with no historyof alcohol problems as the reference category.

3. Results

3.1. AUD and criteria endorsement information

A total of 3432 (15.5%) of PY drinkers endorsed at least a min-imum of 2 criteria, qualifying under the DSM-5 proposed criteria

M. Casey et al. / Drug and Alcohol Dependence 125 (2012) 75– 80 77

Table 1DSM-5 criteria prevalence.

No. of criteria n PY sample (%) AUDsample (%)

2 1457 6.6 433 748 3.4 224 445 2.0 135 294 1.3 96 169 0.8 57 98 0.4 38 84 0.4 2.59 64 0.3 1.8

10 44 0.2 1.311 29 0.1 0.8

Table 2DSM-5 criterion endorsement patterns.

Criterion endorsed n PY (%) AUDsample (N)

AUDsample (%)

C6 More/Longer 3063 13.8 2376 69C7 Control 2812 12.7 1991 58C2 Hazard 2356 10.6 1743 51C5 Withdrawal 1725 7.8 1447 42C4 Tolerance 1517 6.8 1093 32C10 Health 1129 5.1 1053 31C11 Craving 924 4.2 842 25C8 Time 627 2.8 617 18C3 Interpersonal 532 2.4 518 15C1 Roles 239 1.1 237 7

atTpTnatb

Table 3IRT parameters of the DSM-5 AUD criteria.

Items Discrimination (SE) Severity (SE)

C1 Roles 2.301 (0.19) 2.501 (0.05)C2 Hazard 1.017 (0.04) 1.714 (0.04)C3 Interpersonal 1.986 (0.11) 2.199 (0.04)C4 Tolerance 0.805 (0.04) 2.413 (0.08)C5 Withdrawal 1.377 (0.05) 1.743 (0.03)C6 More/Longer 1.517 (0.05) 1.261 (0.02)C7 Control 1.188 (0.04) 1.512 (0.03)C8 Time 1.849 (0.09) 2.132 (0.04)C9 Activities 2.321 (0.22) 2.536 (0.06)C10 Health 1.917 (0.08) 1.820 (0.03)

depicted in Fig. 1. The item severity parameters of each criterionrepresent the level of AUD with which an individual has a 50%

Fe

C9 Activities 216 1 214 6

s having an AUD, 20.6% of the male sample (n = 2141) and 7.2% ofhe female sample (n = 1291). Table 1 shows criteria prevalence andable 2 shows the rank endorsement order for both the total sam-le and for those meeting an AUD diagnosis. As can be seen fromable 1, two criteria were most often endorsed; almost double theext most common of 3 and only a very small percentage endorsedll 11 criteria. Table 2 shows that the most frequently endorsed cri-

erion was C6 More/Longer and that C9 Activities followed closelyy C1 Roles were the least endorsed. C11, the new craving criterion,

ig. 1. ICCs for proposed DSM-5 AUD criteria. The severity parameter for a criterion mndorsement (y-axis) is 50%. The ICCs also depict the discrimination parameters of items

C11 Craving 1.399 (0.06) 2.116 (0.04)

p < 0.01 on all items.

was the seventh most prevalent criterion, endorsed by 25% of thosewith an AUD diagnosis.

3.2. Factor analysis and unidimensionality

To meet the IRT assumptions prior to analyses, CFA modelswere evaluated using three goodness of fit indices plus factor load-ings. General accepted interpretation of the fit measures is thatfor the root mean square error of approximation (RSMEA), val-ues <0.06 indicate good model fit, whilst for the Comparative FitIndex (CFI) and the Tucker Lewis Index (TLI) values >0.95 indicategood fit. Fit indices for a one and two factor model were identical(RSMEA = 0.014; CFI = 0.995; TLI = 0.994); standardised factor load-ings, with the exception of C1 Roles, were marginally higher in thetwo factor model (0.004–0.017) but with a correlation statistic of0.97, the one factor model was accepted as the best fit to the data.

3.3. IRT model parameters and item characteristic curves

The criteria severity and discrimination estimates are shown inTable 3 and item characteristic curves using these estimates are

probability of endorsing that criterion. Results show that most ofthe criteria of AUD assessed more accurately in the intermediate

ay be determined by identifying the point on the x-axis where the probability of in that those ICCs with steeper slopes have higher discrimination parameters.

