Dr. Zito, Mr. Gardner, and Mr. Ryu are affiliated with the department of pharmaceutical health services research and Dr. Zito is also with the department of psychiatry at the University of Maryland in Baltimore. Dr. Safer is with the departments of psychiatry and pediatrics at Johns Hopkins Medical Institutions in Baltimore. Dr. Soeken is affliated with the University of Maryland School of Nursing in Baltimore. Send correspondence to Dr. Zito at the Department of Pharmaceutical Health Services Research, University of Maryland, 515 West Lombard Street, Room 252, Baltimore, Maryland 21201 (e-mail, firstname.lastname@example.org).
In the early 1990s the use of anticonvulsant medications for psychiatric disorders among youths was quite limited. In an analysis of 647 continuously enrolled Tennessee Medicaid youths aged 0 to 18 years who were new recipients of anticonvulsant medication in 1992, 60 percent had a diagnosis of a seizure disorder and only 11 percent had a diagnosis of a psychiatric disorder only (1). A majority of those with a seizure disorder (66 percent) were given a prescription for phenobarbital or phenytoin, and most (65 percent) of those with a psychiatric diagnosis who were given a prescription for an anticonvulsant received carbamazepine.
Beginning in the early 1990s, anticonvulsant medication began to be used increasingly for youths with bipolar and impulse control disorders. That development paralleled expanding diagnoses of bipolar and pervasive developmental disorders in community-treated youth populations (2,3). In a California health maintenance organization (HMO), the prevalence of anticonvulsant medication use among youths aged five to 17 years doubled from .35 percent in 1994 to .73 percent in 2003 (4). In other large data sets covering youths, increases of anticonvulsant medication prevalence (measured as a prevalence ratio) from 1987 to 1996 were found to vary by treatment program: a 5.9-fold and a 2.2-fold increase in two state Medicaid programs and a 2.5-fold increase in a nonprofit group-model HMO (5). Except for the Tennessee Medicaid research of 1992 by Cooper and colleagues (1) and the Kaiser HMO study by Hunkeler and colleagues (4), none of the incidence or prevalence studies reported diagnostic indications for the use of anticonvulsants among children and adolescents.
The increased use of anticonvulsants for psychiatric indications among youths is occurring despite the drugs' off-label status—that is, limited evidence of efficacy and safety. In view of the relatively stable prevalence of seizure disorders in the United States (6,7), recent sizable increases in the use of anticonvulsants, and very limited diagnostic information about present anticonvulsant use, the primary hypothesis of this research was that the largest proportion of anticonvulsant treatment in recent years would be associated with a psychiatric diagnosis. Consequently, this study appraised year 2000 anticonvulsant patterns for continuous enrollees younger than 18 years in a mid-Atlantic state Medicaid program. Three mutually exclusive diagnostic groups (psychiatric disorders alone, seizure disorders alone, and combined psychiatric and seizure disorders) were formed and examined in relation to anticonvulsant medication use. Additional analyses of anticonvulsant use assessed the influence of age, gender, race or ethnicity, and ICD-9 psychiatric diagnosis, as well as Medicaid eligibility category. Therapy with multiple psychotropic medications and specific medications were also examined.
Administrative claims from outpatient visits and prescriptions were organized for a data set of 258,472 youths who were younger than 18 years and continuously enrolled in Medicaid in a mid-Atlantic state in 2000. Of these, 4,522 had one or more prescription claims for an anticonvulsant during that year. These youths were then grouped into the following ICD-9 diagnostic categories: a psychiatric diagnosis without a seizure disorder (N=3061), a seizure disorder without a psychiatric diagnosis (N=251), and both psychiatric and seizure disorders (N=611). Anticonvulsant medications included carbamazepine, valproic acid (the term is used collectively to refer to all valproic acid-related products), gabapentin, oxcarbazepine, lamotrigine, topiramate, phenytoin, selected barbiturates, succinimides, tiagabine, levetiracetam, and felbamate.
