|Home | About | Journals | Submit | Contact Us | Français|
Scant information exists to guide pharmacological treatment of early-onset schizophrenia. We examine variation across commonly prescribed second-generation antipsychotic medications in medication discontinuation and psychiatric hospital admission among children and adolescents clinically diagnosed with schizophrenia. A 45-state Medicaid claims file (2001–2005) was analyzed focusing on outpatients, aged 6–17 years, diagnosed with schizophrenia or a related disorder prior to starting a new episode of antipsychotic monotherapy with risperidone (n = 805), olanzapine (n = 382), quetiapine (n = 260), aripiprazole (n = 173), or ziprasidone (n = 125). Cox proportional hazard regressions estimated adjusted hazard ratios of 180-day antipsychotic medication discontinuation and 180-day psychiatric hospitalization for patients treated with each medication. During the first 180 days following antipsychotic initiation, most youth treated with quetiapine (70.7%), ziprasidone (73.3%), olanzapine (73.7%), risperidone (74.7%), and aripirazole (76.5%) discontinued their medication (χ2 = 1.69, df = 4, P = .79). Compared with risperidone, the adjusted hazards of antipsychotic discontinuation did not significantly differ for any of the 4-comparator medications. The percentages of youth receiving inpatient psychiatric treatment while receiving their initial antipsychotic medication ranged from 7.19% (aripiprazole) to 9.89% (quetiapine) (χ2 = 0.79, df = 4, P = .94). As compared with risperidone, the adjusted hazard ratio of psychiatric hospital admission was 0.96 (95% CI: 0.57–1.61) for olanzapine, 1.03 (95% CI: 0.59–1.81) for quetiapine, 0.85 (95% CI: 0.43–1.70) for aripiprazole, and 1.22 (95% CI: 0.60–2.51) for ziprasidone. The results suggest that rapid antipsychotic medication discontinuation and psychiatric hospital admission are common in the community treatment of early-onset schizophrenia. No significant differences were detected in risk of either adverse outcome across 5 commonly prescribed second-generation antipsychotic medications.
Little information on treatment effectiveness exists to guide the pharmacological management of early-onset schizophrenia that is often severe,1,2 persistent3,4, and impairing.1,5 As compared with adult-onset schizophrenia, onset before age 18 years is often regarded as a more malignant form of the disorder6 with a poorer long-term course.3,7 Although short-term clinical trials suggest that the efficacy of olanzapine,8 aripiprazole,9 and risperidone10 is each superior to placebo in early-onset schizophrenia, data regarding the comparative effectiveness of these and other commonly prescribed second-generation antipsychotic medications remain scarce. The Treatment of Early-Onset in Schizophrenia Study (TEOSS), a randomized controlled 8-week trial comparing olanzapine (n = 35), risperidone (n = 41), and molindone (n = 40) for early-onset schizophrenia or schizoaffective disorder, found no significant group differences in response rates or magnitude of symptom reduction,11 although CIs in this relatively small study population were wide. It is not known how results from these prospective clinical trials, which involve carefully selected research participants, relate to outcomes experienced by relatively unselected young people diagnosed with schizophrenia in community practice. The TEOSS study, eg, excluded youth with mental retardation, current major depression, active substance abuse, and acute risk of self-harm.11 The co-occurrence of these comorbid conditions may moderate response to antipsychotic medications.
One important measure of antipsychotic effectiveness is the duration with which patients with schizophrenia continue their medications. All source antipsychotic discontinuation integrates physician, patient, and family perceptions of medication efficacy, tolerability, and safety into a single global, clinically meaningful effectiveness measure.12,13 The high frequency with which children and adolescents with schizophrenia or related disorders discontinue antipsychotic medications in short-term controlled trials raises important questions concerning the clinical effectiveness of these treatments. In TEOSS, 51% of youth treated with olanzapine, 38% of molindone patients, and 32% of those treated with risperidone did not complete 8-weeks of treatment.11 In 6-week trials, 32% of young people assigned olanzapine8 and 17% of those assigned aripiprazole9 or risperidone10 discontinued treatment before completing the study. Because these clinical trials exclude some patients at high risk of early treatment discontinuation, such as those with substance use disorders,14,15 the rates of early treatment discontinuation in clinical practice might be even greater.