78 M. Casey et al. / Drug and Alcohol Dependence 125 (2012) 75– 80

criter

sohtMlc

aAfscCf

3

a(htph

3

ssrarfimtctftsat

Race assessment found DCF between Whites and Blacks, withBlacks more likely to endorse both C4 Tolerance and C7 Control,and between Whites and Hispanics, with Hispanics less likely toendorse C2 Hazard. SES assessment showed that compared to those

Table 4Differential functioning of AUD criteria.

Variable Comparisongroup

Criterionendorsement

X2 value N

Ref. group

Gender Male C2 Hazard 130.91 10,395Female

Race Black C4 Tolerance 94.51 3426White C7 Control 89.21

Hispanic C2 Hazard 47.51 3852

SES <$12,999 C2 Hazard 30.34 1999>$80,000 C6 More/Longer 31.39

Age 40–49 years C7 Control 28.62 5250

Fig. 2. Proposed DSM-5

everity range, falling mainly within plus 1.3 standard deviationsf the mean. C9 Activities, C1 Roles and C4 Tolerance possessedigher severity values than the other criteria, indicating that onlyhose with high levels of AUD are likely to endorse these criteria. C6

ore/Longer was the least severe, indicating that those with lowestevels of AUD are likely to first endorse this criterion. The cravingriterion (C11) assessed in the middle (sixth) of the severity range.

Item discrimination parameters represent the items’ (criteria)bility to differentiate among individuals with varied levels of AUD.s can be seen from Table 2, the discrimination parameters ranged

rom 0.779 to 2.246. C9 Roles and C1 Activities criteria demon-trated higher discrimination abilities compared with the otherriteria. The lowest discriminatory criteria were C4 Tolerance and2 Hazard. Inspection of the craving criterion (C11) shows that it

alls around midway (seventh) in terms of discrimination.

.4. Criterion information curves

To establish where criteria provided the most informationcross the AUD latent trait continuum, criterion information curvesCICs) were calculated for each criterion as shown in Fig. 2. Theighest level of reliability for each criterion is where it peaks onhe continuum. C9 Activities followed by C1 Roles (highest peaks)rovided the most information, C4 Tolerance and C6 More/Longerad the lowest peaks, whilst C11 Craving had the fifth lowest peak.

.5. Comparative analysis: DSM-IV

To supplement our previous analyses we conducted IRT analy-es on the current DSM-IV criteria for comparative purposes. Theame symptom items were used to formulate the current crite-ia with two changes. The items assessing craving were removednd the item assessing legal problems, coded as Cleg, was added,esulting in 11 criteria (C1–C10 plus Cleg). Two CFA models weretted (details similar to that outlined previously), a one factor (1F)odel using the current AA and AD criteria collectively, and a

wo factor model (2F) with the AA (C1–C3, Cleg) and AD (C4–C10)riteria hypothesised as two separate factors, f1 and f2, respec-ively. Results (not shown) indicated that both the one and twoactor models were almost identical in comparative fit, whilst fac-

or loading ranges varied only minimally. The two factor modelhowed a high correlation between the AA and AD factors (0.96),ccordingly, the one factor model was considered the best fit tohe data.

ion information curves.

Using the one factor model, IRT analysis for severity parametersshowed that Cleg Legal (3.391), C9 Activities (2.581) and C1 Roles(2.535) had the highest severity estimates (in descending order)whilst the least severe (in ascending order) were C6 More/Longer(1.299), C7 Control (1.502) and C2 Hazard (1.738). C1 Roles (2.143)followed by C9 Activities (2.126) had the highest discriminationwhilst C4 Tolerance (0.796) followed by C2 Hazard (1.027) were theleast discriminating. Cleg was the third least discriminating (1.047),with the rank order of the other criteria remaining unchanged. Cri-terion information curves were also assessed and the order of theirinformation placement on the latent model continuum showed thatC1 Roles and C9 Activities had the highest peaks, a reversed order ofthe DSM-5 model, whilst C4 Tolerance had the lowest peak, similarto its performance in the newly proposed model, with Cleg havingthe third lowest peak.

3.6. Differential criterion functioning

Statistically significant DCFs (Table 4) were found across somegroupings of all variables except a family history of alcohol prob-lems, where no measurement non-invariance was evidenced.Males were more likely than females to endorse C2 Hazard.

