INTRODUCTION

When I was invited to serve as the discussant for this technical review, I was asked to attune my ear, which is accustomed to alcoholism treatment research, and comment on what I heard from an outsider’s perspective. After listening to these 2 days of papers, I feel not at all like an outsider, for there are many familiar issues and problems for me here.

COMMONALITIES IN ADDICTIVE BEHAVIORS

Obviously, there are many overlaps between alcohol and other drug abuse. Our clients, in fact, do not seem to realize that there are two separate Institutes. We treat and study substantially overlapping populations. It is rare these days to find a client who has problems only with alcohol, and we have heard here that perhaps half of methadone-maintained people have active drinking problems.

Relapse is a familiar phenomenon to us all (as Dr. Marlatt’s writings have emphasized), as is the issue of impaired control of behavior.

The etiology of alcohol and other drug problems is clearly complex, involving biological, psychological, and social factors and (some of us would add) spiritual dimensions. The papers presented here suggest, not surprisingly, that the general treatment strategies that work well with drug problems resemble those with demonstrated efficacy for alcohol problems.

We even seem to make the same mistakes in treatment and research. I had rather hoped to find that the treatment of drug abuse is less mired in a dispositional disease model, but I see the same wide gap between science and practice that has plagued the alcohol field. The popular disease model posits that addicts (or alcoholics) are qualitatively different from normal human beings, not only in their behavior but in genetics, physiology, and character, and that this is why they have the problems they do.

In this way, the dispositional disease model is oddly like the moral model that creates “them” and “us.” In a recent dissertation in our lab, Moyers (1991) studied the factor structure of treatment providers’ beliefs about alcoholism; she found a robust first factor reflecting all of the traditional beliefs of the disease model.

The moralistic items included on the questionnaire (e.g., “Alcoholics are liars and cannot be trusted’) also loaded on this primary factor, as did characterologic attributions. The essence of the factor seemed to be that alcoholics are all like each other but different–biologically, genetically, morally, and characterologically– from normal human beings.

The papers presented here, in contrast, suggest that drug abusers are fundamentally like other people except that they use drugs and suffer the consequences. This is the same picture that emerges with alcohol. “Alcoholics” are as unique and different from one another as snowflakes.

No replicable prealcoholic personality has been found after half a century of searching for it (e.g., Vaillant 1983), and behavioral precursors are limited to the same childhood conduct and school problems that are related to drug abuse more generally (Miller and Brown 1991). People with alcohol and other drug problems do not respond any more favorably to being confronted than do the rest of us (Miller et al., in press).

DRUG USE AS BEHAVIOR

Another familiar picture from this conference is that drug use-even with supposedly “out of control” drugs and people–responds to operant contingencies. This was demonstrated in the early 1970s with alcoholics (for a review, see Heather and Robertson 1981). Drug use is, first and foremost, behavior, shaped by and responsive to principles of learning such as reinforcement, punishment, classical conditioning, and modeling.

This is one of the messages that I seek to convey to psychologists and other mental health professionals: “Alcohol and drug abusers are not Martians, governed by different laws of behavior and requiring referral only to an initiated inner circle of experts with unique knowledge and techniques. Everything that you have learned in your training is directly applicable because use is behavior, and users are people like everyone else.”

In 10 years, I hope, this statement will be seen as an uninformative and obvious observation. Today, in the United States, it remains embarrassingly controversial.

A common behavioral strategy for which I have heard many applications here is successive approximation.

Dr. Grabowski advocated reinforcing successive reductions in drug use, and Dr. Marlatt pointed out, in arguing for a harm reduction perspective, that steps in the right direction are just that. Dr. Schuster also reinforced this point: if an addict gives up shooting heroin, prostitution, crime, and smoking crack cocaine but continues to use marijuana, both the individual and society benefit nonetheless.

Again, one hopes that readers in 10 years will ask how anyone could not see this. In a recent article (Miller and Page 1991), we described a variety of “warmturkey” alternatives for clients unwilling to accept immediate, permanent, cold-turkey abstention.

These include a trial period of abstinence (sobriety sampling), gradual fading of dosage toward abstinence, or a trial span of moderation to reduce or eliminate problems and dependence.

In a long-term followup study (Miller et al. 1992), we found that drinkers treated with a goal of moderation more often opted, in the long run, for abstinence-about twice as often, in fact, as those maintaining problem-free drinking.

Even though the practitioner’s ultimate goal for the client may be abstinence, insisting on immediate, total, and permanent abstention from all psychoactive drugs is not necessarily the most effective way to achieve that goal.

In a way, this has been acknowledged more readily in drug abuse treatment (e.g., methadone and nicotine substitution) than in alcoholism treatment. One speaker here did advocate excluding from treatment those clients who say they want only a drug holiday–a view at odds with a harm reduction approach that provides people with whatever treatment and degree of improvement they are ready to accept.

