- Email: [email protected]
Polypharmacy prevalence rates in the treatment of unipolar depression in an outpatient clinic
Journal of Affective Disorders 117 (2009) 18–23
Contents lists available at ScienceDirect
Journal of Affective Disorders j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a d
Research report
Polypharmacy prevalence rates in the treatment of unipolar depression in an outpatient clinic Anna Glezer a,⁎, Nancy Byatt b, Richard Cook Jr. b, Anthony J. Rothschild b,c a Massachusetts General Hospital/McLean Hospital Psychiatry Residency Training Program; University of Massachusetts Medical School. 15 Parkman Street, Wang 812 Boston, MA 02114, United States b University of Massachusetts Medical School; UMass Memorial Health Care, United States c University of Massachusetts Medical School, 361 Plantation St., Worcester, Massachusetts, 01605, United States
a r t i c l e
i n f o
Article history: Received 26 November 2007 Received in revised form 16 November 2008 Accepted 18 November 2008 Available online 21 December 2008 Keywords: Psychopharmacology Polypharmacy Depression Treatment
a b s t r a c t Objective: Unipolar depression is a complex illness with an ever increasing number of pharmacological treatments available. As the number of available medication options increases, so does the potential for polypharmacy, a practice with possible complications. Such complications include a greater number of side effects with the initiation of additional medications and the consequences of drug–drug interactions. Therefore, it is important to understand the effects of polypharmacy on efficacy of treatment as well as whether polypharmacy is instituted after appropriate initial monotherapy trials. This study attempts to answer these questions. Methods: Patient names for this investigation were provided by residents in the UMass Medical School Psychiatry Residency program. The charts of these patients were analyzed to collect data regarding demographics, clinical data about the illness, medication names, types, duration of use and to access efficacy using the Clinical Global Impression (CGI). Results: Of 160 reviewed charts, 135 subjects were included in the final analyses (others excluded due to incomplete data or no depression diagnosis). Patients were on average on 2.0 medications (SD = 0.5) with 2.1 past trials. A greater number of current antidepressant medications did not correlate with a greater improvement in CGI, implying that polypharmacy did not necessarily lead to greater efficacy. The impact of illness severity, as measured by comorbidities, substance abuse, and trauma history, were included in the analysis, and did not significantly change this outcome. Additionally, prescription patterns themselves were examined, including reasons for termination of medication trials, the impact of side-effects, and the number of patients that had complete monotherapy trials before transitioning to polypharmacy. Study limitations: Small sample size, retrospective review, one CGI rater. Conclusions: First, many patients did not receive an adequate monotherapy trial, as defined by dose and duration, before progressing to polypharmacy. Secondly, the use of two or more medications concurrently did not appear to increase efficacy as measured by CGI. Due to study limitations, direct comparison of the efficacy of one and two medications was not significant; however the study did demonstrate a general correlation between polypharmacy and greater depressive symptomatology. This suggests the need for optimizing monotherapy regimens before initiating polypharmacy that may not lead to improvement in symptoms. © 2008 Elsevier B.V. All rights reserved.