30–39 years 60 plus years C7 Control 48.31 4360

Education High school C4 Tolerance 28.03 6420>Bachelor degree GED C4 Tolerance 20.91 5902

p < 0.001 on all X2 values.

hol D

ilaltw

4

pssuasfoecsdfifiw

alwshtAa

ctionwdttKimfKsrKtnlctrccd

ysn

M. Casey et al. / Drug and Alco

n the >$80,000, those in the lowest income group (<$12,999) wereess likely to endorse C2 Hazard and C6 More/Longer. With regard toge, those in the 40–49 and the ≥60 years of age groups were moreikely to endorse C7 Control than the referent group. Compared tohose with >bachelor’s degree, those at GED and high school levelsere more likely to endorse C4 Tolerance.

. Discussion

In this paper, we examined the performance of the newlyroposed DSM-5 criteria for AUDs within a large, representativeample of the U.S. population. To our knowledge, only one othertudy has conducted a similar type of analysis (Mewton et al., 2011)sing an Australian sample (1997) and whilst similar diagnosticssessment methodology was used (modified version of CIDI ver-ion 2.0) it was not identical to that employed in this study. Theew other studies available to date that assessed craving as partf the DSM diagnostic criteria included the legal criterion; how-ver these studies may be useful for comparative purposes. In theurrent study, the DSM-5 CFAs, conducted to establish unidimen-ionality showed that a one factor model was an excellent fit to theata. Additionally, the fit indices of this model showed an improvedt when compared to the DSM-IV model. This provides support

or the provisional amalgamation of the current AA-AD categories,nclusive of the above changes, recently suggested by the DSM-5

orking group (2010).The IRT results showed that the diagnostic criteria were arrayed

long a continuum of severity which assessed in an intermediateevel of AUD. Of interest is the performance of the C1 Roles criterion

hich had the second highest severity value and the C3 Interper-onal criterion (fourth highest), two of the current AA criterion. Asas been reported in other research, the placement of AA criteriaowards the higher end of the severity scale suggests that a currentA diagnosis is not prodromal to a current AD diagnosis (Kahlernd Strong, 2006; Saha et al., 2006).

As so little is known about the psychometric properties of theraving criterion as part of the diagnostic classification under inves-igation, results relating specifically to it are particularly germanen the current study. The performance of the craving criterionn the discrimination scale, in seventh place, shows that it haso more than a mediocre ability to differentiate among thoseith varying levels of AUD, suggesting that its inclusion on aiscriminating basis is largely redundant. This result differs fromhat of Mewton et al. (2011) who reported that craving had thehird highest discriminatory value in their Australian sample whilsteyes et al. (2010) reported it in second place. The finding that it is

n sixth place on the severity continuum shows that it is assessingore towards an intermediate level of severity, similar to its per-

ormance in the Mewton et al. study but again this differs from theeyes et al. study which reported it as having the second highesteverity estimate. The differences reported on the craving crite-ion’s performance in IRT analyses between this study and that ofeyes et al. are pertinent, as the Keyes et al. study used data from

he National Longitudinal Alcohol Epidemiologic Survey, anotherationally representative US adult survey. These differences are not

imited to craving, but are evident in the rank order of almost all theriteria and it is unclear why such differences are present, howeverhe NLAES data is now twenty years old and the differences may beeflective of time changes. Additionally the Keyes et al. study fittedraving as part of a DSM-IV model and it may be a partial or totalombination of these factors that are responsible for the observedifferences.

As craving is in effect replacing the legal criterion, which in anal-ses of the current classification was found to have the highesteverity estimate, the addition of the craving criterion will aid diag-osis of those within the mid-range of intermediate severity. Whilst

ependence 125 (2012) 75– 80 79

this is not detrimental to a diagnostic system, research on the cur-rent DSM-IV has shown that new criteria that captured those at themilder end of the severity spectrum would be most beneficial (Sahaet al., 2006; Martin et al., 2008).

When the endorsement patterns are considered, craving fell inseventh place, with 25% of those classified under the proposedDSM-5 criteria positively endorsing it. Whilst this is a relativelysmall number when other criteria are considered, for example C6(69%) and C7 (58%), it does translate that one in four of those whowould meet an AUD diagnosis under the proposed revisions had aproblem with craving symptoms. This may have important impli-cations for the design and delivery of treatment programmes ofthose with AUDs, as there is extensive existent literature showingthat craving plays a significant role in relapse. Several longitudi-nal studies have reported that participants who relapse at eithersix or twelve month follow up have higher scores of craving eitherat baseline and/or during treatment and post-treatment and thatcraving is a significant predictor of future relapse (Bottlender andSoyka, 2004; Evren et al., 2010). Additionally, one comparison studythat assessed the impact of targeting craving during treatmentfound that significantly less heavy drinking days were reported bythose who received the craving treatment sessions than those whodid not, both during the treatment period and twelve months later(Witkiewitz et al., 2011).

