In February 2000, our work group at the University of Maryland Pharmacoepidemiology Laboratory reported findings 1 that received widespread publicity 2 and generated strong public and professional 3,4 reactions. Calling it a “firestorm” may not be too strong a description of the activities of the subsequent months. The invitation from Dr. Gorski to participate in the Wellchild conference entitled “Too Young for ADHD?” was at least partly related to the widespread media reaction to our study. The goal of the present discussion is to review the findings of our study of psychotropic use in preschoolers and to stimulate further dialogue on the most critical follow-up questions based on the study findings.
Briefly, our study examined data on preschool-aged youngsters (2–4 yr old) from three large health systems—a midwestern Medicaid (MWM), a mid-Atlantic Medicaid (MAM), and a group model health maintenance organization (HMO)—for the years 1991, 1993, and 1995. The cross-sectional survey comprised data from computerized administrative records that reflect the usual prescribing practices of community-based physicians, enrolled youths, ambulatory care services, and paid prescription claims. In total for a given year, approximately 200,000 youths were potential recipients of these ambulatory services. The extent of psychotropic treatment was reported in terms of the prevalence for selected medications, which was defined as the number of youths receiving one or more prescriptions for the selected medication or medication group per 1000 enrolled youths.
The leading medications in this age group in 1995 for MWM youths were reported in terms of total prevalence (per 1000) as follows: stimulants (12.3), 90% of which represented methylphenidate (11.1); antidepressants (3.2); clonidine (2.3); and neuroleptics (0.9). A similar rank order occurred for MAM, although at a lower prevalence. The HMO prevalence was approximately half that of MWM except that its prevalence of clonidine was nearly 3 times greater than that of antidepressants (1.9 vs 0.7). Sizable increases in prevalence were noted between 1991 and 1995 across the three sites for clonidine, stimulants, and antidepressants, whereas neuroleptic use increased only slightly. Methylphenidate prevalence in 2- to 4-year-old children increased at each site: MWM, 3-fold; MAM, 1.7-fold; and HMO, 3.1-fold. Antidepressants increased approximately 2-fold; clonidine increased most dramatically, 7- to 28-fold. Neuroleptic use increased only slightly, ranging from a 1.2-fold to a 1.5-fold increase. These quantitative estimates suggest sizable increases in a relatively brief time interval (5 yr) for three of the four drug groups we studied. A few commentators suggested that the increases probably were justified to correct for previous “undertreatment” in the community. For the most part, the findings were used widely to criticize community practice because they suggested “inappropriate” treatment. The fact that the patients who were getting medications for emotional and behavioral disorders were only 2 to 4 years old, and that these medications have insufficient research evidence in this age group regarding efficacy, effectiveness, and safety, undoubtedly fueled the firestorm.
We suggest there is more to the dialogue than extremist positions, which are best left to high school debate clubs. The present public focus on stimulant prescription patterns provides us with an opportunity to bring “new knowledge” into the mainstream of scientific assessment that aids the development of a comprehensive model for the evaluation of drug therapy for long-term care. Our study findings brought to clinical readers’ attention a relatively little-used approach to documenting clinical practice patterns—prevalence data on the specifics of medication use in several community-based populations. Its novelty to the clinical science community is both its strength and weakness. Its strength is we have in one or two simple measurements (prevalence or trends) quantitative data that are evidence something dramatic has been happening that cannot be dismissed as unrepresentative or anecdotal. Its weaknesses include the lack of “benchmarks” against which to measure these changes and, most crucial, the lack of information about the diagnostic evaluations, as well as the health outcomes of the expanded pool of youths receiving these treatments. So at the very least, our findings are provocative and point to additional research questions—which questions we choose to focus on, will, as the saying goes, determine the answers.
Certainly, there are many viewpoints on the priority listing of the relevant research questions. The conference proceedings give me a welcome opportunity to crystallize my thinking in regard to our top priorities, which will be presented here as the five burning questions our findings sparked. I focus the questions on attention-deficit hyperactivity disorder (ADHD) and its treatment for all youths—not just preschoolers—because this condition is the most prevalent and has received the most scientific scrutiny of all emotional disorders in United States children. In the following sections, each question is followed by a brief discussion of its rationale, the goal of related efforts, and suggestions for further research applicable to clinical practice. Broad inclusion of consumer, educational, juvenile justice, and political interests and implications are either explicitly stated or implied within the issues that are raised.
Question 1: What are the variations in drug prevalence and utilization patterns for youths with attention deficit hyperactivity disorder (ADHD) and what do they mean?