1. Introduction
⁎ Corresponding author. Tel.: +1 617 724 6300 #134 0350. E-mail addresses: [email protected] (A. Glezer), [email protected] (N. Byatt), [email protected] (A.J. Rothschild). 0165-0327/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jad.2008.11.016
Major Depressive Disorder is a complex psychiatric illness affecting 15–20% of the American population. Treatment often entails a combination of psychopharmacologic and psychotherapeutic interventions. In recent years, the number
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
of psychotropic medications available for the treatment of depression has significantly increased, providing more treatment options for patients. With the development of psychopharmacologic treatments for Major Depression, a new trend may be emerging: increased polypharmacy – the use of multiple psychiatric medications concurrently – for the treatment of depression. One investigation looking at prevalence rates over the past decades found that the percentage of patients with a diagnosed mood disorder discharged from the NIMH hospital on three or more medications was 3.3% in 1974–1979, then 9.3% in 1980–1984, rising to 34.9% in 1985–1989, and jumping to 43.8% in the period of 1990–1994 (Frye et al., 2000). Most clinicians would assert that the purpose of polypharmacy is to increase efficacy, and other reasons, including management of side effects or treating co-morbid conditions. The link, however, between polypharmacy and efficacy has not been well established. A goal of this study was to assess this relationship between increased number of antidepressant medications used concurrently and successful reduction of depressive symptoms. This question has been studied in inpatient facilities (Barbui et al., 2005) and international settings (Centorrino et al., 2002), and by examining antipsychotic medications (Ganguly et al., 2004; Janseen et al., 2004). This investigation focuses on patients diagnosed with unipolar depression at an outpatient clinic, as this environment and patient population has not previously been targeted. The out-patient clinic chosen was a resident-run clinic. Analysis of resident prescribing practices may speak to the current training they are receiving, whether polypharmacy practices are being perpetuated in the next generation of psychiatrists. Several potential complicating factors may contribute to polypharmacy. A longer duration of a major depressive episode or illness complicated by comorbid diagnoses may necessitate the use of multiple medications. These confounding factors were addressed in the development of this study. The effects of an abuse history, substance abuse, and age of illness onset were also examined as markers of severity. Other key contributors to polypharmacy, such as side-effect management, previous medication failure or inadequate response are also important, but are not directly addressed by this investigation, and this is noted under “limitations”. A second goal of this investigation was to determine whether polypharmacy was being utilized after exhaustion of monotherapy options. Several algorithms exist (Osser, 1996– 2003; TMAP, 2007) making recommendations regarding psychotropic medication therapy initiation and progression. These guidelines (and others such as those created by the APA) currently recommend starting with an SSRI to treat unipolar depression (Simon, 2001). These medications are effective for symptom reduction while having one of the lowest side-effect profiles of currently available anti-depressants and little abuse or overdose potential. One major recent investigation addressing this issue has been the STAR⁎D (Rush et al., 2006) trial, which demonstrated almost one third remission using monotherapy (citalopram), and emphasized the need for longer trials before changing treatment. Several potential downfalls exist in using complex medication regimens in the treatment of depression. Increasing health care costs and the expense of psychotropic medications (Kendrick et al., 2006; Nurnberg et al., 1999) may be prohibitive
19
to some patients and drive up costs in general. Additionally, it has been shown that as the number of prescribed medications increases, patient compliance decreases (Haynes,1976). Adding multiple medications may also increase the number of side effects a patient experiences and lead to potentially dangerous drug–drug interactions. These factors, in turn, may engender a worse outcome in the treatment of this illness. 1.1. Methodology Approval for the chart review study was obtained from the University of Massachusetts Medical School Institutional Review Board. Names of patients were provided by all third and fourth year residents in the University of Massachusetts Medical School Psychiatry Residency Program. A total of 169 names were provided by residents, having been asked for lists of current outpatients with diagnosis of unipolar depression. Charts were reviewed via the electronic medical record and paper documents. After review, 126 patients were included for analysis. Subjects were excluded if they did not have a diagnosis of unipolar depression, or if there were no records of follow-up visits after the initial intake. Each chart was assessed for demographic information – patient age, race, and gender – and to confirm diagnosis. The current or most recent Major Depressive Episode was reviewed to determine which medications the patient was taking and for what duration. Efficacy was assessed using the Clinical Global Impression Scale (CGI). Initial CGI data reflected illness severity at the start of the current episode and the final CGI reflected change by October, 2006 (when this review was conducted) or resolution of episode and symptom remission, whichever came first. Resolution and symptom remission were determined by either empiric data (ie. rating scales) or by the use of terms “in remission” found in progress notes. All CGI ratings were done by the first author. Information was gathered regarding past medication trials and side-effects experienced. The Antidepressant Treatment History Form (Oquendo et al., 2003) was used to document whether a medication trial was complete (ie. fluoxetine at 40 mg for 6 wks) or incomplete (ie. fluoxetine at any dose for less than 4 weeks). The ATHF takes into account both duration and dose. Medications were classified as antidepressants (ie. SSRIs, TCAs, mirtazepine, buproprion) or general psychotropics (antipsychotics, benzodiazepines, sleep aids). Finally, the number of current psychiatric diagnoses, substance abuse history, and a history of physical or sexual abuse were all noted if that information was available in the charts. The addition of psychotherapy to treatment regimens was not included in the review. Statistics were performed using Excel and SPSS software. Pearson correlations and linear regression were conducted to determine correlations among various ordinal and continuous factors. A statistical significance level of 0.05 was used. 2. Results 2.1. Patient descriptives Patients included in this study were on average 45.6 years old (SD 13.9), with slightly more women than men (73 vs. 53). The population was predominantly Caucasian (N = 104, 83%).
