Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
August 2004

Age Effects on Antidepressant-Induced Manic Conversion

Author Affiliations

From the Child Study Center (Drs Martin, Young, Leckman, and Rosenheck), the Department of Psychiatry (Drs Martin, Leckman, Rosenheck, and Leslie), and the School of Epidemiology and Public Health (Drs Mukonoweshuro, Rosenheck, and Leslie), Yale University School of Medicine, New Haven, Conn; Department of Psychiatry, Case Western Reserve University School of Medicine, Cleveland, Ohio (Dr Young); and the Northeast Program Evaluation Center, West Haven, Conn (Drs Rosenheck and Leslie).


Copyright 2004 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2004

Arch Pediatr Adolesc Med. 2004;158(8):773-780. doi:10.1001/archpedi.158.8.773

Background  Antidepressant drug therapy can precipitate mania in vulnerable individuals, but little is known about the effects of age on this phenomenon.

Objective  To pharmacoepidemiologically evaluate the risk of conversion to mania by antidepressant class and patient age.

Design, Setting, and Patients  Using an administrative national database of more than 7 million privately insured individuals, linked outpatient and pharmacy claims were analyzed for mental health users aged 5 to 29 years (N = 87 920).

Main Outcome Measures  The proportion and cumulative hazard of manic conversion were analyzed by antidepressant class and subject age among children, adolescents, and young adults with an anxiety or nonbipolar mood disorder in the United States between January 1, 1997, and December 31, 2001. Manic conversion was defined as a new diagnosis of bipolar illness.

Results  During median follow-up of 41 weeks (range, 8-251 weeks), manic conversion occurred in 4786 patients (5.4%). Multivariate analyses using time-dependent Cox proportional hazards models indicated that an increased risk of manic conversion was associated with antidepressant category vs no antidepressant exposure (hazard ratios: 2.1 for selective serotonin reuptake inhibitors, P<.001; 3.8 for "other" antidepressants, P<.001; and 3.9 for tricyclic antidepressants, P = .002). Antidepressant × age interactions revealed inverse age effects for selective serotonin reuptake inhibitors and other antidepressants (β = −.05; P<.001 for both) but not for tricyclic antidepressants (β = −.02; P = .25). Peripubertal children exposed to antidepressants were at highest risk of conversion (number needed to harm: 10 [95% confidence interval, 9-12] among 10- to 14-year-olds vs 23 [95% confidence interval, 21-25] among 15- to 29-year-olds).

Conclusions  Patient age is an effect modifier on the risk of antidepressant-associated manic conversion. Treatment with antidepressants is associated with highest conversion hazards among children aged 10 to 14 years.

It has long been known that antidepressant medications can precipitate mania in vulnerable individuals. Indeed, the monoamine oxidase inhibitor iproniazid, one of the first known antidepressants, was serendipitously discovered as a result of its euphoric and mood-elevating effects on patients treated for tuberculosis.1 Antidepressant-induced switching (the transition from depression into mania) has been studied in several naturalistic and controlled studies, with rates ranging from 7%2 to 10%3,4 and as high as 67%5 among bipolar patients. A meta-analysis6 comparing the propensity of different antidepressant drug classes to induce mania among bipolar depressive patients found that tricylic antidepressants (TCAs) were associated with considerably higher rates (11.2%) than selective serotonin reuptake inhibitors (SSRIs), for which rates were comparable to placebo (3.7% and 4.2%, respectively). In contrast, rates among unipolar depressive patients were low (approximately 1%).

In addition to manic switching, other antidepressant adverse effects on the natural course of mood disorders have been described.7 Among these, drug-induced (as opposed to spontaneous) cycling8,9 and cycle acceleration10,11 are the most widely recognized, even in the context of concurrent treatment with mood stabilizers.12 Switching and cycle acceleration are characteristics that have been linked to poor long-term outcome in bipolar disorder.13 Despite these lines of evidence negatively implicating antidepressant drug use in the long-term course of affective illness, the issue is far from settled, as exemplified by at least 2 studies2,14 that have found no evidence supporting TCA use as a contributing factor to affective instability in bipolar disorder.