Rationale: Little community-based epidemiologic research on mental health treatment takes place across the United States except for reporting from states to the Center for Mental Health Services of the Substance Abuse and Mental Health Services Administration, and this rarely includes drug therapy use. 5 Incomplete information on national patterns can be gleaned from federal surveys, for example, the Medical Expenditure Panel Survey (MEPS) (person-based longitudinal survey since 1996) and National Ambulatory Medical Care Survey (NAMCS) (physician-reported treatment visit data), but limited numbers make the reliability of these data inadequate to study drug treatment patterns in children and the data lack clinical detail. 6
Although our study used broader data sources than existing ones, we need a national surveillance system for clinical pharmacology to provide ongoing comprehensive data. This is not a new concept. Over two decades ago the Joint Commission on Prescription Drug Use convened a distinguished panel of national medical care experts who urged us to create a center for clinical pharmacology so that public health based research questions related to the extent of use, effectiveness, safety, and postmarketing surveillance of drugs could be undertaken. 7 Descriptive pharmacoepidemiology is the engine that can drive the whole process—showing us rates of use, trends over time, and generating hypotheses to address the research questions raised by inspecting population-based statistics on utilization.
A short list of the needs for routine descriptive data on the utilization of psychotropic treatments includes the following:
1. Sampling from a universe of health care coverage systems. Our study is particularly limited by the absence of data from employed families with commercial insurance, for example, Blue Cross and Blue Shield. After all, treatment varies according to socioeconomic status. Access to such proprietary data on health outcomes is quite limited and not easy to process when access is provided. Added pressures come from recent battles to restrict access to records because of patient confidentiality, which dovetail with the lack of incentives by proprietary interest groups to closely monitor the quality of care.
2. Creating the infrastructure to establish prevalence and practice patterns data in a routine, clinical, nonprofit reporting system is vital. Analysis of millions of health encounters is tedious, time-consuming, and requires considerable procedural work to “clean” the data before it can be used as a dataset. Variations in data sources across states and health systems are a major challenge to a system already overburdened by instability, mergers, limited access, and treatment discontinuities of all sorts caused by a managed-care model of health-service delivery. Furthermore, those analyses that have been done were driven by cost containment or medication promotional motivation, rather than by a clinical or public health concern.
3. Continuing medical education for practitioners in training, experienced clinicians, and academic researchers to understand the meaning of the variations identified in population-based studies. More than three decades have elapsed since John Wennberg brought statistical techniques to assess the appropriateness of surgical procedures. 8 Still, his disciples are continually struggling to apply the principles of health services research to more clinically complex situations. 9 During and after the current firestorm over medication for emotional disorders of youths, there were disparate responses but few calls for more outcomes research in community-practice settings. 10 Confusion reigns when academic specialists try to judge the meaning of the large-scale community medication prevalence data without benefit of a context in which to place this new knowledge. For example, a 10% methylphenidate treatment rate among youths of all ages was arbitrarily offered as the upper limit of appropriateness in a recent pediatric editorial. 11 Given the absence of standards for prevalence in regard to age, gender, diagnostic grouping, or other indications for use, it is difficult to see how this or any number could be used to assess appropriateness. Another problem in terms of meaningfulness relates to how we interpret small numbers. Some commentators of our findings focused on the low prevalence (2/1000 [0.2%] for clonidine) and inferred that it was acceptable because it was so small. This opinion ignores the 7- to 28-fold increase in a relatively short period of 5 years for a drug that lacks efficacy and safety information in this age group. It is a common human fallacy that we look at statistical probabilities and make biased inferences depending on the heuristic or belief we hold concerning the likelihood of uncertain events such as an adverse drug event. 12 For example, it may be easy to assume that infrequent events (e.g., 1/1000 or 1/10,000) are insignificant or not serious because of a lack of familiarity with them. 12 We conclude that the interpretation of drug prevalence and patterns will require routine monitoring of national treatment patterns and ongoing assessments to generate clinically nuanced hypotheses from the data.
Goal: The primary goal is to produce useful, comprehensive data on psychotropic prevalence according to host (e.g., age, gender, socioeconomic [education and income] status, race/ethnicity, other cultural factors, and medical history);environment (e.g., health insurance coverage, physician specialty, physician-reported diagnosis, school system, culture, and special population status [foster care/supplemental security income/state children’s health insurance program/low income]); and medicinal agent (including off-label use, promotional factors, concomitant use of more than one medication, dosing patterns, and duration of treatment).
Recommendation: Standardized pharmaceutical reporting (annual or biennial) based on a national sampling from usual practice settings is needed. The analyses should be conducted by public sector investigators who are funded by an independent, nonprofit agency, for example, a center for clinical psychopharmacology. Funding for these studies should be garnered from federal health agencies. A recent effort along these lines is the Center for Research and Therapeutics (CERTS), although the funding has been far below that which is needed to produce the surveillance system and the expertise that is needed.
Question 2: What are the outcomes of community-based treatment for youths with ADHD?