With the exception of a family history of alcohol problems, fourof the criteria exhibited substantial measurement bias across atleast one demographic subgroup in terms of severity. Some resultsextend the existing literature, the presence of criteria bias acrossSES and education levels, whilst some support previous studiesshowing DCF/DIF across age, race and gender (Kahler and Strong,2006; Saha et al., 2006; Keyes et al., 2010; Mewton et al., 2011).The absence of craving differential functioning in this study is con-trary to previous findings (Keyes et al., 2010; Mewton et al., 2011)warranting further investigation. The information collected in theoriginal survey does not allow for causes of DCF to be examined,however results show some criteria are performing more poorlythan others and this is problematic in an international classification.

Some limitations apply to this study. The aim of the study wasto examine the newly proposed AUD DSM-5 diagnostic classifica-tion and to this end we utilised data drawn from a U.S. population,which we acknowledge is not representative of all drinking cul-tures. However the findings can be generalised to cultures withsimilar drinking patterns and will be informative for studies ofcraving assessment in alcohol ‘comparative culture’ studies wherepatterns mentioned above are different. Measures of past yearexperiences were used in the study in keeping with the require-ments of the DSM diagnostic system and whilst this may reducerecall bias, some measure of it may still remain. Craving was oper-ationalized using two SI questions and given the differences inperformance that appear to be linked to how craving is operational-ized and assessed, the study does not claim to have captured, inits entirety, the extremely complex subjective experience that iscraving, which appears to incorporate phenomenological, biolog-ical, cognitive and affective states as well as being influenced byenvironmental factors. Future research for this area could focus onthe development of a brief measurement of craving that better cap-tures these multi-dimensions, improving the discriminatory abilityof a diagnostic craving criterion.

In summary, our analyses have shown that the proposed crav-ing criterion fits well into a unidimensional AUD construct whichdoes not include a legal criterion. A craving criterion will enhancediagnosis of those falling in the intermediate severity range of AUD,

but adds little to the existing discrimination level of the collectivecriteria. With a 25% endorsement level in those with a potentialAUD diagnosis, craving improves the application of the diagnos-tic criteria, covering a previously unrepresented problem area and

8 hol D

iisuts

R

C

ppeyDA

C

A

aadwp

R

A

A

A

A

A

A

A

B

C

D

E

F

G

G

0 M. Casey et al. / Drug and Alco

dentification of this subsample may have important implicationsn terms of treatment delivery and outcomes. The analyses alsohowed that criteria measuring the milder end of the AUD contin-um remain noticeably absent. In addition, the identification of cri-eria measurement non-invariance across demographic subgroupshows further refinement of the proposed criteria is required.

ole of funding source

Nothing declared.

ontributors

G. Adamson and M. Shevlin designed the study and wrote therotocol. M. Casey conducted the literature searches, summaries ofrevious related work and wrote the manuscript draft, with DCFxcepted. M. Casey and G. Adamson undertook the statistical anal-sis, with DCF excepted. A. McKinney and M. Shevlin undertook theCF literature search, statistical analysis and DCF manuscript draft.ll authors contributed to and have approved the final manuscript.

onflict of interest

No conflict declared.

cknowledgements

This study was funded as part of a PhD placement through Department of Education and Learning (DEL) Northern Irelandward to the corresponding author. DEL had no further role in studyesign; in the collection, analysis and interpretation of data; in theriting of the report; or in the decision to submit the paper forublication.

eferences

ddolorato, G., Leggio, L., Abenavoli, L., Gasbarrini, G., 2005. Neurobiochemical andclinical aspects of craving in alcohol addiction: a review. Addict. Behav. 30,1209–1224.

merican Psychiatric Association, 1980. Diagnostic and Statistical Manual of MentalDisorders, 3rd ed. American Psychiatric Association, Washington, DC.

merican Psychiatric Association, 1987. Diagnostic and Statistical Manual of MentalDisorders, 3rd ed., text rev. American Psychiatric Association, Washington, DC.

merican Psychiatric Association, 1994. Diagnostic and Statistical Manual of MentalDisorders, 4th ed. American Psychiatric Association, Washington, DC.