The role of age (particularly of young age) on antidepressant-induced manic switching has received relatively little attention but is of clinical interest—and of potential concern—for several reasons. First, the prevalence of antidepressant drug use among children and adolescents has risen steeply during the past decade,1517 even among children as young as 2 to 4 years.18 Second, the recognition and diagnosis of bipolar disorder, once considered rare earlier than age 18 years, has steadily increased during the same period,19 as has the use of mood stabilizers in this age group.17 Although late adolescence is a clear epoch for the onset of bipolar disorder, it is not clear whether this or younger ages may be associated with heightened sensitivity to the untoward effects of antidepressant agents. Third, safety data for the use of antidepressants in children and adolescents largely derive from short-term trials (typically 6-12 weeks); adverse events occurring outside of this time window are not routinely identified.20 Finally, there has been recent controversy over the potential for antidepressants (such as paroxetine and venlafaxine hydrochloride) to exacerbate suicidal ideation among children.21,22 Such concerns, which have led to action from British and American regulatory agencies,23 highlight the fact that serious adverse effects associated with antidepressant drug use may be behavioral and potentially life threatening.

Using a large administrative database, we previously described how, between January 1, 1997, and December 31, 2000, (1) the diagnosis of bipolar disorder increased among mental health service users aged 6 to 17 years from a 1-year prevalence of 1.2% to 1.8% (a 54% change)24 and (2) the use of antidepressants and mood stabilizers increased from 25% and 5.5% to 28% and 6.2%, respectively (shifts of 12.4% and 13.1%, respectively; P<.001 for trend for all comparisons).25 In this study, we use the same administrative database to examine age effects on antidepressant-associated manic conversion. Given that our main outcome of interest (defined herein as the onset of a new bipolar disorder diagnosis) takes place against the backdrop of these time trends, we used time-to-event statistical modeling to track continuously enrolled individuals in an effort to determine the differential contributions of age, antidepressant drug exposure, and their interactions over time on the likelihood of manic diagnostic conversion.

Because our outcome of interest was based on diagnostic information of an administrative nature, we use the term manic conversion throughout rather than the more narrowly defined term manic switch, which would have required specific clinical information that is unavailable in this database.


Subjects were included in the MarketScan Research Database (Thomson Medstat, Ann Arbor, Mich), a publicly available fee-for-service medical and prescription claims resource that contains information for individuals nationwide who are insured through the benefit plans of large employers and that includes more than 200 different insurance companies. The working sample consisted of patients aged 5 to 29 years with a primary diagnosis of a depressive or anxiety disorder who had pharmacy claims data available for the 5-year study (January 1, 1997, to December 31, 2001). Individual information was stripped of any personal identifiers, stored anonymously, and exclusively referenced through study-specific unique identifiers, following published guidelines.26


Any claim with an International Classification of Diseases, Ninth Revision (ICD-9),27 diagnosis code of 290.00 to 319.99 was considered a mental health claim, regardless of the setting of care. Each individual was assigned to a primary diagnostic category, defined as the one responsible for most of the mental health services during follow-up, as measured by the total number of encounters. We defined 5 mental health disorder categories based on a modified coding scheme24: (1) severe depression (ICD-9 codes 296.20-296.39 and 296.82), (2) mild depression (ICD-9 codes 300.40-300.59, 301.10, 309.00-309.19, 309.28, 301.12, and 311.00-311.99), (3) anxiety disorders (ICD-9 codes 300.00-300.39, 307.20-307.23, 308.00-308.99, 309.20-309.24, and 313.00-313.29), (4) bipolar disorder (ICD-9 codes 296.00-296.19, 296.40-296.81, and 296.89-296.99), and (5) other mental health conditions (the remaining ICD-9 codes in the mental health range).

Independent variables included patient demographic characteristics (sex and age grouped into five 5-year categories) and 3 measures of illness severity: (1) primary diagnosis (severe depression, mild depression, or anxiety), (2) number of different mental diagnoses in the year, and (3) whether inpatient mental health episodes were incurred during the period of observation.

Only patients with linked pharmacy data were included in the study. Patients with any claims for psychotropic drugs were identified from a comprehensive National Drug Code registry. Psychotropics were assigned to 1 of 6 drug classes following a previous classification scheme.24,25 For each patient, start and end dates (the latter based on the prescription's day supply) were calculated for all medications. Finally, the sum of different medication categories used (range, 0-8) served as a fourth proxy of illness severity.