Rationale: The treatment of ADHD with psychostimulants has been subjected to more than 200 controlled clinical trials. Evidence for the efficacy of psychotherapies for ADHD is also well established 13 based on criteria met by 29 published studies and probably efficacious based on another six studies. The leading interventions are behavioral-parent training and behavioral modification in classroom settings. 14 Despite a substantial body of evidence of short-term efficacy on these treatments over the last two to three decades, unanswered questions about the use of combination therapy remained. These were taken up in the Multimodal Treatment Assessment (MTA) study, the largest and most sophisticated clinical trial for the assessment of emotional and behavioral treatment interventions in children. 15 The trial enrolled 579 youths randomly assigned to one of four treatment conditions (medication, behavioral therapy, combination, or community control) and assessed them at various points up to a 14-month endpoint. Nineteen dependent measures were assessed including both symptoms and associated impairment (e.g., parent-child relationships, social skills, and academic achievement). The major findings at the 14-month endpoint showed improvements in all four groups based on parent and teacher symptom ratings, specifically, in terms of effect sizes (SD units of improvement): .9 to 1.3 for behavioral therapy and community controls and 1.5 to 1.8 for the combined and medication groups.
Much has been written about the widely promulgated results of this important study. Distinct interpretations of the findings vary in terms of the added value of combination treatment 14 or whether it is worthwhile to evaluate response to behavioral therapy before initiating stimulants. Efforts to clarify the misimpressions of the study findings emphasized the following points:16
1. Medication was helpful in all groups. Nevertheless, the medication alone group actually received more medical contacts than is typical for the usual practice setting so the intervention represented more than medication alone.
2. Fourteen months is a longer evaluation period than previous studies but does not represent long-term effectiveness.
3. Combination therapy resulted in somewhat lower doses of medication. The average daily dose was 20% lower at 14 months among those receiving combined treatment (31 mg/d) than in the medication group (38 mg/d).
4. The behavioral intervention was delivered early in the study period, and drug therapy was delivered throughout the period. Thus, at endpoint the medication was active, whereas the behavioral intervention may have faded because the therapy had been completed 4 to 6 months earlier. 17 This factor may have disadvantaged behavioral therapy in the comparison. Stimulant medication is known to have only temporary effects in controlling symptoms, and other goals are vital to successful treatment. Dr. Vitiello, an MTA coinvestigator, emphasizes that “it is not known whether suppression of these symptoms in childhood results in better educational achievement or social, occupational, and mental health status.”18 The MTA has spawned a great deal of sophisticated analysis on the data, and their interpretation is the subject of continuing study among the various disciplines.
In the midst of the internecine war between the medication trialists and the psychosocial trialists came societal concerns raised by the finding of preschoolers’ increased use of stimulants. One major response by the National Institute of Mental Health (NIMH) was to answer a question about the efficacy of methylphenidate in preschoolers by conducting the Preschool ADHD Treatment Study (PATS). Irony abounds in The Washington Post headline “Scandal! They haven’t tested Ritalin on the children it’s prescribed for! Scandal! They’re going to test Ritalin on the children.”19 Most leading experts in government and academia see a clinical trial as the most needed and immediate agenda. Other experts do not see the further establishment of methylphenidate’s efficacy for symptom control as the most critical issue regarding ADHD in the preschool-aged population. 20 Instead, they argue that effectiveness 10,13 and the community-based context of our therapeutic interventions need to be researched. What is unfortunate is that the results of clinical trials receive major public attention whereas equally important concerns regarding community practice and long-term outcome receive a far more muted professional discourse.
To be fair, a comprehensive model of psychopharmacologic evaluation requires both types of information. Clinical trial data comprise Phase 3 information in the Food and Drug Administration’s (FDA) conceptualization of the drug development process (FDA website). However, the well-known limitations of clinical trials methodology (particularly volunteer bias, selection bias caused by exclusion and inclusion criteria, and artifacts resulting from the research design and environment) limit the extent to which findings generalize back to the target population they are intended to treat. Extending a clinical trial has been viewed by some as an effectiveness trial and is conducted as a Phase 4 add-on to a Phase 3 efficacy study, but the restrictiveness of the clinical trial remains. Proposals to conduct large simple trials to evaluate the long-term outcome of psychotropic treatment have been described, 21 but these appear to be larger and more naturalistic versions of extended clinical trials. Unfortunately, they will not produce the greater generalizability that is needed. Effectiveness of long-term treatment, as yet, has no good models because of methodological, cost, and infrastructure issues, although interesting efforts to produce these data from psychiatric populations come from professional association-sponsored practice research networks, for example, Pediatric Research in Office Settings (PROS) of the American Academy of Pediatrics, Ambulatory Sentinel Practice Network (ASPN) of the American Society of Family Medicine, 22 or the Practice Research Network (PRN) of the American Psychiatric Association. 23 These pioneering studies in the development of clinically rich treatment data for long-term care lead the way toward outcomes research. Although of value, they do not generalize to the larger community and suffer from volunteer bias and subjective time-consuming reporting.