Such flexibility in treatment goals may lead to better retention and improved outcomes (Sanchez-Craig and Lei 1986). Several examples were provided here that behaviors commonly believed to be nearly intractable or to lie beyond the person’s conscious control behave as operants and respond to reinforcement contingencies: cocaine use, illicit drug use to supplement methadone, and the parasuicidal and therapy-disrupting behavior of borderline patients.

In Dr. Higgins’ data, I saw a suggestion that the community reinforcement approach was less successful with those who might be judged to have the most impaired control: daily, heavy cocaine users.

Perhaps we will find parameters of drug use that predict differential modifiability by operant and volitional efforts (cf., Miller and Brown 1991). It is likewise clear that the social environment more generally exerts strong influence on drug use. Dr. Higgins reported that employment and spouse involvement–two longstanding predictors of alcohol outcomes–were prognostic of cocaine outcomes as well.

Dr. Henggeler provided a causal model in which neighborhood and peers exert separate and direct effects on the likelihood of adolescent drug use. Clearly, we should not underestimate the “social” in “psychosocial.”

Skill training strategies, as I indicated in my review, have accounted for at least half of the treatment methods with demonstrated efficacy for alcohol problems.

Here I note the findings of Dr. Childress that active, not passive, coping strategies reduced craving. Dr. Higgins’ encouraging success with the community reinforcement approach, which relies heavily on skill training, mirrors its reported strong effects with alcohol abuse and dependence.

BEHAVIORAL INTERVENTION RESEARCH

Although the National Institute on Drug Abuse (NIDA) and National Institute on Alcohol Abuse and Alcoholism (NIAAA) have sponsored both biomedical and psychosocial intervention research, I am struck by the seemingly different patterns that have arisen–surely from historical and political factors rather than from inherent differences in our populations.

In prevention research, the alcohol field has emphasized “the quest for the test”–the search for pathognomonic biological factors that differentiate alcoholics from normal people. The preventive implications of this approach are limited at best, implying a program of identification and exhortation of alcoholics possessing the tainted gene or physiology. NIDA, in contrast, seems to have emphasized, to the dismay of some researchers, a supply-and-demand reduction approach that relies on interdiction and psychosocial strategies.

The idea of limiting the supply of (or access to) alcohol in our society, however, is difficult to sell to legislators, and the Norwegian Government’s official slogan, “Drink less alcohol,” seems as alien here as goat cheese. When it comes to treatment, oddly enough, the roles have been reversed.

NIDA has invested heavily in biomedical interventions for addicts, as shown by the proposed inauguration of a new $65 million initiative for medication development. The occasion of this conference witnesses a new interest at NIDA in strengthening behavioral treatment research.

At NIAAA, behavioral treatment research has a long and strong history, and psychosocial and biomedical studies share roughly equal proportions of the research budget. It was in a 1990 program announcement that NIAAA solicited new pharmacologic trials for alcoholism treatment. Again, we are dealing with heavily overlapping populations.

In treatment studies in our center, we now routinely include other drug use in our outcome measures. Alcohol use is obviously a significant factor in drug abuse treatment and trials. It is time, I believe, to explore how NIDA and NIAAA can cooperate in facilitating, coordinating, and funding treatment trials that increasingly bridge the missions of these Institutes.

It is time to develop consensus state-ofthe- art followup measures for alcohol and other drug use, the central dependent variables of our research. There are also many common issues in assessment methodology. Urine testing represents a gold standard in drug use assessment, a technology for which there is as yet no good parallel to verify recent drinking.

Consequently, alcohol treatment researchers have refined other methods for corroborating client self-reports–particularly collateral interviews, which have received curiously little attention in drug abuse treatment studies.

It remains to be seen how well the reports of significant others might serve to verify use or nonuse of cocaine, heroin, or marijuana, for example. The validity of collateral reports may vary from drug to drug. Heroin use, for example, is not as likely to be observed by nonusing spouses, and its effects may be harder to detect.

Cocaine runs, as Dr. Higgins commented, have more obvious effects observable by significant others. Marijuana use may be more readily observed directly by others. Collateral reports could serve as a further check on self-report, particularly in light of the known false positive rate (for recent drug use) inherent in urine testing due to residual traces of prior use.

NIDA and NIAAA researchers face similar assessment problems and could work profitably together to forge common treatment outcome measurement strategies. The growing awareness of multidiagnosis cases suggests that similar interface with the National Institute of Mental Health (NIMH) is overdue, and it is likely that quality alcohol and drug assessment will be of equal importance to research in other parts of the National Institutes of Health (NIH).

The reasons for joint measurement of alcohol and other drug dimensions are not limited to assessment concerns. Dr. Higgins reported that disulfiram, in essence, constitutes an effective treatment for cocaine use, specifically among those who also drink heavily.