The primary outcome of interest was a new diagnosis of bipolar disorder as assessed by 2 or more claims. Whenever the outcome was preceded by the introduction of a mood stabilizer (lithium or an antiepileptic drug other than phenytoin), the earlier event was used to mark the conversion date in subsequent time-dependent analyses. Patients with either outcome occurring within the first 2 months of observation were dropped from the working sample based on the rationale that we were interested in only 1 of 2 main types of adverse behavioral reactions to antidepressant drug treatment that have been described in youngsters.28 The first type occurs early in treatment (median onset, 28 days), is characterized by anxiety or akathisia, and dissipates soon after drug discontinuation; the second type occurs later in treatment (median onset, 91 days), can more closely resemble manic activation, and often lags after drug discontinuation. By excluding from analysis individuals who converted within the first 2 months of observation, we sought a more stringent definition centered on the latter type, thus minimizing misclassification errors, for example, of akathisia into mania.


Standard univariate analyses (χ2 and 2-tailed, unpaired t tests) were performed for descriptive purposes. The interval between study entry (the first visit or pharmacy claim available) and diagnostic conversion or last censored study observation was calculated and expressed in weeks or years. Observations that did not meet the end point definition were censored for analytic purposes. Conversion rates were calculated by dividing the number of outcome events by the number of person-years of observation. Confidence intervals (CIs) for conversion rates were calculated using standard epidemiologic formulas,29 and the difference in conversion rates across groups was expressed using the evidence-based metrics of number needed to harm (NNH) and associated CIs.30

The Kaplan-Meier method was used to examine crude survival rates, and Cox proportional hazards regression analysis was used to adjust for potential confounders. Because of the changing nature of the medications used during the observation period, time-dependent covariates were created for these variables, using prescription start and end dates to construct the required intervals,31 yielding a total of 397 987 observation periods in the final working data set, a mean of 4.4 per participant. In this way, adjusted hazard ratios and 95% CIs were calculated to establish the contribution of different antidepressant agents on conversion (compared with a nonantidepressant reference category) while controlling for age and confounders, especially illness severity. The stratified nature of the Kaplan-Meier and NNH analyses dictated that only individuals prescribed a single antidepressant class could be included in them, but the Cox models provided adjusted estimates for all individuals, including those taking 2 or more antidepressants during their time of observation.

Given our a priori hypothesis that the effects of antidepressant drug use on manic conversion would differ according to patient age, we included interaction terms to ascertain the role of age as an effect modifier on conversion hazards. To that end, in addition to the main effect models (of age and antidepressant class, models I and II, respectively), antidepressant × age terms were added to interaction models. Age in years was first used in interaction terms for all antidepressants and for each class (models IIIa and IVa, respectively), testing for significance using the Wald χ2 test. Interactions were then deconstructed by using age as a 5-year category rather than a continuous variable (models IIIb and IVb, respectively). The oldest age group (24-29 years) was selected as the reference category.


There were 4786 converters among the 87 920 patients in the sample, an overall period prevalence of 5.4%. Participants were followed during the 5-year study for a median of 41 weeks (range, 8-251 weeks), for a cumulative total of 79 408 person-years and a conversion rate of 6.0% per year (95% CI, 5.9%-6.2%). The primary outcome was preceded by the introduction of a mood stabilizer in 2766 converters (57.8%), but, as noted earlier, all converters went on to have 2 or more claims with a diagnosis of bipolar disorder.

The study sample was predominantly female (60%) and older (72% of participants were aged 15-29 years). Mild depression was the most common diagnosis (in 54% of participants), followed by anxiety (in 28%), and severe depression (in 19%). Only 201 individuals (2.3 per 1000) had had an inpatient psychiatric hospitalization. A total of 32 864 patients (37.4%) were treated with at least 1 antidepressant medication; although the modal use among the medication-exposed group was of a single antidepressant, 9093 of them (27.7%) used 2 or more different medications (antidepressant or other psychotropic). Demographic and clinical characteristics of the study sample according to conversion status are summarized in Table 1.