20
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
2.2. Illness descriptives The average duration of the current or most recent Major Depressive Episode was 1.90 years (SD 2.10), ranging 0.08– 16.0. The most common age of illness onset was 20–40 yrs (40%, N = 48), followed closely by diagnosis before age 20 (38%, N = 46). Nineteen percent (N = 27) were diagnosed in middle age (40–60 yrs.) and rarely (3%, N = 4) were subjects diagnosed with depression after age 60. The majority of patients had diagnoses of unipolar depression without psychosis; 16 had MDD with psychosis. Fifty percent (N = 62) had current or past histories of substance abuse and more than half of those for whom information was available regarding physical or sexual abuse disclosed having been abused (54%, N = 55). The majority of patients had at least one comorbid diagnosis – commonly an anxiety disorder. The mean CGI score at baseline was 4.26 ± 0.96, which corresponds to a “moderate” severity of illness. 2.3. Treatment impact Overall, there was an improvement in depressive symptoms with treatment. The average final CGI score was 2.56 ± 1.16, which is in the “improved” range, with 76% (N = 96) of patients reporting improvement in depression symptomatology. More specifically, 20% (N = 25) achieved symptom remission, defined by a CGI score of 1, 33% (N = 42) demonstrated marked improvement (CGI = 2), and 23% (N = 29) showed minor improvement (CGI = 3). Twenty percent (N = 25) of all patients did not show any change in CGI score. Three percent (4 patients) did show minor worsening (CGI = 5), and 0.8% (1 patient) had significantly worse symptoms (CGI = 6). The amount of improvement, or final CGI score, was independent of the initial severity, the starting CGI score. 2.4. Impact of illness severity T-tests determined no statistically significant difference in the final CGI score between patients with and without substance abuse (p = 0.09). There was also no difference between those who reported and those who denied a history of abuse (p = 0.26). Mean final CGI scores did not significantly differ between those with MDD with psychosis and those without psychosis (p = 0.06). The majority of subjects had one or more comorbid diagnoses (78%). The mean final CGI score comparing those
Fig. 1. Statistically significant positive Pearson correlation (0.28) between the number of total psychotropic medications a patient is taking and the final CGI score p = 0.036, implying that greater number of medications is associated with worse outcomes. ANOVA not statistically significant p = 0.181.
Fig. 2. Statistically significant positive Pearson correlation (p = 0.009) between the number of antidepressant medications a patient is taking and the final CGI score. ANOVA significant at p = 0.013. Post-hoc analysis did not demonstrate significant differences between none, one, and two medications.
with and without comorbid diagnoses did differ (one-way ANOVA p = 0.035). Patients with two or more comorbid diagnoses had significantly higher mean final CGI scores (2.92 ± 1.14) when compared with patients without any comorbid diagnosis (2.31 ± 1.15) (p = 0.04). There was no significant correlation between the age of illness onset and the starting CGI, the marker of illness severity (p = 0.50). No statistically significant relationship was noted between baseline CGI scores and final CGI scores (p = 0.24), implying that the severity at the start of the depressive episode did not impact the final outcomes. 2.5. Medication impact Medication regimens were assessed based on total number of psychotropic medications a patient was taking, the number of antidepressant medications, and the number of antidepressant medications at full dose, as defined by the ATHF. Patients were prescribed a mean of 1.98 (SD 0.94) psychotropic medications, 1.15 (SD 0.52) antidepressants, and 0.64 (SD 0.59) full dosed antidepressants. Fifty-six percent (N = 70) of patients were on one full dose of an antidepressant medication. Forty percent (N = 51) did not have a full dose of any antidepressant medication.