Medication nonadherence is a prevalent problem in the treatment of adult schizophrenia,16 though less is known regarding community treatment patterns in early-onset schizophrenia. Antipsychotic nonadherence is related to several factors including the efficacy and tolerability of the medications12 and other treatment17 as well as patient16 characteristics. A common consequence of antipsychotic medication nonadherence is relapse and psychiatric hospitalization.18
Psychiatric hospital admission is often a painful experience for young people with schizophrenia and for their families. Although hospital admission represents only one aspect of outcome in schizophrenia, it is a clinically important and financially significant event that commonly indicates that the patient's symptoms cannot be safely managed outside of an institutional setting. Because decisions to admit patients with schizophrenia for inpatient care typically turn on a combination of clinical, behavioral, social, and resource factors, hospital admission is a broad and ecologically valid measure of clinical course. Given the uncommonness of early-onset schizophrenia19 and the relatively low frequency of inpatient admission, however, no previous study has been adequately powered to examine whether individual antipsychotic medications differentially reduce the risk of psychiatric hospital admission in this patient population.
In the following analysis, we examine community treatment patterns of children and adolescents initiating antipsychotic monotherapy following a clinical diagnosis with schizophrenia or a related mental disorder. Data from Medicaid claims are used to compare time until all source medication discontinuation and risk of psychiatric hospital admission following initiation of treatment with risperidone, olanzapine, quetiapine, ziprasidone, or aripiprazole. A greater understanding of these treatment patterns may help inform antipsychotic medication selection by psychiatrists in the management of early-onset schizophrenia.
Service and pharmacy claims data were examined from 45-state Medicaid analytic extract files (2001–2005) that did not include Arizona, Delaware, Nevada, Oregon, Rhode Island, or the District of Columbia. These files include pharmacy and service utilization information from more than 40 million individuals enrolled in the Medicaid program. Approximately 39.3% of the nations’ children were enrolled in the Medicaid program at some point during 2005.20,21 In interpreting the study results, it is important to bear in mind that variations in state Medicaid mental health care policies may influence schizophrenia treatment patterns.22 The deidentified data that were used for this study have been determined to be exempt from human subjects review by the New York State Psychiatric Institute and Rutgers University Institutional Review Boards.
We focus on youth who initiate antipsychotic treatment following an outpatient clinical diagnosis of schizophrenia, schizoaffective disorder, or schizophreniform disorder (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM]: 295). Selected patients were 6–17 years of age, enrolled in fee-for-service Medicaid plans, not enrolled in Medicare, and free of any antipsychotic prescriptions for at least 180 continuous days before filling a risperidone, olanzapine, aripiprazole, quetiapine, or ziprasidone prescription of ≤30 days supply. Patients were also required to have received at least 1 primary diagnosis for schizophrenia during the 60-day period leading up to and including the date of the new antipsychotic prescription and to be eligible for Medicaid for ≥180 days after antipsychotic initiation. The mean number of days between the last schizophrenia diagnosis and the index antipsychotic medication was 7.6 days (SD = 12.2), and for 74.7% of the sample, schizophrenia was the last mental disorder diagnosis before the index antipsychotic prescription.
Such a new treatment episode design limits bias from depletion of susceptibles and selection bias among prevalent users related to early discontinuation for drug intolerance or lack of effectiveness,23,24 while it excludes patients who are already engaged in mental treatment and are switching antipsychotic medications.