Table 1. 
Image not available
Demographic and Clinical Characteristics of the 87 920 Study Participants*

There were clear differences in age composition between converters and nonconverters. In unadjusted analyses, and as depicted in Kaplan-Meier curves, converters were most likely to belong to the 15- to 19-year-old age category and least likely to belong to the youngest group (rate ratio, 3.2; 95% CI, 3.1-3.4) (Figure 1).

Figure 1.
Image not available

Kaplan-Meier estimates of the cumulative hazard of conversion according to age category at study entry. An outcome event (manic conversion) was defined as a new diagnosis of bipolar disorder as assessed by 2 or more claims. Whenever the outcome was preceded by the introduction of a mood stabilizer (lithium or an antiepileptic drug other than phenytoin), the earlier event was used to mark the conversion date. There was a significant difference in hazard rates across the age groups (P<.001, log-rank test). CI indicates confidence interval.

The conversion rate among antidepressant-treated patients (7.7% per year) was 3-fold that among unexposed patients (2.5% per year; rate ratio, 3.1; 95% CI, 3.0-3.2). Conversion rates differed by medication type: higher for TCAs and other antidepressants and lower for SSRIs (rate ratios vs the no-antidepressant category: 3.7 and 3.0, respectively) (Figure 2). Although the TCA and other antidepressant categories did not differ significantly from one another, they were kept separate rather than combined into a single category in subsequent analyses.

Figure 2.
Image not available

Kaplan-Meier estimates of the cumulative hazard of conversion according to medication category. An outcome event (manic conversion) was defined as a new diagnosis of bipolar disorder as assessed by 2 or more claims. Whenever the outcome was preceded by the introduction of a mood stabilizer (lithium or an antiepileptic drug other than phenytoin), the earlier event was used to mark the conversion date. There was a significant difference in hazard rates across the antidepressant categories (P<.001, log-rank test). Patients taking more than 1 antidepressant agent were excluded from analysis, yielding smaller totals for number at risk and number of converters. CI indicates confidence interval; SSRI, selective serotonin reuptake inhibitor; TCA, tricyclic antidepressant.

Age × medication interactions were first apparent in unadjusted analyses stratified by age category and antidepressant medication exposure (Figure 3). The conversion ratio between the antidepressant-exposed and antidepressant-unexposed groups was significantly higher in the younger age group (5- to 14-year-olds; rate ratio, 2.9; 95% CI, 2.8-3.1) than in the older age group (15- to 29-year-olds; rate ratio, 1.4; 95% CI, 1.3-1.5; log-rank test, 160.8; P<.001). This difference was largely due to the low conversion rate among the younger patients in the antidepressant-unexposed group.

Figure 3.
Image not available

Kaplan-Meier estimates of the cumulative hazard of conversion according to age category and antidepressant exposure. An outcome event (manic conversion) was defined as a new diagnosis of bipolar disorder as assessed by 2 or more claims. Whenever the outcome was preceded by the introduction of a mood stabilizer (lithium or an antiepileptic drug other than phenytoin), the earlier event was used to mark the conversion date. There were significant differences in hazard rates (1) across the antidepressant-exposed and antidepressant-unexposed categories; (2) among the antidepressant-unexposed group, across those younger than 14 years/older than 15 years; and (3) across the 4 categories (P<.001, log-rank test for each comparison). CI indicates confidence interval.

The effects of age and antidepressant exposure on conversion status were further examined through adjusted models that controlled for confounding variables. In all final time-dependent Cox proportional hazards models, there were no sex differences (P = .08), but main effects emerged for the diagnoses of severe and mild depression (hazard ratios compared with the anxiety group: 3.4 and 1.3, respectively) and for inpatient status, number of diagnoses, and number of different medications (hazard ratios: 6.7, 1.5, and 1.1, respectively; P<.001 for all except where noted otherwise). In parallel to results from the unadjusted analyses, main effects of medication use were highest for TCAs (hazard ratio, 3.9; P = .002), followed by other antidepressants (hazard ratio, 3.8) and SSRIs (hazard ratio, 2.1) (Table 2, model IVa). Although in all models there was a main effect of age (hazard ratio, 1.1 for each 5-year increase) that was consistent with the unadjusted analyses, several age × medication interactions were also apparent (Table 2, models IIIa-IVb).