Fig. 3. Positive Pearson correlation coefficient (0.15) between the number of antidepressant medications at full trial dosing a patient is taking and the final CGI score. ANOVA significant at p = 0.019. Post-hoc analysis did not demonstrate significant differences between none, one, and two medications.
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
There was a statistically significant positive correlation (p = 0.028) between the number of total medications a patient was taking and the final CGI, and a statistically significant positive correlation (p = 0.009) between the number of antidepressant medications and a higher final CGI score. There was no statistically significant correlation between number of antidepressants at full dose a patient was prescribed and the final CGI, but the correlation trended positively (correlation coefficient = 0.15). No relationship existed between total number of psychotropic medications and final CGI score (ANOVA p = 0.181) Fig. 1. However, there was a statistically significant association between the final CGI score and total number of antidepressant therapies (ANOVA p = 0.013) (Fig. 2). There also was a statistically significant association between the final CGI score and the total number of full-trial antidepressant therapies (ANOVA = 0.019) (Fig. 3). The relationship between illness severity, as measured by starting CGI score, and the number of current and past medication trials was analyzed. While there was no significant relationship between starting CGI and the number of current antidepressant medications prescribed (p = 0.48), there was a highly significant correlation between the starting CGI and both the number of full dose antidepressant medications a patient was currently prescribed (p = 0.004) and the total number of psychotropic medications a patient was currently prescribed (p b 0.001). There were no relationships between starting CGI scores and number of past medication trials of full dose or incomplete dose antidepressants. Furthermore, the presence of psychosis in the diagnosis did significantly correlate with an increased total number of active psychotropic medications (p = 0.024), explained the by the addition of an antipsychotic medication to the regimen. An increased number of comorbid diagnoses was also significantly correlated with a higher number of past antidepressant complete trials (p = 0.041). The number of comorbid diagnoses did not impact the final CGI score. When controlling for the number of comorbid diagnoses, there remained a significant positive correlation between the final CGI score and both the number of prescribed antidepressants (p = 0.009) and the total number of psychotropic medications (p = 0.036). This implies that the prior finding of a worse outcome with greater numbers of medications is not explained by the presumption that patients with more comorbid diagnoses, a marker of severity, are prescribed more medications. Lastly, medication prescription patterns themselves were assessed, such as number of complete or incomplete trials, the use of SSRIs versus other antidepressants, and the impact of side effects on medication use. The mean number of incomplete or unknown trials reported was 1.71 (SD 1.65), ranging from 0–7, with 74% (N = 88) of patients having one or more incomplete trials of antidepressant medications or trials without known outcomes. However, 33% of patients (N = 44) had had at least one complete past medication trial (adequate dose and duration). Nineteen percent of patients (N = 24) in this study had no previous medication trials. Additionally, patients had an average of 2.17 (SD 1.16) trials (complete or incomplete) of an SSRI/SNRI medication and 1.21 (SD 1.29) trials of other antidepressant medications. Only 2 patients did not have any past or active trials of an SSRI/SNRI medication.