The lower age bound is based on the paucity of evidence for antipsychotic efficacy in children with schizophrenia who are younger than age 6 years,25 and the upper bound of 17 years is set in accord with the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition-TR, definition for early onset.26 Because of concerns over the consistency and validity of Medicaid managed care claims,27 the analysis was limited to patients enrolled in fee-for-service plans. The requirement for 180 days without antipsychotic prescriptions generates a new antipsychotic user cohort that permits estimation of antipsychotic treatment duration and time to hospital admission and allows for assessment of patient characteristics before treatment initiation. 28 A new antipsychotic user design also prevents bias introduced by prevalent users who may be more likely to already be tolerating an antipsychotic drug at the start of follow-up and therefore less prone to subsequent early medication discontinuation or hospital admission. Because the supply of the first antipsychotic medication may directly affect the length of the measured antipsychotic treatment continuity, study subjects were limited to first antipsychotic prescriptions with a supply of ≤30 days. A primary clinical diagnosis of schizophrenia helps to ensure that there is a recognized clinical need for continuous antipsychotic treatment; minimizes differences in disease severity between new users of the five study antipsychotic medications risperidone, olanzapine, quetiapine, aripirazole, and ziprasidone; and excludes the large number of youth who receive antipsychotic medications for diagnoses other than schizophrenia. 24
Antipsychotic medication was the independent variable of interest. On the basis of their first prescribed antipsychotic medication, patients were assigned to risperidone, olanzapine, quetiapine, aripiprazole, or ziprasidone groups. Other less commonly used antipsychotic medications were not examined. Because risperidone was the most widely prescribed antipsychotic medication in the study sample, it was assigned as the reference group.
Antipsychotic discontinuation and psychiatric hospital admission were selected as indices of diminished clinical effectiveness. We defined antipsychotic discontinuation for any cause as a gap of ≥15 days in the prescription record of the first antipsychotic medication18,29,30 or starting a second antipsychotic medication. Psychiatric hospital admission, which is a marker of clinical relapse, was defined as any inpatient treatment episode with a primary discharge diagnosis of a mental disorder (ICD-9-CM: 290–319) following the first use of the antipsychotic medication.
Several patient demographic and clinical characteristics were considered as potential confounding factors of the strength of associations between individual antipsychotic treatment and the outcomes of interest. Demographic factors include patient age (child: 6–12 years; adolescent: 13–17 years), sex (male and female), and race/ethnicity (white non-Hispanic, black non-Hispanic, and Hispanic). American Indian, Asian Pacific Islander, and patients who claim 2 or more races were considered within the white non-Hispanic group. Clinical characteristics of youth may vary across Medicaid eligibility categories. For example, children eligible for Medicaid through foster care have higher rates of psychiatric hospital admission than those eligible through supplemental security income disability or low income.31 For this reason, Medicaid eligibility category (low income, disability, or foster care placement) was also considered as a potential confounder.
Several potential confounders were defined from service and prescription claims during the 180-day period before new antipsychotic use. These include a clinical diagnosis of co-occurring psychiatric disorders under broad groups of substance use (ICD-9-CM: 291, 292, 303, 304), disruptive behavior disorders (312.0–312.4, 312.81, 312.82, 312.89, 312.9, 313.81), depressive disorders (293.83, 296.2, 296.3. 296.9, 298.0, 300.4, 311) anxiety disorders (293.84, 300.0, 300.2, 300.3, 308.3, 309.21, 309.81, 313.0, 313.2, 313.89), and pervasive developmental disorders or mental retardation (299, 317–319). Separate variables code presence or absence of 1 or more filled prescriptions for antidepressants, anxiolytics, stimulants, and mood stabilizers (lithium, carbamazepine, divalproex sodium/valproic acid/valproate sodium, lamotrigine, and oxcarbazepine) in the absence of a seizure disorder diagnosis (code: 345) during the 180-day period before new antipsychotic use. Other clinical covariates indicate whether patients received any inpatient stays, any emergency department visits, or any psychotherapy visits (current procedural terminology: 90804–90829, 90841–90847, 90849, 90853, 90855, 90857, 90875, 90876) for the treatment of mental disorders (290–319) during the 180-day prestudy period.