What then is the burning question that these complex matters suggest? The burning issue is to produce Phase 4 information consistent with the FDA drug development model, that is, postmarketing studies on the outcomes of therapy in the usual practice setting. Treatment outcome studies should be designed and developed with the following criteria: (1) stem from the usual practice setting; (2) be representative of the major treatment populations in the community; and (3) incorporate computerized clinical encounter data that will quantify outcomes for a host of outcome measures, particularly the natural course of the disorder and treatments across time.
Goal: The aim is to produce studies that measure outcome for all interventions (drug therapy, psychosocial interventions, combined) from naturalistic data using research-validated measures to assess the following:
* Symptom control
* Functional improvement in three settings: home (family relations), neighborhood (social development, juvenile justice), and school (academic performance)
* Satisfaction with treatment
* Reasons for terminating treatment: sufficient improvement, adverse events, refusal/dropouts.
Recommendation: Assessment models need to be created within various settings to evaluate youths (e.g., with ADHD) in longitudinal studies emphasizing time in treatment and the complexity of drug therapy. School and community settings should be included, validation of physician-reported diagnoses should be undertaken, and a measure of general health ascertained.
Question 3. How is appropriateness best established for widely used, marketed ADHD medications?
Rationale: Appropriateness is a judgment based on integrating findings from empirical studies with accepted practice standards. Two recent studies garnered considerable attention when they took up the nettlesome question of the appropriateness of ADHD drug therapy in the usual practice settings that are the locus of expanded stimulant use. 24,25 In the first paper, 26 data were examined from the Methods for the Epidemiology of Child and Adolescent Mental Disorders (MECA) study in which 9 to 17 year olds and their parents were interviewed across four U.S. communities. Table 1 shows the results.
Among the whole sample, only 16 youths aged 9 to 17 were identified in 1992 as recipients of stimulant treatment. The estimate of 1.4% is unstable: The confidence interval gives 95% confidence, indicating that the rate may be as high as 2.3%. Among the youths assessed by lay interviewers using the Diagnostic Interview Schedule for Children (DISC), a structured version of the DSM-III-R, 66 met full criteria for ADHD (DISC-ADHD) and eight (12.1%) reported receiving a stimulant in the past 12 months. The bulk of the treatments reported for DISC-ADHD in this age group comprised school-based (24.2%) or office-based (31.8%) psychosocial treatments. These data were interpreted as reassurance that widespread overmedication is not happening: Only 1.4% of 9 to 17 year olds in 1992 received a stimulant. Community standards were adequate because on rediagnosis by the DISC in a community-based survey, only 12.1% of those meeting ADHD criteria received a stimulant, whereas a majority received school- or office-based psychosocial interventions.
Several limitations of the study design include the following:
1. Volunteer bias skews the data toward those willing to report on their families’ experience and limits generalizability.
2. Higher income respondents (one third with income in 1992 greater than $55,000 per year) may have access to psychosocial treatments (31.8% of those with DISC-ADHD) that probably do not apply to the majority of community-based youths.
3. The age of the surveyed youths excludes those most likely to be diagnosed and/or treated with stimulants in community settings, typically 6 to 8 year olds.
4. As a consequence of the age group selected, comparatively fewer subjects had the outcome of interest to ensure reliable findings. Moreover, the data contrast sharply with proportional relationships between clinician-reported diagnosis and the rate of stimulant prescriptions. In contrast with the 12% rate reported here, during the period from 1989 to 1996 national data on physician visits for ADHD among 5 to 14 year olds yielded stimulant treatment rates ranging from 62.6% to 76.6%. 27
The second study to address appropriateness comes from an experienced research team devoted to epidemiologic measures of children with developmental and emotional disorders. From longitudinal research on the development of psychiatric disorders and the need for mental health services in rural and urban North Carolina youths (Great Smoky Mountains Study), Angold et al 28 studied 4500 9 to 16 year olds; they then applied research diagnoses to this community sample recruited through the public school systems of 11 western counties using a household equal probability design. After an initial screening of parents for youth behavioral problems, those meeting the criteria were continued into a second phase in which 80% responded (n = 1422). Contacts occurred every 3 months by telephone and by four annual observations. A scale with published validity and reliability (Child and Adolescent Pyschiatric Assessment, CAPA) was used to assess symptoms that contribute to a wide range of diagnoses. Parent-reported ADHD diagnostic information was used to create the research diagnosis for this study (CAPA-ADHD). Parent and youth reports of symptoms were collected at each annual observation, but an incomplete number of teacher reports were only available at waves 2 through 4.
Table 2 shows the relationship between youths receiving stimulants according to the CAPA-ADHD grouping. When the diagnostic groups are examined proportionally, 57.5% (482/838) of those receiving stimulant prescriptions had neither full nor partial ADHD symptoms, and only 33.7% (283/838) met full criteria. The data are interpreted as evidence of inappropriate use because 57.5% of youths receiving a stimulant did not meet research diagnostic criteria for ADHD. 29
Similar to the previous epidemiologic study, this study has some design limitations:
1. The pattern of measuring diagnosis and drug therapy at various time points across a 4-year period creates questions about temporal relationships that make it difficult to judge appropriateness. If symptoms are controlled and the medication is being used for this purpose, should a child continue to meet full criteria for ADHD?