21
Data was collected regarding number of antidepressant medication trials stopped due to side effects. Almost 40% (N = 50) of subjects had a history of discontinuing one or more medications secondary to side-effects. Of those that did terminate a trial, 62% (N = 31) had one truncated medication trial due to side-effects, 16% (N = 8) had discontinued two trials, and 22% (N = 11) had stopped three or more medication trials. Overall, 215 past trials of medications were discontinued without a known complete trial, and 88 (41%) total trials were discontinued secondary to side-effects. While the final outcome as measured by CGI did not correlate with number of medications discontinued in the past due to sideeffects (p = 0.59), an increased number of trials terminated due to side-effects did significantly correlate with an increased number of past medication trials (p b 0.001). Clearly, an increased number of side-effects may lead to greater frequency of premature medication trial discontinuation, thereby increasing the numbers of incomplete medication trials. However, there was insufficient data to determine whether side-effects were more likely to be experienced as the number of concurrent psychotropic medications prescribed increased. 3. Discussion Rates of polypharmacy prescription in the treatment of depression have been increasing. This study was conducted to elucidate the pattern of transition from first line treatments to polypharmacy and to evaluate the relationship between the use of multiple concurrent psychotropic medications and treatment efficacy. The first main goal of the investigation – to address prescribing patterns – was attained. The hope was to determine if patients are undergoing trials of first-line therapies like SSRIs before progressing to more complex regimens. In this study sample, the majority of patients (98%) had undergone a trial of an SSRI/SNRI, which is encouraging as these medications are often recommended as first-line choices. On the other hand, almost three quarters of patients (74%) had experienced at least one incomplete trial or one without a known outcome. More startling was that almost half (40%) of patients did not receive a full dose of any medication. Coupled with these results was that patients were, on average, prescribed two (mean = 1.98) psychotropic medications and more than one (mean = 1.15) antidepressant. This suggests that patients had medications added to their regimen without optimizing the initial treatments, leading to polypharmacy. While possible that some patients were started on augmentation therapies to boost response, as reported to be efficacious in other clinical trials like STAR⁎D (Trivedi et al., 2006), these strategies are recommended after exhaustion of monotherapy. The experience of side-effects is a common reason for terminating a medication trial. However, less than half of all incomplete trials (41%) were those stopped secondary to side-effects. Other reasons for premature trial cessation included self-discontinuation by the patient due to cost, lack of desire to continue medication, or running out of refills. Oftentimes, the reason was unknown or the data unavailable. Additionally, the data suggest a relationship between treatment outcome and number of prescribed medications.
22
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
There was a significant correlation between a worse treatment outcome (as defined by the CGI score) and both the total number of psychotropic medications and the number of antidepressants. This may be evidence that polypharmacy does not improve clinical outcomes. One question introduced with such investigations is how to distinguish between patient severity and polypharmacy as both affect treatment response. That is, patients prescribed more medications may inherently have a more complicated course of illness, thereby explaining the worse outcome. This study attempted to address this issue by isolating certain features of clinical illness that lead to a more severe picture. A history of substance abuse, physical or sexual abuse, and presence of psychosis were not associated with worse outcomes. Nor were these factors associated with greater numbers of prescribed medications – except that a diagnosis of MDD with psychosis correlated with greater number of total psychotropic medications. Controlling for the presence of comorbid diagnoses led to similar conclusions, namely that worse outcomes were associated with greater numbers of antidepressants and psychotropic medications in general. It has been suggested by prior investigations that patients suffering from treatment resistant depression benefit from polypharmacy prescription at the outset. However, this was not the targeted population in this study. One potential future direction for investigation is delving more deeply into treatment resistant depression and how other measures of severity impact treatment outcome and the use of multiple medications.
advantage to polypharmacy in the treatment of unipolar depression versus an adequate monotherapy trial of a different FDA-approved medication. Until those studies are conducted, the results of this investigation should prompt clinicians to assess their prescribing patterns and evaluate the medication regimens of patients on polypharmacy and those whose monotherapy regimens may be suboptimal. Role of funding sources There was no funding provided for this research as it was primarily done during the primary investigator's time in medical school. Conflicts of interest and author disclosures Dr. Anthony J. Rothschild has received research support from NIMH, Wyeth, Novartis, and Cyberonics; served as a consultant to Pfizer, Glaxo, Forest, and Lilly; and has received honoraria and served on the speakers bureaus of Bristol-Myers, Lilly, Pfizer, and Forest. Drs. Anna Glezer, Nancy Byatt, and Richard Cook do not have any disclosures.
Acknowledgements The authors of this research paper would like to acknowledge the support staff of the Center for Psychopharmacologic Research and Treatment at the University of Massachusetts Medical School as well as those involved in the Senior Scholars Research Program at the University of Massachusetts Medical School. We would also like to thank the residents of the Psychiatry Training Program who provided patient names and data for this project.