The 5 antipsychotic medication groups were compared with respect to each of the categorical potential confounding variables (covariates). Significant omnibus χ2 differences (P < .05, 2 tailed) were partitioned to test each pairwise group comparisons. In bivariate analyses, a difference-in-proportions test was used to compare the proportion of patients in each medication group who discontinued their antipsychotic medication during the 180 days following the index prescription. Among patients discontinuing medications during this period, ANOVA was used to test mean group differences in duration of antipsychotic treatment. Unadjusted and adjusted hazard ratios of all cause discontinuation for individual antipsychotic medications were then estimated with Cox proportional hazard regressions. The adjusted models control for variables with a priori evidence of association with antipsychotic medication discontinuation, including age, gender, comorbid substance use disorder,32–34 as well as covariates that were differentially distributed across study groups in the previous analyses (P < .1), and therefore might confound associations with antipsychotic discontinuation. Follow-up time started at the index antipsychotic prescription and stopped at discontinuation of index antipsychotic medication, psychiatric hospital admission, or day 180 after antipsychotic initiation, whichever came first. The adjusted proportional hazards model is the primary analysis because it takes account in a single model of probability of discontinuation, time to discontinuation, and differences in diagnostic histories and other observed characteristics between the treatment groups, as well as censoring due to hospitalization.
In bivariate analyses of time to hospitalization, the cumulative percentage of patients with ≥1 psychiatric inpatient admission was determined during the period of antipsychotic medication treatment. Among patients with psychiatric hospital admissions during this period, ANOVA was used to compare mean group differences in time to inpatient admission. As a sensitivity analysis, the cumulative percentage of patients with psychiatric inpatient admissions and time to first admission were determined carrying the first antipsychotic treatment period forward for the 180-day period following new antipsychotic use, regardless of treatment discontinuation. This is analogous to an intent-to-treat analysis of clinical trial data.
We estimated unadjusted and adjusted Cox proportional hazard regressions that modeled the effect of individual antipsychotic medications on time to psychiatric hospitalization. Follow-up time started at the index antipsychotic prescription and stopped at psychiatric hospital admission, discontinuation of the first antipsychotic medication, initiation of a second antipsychotic medication, or day 180 after antipsychotic initiation, whichever came first. This analysis was also then replicated using a first treatment carried forward approach to examine whether post index selection effects (potentially introduced by censoring at treatment discontinuation) may have biased the results of the primary model.
In relation to the quetiapine or aripiprazole groups, the olanzapine group included a significantly larger percentage of male patients. In addition, the olazanpine group included a significantly greater proportion of adolescents (ages 13–17 years) than the risperidone group (table 1). All 5 medication groups were racially and ethnically diverse. A significantly larger proportion of the quetiapine group than the risperidone or olanzapine groups was of white non-Hispanic ancestry. Hispanics comprised a significantly larger percentage of the aripiprazole than ziprasidone group. Roughly one half of each medication group was eligible for Medicaid on the basis of low income, and approximately one-third of each group was eligible on the basis of a disability.
Approximately one half of the patients in each medication group were diagnosed with schizophrenia with the remainder diagnosed with either schizoaffective or schizophreniform disorder. Co-occurring psychiatric diagnoses were common, especially disruptive behavior disorders and depressive disorders. A significantly larger proportion of the risperidone than olanzapine patients were also diagnosed with a disruptive behavior disorder during the 180-day period before the index antipsychotic prescription. During this period, anxiety disorders were more commonly diagnosed among patients treated with quetiapine than those treated with aripiprazole or ziprasidone and among those treated with olanzapine than aripiprazole (table 2).
During the 180-day period prior to initiating antipsychotic treatment, study patients also frequently received treatment with other psychotropic medications. Antidepressants were the most common class followed by mood stabilizers and stimulants. Significant variation was evident among antipsychotic medication groups in the percentage treated with mood stabilizers during the 180 days before the index prescription. During this period, however, the 5 antipsychotic groups did not significantly differ with respect to the percentage treated in a hospital, an emergency department, or with outpatient psychotherapy for a mental disorder (table 2).