2. If only parent ratings are used to assess behavior, are we relying too heavily on biased ratings (either for or against medications) that may relate to compliant behavior rather than to broader measurements of social and cognitive development?
3. The 9- to 16-year-old age group misses the most prevalent age for stimulant treatments.
4. As was true of the previous report, systematic assessment of the behavior in the school environment is not included. In the final analysis, using epidemiologic study designs to assess appropriateness for chronic disorders is problematic. The primary limitation is rediagnosis in the absence of broad clinical assessment and a fuller medical accounting of the treatment history. As a consequence, we risk producing a feasible, more reliable, but artifactual information base according to the merits of the study design and selected criteria.
Several important findings did emerge from this epidemiologic study. For example, the prevalence of tics associated with stimulants (4 vs 0.4%) provides a rough estimate of the relative risk of tics based on lay observation of the disorder in a community sample. The implication for the purpose of family education and informed consent is valuable. Similarly, it is crucial to know that many 9- to 16-year-old youths were being treated with stimulants for other disruptive disorders, for example, conduct disorder (CD) or opposition defiance disorder (ODD). This represents important new knowledge concerning community practice patterns and begs for attention in regard to subsequent clinical trials of effectiveness.
What alternative, if any, could better address the important question of appropriateness? Wennberg’s commentaries on the appropriateness of treatments for cardiovascular disorder may apply here as well. He suggests that variations in treatment practice may not derive from bad quality but from underevaluated theory. 30 Thus, the variations in treatment practice illustrated by the Angold study are a call for an evaluation of the outcomes of community practices to answer questions about what behaviors are being treated without meeting diagnostic criteria and with what consequences (effectiveness in regard to improved functioning and safety). Community settings in both primary care and psychiatry should be studied as a source of clinical epidemiological data on these treatment practices. Building on the infrastructure proposed in Question 1, my recommendation for the study of appropriateness and outcome is to build ad hoc studies into the clinical settings from which routine reporting is conducted. With specialized infrastructure to assess social, biological, and academic functioning, it would be possible to measure appropriateness within a richer clinical data set. The studies can address off-label prescribing, drug combinations, long-term effectiveness, dosing issues, 31 and satisfaction with treatment and outcomes across the range of dimensions (symptoms, family and peer social functioning, and school adjustment). 32 An example of the application of field research, despite several limitations, can be found in our study of clozapine among early recipients of this important drug among New York State psychiatric inpatients. 33
Goal: The outcomes of care for ADHD need to be measured within representative treatment settings to generalize to “real world” youths, their usual medical care providers, and their functioning at home, school, and in the community.
Recommendation: Infrastructure needs to be developed in relation to treatment intervention so as to collect and quantify routine clinical encounters data including provider-reported diagnosis and outcome. Instruments for comparing outcomes based on research reassessment of provider-reported diagnosis and outcome need to be developed and validated, including the spectrum of major health providers (e.g., Medicaid and HMO settings and selected commercially insured populations). Within this system, generate hypotheses for ad hoc research studies. Use research-diagnosis data to clarify the validity of the treatment outcome data. Follow cohorts over extended time periods so that long-term outcomes will be ascertained.
Question 4. What are the major safety concerns regarding medications used to treat ADHD and how will they be addressed?