3.1. Study limitations References Several methodological shortcomings limit this investigation. First, this was a retrospective chart review with a relatively small N, making it less ideal for deciphering the roots behind the relationship between polypharmacy and treatment efficacy, although valid to describe medication prescription patterns. Additionally, only one rater was used to gauge CGI, which promotes consistency, but cannot speak to inter-rater reliability of the scores. Finally, several confounds were addressed to better elucidate the relationship between polypharmacy and treatment efficacy, such as the presence of psychosis, substance abuse, and trauma history. However, this information was not present on every patient, and other confounding factors may exist that could be addressed in future investigations, such as additional medical or psychiatric co-morbidities. The impact of psychotherapy was not dealt with in this study, and can be another avenue of future investigation. 4. Conclusions The increasingly common practice of polypharmacy for treatment of mental illness is concerning given the potential increased risk of side-effects, decreased compliance, and increased costs. This investigation demonstrated that patients often receive incomplete trials of medication before advancing to more complicated regimens. Additionally, greater numbers of antidepressants and psychotropic medications in general are not necessarily correlated with improved outcomes in the outpatient treatment of unipolar depression. What requires further investigation is whether there is any
Barbui, C., Ciuna, A., Nose, M., Levi, D., Andretta, M., Patten, S.B., Amaddeo, F., Tansella, M., 2005. Drug treatment modalities in psychiatric inpatient practice – a 20-year comparison. Eur. Arch. Psychiatry Clin. Neurosci. 255, 136–142. Centorrino, F., Eakin, M., Bahk, W., Kelleher, J., Goren, J., Salvatore, P., Egli, S., Baldessarini, R., 2002. Inpatient antipsychotic drug use in 1998, 1993, and 1989. Am. J. Psychiatry 159, 1932–1935. Frye, M.A., Ketter, T.A., Leverich, G.S., Huggins, T., Lantz, C., Denicoff, K.D., Post, R., 2000. The increasing use of polypharmacotherapy for refractory mood disorders: 22 years of study. J. Clin. Psychiatry 61, 9–15. Ganguly, R., Kotzan, J.A., Miller, L.S., Kennedy, K., Martin, B.C., 2004. Prevalence, trends, and factors associated with antipsychotic polypharmacy among medicaid-eligible schizophrenia patients, 1998–2000. J. Clin. Psychiatry 65 (10), 1377–1388. Haynes, R., 1976. A critical review of the ‘determinants’ of patient compliance with therapeutic regimens. In: Sackett, D., Haynes, R. (Eds.), Compliance with Therapeutic Regimens. Johns Hopkins University Press, Baltimore. Janseen, B., Weinmann, S., Berger, M., Gaebel, W., 2004. Validation of polypharmacy process measures in inpatient schizophrenia care. Schizophr. Bull. 30 (4), 1023–1033. Kendrick, T., Peveler, R., Longworth, L., Baldwin, D., Moore, M., Chatwin, J., Thornett, A., Goddard, J., Campbell, M., Smith, H., Buxton, M., Thompson, C., 2006. Cost-effectiveness and cost-utility of tricyclic antidepressants, selective serotonin reuptake inhibitors and lefepramine. Br. J. Psychiatry 188, 337–345. Nurnberg, H.G., Hensley, P.L., Thompson, P.M., Paine, S.S., 1999. Modeling the pharmacoeconomic cost of three selective serotonin reuptake inhibitors. Psychiatr. Serv. 50 (10), 1351–1353. Oquendo, M.A., Baca-Garcia, E., Kartachov, A., Khait, V., Campbell, C.E., Richards, M., Sackeim, H., Prudic, J., Mann, J., 2003. A computer algorithm for calculating the adequacy of antidepressant treatment in unipolar and bipolar depression. J. Clin. Psychiatry 64, 825–833. Osser, D.N., 1996–2003. Algorithm for the pharmacotherapy of depression. Mental Health Connections, Inc., Lexington, MA. http://www.mhc.com/ Algorithms/Depression. Rush, A.J., Trivedi, M.H., Wisniewski, S.R., Nierenberg, A.A., Stewart, J.W., Warden, D., Niederehe, G., Thase, M.E., Lavori, P.W., Lebowitz, B.D., McGrath,
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23 P.J., Rosenbaum, J.F., Sackeim, H.A., Kupfer, D.J., Luther, J., Fava, M., 2006. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR⁎D report. Am. J. Psychiatry 163 (11), 1905–1917. Simon, G., 2001. Choosing a first-line antidepressant. JAMA 286, 3003–3004. Texas Medication Algorithm Project, 2007. http://www.dshs.state.tx.us/ mhprograms/TMAP.htm.
