Attention-Deficit/Hyperactivity Disorder Treatment and Later Drug Use
Lydia Furman, MDDivision of General Academic Pediatrics
Rainbow Babies & Children’s Hospital
Cleveland, OH 44106, USA
To the Editor.—
I was fascinated with the article by Barkley et al1 on stimulant treatment for attention-deficit/hyperactivity disorder and risk of later drug abuse. The authors have tried to answer a question that worries many.
I am quite confused by their conclusion, which seems at clear odds with their data. They found, unfortunately, that treatment with stimulant medication in high school is significantly associated with the risk of ever trying cocaine. That this is a valid conclusion is supported by their finding that treatment with stimulant medication in childhood is associated with cocaine use in adulthood, although after adjusting for "disease severity" and the presence of conduct disorder, the association was no longer significant.
The authors work extraordinarily hard to explain to the reader why the first finding, which is statistically significant, is actually not real. The naïve reader may wonder why the authors are so invested in undermining their own carefully performed results.
It would make more sense to say that patients with conduct- disorder symptoms, or very severe attention-deficit symptoms, should not be treated with stimulant drugs because it appears that, in this subgroup, such treatment may be associated with an increased risk of cocaine use in adulthood. Or the authors could conclude that their population cannot answer the question that is being asked, because they did not compare a "treated with behavioral/psychotherapy plus stimulants" group versus a "treated with behavioral/psychotherapy without stimulants" group. In support of this conclusion, the authors do note that one limitation of their study is that the number of children with disorders not treated with stimulants is very small, limiting the power of the study.
It is tempting to conclude that the authors are biased in favor of treatment with stimulant medication, and that even the worrisome results obtained did not alter their viewpoint. The concluding paragraph is strongly worded in favor of such treatment and does not suggest additional study is needed, which actually seems a more appropriate conclusion if the data are examined dispassionately.
REFERENCE
- Barkley RA, Fischer M, Smallish L, Fletcher K. Does the treatment of attention-deficit/hyperactivity disorder with stimulants contribute to drug use/abuse? A 13-year prospective study.
Pediatrics.2003; 111
:97
–109
[Abstract/Free Full Text]
Russell A. Barkley, PhD
Office of Research
College of Health Professions
Medical University of South Carolina
Charleston, SC 29425, USA
In Reply.—
My colleagues and I appreciate this opportunity to address the accusations raised by Dr Furman about our research article.1 Dr Furman believes that our conclusions are at odds with the results that actually appeared in our article; a belief we find both puzzling and unfair. We concluded that there was no compelling evidence from our study (and 11 of 12 previous ones) that would support the conclusion that stimulant therapy for attention-deficit/hyperactivity disorder (ADHD) in childhood or adolescence is associated with risk for drug use or abuse. She disagrees, grounding her belief in the only significant finding to come out of all of our excessive analyses while seeming to overlook the place of this one finding within the larger context of our article. This finding was the association of having ever been treated with a stimulant in high school (expressed categorically, yes/no) and whether the individual had ever tried cocaine at least once (also expressed categorically). She asserts that this is "a valid conclusion" from our results and believes that it warrants the warning that patients with conduct disorder or severe ADHD should not be treated with stimulants. Surely this ignores the place of this one finding in the totality of all the other analyses we conducted not to mention the numerous prior research articles consistent with both our results and conclusions. And it is certainly not necessarily true that a significant result of a statistical test leads, prima facie, to a "valid conclusion," much less a clinical admonition not to treat. One must consider the various reasonable ways in which that finding can be interpreted as well as the simpler possibility that it occurred by chance alone.
As we noted, if a relationship existed between childhood stimulant therapy and later illicit drug use or abuse (cocaine in this case), it should have emerged as significant consistently across the various ways we chose to analyze (if not overanalyze) our data. Yet it did not. The simplest analyses we reported were the most robust and compelling against that conclusion. We found no significant association of duration of stimulant treatment with later frequency of use of any substance, including cocaine. We could (and perhaps should) have ended our article with that approach as there was no real need for additional analyses. The sensitization hypotheses predicted such a significant relationship, as we noted, and we did not find it. Case seemingly closed.
But we persisted with other analyses nonetheless, although it increased the risk of type I errors (finding a result to be significant by chance alone). We recast the analytic approach by comparing those who had and had not been treated with stimulants and examined whether they had ever tried various drugs as teens. Same result—no relationship to cocaine use. We compared these same groups for whether they had ever tried cocaine by adulthood. Same result. We also compared these same groups in the proportion qualifying in adulthood for and Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised drug dependence or abuse disorder including cocaine. No association with cocaine dependence or abuse. We then compared these groups in their frequency of drug use, including cocaine. Nada. Given that the one previous study by Lambert and Hartsough1 claiming to have found such a relationship had analyzed their data quite differently, we used their analytic approach as well, recasting our subjects again into those who had and had not been treated for a year or more with stimulants. No difference in the frequency of cocaine use. We then looked at whether treatment for a year or more increased the risk of ever trying cocaine as a teen. Zip. But a relationship was initially found with ever having tried cocaine by adulthood. Yet even that one became nonsignificant after controlling for an obvious confounding source for it, that being severity of conduct disorder. Eight tests, eight strikes. An additional analysis even found that treatment with stimulants for more than a year significantly reduced the risk of a cocaine abuse disorder, something Dr Furman fails to note. But that one too became nonsignificant after controlling for appropriate confounding variables. Still, we pressed on with 4 more analyses. Three more were not significant. We then come to the only significant association noted by Dr Furman, and even it was not significant if stimulant therapy was treated dimensionally as a duration measure (as it should be). Only when treated categorically (ever treated) was it associated with ever using cocaine, also treated categorically (yes/no). One significant and rather arcane effect in at least 12 different ways of examining that relationship. Forgive us for not finding this one effect to be "a valid conclusion" in this context. Dr Furman overlooks the very real possibility that given the >100 statistical tests conducted in this article, this one might be significant by chance alone (indeed 
5 would be so). To reiterate, context is everything in construing a statistical result as a "valid conclusion." We did not have to work "extraordinarily hard to explain to the reader why that one finding ... is actually not real." On the contrary, we had to work extraordinarily hard to even find it. As for our being biased in favor of stimulant treatment, let readers draw their own conclusion from the foregoing surely overanalyzed data set. No convincing association of stimulant therapy with drug use, dependence, or abuse emerged in our study.
REFERENCE
- Barkley RA, Fischer M, Smallish L, Fletcher K. Does the treatment of attention-deficit/hyperactivity disorder with stimulants contribute to drug use/abuse? A 13-year prospective study. Pediatrics.2003; 111 :97 –109
PEDIATRICS (ISSN 1098-4275). ©2003 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