Rationale: Concerns about the increased exposure of preschoolers to stimulants and other psychotropics need to be extended to questions about safety because of concerns about rare side effects and presently unknown effects on the developing brain. 3 Among the immediate federal responses to this concern were plans for a second conference dealing with the safety of psychotropic medication in preschoolers. An earlier meeting in December 1999 was devoted to long-term safety (and abuse) questions raised by increasing exposure to stimulants among all ages of youth. The second meeting occurred in September 2000 (Forum on Adverse Events Monitoring: Recommended strategies for developing and implementing adverse events monitoring in child psychopharmacology) and included representatives from the pharmaceutical industry, academe, and the relevant federal agencies. A third meeting addressed the preschool-aged group in particular. A range of follow-up studies will be solicited from a multidisciplinary work group that is being formed. 34
Some salient points gained from these meetings relate to a burning question on the safety of medications for any medical or mental health treatment, especially when used for long-term symptomatic management. Evidence of drug safety is based on rules that are quite distinct from the criteria we apply to drug efficacy questions. 35 Drug safety methodology does not rely on randomized, double-blind, clinical trials. Attribution of causation is far more difficult to establish; it requires evidence according to long-held scientific criteria and logical inferencing, that is, Koch’s postulates for the causal association of infectious disease with bacteria or Sir Austin Bradford Hill’s application of classic criteria to occupational exposures and the development of disease. 36 Three problems limit the ability of clinical trial methods to answer safety questions. First, humans cannot be subjected to a test to assess harm. Second, the search for rare events requires counts of the drug-exposed population far in excess of what is feasible for a clinical trial. Third, adverse drug event (ADE) monitoring is often the search for the unknown, thus critical measures of outcome usually cannot be anticipated in advance. Several recent compelling examples involving rare but life-threatening events include fenfluramine associated with cardiac valvular changes and pemoline producing liver toxicity and deaths in youngsters. 37
Other broader social factors may contribute to the limited state of methodology for ADEs, which stem from the stakeholders. Efficacy studies are mandated in Phase 3 of the drug development process by FDA regulation before a pharmaceutical manufacturer can bring a drug to market. However, the tasks for ADE monitoring belong to Phase 4, which is after the marketing of a drug and encompasses the whole future drug-exposed population. Thus, the task is quite large, and there is little funding or public health mission to build an active surveillance system that goes beyond the current system. 38MedWatch is the current publicly funded passive surveillance system for detecting possible ADEs, but it is sorely limited by (1) volunteer bias, (2) lack of denominator data from which to project risk assessment, and (3) inadequate staff to assess the trends or signals that may be occurring in these reports.
These issues are not new, but there is little national support for a bold new approach to monitoring safety. ADE monitoring of child psychopharmacology could be done in the community and rates then projected based on sampling routine practice patterns. The hesitation to call for action may reflect the lack of methodology to approach this monumental task. However, if an adverse drug event monitoring system is imbedded into the national monitoring system (to address Questions 1, 2, and 3), then a comprehensive, data-based, drug monitoring system would be established with the capacity to project estimates of risk assessment from routine sampling for all but the most infrequent drug exposures. The collection of descriptive data for prevalence reporting would serve also as baseline exposure rates from which to project case control or cohort studies to assess various questions of drug safety.
Consider the case of pemoline and the risk of liver failure. It took more than 20 years to establish an association between this serious outcome and the use of this stimulant for ADHD. 37 The information regarding an association was delayed in part by an inadequate broadcasting of suspicious signals from the FDA passive surveillance system.
Goal: We need to reduce the time it takes to evaluate serious unexpected adverse events. Even more challenging are the safety issues surrounding concomitant medications, high-dose exposure, and drug-drug interaction monitoring among prescription drugs. Herbal remedies and problems that are likely to increase as a function of the use of more psychotropic medications should also be evaluated.
Recommendation: An infrastructure to launch a national safety surveillance system needs to be built using (1) case registries for rarely used or high risk drug therapies, and (2) probability sampling from a national network of funded reporting sites within representative clinical practice settings, as was described in response to Questions 1, 2, and 3. Given the limited infrastructure and methods for ADE monitoring in the United States, substantial funding is needed from the NIMH and other federal agencies, foundations, and, hopefully, insurer interests. Additional stakeholders include the pharmaceutical industry and the FDA. The development of an independent drug safety board has been proposed as a means of moving beyond the current inadequate system. 39 Some critics have argued that ongoing support for postmarketing surveillance should come from pharmaceutical industry fees for new drug evaluations after the entry of the drug into the U.S. market, but recent debate regarding reenactment of Prescription Drug User Fee Act (PDUFA) legislation suggests numerous unresolved issues (FDA webpage).
Question 5. What social factors influence the designation of behavioral problems as ADHD or a disruptive disorder (CD or ODD)?
* Medical practice changes, for example, DSM-IV criteria
* School policies, Individuals with Disabilities Education Act (IDEA), and school-based health programs
* Law enforcement policies
* Supplemental Security Income criteria
* Family stressors (income, housing, family stability, physical health)
* Cultural diversity (immigration, racism, cultural values)
Rationale: The research mission of the NIMH in the 1990s brought a strong emphasis on service patterns and access to care to complement a long-standing commitment to measure psychopathology reliably, search for its biological correlates and determinants, and to evaluate interventions to prevent or treat it effectively. As a result, more clinical and lay readers are aware of the social context of mental disorders today than would have been true a decade ago. If this statement were true we would expect to avoid the following:
1. Reducing the interpretation of the MTA study findings to a “bottom line” that drugs alone do just as well as combined therapy.
2. The ecological fallacy caused by assuming that group data findings can be applied to each individual patient who meets the study criteria.
3. Inferences that studying the biological basis of brain disorders reduces the need to study nonbiological treatment interventions—a fallacy of biological reductionism.
4. Inferences that the genomic risk for an emotional disorder cannot be influenced by environmental factors and nonbiological interventions. 40
5. Inferences that the mechanism of action of a drug that symptomatically improves an emotional or behavioral disorder explains the etiology of that disorder.