Outcome measures included annual prevalence of anticonvulsant treatment among Medicaid enrollees younger than 18 years; annual treated prevalence of seizure or psychiatric disorders among those dispensed anticonvulsants; use patterns of psychotropic classes and anticonvulsant drugs; anticonvulsant use differences in relation to age, gender, race or ethnicity; Medicaid eligibility category and ICD-9 psychiatric diagnosis; and multiple psychiatric medication patterns in relation to diagnoses. Four major Medicaid eligibility categories were identified: Temporary Assistance for Needy Families (TANF), Supplemental Security Income (SSI), State Children's Health Insurance Program (SCHIP), and foster care.
The study protocol was approved by the institutional review boards of the University of Maryland, Baltimore, and the mid-Atlantic state's department of health and mental hygiene. Analyses were conducted with SAS Unix version 7.0.
Prevalence in relation to diagnosis
The annual prevalence of anticonvulsant prescription for all youths in our sample was 1.75 percent (4,522 of 258,472 youths). Among the 258,472 youths, 862 youths who received such a prescription had a seizure disorder (with or without a psychiatric diagnosis; .3 percent), and 3,672 youths who received such a prescription had a psychiatric disorder (with or without a seizure disorder; 1.4 percent). Of the 4,522 youths who received an anticonvulsant, 251 (5.6 percent) had a seizure disorder only, 611 (13.5 percent) had both seizure and psychiatric disorders, and 3,061 (67.7 percent) had a psychiatric disorder only. Of the 3,923 youths who received a diagnosis and were dispensed an anticonvulsant, 862 (19 percent) had a seizure disorder and 3,672 (81 percent) had a psychiatric disorder. Seventy-one percent of those with a seizure disorder (611 of 862 youths) also had a psychiatric disorder.
Patterns related to age, gender, race or ethnicity, and eligibility
Table 1 presents the demographic and clinical factors according to the three mutually exclusive diagnostic groups. Among the 3,061 youths with only a psychiatric disorder who received a prescription for an anticonvulsant, 2,971 were aged five to 17 years (97.1 percent; 95 percent confidence interval [CI]=96.4 to 97.6 percent) and 90 were younger than five years (2.9 percent; CI=2.4 to 3.6). Among the 251 youths with only a seizure disorder with such a prescription, 180 were aged five to 17 years (71.7 percent; CI=65.1 to 78.3 percent) and 71 were younger than five years (28.3 percent; CI=17.8 to 38.8). Among youths younger than five years who received an anticonvulsant, relatively more youths were given a diagnosis of a seizure disorder than a psychiatric disorder (28.3 percent compared with 2.9 percent).
Males with a psychiatric disorder alone or with a co-existing seizure disorder diagnosis were 1.9 times more likely than females with these diagnoses to receive a prescription for anticonvulsants. By contrast, the prescription of anticonvulsants differed only slightly by gender among those with only a seizure disorder; in this group, males were 1.2 times more likely than females to receive such a prescription.
In a comparison between whites and blacks, significant differences were observed between youths who had a psychiatric disorder only and those who had a seizure disorder only (Table 1). Whites constituted a majority of those who had a psychiatric disorder only and received a prescription for an anticonvulsant (55.7 percent), whereas only 34.7 percent of whites who had only a seizure disorder received such a prescription. A majority of youths who had a seizure disorder only and received a prescription for anticonvulsants were black (65.3 percent), which approximates the racial distribution of the Medicaid population in the mid-Atlantic state studied.