Across the 5 groups, approximately three-quarters of patients discontinued their initial antipsychotic medication during the first 6 months of treatment (Figure 1). Among those who discontinued, the mean time to discontinuation fell within a relatively narrow range from 51.03 days for the ziprasidone group to 57.77 days for the aripiprazole group. As compared with patients treated with risperidone, there were no significant group differences in the unadjusted or adjusted time to antipsychotic discontinuation (table 3). Among patients with at least 30 days of follow-up beyond the initial drug discontinuation (N = 1291, 99.9% of all patients who discontinued medications), approximately 32.6% (421 of 1291) of patients restarted an antipsychotic medication within 30 days (data not shown).
Psychiatric hospital admissions were relatively common during the period of treatment with initial antipsychotic medication ranging from 7.19% (aripiprazole) to 9.89% (ziprasidone). None of the hazard ratios for psychiatric admission in the unadjusted and adjusted models were significantly different from the reference group (risperidone) (table 3).
As a sensitivity analysis, the percentage of study patients and time to psychiatric hospital admission was assessed carrying the first antipsychotic medication treatment period forward 180 days from the first prescription (data not shown). The percentage of patients in each medication group with a psychiatric inpatient admission during this time period ranged from 14.21% for the risperidone group to 16.06% for the quetiapine group (χ2 = 0.44, df = 4, P = .98, data not shown). In accord with the as-treated analysis, none of the hazard ratios in the unadjusted or adjusted models significantly differed from risperidone in the first treatment carried forward analysis (data not shown).
This claims-based comparative effectiveness analysis found no significant differences in time to antipsychotic treatment discontinuation or hospital admission for olanzapine, quetiapine, aripiprazole, or ziprasidone as compared with risperidone in early-onset schizophrenia. Although a failure to find significant group differences in these outcomes within an observational study should not be interpreted as establishing clinical equivalence among the 5 antipsychotic medications, the relatively large sample sizes and narrow CIs decrease the likelihood that large medication group differences exist with respect to either measure of clinical effectiveness.
No previous study has compared antipsychotic medications with respect to time to hospital admission and treatment discontinuation in early-onset schizophrenia. In one Medicaid claims-–based study of adults diagnosed with schizophrenia, no significant differences in time to psychiatric hospital admission were found among adult schizophrenia patients initiating treatment with risperidone, olanzapine, quetiapine, or ziprasidone.35
An absence of evidence for superior clinical effectiveness of one second-generation antipsychotic medication over another in early-onset schizophrenia may focus increased attention on emerging evidence from other studies on the comparative differences in the safety of antipsychotic medications in this patient population. In the TEOSS study, olanzapine resulted in significantly more weight gain than risperidone and in larger increases in serum cholesterol.11 A short-term observational study with diagnostically mixed antipsychotic-naive youth found that olanzapine had the most marked adverse effects on weight, glucose levels, and several lipid parameters, while quetiapine and risperidone increased serum triglycerides but did not result in significant deterioration of glucose or insulin measures.36 The results of the TEOSS,11 the recent observational study,36 and the present study suggest that early-onset psychosis patients are sensitive to the side effects of the second-generation antipsychotic medications. Two factors could account for this. The first is the young age of early-onset patients that renders them more susceptible and the other is that they have little or no prior antipsychotic medication treatment exposure.
The current findings may have implications for the management of Medicaid pharmacy benefits. Antipsychotic medications have become the highest cost drug class for Medicaid programs.37 As a result, several states have implemented policies to manage antipsychotic medication costs. Wide variation exists in the extent to which individual state Medicaid programs rely on prior authorization,38,39 limit the number of prescriptions, and mandate the use of generics as first line treatment40,41 to control antipsychotic medication cost. A lack of significant differences in clinical effectiveness among the 5 antipsychotic drugs in the current study may spur further efforts to manage access to high cost–branded antipsychotic medications for younger patients. Generic risperidone first became available in 2008, and as patents expire for olanzapine in 2011 and quetiapine in 2012, the number of lower cost generic options of second-generation antipsychotic medications will likely increase in the near future.