23
Trivedi, M.H., Fava, M., Wisniewski, S.R., Thase, M.E., Quitkin, F., Warden, D., Ritz, L., Nierenberg, A.A., Lebowitz, B.D., Biggs, M.M., Luther, J.F., ShoresWilson, K., Rush, A.J., 2006. Medication augmentation after the failure of SSRIs for depression. N Engl J Med 23;354(12):1243–52.
Contents lists available at ScienceDirect
Journal of Affective Disorders j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j a d
Research report
Polypharmacy prevalence rates in the treatment of unipolar depression in an outpatient clinic Anna Glezer a,⁎, Nancy Byatt b, Richard Cook Jr. b, Anthony J. Rothschild b,c a Massachusetts General Hospital/McLean Hospital Psychiatry Residency Training Program; University of Massachusetts Medical School. 15 Parkman Street, Wang 812 Boston, MA 02114, United States b University of Massachusetts Medical School; UMass Memorial Health Care, United States c University of Massachusetts Medical School, 361 Plantation St., Worcester, Massachusetts, 01605, United States
a r t i c l e
i n f o
Article history: Received 26 November 2007 Received in revised form 16 November 2008 Accepted 18 November 2008 Available online 21 December 2008 Keywords: Psychopharmacology Polypharmacy Depression Treatment
a b s t r a c t Objective: Unipolar depression is a complex illness with an ever increasing number of pharmacological treatments available. As the number of available medication options increases, so does the potential for polypharmacy, a practice with possible complications. Such complications include a greater number of side effects with the initiation of additional medications and the consequences of drug–drug interactions. Therefore, it is important to understand the effects of polypharmacy on efficacy of treatment as well as whether polypharmacy is instituted after appropriate initial monotherapy trials. This study attempts to answer these questions. Methods: Patient names for this investigation were provided by residents in the UMass Medical School Psychiatry Residency program. The charts of these patients were analyzed to collect data regarding demographics, clinical data about the illness, medication names, types, duration of use and to access efficacy using the Clinical Global Impression (CGI). Results: Of 160 reviewed charts, 135 subjects were included in the final analyses (others excluded due to incomplete data or no depression diagnosis). Patients were on average on 2.0 medications (SD = 0.5) with 2.1 past trials. A greater number of current antidepressant medications did not correlate with a greater improvement in CGI, implying that polypharmacy did not necessarily lead to greater efficacy. The impact of illness severity, as measured by comorbidities, substance abuse, and trauma history, were included in the analysis, and did not significantly change this outcome. Additionally, prescription patterns themselves were examined, including reasons for termination of medication trials, the impact of side-effects, and the number of patients that had complete monotherapy trials before transitioning to polypharmacy. Study limitations: Small sample size, retrospective review, one CGI rater. Conclusions: First, many patients did not receive an adequate monotherapy trial, as defined by dose and duration, before progressing to polypharmacy. Secondly, the use of two or more medications concurrently did not appear to increase efficacy as measured by CGI. Due to study limitations, direct comparison of the efficacy of one and two medications was not significant; however the study did demonstrate a general correlation between polypharmacy and greater depressive symptomatology. This suggests the need for optimizing monotherapy regimens before initiating polypharmacy that may not lead to improvement in symptoms. © 2008 Elsevier B.V. All rights reserved.