merican Psychiatric Association, 2010. DSM-5. http://www.dsm5.org/ProposedRevisions/pages/proposedrevision.aspx?rid=452 (accessed 20.02.10).

nton, R.F., Drobes, D.J., 1998. Clinical measurement of craving in addiction. Psychi-atr. Ann. 28, 553–560.

nton, R.F., Moak, D.H., Waid, L.R., Latham, P.K., Malcolm, R.J., Dias, J.K., 1999.Naltrexone and cognitive behavioral therapy for the treatment of outpa-tient alcoholics: results of a placebo-controlled trial. Am. J. Psychiatry 156,1758–1764.

ottlender, M., Soyka, M., 2004. Impact of craving on alcohol relapse dur-ing, and 12 months following, outpatient treatment. Alcohol Alcohol. 39,357–361.

herpitel, C.J., Borges, G., Ye, Y., Bond, J., Cremonte, M., Moskalewicz, J., Swiatkiewicz,G., 2010. Performance of a craving criterion in DSM alcohol use disorders. J. Stud.Alcohol Drugs 71, 674–684.

rummond, D.C., 2001. Theories of drug craving, ancient and modern. Addiction 96,33–46.

vren, C., Cetin, R., Durkaya, M., Dalbudak, E., 2010. Clinical factors associated withrelapse in male alcohol dependents during six-month follow-up. Behav. Cogn.Psychother. 20, 14–22.

lannery, B.A., Poole, S.A., Gallop, R.J., Volpicelli, J.R., 2003. Alcohol craving predictsdrinking during treatment: an analysis of three assessment instruments. J. Stud.Alcohol Drugs 64, 120–126.

rant, B.F., Dawson, D.A., Hasin, D.S., 2001. The Alcohol Use Disorder and AssociatedDisabilities Interview Schedule-DSM-IV Version. National Institute on AlcoholAbuse and Alcoholism, Bethesda, MD.

rant, B.F., Dawson, D.A., Stinson, F.S., Chou, P.S., Kay, W., Pickering, R., 2003.The Alcohol Use Disorder and Associated Disabilities Interview Schedule-IV

ependence 125 (2012) 75– 80

(AUDADIS-IV): reliability of alcohol consumption, tobacco use, family history ofdepression and psychiatric diagnostic modules in a general population sample.Drug Alcohol Depend. 71, 7–16.

Grant, B.F., Dawson, D.A., Stinson, F.S., Chou, P., Dufour, M.C., Pickering, R.P., 2006.The 12-month prevalence and trends in DSM-IV alcohol abuse and dependence.Alcohol Res. Health 29 (2), 79–91.

Grant, W.J., Goldstein, R.B., Chou, S.P., Smith, S.M., Saha, T.D., Pickering, R.P., Daw-son, D.A., Huang, B., Stinson, F.S., Grant, B.F., 2008. The Alcohol Use Disordersand Associated Disabilities Interview Schedule (AUDADIS-IV): reliability of newpsychiatric diagnostic modules and risk factors in a general population sample.Drug Alcohol Depend. 92, 27–36.

Grant, B.F., Harford, T.C., Dawson, D.A., Chou, P.S., Pickering, R., 1995. The Alcohol UseDisorder and Associated Disabilities Schedule (AUDADIS): reliability of alcoholand drug modules in a general population sample. Drug Alcohol Depend. 39,37–44.

Grant, B.F., Kaplan, K.D., 2005. Source and Accuracy Statement for the Wave 2National Epidemiologic Survey on Alcohol and Related Conditions (NESARC).National Institute on Alcohol Abuse and Alcoholism, Rockville, MD.

Hasin, D.S., Grant, B.F., 2004. Co-occurrence of DSM-IV alcohol abuse in DSM-IValcohol dependence: results from the National Epidemiologic Survey on Alcoholand Related Conditions on heterogeneity that differs by population subgroups.Arch. Gen. Psychiatry 61, 891–896.

Kahler, C.W., Strong, D.R., 2006. A Rasch model analysis of DSM-IV alcohol abuseand dependence items in the National Epidemiological Survey on Alcohol andRelated Conditions Alcohol. Alcohol. Clin. Exp. Res. 30, 1165–1175.

Keyes, K.M., Krueger, R.F., Grant, B.F., Hasin, D.S., 2010. Alcohol craving and dimen-sionality of alcohol disorders. Psychol. Med. 41, 629–640.