Within the first 6 months of treatment, approximately three-quarters of children and adolescents diagnosed with schizophrenia discontinued their initial antipsychotic medication, typically after about 7 or 8 weeks of treatment. High rates of treatment discontinuation and treatment nonresponse have also been reported in several clinical trials of children and adolescents with schizophrenia.8,11,42,43 The consistency of this pattern in research and community samples underscores the difficult treatment challenges posed by early-onset schizophrenia and parallels those seen in adult schizophrenia patients.12,44
The current study has several limitations. First, given the observational design, the study groups may vary in several unmeasured and therefore uncontrolled factors that influence time to antipsychotic discontinuation or psychiatric hospital admission. These include, eg, symptom severity, illness insight, social support, untreated substance use problems, and several other factors. Second, diagnoses are based on practitioner judgment and are not subject to expert validation. Research on the validity of Medicaid diagnoses of schizophrenia has been confined to adults.45,46 Third, because patients prescribed clozapine may have especially severe or treatment-resistant symptoms and more intensive mental health service use,47 we elected not to include clozapine in the current observational analysis. In this regard, it should be noted that clinical trials indicate superior antipsychotic efficacy of clozapine over haloperidol48 and over olanzapine42 in treatment-resistant early-onset schizophrenia. Fourth, there is a potential for misclassifying time of discontinuation; increasing or decreasing the required gap in the prescription record might have yielded different results. However, a sensitivity analysis in which antipsychotic medication discontinuation was defined as a 30-day gap yielded a similar pattern of results (data not shown). Claims data indicate when medication supply is scheduled to run out rather than when medications are discontinued which may occur earlier, though residual misclassification would be expected to be small and nondifferential with respect to medication group. Fifth, no attempt was made to examine antipsychotic dose in relation to the study outcomes. Sixth, although all selected subjects had at least 180 days prior to their qualifying antipsychotic medication prescription without a filled antipsychotic prescription, it is not possible to determine prior lifetime exposure to antipsychotic medications. Seventh, the study was confined to patients enrolled in the Medicaid program and may not generalize to privately insured young people and there is not sufficient power to examine outcomes within subgroups of interest such as racial/ethnic minorities.
These limitations exist in relation to advantages that result from the large sample sizes available in claims-based analyses. Because of the low prevalence of early-onset schizophrenia in clinical practice, it is unlikely that clinical trials will ever be conducted with adequate power to assess meaningful differential risk of psychiatric hospital admission. In smaller trials, negative findings may readily conceal meaningful group differences that might only become apparent in a larger study.
Little is known about the extent, timing, and distribution of risk across antipsychotic medications in treatment discontinuation and psychiatric hospital admission in early-onset schizophrenia. Because mental health care in the community markedly differs from treatment within clinical trials, questions remain regarding the extent to which efficacy as measured in clinical trials translate into clinical effectiveness as experienced in routine care. Treatment with each of the commonly prescribed second-generation antipsychotic medications was associated with a high risk of early treatment discontinuation and psychiatric hospital admission. A clear need exists to develop more effective pharmacological treatments for children and adolescents with early-onset schizophrenia and related disorders.
Agency for Healthcare Research and Quality awards U18 HS017918 and U18 HS016097; the Center for Education and Research on Mental Health Therapeutics.
The authors wish to thank Cynthia Kornegay, PhD, of the US Food and Drug Administration for her helpful comments. Disclosure: Dr Olfson has received research grants to Columbia University from Bristol Myers Squibb and Eli Lilly & Company. Dr Lieberman serves on the Advisory Board of Bioline, GlaxoSmithKline, Intracellular Therapies, Eli Lilly, Pierre Fabre, and Psychogenics. He does not receive direct financial compensation or salary support for participation in research, consulting, or advisory board activities. He receives grant support from Allon, GlaxoSmithKline, Eli Lilly, Merck, Novartis, Pfizer, Sepracor (Sunovion), and Targacept; and he holds a patent from Repligen.