DSM‐5 substance use disorder: how conceptual missteps weakened the foundations of the addictive disorders field

DSM-5, following the recommendations of its substance-related disorders workgroup, radically reformulated the diagnostic criteria for substance use disorder (SUD). The changes have generated considerable controversy and concern, and are here reviewed and evaluated. DSM-IV included two substance use disorders, substance dependence and substance abuse. Dependence attempted directly to capture the concept of addiction as a psychiatric disorder, understood as impaired-control use. (The label addiction itself was narrowly rejected in earlier DSM revisions due to its purportedly pejorative nature.) Dependence diagnosis required any three or more of seven possible symptoms indicating impaired-control substance use, including two physiological dependence symptoms (tolerance, withdrawal) and five behavioural symptoms (giving up activities, continuing use despite harmful physical or psychological effects, trying to stop but cannot, taking larger amounts than intended, spending much time obtaining and taking the substance). Abuse diagnosis represented problematic use and required any one or more of four possible substance-related symptoms, including hazardous use (most commonly, driving while intoxicated), interpersonal problems (e.g., arguing with spouse), failure to fulfill role obligations at work, school, or home, and legal problems. ‘Abuse’ has long been a questionable disorder category. The ‘dependence syndrome’ model of addiction as impaired control over use 1, 2 is generally accepted as the most plausible conceptual justification for understanding addiction as a true psychiatric disorder rather than as moral weakness, social deviance, or rational choice. Since DSM-III-R, DSM diagnostic criteria for substance use disorders have been based on this model. Consistent with the model, the original intention was for ‘dependence’ to be the sole addiction disorder category 3. However, pragmatic concerns about having a diagnostic label for anyone needing help with substance use overrode validity concerns, and the abuse category was included in successive DSM editions until DSM-5. DSM's ‘abuse’ criteria clearly conflated social deviance with addictive pathology. ICD's parallel but narrower ‘harmful use’ category rejects face-invalid abuse criteria such as hazardous use and insists on substance-caused physical or mental health problems, thus at least ensuring medical need for intervention. Two rationales have been used to justify the DSM's ‘abuse’ category despite its face invalidity. First, the dependence syndrome model portrays addiction as a ‘biaxial’ concept with two components, an impaired-control motivational dysfunction and a harmful consequence component involving negative social, psychological, and physical consequences of excessive use 1. The model asserts that harmful consequences by themselves do not imply addiction (e.g., heavy recreational drinking without alcoholism can cause liver cirrhosis or automobile accidents). Nonetheless, the components are sometimes interpreted as distinct categories of psychiatric disorder: impaired-control use (dependence/addiction) and a residual category of harmful use without dependence (abuse). However, as both the DSM-5 definition of mental disorder and the ‘harmful dysfunction’ analysis of mental disorder 4, 5 make clear, harm without an underlying dysfunction is not a medical disorder. The biaxial conception, properly understood, requi-res both impaired control and consequent harm as components of every diagnosis of substance use disorder 6 and thus provides no support for an ‘abuse’ disorder category. Second, it is commonly claimed that abuse is mild or prodromal dependence, thus justifying its disorder status. This empirical claim has been amply disconfirmed. Substance abuse—however worthy of intervention—does not predict later dependence at substantial rates and does not correlate with validators in the same pattern as dependence 7. Abuse is also highly contextually anchored; for example, a large percentage of abuse diagnoses are due to driving under the influence of alcohol, which occurs at increased rates among those of higher socioeconomic status simply because they have automobiles 8, 9. To its credit, DSM-5 finally eliminated the substance abuse category. However, rather than acknowledging that the abuse category lacks conceptual validity as an addictive disorder and placing it in the Z Codes for non-disordered problems often treated by clinicians, DSM-5 took another route. Much of the abuse category was assimilated to dependence, in keeping with the empirically discredited ‘mild or prodromal dependence’ account of abuse. The workgroup relied heavily in its deliberations on factor analyses and item response theory analyses of abuse and dependence symptom items, most of which fail to discriminate abuse symptoms from dependence symptoms as distinct clusters or syndromes. However, researchers have long noted that the DSM dependence items are poorly worded to draw the dependence/abuse distinction among the highly correlated dependence and abuse items 10. Nonetheless, rather than taking the conceptual implausibility of the technical results as confirming the need for rewording and reevaluating the items and strengthening the validity of the dependence criteria, the workgroup interpreted the technical results as refuting the idea of there being abuse that is not a form of dependence: ‘The “abuse category” has been eliminated from the proposed structure because of the lack of data to support an intermediate state between drug use and drug addiction’ (11, p. 867). The implication was that abuse symptoms (including, e.g., hazardous