Youths who were medicated with anticonvulsants and had a psychiatric disorder were more likely to be in a Medicaid eligibility category reflecting severe impairment or social disability. Specifically, 41.3 percent of the youths with only a psychiatric disorder received SSI compared with 28.0 percent of youths with only a seizure disorder. Likewise, 23.2 percent of those with only a psychiatric disorder were in foster care compared with 5.5 percent of those with a seizure disorder only. The disparities between these two groups were reversed in the categories reflecting poverty or low family income. Compared with the population with seizure disorder alone, a significantly lower percentage of youths with a psychiatric disorder alone had TANF eligibility (20.7 percent for a psychiatric disorder only compared with 30.3 percent for a seizure disorder only) and SCHIP (14.8 for only a psychiatric disorder compared with 36.2 percent for a seizure disorder only).
Anticonvulsant medication patterns
The rank order of leading anticonvulsant medication use for youths with only a psychiatric diagnosis was valproic acid (61 percent), gabapentin (29 percent), and carbamazepine (16 percent). For those with a seizure disorder only, the leading anticonvulsant medications were carbamazepine (47 percent), valproic acid (33 percent), and phenytoin (38 youths, or 15 percent).
In addition to receiving an anticonvulsant, youths with only a psychiatric disorder were given prescriptions for the following psychotropic drug classes: antidepressants (52 percent), antipsychotics (52 percent), and stimulants (47 percent). For those with both a psychiatric and a seizure disorder, relatively fewer youths were given drugs from other psychotropic classes: anxiolytics (169 youths, or 28 percent), stimulants (16 percent), antipsychotics (13 percent), and antidepressants (10 percent).
For those who received an anticonvulsant and had only a psychiatric disorder, the axis I psychiatric diagnoses that were most common were depression (1,327 youths, or 43 percent), attention deficit hyperactivity disorder (1,214 youths, or 40 percent), oppositional defiant disorder or conduct disorder (898 youths, or 29 percent), and adjustment disorder (488 youths, or 16 percent). For those who received anticonvulsants and who had both a psychiatric and a seizure disorder, the leading axis I psychiatric diagnoses were mental retardation (264 youths, or 43 percent), attention-deficit hyperactivity disorder (95 youths, or 16 percent), oppositional defiant disorder or conduct disorder (68 youths, or 11 percent), and depression (47 youths, or 8 percent).
In addition to an anticonvulsant, two or more psychotropic classes were dispensed in a one-year period for 2,564 youths with a psychiatric disorder only (84 percent), for 43 with a seizure disorder only (17 percent), and for 273 with both diagnoses (45 percent).
The major finding of this study is that in a continuously enrolled Medicaid population of youths in 2000, youths with a psychiatric diagnosis were over four times more likely than those with a seizure disorder diagnosis to be dispensed an anticonvulsant medication. These recent mid-Atlantic Medicaid findings contrast with the 1992 Tennessee Medicaid study, which indicated that seizure diagnoses were most commonly associated with anticonvulsant medication use among youths (1). This difference parallels the prominently increased rate of diagnosing bipolar disorder and pervasive developmental disorder among youths during the past decade.
It is of note that the major drugs in the anticonvulsant class are prescribed off-label (without U.S. Food and Drug Administration [FDA] approved labeling) for childhood psychiatric disorders. Carbamazepine was approved for acute mania by the FDA in 2005, but only for adults. Gabapentin is not approved for this indication. Divalproex has an indication from the FDA for the treatment of acute mania among adults, with the cautionary statement that controlled research for this use beyond three weeks has not been conducted.
The use of anticonvulsants for psychiatric indications was higher for white youths than for black youths, which is consistent with other studies showing similar racial or ethnic disparities in psychotropic treatment (1,8,9,10). The use of anticonvulsants for youths younger than five years was proportionally far greater for seizure control than for psychiatric indications. This differential use pattern by age was also reported by Cooper and colleagues (1).
