Motivational interviewing and the incredible shrinking treatment effect

Abstract

McCambridge & Strang (2005) report that the effects of their brief intervention had disappeared by 12-month follow-up, and speculate that ‘deterioration of effect is the most likely explanation’. Before reflecting on reasons, I would first question whether their effects had actually ‘disappeared’. Their design compared a 1-hour motivational interview (MI) with an assessment-only control. Abstinence rates at 12 months in the MI group were double those of the control group for all three drugs: tobacco (20% versus 10%), alcohol (9% versus 2%) and marijuana (17% versus 8%). The authors estimate the odds ratio for cessation of any drug as 2.55, with a confidence interval of 1.23–5.25. Most practitioners would regard a doubling of quit rate as clinically meaningful, particularly after such a brief intervention. An odds ratio for which the confidence interval does not overlap 1.0 is also usually regarded as significant. Furthermore, among pretreatment smokers, cigarette consumption remained significantly lower at 12 months relative to baseline in the MI group, but not in the control group. From my perspective, these are surprising 12-month residual effects of one counseling session. Nevertheless, the between-group effects of brief MI do diminish over time. In a new meta-analysis of 72 clinical trials (Hettema, Steele & Miller, in press) we found that the mean effect size of MI averaged across all reported outcome variables was d = 0.77 up to 1 month later, 0.39 for 1–3 months, 0.31 at 3–6 months, 0.30 in months 6–12 and 0.11 at follow-ups longer than 12 months. The McCambridge & Strang findings are consistent with this trend. Why, then, does the effect of MI diminish over time? Seldom is it explained by a return of the MI group to baseline values. In most studies, within-group mean behavior change observed after MI is largely maintained across a year of follow-up, and does not resemble the reversal that might occur with, for instance, discontinuation of an antihypertensive medication. What happens more often is that the comparison group catches up with the intervention group over time. Combined with high variability, this tends to result in a gradual diminution of between-group effects across follow-up points. This is not unique to MI, of course, but is a common finding with other target problems, treatments and populations. In behavioral trials, control groups tend to get better over time. What continues to surprise me is that there is any 12-month effect of such a brief intervention. To me, it is the presence, not absence of enduring effects that begs for explanation. If the principal effect of a brief intervention is to produce a marked reduction in drinking 6 or 12 months earlier than might otherwise occur, that is still a good outcome. The adverse effects of alcohol and other drugs, particularly in young people, are frequently the consequence of acute events during and after use. Reducing risk sooner rather than later seems a worthy goal, albeit subject to cost–benefit analysis.