A cocaine relapse may begin with drinking alcohol, suggesting a new meaning for the concept of “gateway” drugs. Similar attention should be paid to tobacco use, which is being explored in current NIAAA research as a correlate of alcohol use and relapse.

Suppression of only one drug, without paying attention to the impact on and effects of other drug use, makes little sense in a population in which polydrug abuse is normative.

Further, research should seek to disaggregate the effects of treatment modalities from settings. Residential and inpatient treatment have consumed the lion’s share of treatment dollars, despite the fact that nearly every literature review of the past two decades has concluded that such settings offer no overall benefit above that afforded by outpatient treatment (Annis 1985; Kiesler 1982; Miller and Hester 1986; U.S. Congress 1983).

If a specifiable subgroup does benefit differentially from more intensive and expensive care (a sensible possibility), this remains to be demonstrated, and the characteristics of this group should be documented and replicated.

Treatment process also deserves much greater attention. Dr. Borkovec’s call for research to delve deeply into treatment processes is well taken. Careful process research can help us understand how and why change occurs and elucidate the nature of the very problems we are treating.

Yet, such depth is not logically precluded in comparative clinical trials. It is possible to study two or more treatments deeply and simultaneously, gaining both process and relative outcome knowledge.

The comparison of different strategies is, I believe, entirely appropriate at this stage of knowledge development. The “horse race” pejorative unfairly oversimplifies the modem welldesigned clinical trial.

New insights into the nature of a disorder can arise from main effects and interaction (matching) effects as well as from the same within-treatment analyses that are possible in a singletreatment study.

The randomized trial also provides a level of causal inference not achieved readily through other designs. “Comparative” and “depth” are not alternative designs but different possible aspects within designs. Like process and outcome, both types of knowledge can be obtained from a well-designed trial.

Dr. Howard also was eager to hobble and humble randomized trials, or at least, by polemic attack, to inspire a defensive improvement in them. I surely favor the latter goal, and the methodologic criticisms he raised are worthy considerations, but they are worries for which remedies already exist.

His concern that randomization can result in nonequivalent groups on critical pretreatment variables is mitigated in larger samples, and it can be addressed by a variety of methods for ensuring balance while retaining essential randomization such as L.J. Wei’s (1978) urn randomization procedure.

The problem of data attrition is not unique to randomized trials, but, as several presenters here have shown, there are various effective ways to improve the retention of subjects in treatment and research. Dr. Howard’s worry that the distributions of experimental and control groups may overlap is real enough, but it is not properly solved by eliminating the control group from one’s design!

Noncomparative designs simply ignore the problem of relative outcomes. Having reviewed the alcoholism outcome literature over the past two decades, it is my experience that the relative contrasts of comparative trials yield a much more consistent picture across replications than do uncontrolled trials (cf., Holder et al. 1991; Miller and Hester 1980).

At the same time, I hasten to agree that randomized trials are not always the most appropriate method for generating new knowledge about treatment. Consider, for example, the need for new research with Alcoholics Anonymous (AA) and other 12-step groups, a priority highlighted by the National Academy of Sciences (1989).

Twelve-step group participation is too often regarded by researchers as a nuisance variable to be minimized.

Yet, the vast majority of alcohol and drug treatment programs in the United States espouse a 12-step philosophy and commend or require group attendance, and the 12-step fellowships serve far more individuals every year than do all treatment efforts combined. What happens to people over a course of involvement in such groups?

Why do some stay and others fail to return? What attracts people to 12-step fellowships? How, when, and for whom does change occur? These and many other questions can be answered by research (McCrady and Miller, in press), but the 12-step groups do not readily lend themselves to randomized trials because of their ubiquity.

Incidentally, contrary to statements made here, there is no official institutionalized resistance to research in AA. To the contrary, Bill Wilson wrote a memorandum encouraging AA members to participate in scientific research that continues to be circulated by the AA central offices.

My point though, for which this is but an illustration, is that there are behavioral outcome questions for which designs other than randomized trials are optimal. Both types of research are needed. On the issue of studying heterogeneous versus homogeneous samples, again it depends upon one’s purpose.

In seeking to discover client-bytreatment matching interactions, it is essential to start with a sample that is heterogeneous with regard to the predictor variables.

Recent Federal requirements to represent females and minority groups in study samples whenever feasible also favor heterogeneity. The limiting of a treatment study to a homogeneous sample may be warranted once there is already persuasive evidence that the intervention(s) under study will be differentially appropriate for that subsample.

We have little such knowledge regarding drug abusers. Had the NIMH collaborative trial been limited to “exogenous” depression, for example, the field would have been deprived of the important information that “endogenous” depression responds comparably to pharmacotherapy and cognitive therapy–a finding that in itself raises questions about the exogenous/ endogenous distinction and related etiologic assumptions.