use, despite all the contrary considerations) could be considered variants of dependence symptoms and could validly be retained as impaired-control measures. Conceptual grounding in the dependence syndrome model was thus abandoned in favor of obeisance to a literal reading of the technical results. The consequences for the decision process are vividly presented in the following description by a workgroup member of a scene that could be right out of one of Solomon Asch's social psychology experiments in which group pressure causes an individual to affirm things that are plainly false: ‘I was one of the last holdouts against combining the abuse and dependence criteria because, to me, the dependence process and its consequences do seem conceptually distinct. However, as I looked around the table at my colleagues in our in-person meeting on this issue and considered the overwhelming abundance of evidence in favor of combining the abuse and dependence criteria, I found I no longer had grounds to hold to my position’ (12, p. 703). In fact, the evidence for homogeneity of dependence and abuse was not ‘overwhelming’. The failure of a technical method to clearly discriminate two things, or the demonstration that two things are intercorrelated in ways that allow them to fit a unidimensional mathematical model, is not a demonstration that two things do not exist. It is only a demonstration that the specific method of analysis using the specific selected measures does not discriminate them if they do exist—and in this case, we have strong independent clinical and personal evidence that the two entities, addiction and harmful use, do exist. Unidimensionality of item response analysis results can mean many things other than that there exists only one entity varying in severity 13, and no far-reaching conclusion can be drawn before alternative hypotheses are systematically tested, a task not undertaken by the workgroup. Moreover, there are many other sources of potential evidence that were not given weight. Observing that DSM-5′s decisions were based on a highly selective citing of evidence, Edwards concluded that the workgroup's reconceptualization ‘goes against clinical experience, which suggests that people can develop destructive and disruptive drinking behavior without clinical symptoms of dependence’ (14, p. 701), as for example in college students’ harmful binge drinking independent of addiction. The workgroup's premature embrace of specific technical analyses to override compelling clinical and conceptual realities brings to mind Wittgenstein's famous admonition: ‘In psychology there are experimental methods and conceptual confusion… The existence of the experimental method makes us think we have the means of solving the problems that trouble us; though problem and method pass one another by’ (15, p. 232; emphasis in original). DSM-5 thus moved the abuse criteria, with all their validity problems, into an expanded category renamed ‘substance use disorder’ (SUD). Only the ‘legal problems’ criterion was abandoned because it was rarely used and plainly a matter of social deviance, not pathology. Additionally, a new ‘craving’ criterion was added both because craving is intuitively a phenomenological counterpart to impaired control and because it is an attractive target of pharmacological intervention. With 11 possible symptoms rather than 7, the idea was that SUD represents an enlarged and more valid impaired-control category with dimensional structure ranging from mild to severe cases; 2 or 3 symptoms is ‘mild’, 4 or 5 ‘moderate’, and 6+ ‘severe’ SUD. Mechanically combining existing abuse and dependence criteria in SUD without reformulation exacerbated a problem of symptom criterion redundancy, weakening validity. Criteria overlap in meaning so that it is possible to satisfy multiple criteria and spuriously reach diagnostic threshold due to just one actual feature. For example, one can imagine satisfying the three symptoms of failure to fulfill role obligations, use despite social/interpersonal problems, and activities given up or reduced, all based on the same problem, lessened marital engagement due to time spent drinking with one's friends. Yet this one feature has low face validity for indicating impaired-use disorder. With so many more symptom options in SUD than in the former dependence category, it becomes much easier for an individual to meet three criteria. Thus, one might think that to maintain validity, the threshold number of symptoms needed for diagnosis would be raised to a higher number. This would be especially true because the added abuse symptoms appear to have less validity than the former dependence symptoms in identifying impaired-control dysfunction. Thus, the number of symptoms needed to have confidence that there is a dysfunction should be increased. The workgroup did the opposite; they lowered the number of symptoms required for diagnosis to 2. This is probably the most controversial substance-related change made by the DSM-5. If one considers SUD an expanded impaired-control addiction category analogous to dependence, then this change dramatically increases—indeed, more than doubles—the estimated prevalence of addiction. DSM-5′s SUD revision with its 2-symptom diagnostic threshold allows continued diagnosis of those with two abuse symptoms (e.g., drives under the influence and argues with spouse about it) and those with one abuse and one dependence symptom (e.g., has developed some tolerance to alcohol and drives under the influence). Moreover, diagnosis newly applies to those with two dependence symptoms (e.g., increased tolerance and drinks larger amounts than intended), dubbed by the literature as ‘diagnostic orphans’ as if they had been incorrectly abandoned rather than purposely placed below a plausible diagnostic threshold. However, studies show that these groups resemble former abuse cases more than dependence cases in prognosis. In terms of addiction as impaired control, the evidence is that probably they are mostly false positives, as validity evaluation of DSM-5 SUD suggests. DSM-5 has been heavily criticized for the lowered SUD diagnostic threshold. Before looking at the rationales offered for this change, one might ask: ‘What did this effort yield in terms of increases in diagnostic validity and clinical utility?’ Clinical utility for now appears weakened, primarily because virtually all treatment studies have used dependence samples. Consequently, nothing is known about what works with the much larger and more heterogeneous SUD category. Predictive studies have generally shown quite different long-term outcomes for members of the DSM-IV dependence category and the added SUD cases. Nor do the ‘mild’ SUD cases correspond to the former abuse cases, so that no simple translation is possible. More importantly for long-term scientific purposes, recent analyses present a markedly disappointing picture regarding validity 6. Looking at alcohol use disorder (AUD), which is by far the largest SUD category, DSM-5 AUD provides no gain at all in predictive validity or correlation with crucial validators relative to DSM-IV combined alcohol dependence and abuse. For example, DSM-5 AUD is not significantly different from DSM-IV combined dependence and abuse in levels of standard lifetime validators including episode duration, number of episodes, frequent heavy drinking, number of symptoms, suicidal ideation, service use, and family history. One-year validator results are similarly without validity gains. Alternatively, if one construes SUD as an expanded impaired-control category and compares DSM-5 AUD to DSM-IV alcohol dependence, then DSM-5 AUD represents a major loss of validity 6. DSM-5 AUD is significantly lower than DSM-IV alcohol dependence on every one of the above-mentioned validators except episode duration. Depending on the baseline of DSM-5 AUD no validity gain or validity relative to that the threshold of two or more criteria is that it an heterogeneous group or that those at low severity levels are not against the need to all cases including cases, for example, those in one this is a example of the cases of a disorder are true cases, and no one is to identifying mild cases of addiction. The is that the added are not mild cases of addictive disorder but rather not true cases of addictive disorder at as due to in the is not a mild of a some tolerance to alcohol over time from social and drinking and driving from social drinking is not generally mild addictive disorder. The workgroup's that we need to all cases including the mild the of validity. However, the was probably as a pragmatic that the added cases not be true cases of impaired control, for we be to them as disorders to help The problem with this pragmatic is that it validity as the criterion for category the of needed for more methods for those with true there was a category of the pragmatic was more compelling because the was in its category and scientific could on The present the with no valid target category for scientific The is that no one of criteria is of both identifying the and identifying all could use help for substance use. the group while as if one is identifying the former group a state of than before and an to scientific substance use disorders criteria a dimensional with no a in prevalence without the group a threshold for DSM-5 substance use disorders that would yield the with the prevalence of DSM-IV substance abuse and dependence disorders This an example of the or of from the there is no between a concept and its that the distinction can be drawn one The that the between addiction and is does not imply that the threshold is and can be placed to such as prevalence rates generated by other criteria that have been to be concept can have many cases, and the should be drawn with an to the of conceptually cases on the of the into and into and the are but and have cases that make these and despite the there are ways of the that would be If one can draw a one one has in the I is in what DSM-5 did with the concept of it placed the SUD threshold at 2 of 11 symptoms to the of earlier prevalence without to the face conceptual validity of the cases that would on of the new threshold. in of the 2-symptom threshold concerns in service of The lower SUD threshold can be as of DSM-5′s dimensional criteria that allow for diagnosis of prodromal and mild to The is the lower of the severity represents disorders at or the of disorder is to the of to justify the that the new criteria mild the workgroup as can them we can stop them from to the to need liver p. However, this justification is based on an that has shown to be that DSM-IV cases of abuse and dependence that are newly as SUD are an of dependence and have severe outcomes over To example the alcohol factor for liver cirrhosis heavy drinking for which is to at the of the SUD are about cirrhosis all not just of the DSM-5 the SUD label to an of the over and the or so by DSM-IV dependence 6. is no evidence that the cirrhosis cases out of the added group in that such diagnostic is a of the of cirrhosis or identifying those at substance-related symptoms and the a of SUD does not make SUD a disorder. can be severe (e.g., in binge drinking at Moreover, the diagnostic threshold should be the at which the is in a disorder (e.g., impaired rather than a In diagnostic criteria every symptom is given in meeting diagnostic threshold so of dysfunction from the former abuse criteria allows false to the diagnostic