6. Reluctance to accept that secular trends in disease and its treatment can be influenced by social or professional changes in our definitions of the disorders, access to medical care, or differential promotion of some treatments in preference to others.
Additionally, if there is a broader commitment to research and evidence-based practice, we would expect to find the following:
1. Support for a variety of systematic approaches to improve clinical decision, including N-of-1 methodology.
2. Willingness to educate consumer advocates on the difficulties of long-term management of youths with chronic and serious emotional and behavioral disorders and assist community-based advocacy for transitional housing, and special school, work, and rehabilitation programs that support overburdened families.
3. Integration of the best psychosocial methods with the most parsimonious, best practices of psychopharmacology in the search for long-term outcomes that are favorable to families and communities, as well as to researchers and new drug developers.
4. An appreciation of the notion that children inherit, along with their parents’ genes, their parents’ peers and the communities they inhabit. 40
5. An appreciation that chronic and serious mental disorders share with similarly severe medical disorders a need for long-term care and symptomatic treatment based on a common biomedical model and are served best by integrated psychobiological clinical approaches. 41
Goal: Reassess the state of the art in child psychopharmacology in a way that integrates the coming genomic revolution with the available research knowledge base. Use clinical epidemiology as a major arena from which well-established research findings (effectiveness and safety) can be elicited and adapt new models that achieve reliability without sacrificing validity.
Recommendation: (1) Provide a deeper understanding of the context of behavioral problems so that individual, family, and community characteristics related to these problems will be balanced with the diagnostic characterization. (2) Integrate research training from epidemiology, social and psychological sciences, health services research, and policy sciences into the biological sciences more effectively. (3) Provide ongoing interprofessional dialogues to move beyond narrow research interests, particularly among geneticists and clinical scientists. 40 (4) Support training in science and logical inferencing to help empirical scientists interpret complex findings based on individual, group, and population outcomes. (5) Foster independent assessments of the costs and benefits of proposed interventions according to varying viewpoints: individual, vested interest group, community, and societal. 42
These five questions form a framework for the expansion of what is known about leading childhood emotional and behavioral disorders and their treatment. But, in addition, this framework will give us an opportunity to answer related questions that, when taken together with existing methods for the scientific evaluation of drug therapy, form a far more comprehensive model. Ultimately, this approach should yield more complete and useful information on treatments in the usual practice setting used by community-based patients concerning long-term treatment effectiveness, treatment safety, and satisfaction for youths with emotional and behavioral problems.
Dr. Daniel Safer contributed significantly to a critical discussion of the issues raised in this paper and their implications. Dr. James Guevara kindly reviewed a penultimate draft and offered thoughtful remarks.
1. Zito JM, Safer DJ, dosReis S, Gardner JF, Boles M, Lynch F: Trends in the prescribing of psychotropic medications to preschoolers. JAMA 283:1025–1030, 2000
2. Goode E: Sharp rise found in psychiatric drugs for the very young. New York Times, February 23, 2000, pp A1, A14
3. Coyle JT: Psychotropic drug use in very young children. JAMA 283:1059–1060, 2000
4. Koplewicz HS: The role for pills in preschool (op-ed). New York Times, March 1, 2000, p A23
5. Mandersheid RW, Henderson MJ (eds): Mental Health, United States, 2000. Washington, DC, Center for Mental Health Services, DHHS Pub. No. 01-3537, 2001.
6. Zito JM, Safer DJ: Sources of data for pharmacoepidemiological studies of child and adolescent psychiatric disorders. J Child Adolesc Psychopharmacol 7:237–253, 1997
7. Melmon KL: The joint commission on prescription drug use: A ten year perspective. Post Marketing Surveillance 5:3–17, 1991
8. Wennberg J, Gittelsohn A: Small area variations in health care delivery. Science 182:1102–1108, 1973
9. Ashton CM, Petersen NJ, Soucheck J, et al: Geographic variations in utilization rates in Veterans Affairs hospitals and clinics. N Engl J Med 340:32–39, 2001
10. Wolraich ML: The difference between efficacy and effectiveness research in studying attention-deficit/hyperactivity disorder. Arch Pediatr Adolesc Med 153:1220–1221, 1999