Table 2. 
Image not available
Adjusted Hazard Ratios (HRs) for Manic Conversion: Effects of Age Group, Antidepressant (AD) Category, and Their Interactions (4786 Converters)*

Significant negative interactions for age on conversion hazards emerged for the pooled antidepressant, SSRI, and other categories (β = −.05; P<.001 for all). Within each group, age was inversely related to conversion hazards, reaching maximum values in the 10- to 14-year-old category. In contrast, age effects were not apparent in the adjusted analyses for the TCA group at either stratum-specific or overall age levels. The effect of age on conversion status was also seen in the medication- and age-stratified calculations of NNHs (Table 3). To assess the risks of antidepressant-associated manic conversion on the least vulnerable patients, we repeated these analyses using only individuals with diagnoses in either the "mild depression" or "anxiety disorders" clusters. In these analyses, values for 5- to 14-year-olds remained at low levels for SSRIs (NNH, 13; 95% CI, 11-15) and the other antidepressants (NNH, 10; 95% CI, 8-11).

Table 3. 
Image not available
Number Needed to Harm (NNH), by Antidepressant Class and Age Category*

This study revealed that age is an effect modifier on antidepressant-associated manic conversion. Specifically, after adjusting for potential confounders, there was an inverse relationship between age and relative hazards of conversion. The hazard was highest for the peripubertal group of 10- to 14-year-old patients, such that treating as few as 10 (95% CI, 9-12) children with an antidepressant agent could result in at least 1 conversion event, compared with 23 (95% CI, 21-25) among 15- to 29-year-olds.

Our findings are consistent with earlier, although far from definitive, studies describing differential liability to manic induction according to antidepressant class (typically highest for TCAs, lowest for bupropion hydrochloride).6,32 They are also in keeping with studies documenting a peak age at onset of bipolar disorder of 15 to 19 years in recent cohorts,33 and younger age at first treatment as a predictor of vulnerability to antidepressant-induced cycle acceleration.11 Given that hazard rates in this study peaked in the next younger age group (10- to 14-year-olds), our findings suggest that peripubertal children may be especially vulnerable to antidepressant-associated conversion.

Also consistent with our findings, there is an emerging preclinical body of literature that suggests differing vulnerability to the adverse effects of antidepressants according to the developmental stage during which exposure occurs. For example, rodents administered SSRIs prepubertally develop an increased density of serotonin transporters in the frontal cortex that persists into adulthood,34 and their prepubertal (as opposed to adult) response to serotonergic and noradrenergic probes is dramatically different.35,36

In light of this evidence, lack of information regarding pubertal status limited our study. A further limitation was its reliance on administrative rather than clinical data, especially around diagnostic classification and the definition of switching, a shortcoming inherent to other studies that rely on secondary analyses to secure large sample sizes. Other limitations include lack of random assignment in this large sample of convenience, no information on medication compliance, and the potential for confounding by indication, that is, depressed children who are more likely to "naturally" become manic are also those more likely to be prescribed antidepressant drugs. Perhaps the largest limitation and potential threat to the study's validity has to do with the fundamental difficulty in disentangling what may be natural progression of the illness from larger secular trends (such as the more pervasive use of antidepressants and mood stabilizers and of diagnosing bipolar disorder, especially among youths) from true induction of constitutionally vulnerable individuals by means of a pharmacologic "irritant."

Although we cannot entirely rule out the role of these underlying secular trends on our results, the fact that on average subjects were followed for less than 1 year suggests that such effects were probably small. Furthermore, we introduced a variety of measures into our analyses to adjust for confounding, particularly of illness severity, given that the earlier the age at onset, the more severe the illness course is likely to be, and the greater the illness severity, the higher the risk of conversion. Despite these efforts, some residual confounding may well have persisted.

Residual confounding and sample size considerations may explain the lack of a clear conversion age effect for TCAs, in contrast to SSRIs and other antidepressants. Alternatively, the lack of TCA effect modification may not be an artifact. It has been demonstrated that TCAs are no better than placebo in the treatment of depression in children and adolescents,37,38 a negative finding not likely due to failed studies, given an adequately powered recent study.39 Preclinical studies with nonhuman primates have shown different timelines in the maturation of central nervous system pathways, with adult patterns of serotonergic innervation reached by childhood40 as opposed to noradrenergic and dopaminergic pathways, which continue to develop through puberty41 and well into young adulthood.42 Given that most TCAs have predominant effects on adrenergic pathways, the unavailability of a fully mature neural substrate on which to act may explain their lack of efficacy—or of heightened liability to behavioral toxicity—among children and adolescents.