It is of interest that the 1996 one-year prevalence of anticonvulsant dispensing for all youths (continuous and noncontinuous enrollees) in a Northwestern U.S. HMO was .27 percent (5), and the prevalence was .43 percent for similarly defined enrollees in a Northern California HMO (4). These estimates were distinctly lower than Medicaid annual anticonvulsant prevalence findings in a mid-West and mid-Atlantic state for that year (1.1 and 1.3 percent, respectively) (5) and support the finding that psychotropic drug use is higher for youths enrolled in Medicaid than for those served by nonpublic providers (11). In large part, this difference in prevalence is related to eligibility category: when Medicaid datasets are divided by eligibility categories, the SSI and foster care subpopulations have strikingly high rates of medication use for psychiatric disorders (9), as was the case in this study.
The most detailed diagnostic trend data reported in relation to the prevalence of anticonvulsant treatment among U.S. youths comes from the Northern California Kaiser Permanente HMO research team (4). They found that the number of youths who received anticonvulsant treatment for a seizure disorder, when corrected for enrollment, rose only 8 percent from 1995 to 2003. However, over that period, psychiatric diagnoses for those who received anticonvulsant medications far outpaced seizure diagnoses. These increases from 1995 through 2003 were as follows: bipolar (6.5-fold), depression (3.3-fold), attention-deficit hyperactivity disorder (2.0-fold), and anxiety (3.5-fold) (4).
In this research study, diagnoses were lacking in the records of 13.2 percent of anticonvulsant users (599 of 4,522 youths). Similarly, the study by Cooper and colleagues (1) reported that 16 percent of the children given anticonvulsants had no recorded diagnoses "for which anticonvulsants might appropriately be prescribed." Moreover, some loss of linkage of diagnosis to prescription claims may have occurred by limiting the dataset to a one-year interval. However, the disparities in diagnosis are so great that the results are not likely to be affected by truncation of the data at the start and end of the study year.
Another limitation is that the findings from a Medicaid population in a single state do not necessarily generalize to those in other states. However, it is noteworthy that the 1996 prevalence for mood-stabilizing anticonvulsants dispensed to youths younger than 20 years was similar in a mid-Atlantic Medicaid state program and in a mid-Western Medicaid state program (1.28 percent compared with 1.08 percent) (5). Nonetheless, it has been documented that anticonvulsant prevalence findings for youths who are covered by Medicaid clearly do not generalize to those who are privately insured. Finally, this study lacks information about the specialty of the prescribing physicians.
Over the past decade, the use of many anticonvulsant compounds for psychiatric treatment as mood stabilizers has dramatically increased, which prompted the study reported here. This increase occurred while the use of anticonvulsants for the treatment of seizures was relatively stable. The increase was apparent first in adult psychiatric treatment (12), and now anticonvulsant compounds have become a major psychotropic medication class for youths. They are potent and are by no means innocuous (13). For example, valproic acid is associated with polycystic ovaries in females, pancreatitis, liver toxicity and weight gain; lamotrigine has a greater risk of life-threatening allergic reactions than other anticonvulsants. Recently, the American Epilepsy Society's pregnancy outcomes forum panel recommended that young women in the child-bearing years should not be given prescriptions of valproic acid products as a first-line therapy for any indication because of the increased risk of major fetal malformations (14).
Furthermore, the use of anticonvulsants to treat psychiatric disorders among youths is off-label and without adequate support from double-blind, placebo-controlled studies. One double-blind placebo-controlled study of valproic acid for aggressive behavior among 30 youths was negative (15), whereas another with a cross-over design was positive (N=20) (16). A third study using divalproex in a double-blind placebo-controlled design was negative (N=17) (17). No reports of double-blind, placebo-controlled studies of carbamazepine or gabapentin for psychiatric disorders among youths were located in a MEDLINE search.
Clearly, as Wagner (18) has noted, further research is needed. While we await further clarification of the efficacy and long-term safety of anticonvulsants for psychiatric indications, there is a need for professionally defined and validated clinical assessment and monitoring of anticonvulsants in community-treated populations. Methods such as N of 1 protocols could advance practice in terms of reducing inappropriate or unnecessary use, save wasted health care dollars, and reserve the use of anticonvulsants for youths with psychiatric disorders who respond with clear benefit and minimal risk.