11. Accardo P: What’s all the fuss about Ritalin? (invited commentary). J Pediatr 138:6–9, 2001
12. Tversky A, Kahneman D: Judgment under uncertainty: Heuristics and biases. Science 185:1124–1131, 1974
13. Weisz JR, Jensen PS: Efficacy and effectiveness of child and adolescent psychotherapy and pharmacotherapy. Ment Health Serv Res 1:125–157, 1999
14. Pelham WE, Wheeler T, Chronis A: Empirically supported psychosocial treatments for attention deficit hyperactivity disorder. J Clin Child Psychol 27:190–205, 1998
15. MTA Cooperative Group: A 14-month randomized clinical trial of treatment strategies for attention-deficit/hyperactivity disorder. Arch Gen Psychiatry 56:1073–1086, 1999
16. Carey WB: What the multimodal treatment study of children with attention-deficit/hyperactivity disorder did and did not say about the use of methylphenidate for attention deficits. Pediatrics 105:864, 2000
17. Pelham WE: The NIMH Multimodal Treatment Study for attention-deficit hyperactivity disorder: Just say yes to drugs alone? Can J Psychiatry 44:981–990, 1999
18. Vitiello B: On the importance of studying the long-term effects of the stimulant medications used in the treatment of children with AD/HD. Hyper-talk 4: 2000
19. Huget J: Scandal! They haven’t tested Ritalin on the children it’s prescribed for! Scandal! They’re going to test Ritalin on the children. Washington Post, January 2, 2001, P T06
20. Wolraich ML: Attention deficit hyperactivity disorder: The most studied and yet most controversial diagnosis. MRDD Res Rev 5:163–168, 1999
21. Hotopf M, Churchill R, Lewis G: Pragmatic randomised controlled trials in psychiatry. Br J Psychiatry 175:217–223, 1999
22. Wasserman RC, Kelleher KJ, Bocian A, et al: Identification of attentional and hyperactivity problems in primary care: A report from pediatric research in office settings and the ambulatory sentinel practice network. Pediatrics 3:1–7, 1999
23. Zarin DA, Suarez AP, Pincus HA, Kupersanin BA, Zito JM: Clinical and treatment characteristics of children with attention deficit/hyperactivity disorder (ADHD) in psychiatric practice. J Am Acad Child Adolesc Psychiatry 37:1262–1270, 1998
24. Safer DJ, Zito JM, Fine EM: Increased methylphenidate usage for attention deficit disorder in the 1990s. Pediatrics 98:1084–1088, 1996
25. Zito JM, Safer DJ, dosReis S, et al: Ten year prevalence trends for psychotropic medication from two health service systems (Abstract). Presented at the New Clinical Drug Evaluation Meeting, Boca Raton, Florida, June 1–4, 1999
26. Jensen PS, Kettle L, Roper MT, et al: Are stimulants overprescribed ? Treatment of ADHD in four US communities. J Am Acad Child Adolesc Psychiatry 38:797–803, 1999
27. Zito JM, Safer DJ, dosReis S, Magder LS, Gardner J, Zarin DA: Psychotherapeutic medication patterns for youths with attention deficit hyperactivity disorder. Arch Pediatr Adolesc Med 153:1257–1263, 1999
28. Angold A, Erkanli A, Egger H, Costello EJ: Stimulant treatment for children: A community perspective. J Am Acad Child Adolesc Psychiatry 39:1–9, 2000
29. Vitiello B: Commentary on stimulant treatment for children: A community perspective. J Am Acad Child Adolesc Psychiatry 39:992–994, 2000
30. Wennberg JE: The paradox of appropriate care (editorial). JAMA 258:2568–2569, 1987
31. Woosley R: Opportunities in Phase IV to improve drug development. Food and Drug Law Journal 52:188, 1997
32. Jensen PS, Hoagwood K, Petti T: Outcomes of mental health care for children and adolescents. II. Literature review and application of a comprehensive model (Review). J Am Acad Child Adolesc Psychiatry 35:1064–1077, 1996
33. Zito JM, Volavka J, Craig TJ, Czobor P, Banks S, Vitrai J: Pharmacoepidemiology of clozapine in 202 inpatients with schizophrenia. Ann Pharmacotherapy 27:1262–1269, 1993
34. Vitiello B: Psychopharmacology for young children: Clinical needs and research opportunities. Pediatrics 108:983–989, 2001
35. Rogers AS: Unintended drug effects: Identification and attribution, in Hartzema AG, Porta MS, Tilson HH (eds): Pharmacoepidemiology: An Introduction. Cincinnati, OH, Harvey Whitney Books, 1991, pp 64–74
36. Hill AB: The environment and disease: Association or causation? Proc R Soc Med 58:295–300, 2001
37. Safer DJ, Zito JM, Gardner JF: Pemoline hepatotoxicity and post-marketing surveillance. J Am Acad Child Adolesc Psychiatry 40:622–629, 2001
38. Brewer T, Colditz GA: Postmarketing surveillance and adverse drug reactions: Current perspectives and future needs. JAMA 281:824–829, 1999
39. Wood AJJ, Stein CM, Woosley R: Making medicines safer: The need for an independent drug safety board. N Engl J Med 339:1851–1854, 1998
40. Eisenberg L: The social construction of the human brain. Am J Psychiatry 152:1563–1575, 1995
41. Engel GL: The need for a new medical model: A challenge for biomedicine. Science 196:129–136, 1977
42. Zito JM, Provenzano G: Pharmaceutical decisionmaking: Pharmacoepidemiology or pharmacoeconomics: Who’s in the driver’s seat? (Review). Psychopharmacol Bull 31:735–744, 1995