The large sample size of our study allowed us to investigate the effects of age on antidepressant-associated manic conversion and revealed strong age × medication interactions. Children and adolescents younger than 15 years, especially those treated with SSRIs or other antidepressants, are at heightened risk of manic conversion, cautions to be evaluated clinically, scientifically, and ethically43 in the context of a growing evidence base for their use.39,4446 Further pharmacoepidemiologic, pharmacovigilance, and long-term prospective studies are needed to confirm our findings and to establish the long-term safety profile of these drugs.47 Our findings must be considered preliminary associations and must not be taken to equate direct causality, especially in light of the limitations we outlined earlier. Despite this important caveat, judicious prescription practices and thorough risk-benefit analysis at the level of the individual patient are warranted for the age-specific use of antidepressant agents. Caution and restraint may be especially pertinent when treating children whose symptoms are not a major source of impairment, those with a strong family history of bipolar disorder, or those who have not yet tried (and failed) psychotherapeutic interventions of proven efficacy.4851

What This Study Adds

  • It has long been known that antidepressant medications can precipitate mania in vulnerable individuals, but little is known about the effects of age on this clinical phenomenon.

  • Differing response and adverse effect profiles according to developmental stage have been clinically described for antidepressant medications, and a preclinical body of literature suggests that the prepubertal (as opposed to the adult) response to serotonergic and noradrenergic probes can dramatically differ from each other.

  • This study replicates earlier (although still inconclusive) findings that antidepressant drug therapy can induce manic conversion among vulnerable individuals and that antidepressant categories differ in their manic induction propensities (highest for TCAs and the other antidepressants and lowest for SSRIs).

  • To the best of our knowledge, this is the first study to specifically address age effects on antidepressant-induced mania. Significant inverse effects between age and manic induction were found for SSRIs and other antidepressants but not for TCAs.

  • Peripubertal children were found to be at highest risk for conversion (NNH: 10 [95% CI, 9-12] among 10- to 14-year-olds vs 23 [95% CI, 21-25] among 15- to 29-year-olds).

  • Until clinical and prospective studies confirm or refute these findings, judicious prescription practices may be especially pertinent when treating children whose symptoms are not a major source of impairment, those with a strong family history of bipolar disorder, or those who have not yet tried (and failed) psychotherapeutic interventions of proven efficacy.

Back to top
Article Information

Correspondence: Andrés Martin, MD, MPH, Child Study Center, Yale University School of Medicine, 230 S Frontage Rd, New Haven, CT 06520-7900 (Andres.Martin@Yale.Edu).

Accepted for publication April 8, 2004.

This study was supported in part by Scientist Career Development Award 01792 from the National Institute of Mental Health, Rockville, Md.

We acknowledge the statistical consultation provided by Ralitza Gueorguieva, PhD.