Kranzler, H.R., 2000. Pharmacotherapy of alcoholism: gaps in knowledge and oppor-tunities for research. Alcohol Alcohol. 35, 537–547.

Lowman, C., Hunt, W.A., Litten, R.Z., Drummond, D.C., 2000. Research perspectiveson alcohol craving: an overview. Addiction 95 (Suppl. 2), S45–S54.

Martin, C.S., Chung, T., Langenbucher, J., 2008. How should we revise diagnosticcriteria for substance use disorders in the DSM-V. J. Abnorm. Psychol. 117,561–575.

Mason, B.J., 2001. Results of the multicenter study of acamprosate in the treatmentof alcoholism. Biol. Psychiatry 49 (Suppl. 8), S77.

Mason, B.J., Ownby, R.L., 2000. Acamprosate for the treatment of alcohol depen-dence: a review of double-blind, placebo-controlled trials. CNS Spectr. 5,58–69.

Mewton, L., Slade, T., McBride, O., Grove, R., Teeson, M., 2011. An evaluation of theproposed DSM-5 alcohol use disorder criteria using Australian national data.Addiction 106, 941–950.

Muthen, L.K., Muthen, B., 2011. Mplus (Version 6.11). Muthen and Muthen, LosAngeles, CA.

Nelson, C.B., Rehm, J., Ustun, T.B., Grant, B., Chatterji, S., 1999. Factor structures forDSM-IV substance disorder criteria endorsed by alcohol, cannabis, cocaine andopiate users: results from the WHO reliability and validity study. Addiction 94,843–855.

O’Brien, C.P., 2005. Anticraving medications for relapse prevention: a possible newclass of psychoactive medications. Am. J. Psychiatry 162, 1423–1431.

O’Malley, S.S., Jaffe, A.J., Chang, G., Schottenfeld, R.S., Meyer, R.E., Rounsaville, B.,1992. Naltrexone and coping skills therapy for alcohol dependence: a controlledstudy. Arch. Gen. Psychiatry 49, 881–887.

Proudfoot, H., Baillie, A.J., Teeson, M., 2006. The structure of alcohol dependence inthe community. Drug Alcohol Depend. 81, 21–26.

Rosenberg, H., 2009. Clinical and laboratory assessment of the subjective experienceof drug craving. Clin. Psychol. Rev. 29, 519–534.

Rounsaville, B.J., 2002. Experience with ICD-10/DSM-IV substance abuse disorders.Psychopathology 35, 82–88.

Rounsaville, B.J., Bryant, K., Babor, T.F., Kranzler, H., Kadden, R., 1993. Cross sys-tem agreement for substance use disorders: DSM-III-R, DSM-IV and ICD-10.Addiction 88, 337–348.

Saha, T.S., Chou, S.P., Grant, B.F., 2006. Toward an alcohol use disorder continuumusing item response theory: results from the National Epidemiologic Survey onAlcohol and Related Conditions. Psychol. Med. 36, 931–941.

Singleton, E.G., Gorelick, D.A., 1998. Mechanisms of alcohol craving and their clinicalimplications. Recent Dev. Alcohol. 14, 177–195.

Skinner, M.D., Aubin, H., 2010. Craving’s place in addiction theory: contributions ofthe major models. Neurosci. Biobehav. Rev. 34, 606–623.

Tiffany, S.T., 1999. Cognitive concepts of craving. Alcohol Res. Health 23, 215–224.Tiffany, S.T., Conklin, C.A., 2000. A cognitive processing model of alcohol craving and

compulsive alcohol use. Addiction 95 (Suppl. 2), S145–S153.Volpicelli, J.R., Alterman, A.I., Hayashida, M., O’Brien, C.P., 1992. Naltrexone in the

treatment of alcohol dependence. Arch. Gen. Psychiatry 49, 876–880.Witkiewitz, K., Bowen, S., Donovan, M., 2011. Moderating effects of a craving inter-

vention on the relation between negative mood and heavy drinking followingtreatment for alcohol dependence. J. Consult. Clin. Psychol. 79, 54–63.

World Health Organisation, 1992. The ICD-10 Classification of Mental and

Behavioural Disorders. Clinical Descriptions and Diagnostic Guidelines. WorldHealth Organisation, Geneva, Switzerland.

Yoon, G., Kim, S.W., Thuras, P., Grant, J.E., Westermeyer, J., 2006. Alcohol craving inoutpatients with alcohol dependence: rate and clinical correlates. J. Stud. AlcoholDrugs 67, 770–777.