threshold is by of The workgroup the that the new criteria would to disorder that would be with the DSM-5 threshold is p. This the that increased construes the baseline as DSM-IV combined dependence and by DSM-5 SUD in The of have in mind a to DSM-IV dependence, which was generally considered the impaired-control category. many more are by DSM-5 SUD. The change from ‘dependence’ to ‘substance use disorder’ was by the claim that the label ‘dependence’ addiction with dependence’ of tolerance and The workgroup that physiological dependence is an consequence of the to some including and not by itself to imply addiction impaired-control This understanding was in and its dependence threshold which required at least one behavioural symptom of impaired control tolerance and were Nonetheless, the workgroup that the of ‘dependence’ with dependence’ was a problem in and for of physiological dependence that is as addiction. The potential for However, there is no evidence that or about for was generally based on this the of this change to of was by the workgroup's all the changes proposed by the DSM-5 this one is by presented to of with the for the treatment of p. This emphasis in the DSM-5 revision process represented a major the time the workgroup was these a problem of abuse of for was in the with addiction and a number of as as social into it had rarely been that it was to use to were clearly by some and about was not such a The DSM-5 workgroup was to this problem as they was If the had at a change, it would have been on and clinical but it did that DSM-5 lowered the SUD diagnostic threshold from 3 of symptoms to any 2 of 11 symptoms, including tolerance and Thus, the DSM-5 workgroup the problem that they were trying to they made it possible to SUD based on the two physiological dependence symptoms thus physiological with impaired-control use. This problem with the of of for to a diagnosis of DSM-5 substance use disorder with two criteria to the use of for medical purposes, including and in some can tolerance and as physiological used for medical these tolerance and in the of other criteria do not substance use disorders and should not be as that the of symptoms use of and other as in to that symptoms use should not be used as an of SUD a on the meaning of symptoms on is not a in the of and is as a category of DSM disorder. Moreover, disorder often to impaired-control use and so has been a of addiction. is only in the of being an as the to the of the However, many disorders are under such as a a of pressure is to a use the of disorder, and by this does not imply but it and it from SUD criteria that should not be about it in the of addiction in and this is far from the the of the tolerance and symptom criteria only SUD in substance use has two problems. First, it is conceptually as the workgroup tolerance and are not a use disorder the drug is then they are not a use disorder the same drug is not for the same by the they do not by themselves imply impaired-control use. Moreover, the workgroup's the that changes in can cause and For example, the criteria allow the that if a medical having only tolerance and symptoms from a state use is legal and to a state its use is and the same pattern of medical use with the same symptoms, the individual has from being non-disordered to being the results of by due to in the legal status of Second, this the DSM-5′s harmful dysfunction definition of mental disorder, which that the existence of disorder be a matter of deviance or with the only between the and the non-disordered under DSM-5′s SUD criteria be that one is taking the substance in with legal while the other is This makes DSM-5 diagnosis an of social control to disorder the that the definition of mental disorder was to The most implication of these conceptually of DSM-5 SUD diagnosis is with an abuse an individual be an SUD while taking is now more than an individual is tolerance and symptoms are from two behavioural symptoms to be rather than just the one required by The for such changes could not be issue is that there is a of to the workgroup's on physiological dependence as a potential of addiction diagnoses among taking It has long been that taking can to of many of the behavioural SUD symptoms as For example, a individual to much time and up other activities to maintain an if the only is in a It is also that craving with can without impaired-control the workgroup this by craving to with an individual being in the workgroup might have spent more effort systematically the symptom items to conceptual validity. To I have that the concerns about DSM-5 SUD diagnostic criteria are quite and the by the DSM-5 workgroup to concerns generally some including the of the abuse category and the of DSM-5′s substantial changes to SUD diagnosis fail to validity or clinical utility and markedly from criteria. In highly face-invalid criteria such as hazardous use and social to a category that is to a motivational dysfunction of impaired control, and then the diagnostic threshold to a that allows criteria by themselves to for DSM-5 has into the and controversial validity of SUD as a category of psychiatric disorder. The changes and the conceptual of the substance disorders as of analysis that there is such a as impaired control of use that psychiatric This is under from a of If this is then by all means we should on to a different of the of harmful substance use that we now as addiction. The problem is that the DSM-5 workgroup retained the impaired-control while diagnostic criteria in ways with that This the to the validity of the concept of addiction and of the impaired-control model. If the to some of DSM-5′s it a alternative to addictive disorder diagnosis that could the standard for addiction