Kline  NS Clinical experience with iproniazid (Marsilid). J Clin Exp Psychopathol. 1958;19 (supp 1) 72- 79
Lewis  JLWinokur  G The induction of mania: a natural history study with controls. Arch Gen Psychiatry. 1982;39303- 306
Bunney  WE  JrMurphy  DLGoodwin  FKBorge  GF The "switch process" in manic-depressive illness, I: a systematic study of sequential behavioral changes. Arch Gen Psychiatry. 1972;27295- 302
Bunney  WE  Jr Drug therapy and psychobiological research advances in the psychoses in the past decade. Am J Psychiatry. 1978;135 (suppl) 8- 13
Prien  RFKupfer  DJMansky  PA  et al.  Drug therapy in the prevention of recurrences in unipolar and bipolar affective disorders: report of the NIMH Collaborative Study Group comparing lithium carbonate, imipramine, and a lithium carbonate–imipramine combination. Arch Gen Psychiatry. 1984;411096- 1104
Peet  M Induction of mania with selective serotonin re-uptake inhibitors and tricyclic antidepressants. Br J Psychiatry. 1994;164549- 550
Fava  GA Can long-term treatment with antidepressant drugs worsen the course of depression? J Clin Psychiatry. 2003;64123- 133
Boerlin  HLGitlin  MJZoellner  LAHammen  CL Bipolar depression and antidepressant-induced mania: a naturalistic study. J Clin Psychiatry. 1998;59374- 379
Angst  J Switch from depression to mania: a record study over decades between 1920 and 1982. Psychopathology. 1985;18140- 154
Kukopulos  ACaliari  BTundo  A  et al.  Rapid cyclers, temperament, and antidepressants. Compr Psychiatry. 1983;24249- 258
Altshuler  LLPost  RMLeverich  GSMikalauskas  KRosoff  AAckerman  L Antidepressant-induced mania and cycle acceleration: a controversy revisited. Am J Psychiatry. 1995;1521130- 1138
Post  RMAltshuler  LLFrye  MA  et al.  Rate of switch in bipolar patients prospectively treated with second-generation antidepressants as augmentation to mood stabilizers. Bipolar Disord. 2001;3259- 265
Maj  MPirozzi  RMagliano  LBartoli  L The prognostic significance of "switching" in patients with bipolar disorder: a 10-year prospective follow-up study. Am J Psychiatry. 2002;1591711- 1717
Coryell  WSolomon  DTurvey  C  et al.  The long-term course of rapid-cycling bipolar disorder. Arch Gen Psychiatry. 2003;60914- 920
Rushton  JLWhitmire  JT Pediatric stimulant and selective serotonin reuptake inhibitor prescription trends: 1992 to 1998. Arch Pediatr Adolesc Med. 2001;155560- 565
Zito  JMSafer  DJDosReis  S  et al.  Rising prevalence of antidepressants among US youths. Pediatrics. 2002;109721- 727
Zito  JMSafer  DJDosReis  S  et al.  Psychotropic practice patterns for youth: a 10-year perspective. Arch Pediatr Adolesc Med. 2003;15717- 25
Zito  JMSafer  DJdosReis  SGardner  JFBoles  MLynch  F Trends in the prescribing of psychotropic medications to preschoolers. JAMA. 2000;2831025- 1030
Costello  EJPine  DSHammen  C  et al.  Development and natural history of mood disorders. Biol Psychiatry. 2002;52529- 542
Jensen  PSBhatara  VSVitiello  BHoagwood  KFeil  MBurke  LB Psychoactive medication prescribing practices for US children: gaps between research and clinical practice. J Am Acad Child Adolesc Psychiatry. 1999;38557- 565
Connor  DFBrent  DARiddle  MA Paroxetine and the FDA. J Am Acad Child Adolesc Psychiatry. 2004;43127- 130
Brent  DABirmaher  B British warnings on SSRIs questioned. J Am Acad Child Adolesc Psychiatry. 2004;43379- 380Article
Varley  CK Psychopharmacological treatment of major depressive disorder in children and adolescents. JAMA. 2003;2901091- 1093
Martin  ALeslie  D Psychiatric inpatient, outpatient, and medication utilization and costs among privately insured youths, 1997-2000. Am J Psychiatry. 2003;160757- 764
Martin  ALeslie  D Trends in psychotropic medication costs for children and adolescents, 1997-2000. Arch Pediatr Adolesc Med. 2003;157997- 1004
Simon  GEUnutzer  JYoung  BEPincus  HA Large medical databases, population-based research, and patient confidentiality. Am J Psychiatry. 2000;1571731- 1737
Not Available, International Classification of Diseases: Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death.  Geneva, Switzerland World Health Organization1997;
Wilens  TEBiederman  JKwon  A  et al.  A systematic chart review of the nature of psychiatric adverse events in children and adolescents treated with selective serotonin reuptake inhibitors. J Child Adolesc Psychopharmacol. 2003;13143- 152
Rothman  KJGreenland  S Modern Epidemiology. 2nd ed. Philadelphia, Pa Lippincott Williams & Wilkins1998;
Sackett  DLStraus  SERichardson  WSRosenberg  WHaynes  RB Evidence-Based Medicine: How to Practice and Teach EBM.  Edinburgh, Scotland Churchill Livingstone2000;
Therneau  TMGrambsch  PM Modeling Survival Data: Extending the Cox Model.  New York Springer-Verlag NY Inc2000;
Bottlender  RRudolf  DStrauss  AMoller  HJ Mood-stabilisers reduce the risk of developing antidepressant-induced maniform states in acute treatment of bipolar I depressed patients. J Affect Disord. 2001;6379- 83
Chengappa  KNKupfer  DJFrank  E  et al.  Relationship of birth cohort and early age at onset of illness in a bipolar disorder case registry. Am J Psychiatry. 2003;1601636- 1642
Wegerer  VMoll  GHBagli  MRothenberger  ARuther  EHuether  G Persistently increased density of serotonin transporters in the frontal cortex of rats treated with fluoxetine during early juvenile life. J Child Adolesc Psychopharmacol. 1999;913- 24
Carrey  NJDursun  SClements  RRenton  KWaschbusch  DMacMaster  FP Noradrenergic and serotonergic neuroendocrine responses in prepubertal, peripubertal, and postpubertal rats pretreated with desipramine and sertraline. J Am Acad Child Adolesc Psychiatry. 2002;41999- 1006
Leckman  JFYoung  C Commentary: noradrenergic and serotonergic neuroendocrine responses in prepubertal, peripubertal, and postpubertal rats pretreated with desipramine and sertraline. J Am Acad Child Adolesc Psychiatry. 2002;411007- 1009Article
Hazell  PO'Connell  DHeathcote  DRobertson  JHenry  D Efficacy of tricyclic drugs in treating child and adolescent depression: a meta-analysis. BMJ. 1995;310897- 901
Kaufman  JMartin  AKing  RACharney  DS Are child-, adolescent- and adult-onset depression one and the same disorder? Biol Psychiatry. 2001;49980- 1001
Keller  MBRyan  NDStrober  M  et al.  Efficacy of paroxetine in the treatment of adolescent major depression: a randomized, controlled trial. J Am Acad Child Adolesc Psychiatry. 2001;40762- 772
Goldman-Rakic  PSBrown  RM Postnatal development of monoamine content and synthesis in the cerebral cortex of rhesus monkeys. Brain Res. 1982;256339- 349
Woo  TUPucak  MLKye  CHMatus  CVLewis  DA Peripubertal refinement of the intrinsic and associational circuitry in monkey prefrontal cortex. Neuroscience. 1997;801149- 1158
Rosenberg  DRLewis  DA Postnatal maturation of the dopaminergic innervation of monkey prefrontal and motor cortices: a tyrosine hydroxylase immunohistochemical analysis. J Comp Neurol. 1995;358383- 400
Vitiello  BJensen  PSHoagwood  K Integrating science and ethics in child and adolescent psychiatry research. Biol Psychiatry. 1999;461044- 1049
Emslie  GJRush  AJWeinberg  WA  et al.  A double-blind, randomized, placebo-controlled trial of fluoxetine in children and adolescents with depression. Arch Gen Psychiatry. 1997;541031- 1037
Emslie  GJHeiligenstein  JHWagner  KD  et al.  Fluoxetine for acute treatment of depression in children and adolescents: a placebo-controlled, randomized clinical trial. J Am Acad Child Adolesc Psychiatry. 2002;411205- 1215
Wagner  KDAmbrosini  PRynn  M  et al.  Efficacy of sertraline in the treatment of children and adolescents with major depressive disorder: two randomized controlled trials. JAMA. 2003;2901033- 1041
Greenhill  LLVitiello  BAbikoff  H  et al.  Developing methodologies for monitoring long-term safety of psychotropic medications in children: report on the NIMH conference, September 25, 2000. J Am Acad Child Adolesc Psychiatry. 2003;42651- 655
Asarnow  JRJaycox  LHTompson  MC Depression in youth: psychosocial interventions. J Clin Child Psychol. 2001;3033- 47
Harrington  RWhittaker  JShoebridge  PCampbell  F Systematic review of efficacy of cognitive behaviour therapies in childhood and adolescent depressive disorder. BMJ. 1998;3161559- 1563
Mufson  LWeissman  MMMoreau  DGarfinkel  R Efficacy of interpersonal psychotherapy for depressed adolescents. Arch Gen Psychiatry. 1999;56573- 579
Ollendick  THKing  NJ Empirically supported treatments for children with phobic and anxiety disorders: current status. J Clin Child Psychol. 1998;27156- 167