1. Introduction
⁎ Corresponding author. Tel.: +1 617 724 6300 #134 0350. E-mail addresses: [email protected] (A. Glezer), [email protected] (N. Byatt), [email protected] (A.J. Rothschild). 0165-0327/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jad.2008.11.016
Major Depressive Disorder is a complex psychiatric illness affecting 15–20% of the American population. Treatment often entails a combination of psychopharmacologic and psychotherapeutic interventions. In recent years, the number
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
of psychotropic medications available for the treatment of depression has significantly increased, providing more treatment options for patients. With the development of psychopharmacologic treatments for Major Depression, a new trend may be emerging: increased polypharmacy – the use of multiple psychiatric medications concurrently – for the treatment of depression. One investigation looking at prevalence rates over the past decades found that the percentage of patients with a diagnosed mood disorder discharged from the NIMH hospital on three or more medications was 3.3% in 1974–1979, then 9.3% in 1980–1984, rising to 34.9% in 1985–1989, and jumping to 43.8% in the period of 1990–1994 (Frye et al., 2000). Most clinicians would assert that the purpose of polypharmacy is to increase efficacy, and other reasons, including management of side effects or treating co-morbid conditions. The link, however, between polypharmacy and efficacy has not been well established. A goal of this study was to assess this relationship between increased number of antidepressant medications used concurrently and successful reduction of depressive symptoms. This question has been studied in inpatient facilities (Barbui et al., 2005) and international settings (Centorrino et al., 2002), and by examining antipsychotic medications (Ganguly et al., 2004; Janseen et al., 2004). This investigation focuses on patients diagnosed with unipolar depression at an outpatient clinic, as this environment and patient population has not previously been targeted. The out-patient clinic chosen was a resident-run clinic. Analysis of resident prescribing practices may speak to the current training they are receiving, whether polypharmacy practices are being perpetuated in the next generation of psychiatrists. Several potential complicating factors may contribute to polypharmacy. A longer duration of a major depressive episode or illness complicated by comorbid diagnoses may necessitate the use of multiple medications. These confounding factors were addressed in the development of this study. The effects of an abuse history, substance abuse, and age of illness onset were also examined as markers of severity. Other key contributors to polypharmacy, such as side-effect management, previous medication failure or inadequate response are also important, but are not directly addressed by this investigation, and this is noted under “limitations”. A second goal of this investigation was to determine whether polypharmacy was being utilized after exhaustion of monotherapy options. Several algorithms exist (Osser, 1996– 2003; TMAP, 2007) making recommendations regarding psychotropic medication therapy initiation and progression. These guidelines (and others such as those created by the APA) currently recommend starting with an SSRI to treat unipolar depression (Simon, 2001). These medications are effective for symptom reduction while having one of the lowest side-effect profiles of currently available anti-depressants and little abuse or overdose potential. One major recent investigation addressing this issue has been the STAR⁎D (Rush et al., 2006) trial, which demonstrated almost one third remission using monotherapy (citalopram), and emphasized the need for longer trials before changing treatment. Several potential downfalls exist in using complex medication regimens in the treatment of depression. Increasing health care costs and the expense of psychotropic medications (Kendrick et al., 2006; Nurnberg et al., 1999) may be prohibitive
19
to some patients and drive up costs in general. Additionally, it has been shown that as the number of prescribed medications increases, patient compliance decreases (Haynes,1976). Adding multiple medications may also increase the number of side effects a patient experiences and lead to potentially dangerous drug–drug interactions. These factors, in turn, may engender a worse outcome in the treatment of this illness. 1.1. Methodology Approval for the chart review study was obtained from the University of Massachusetts Medical School Institutional Review Board. Names of patients were provided by all third and fourth year residents in the University of Massachusetts Medical School Psychiatry Residency Program. A total of 169 names were provided by residents, having been asked for lists of current outpatients with diagnosis of unipolar depression. Charts were reviewed via the electronic medical record and paper documents. After review, 126 patients were included for analysis. Subjects were excluded if they did not have a diagnosis of unipolar depression, or if there were no records of follow-up visits after the initial intake. Each chart was assessed for demographic information – patient age, race, and gender – and to confirm diagnosis. The current or most recent Major Depressive Episode was reviewed to determine which medications the patient was taking and for what duration. Efficacy was assessed using the Clinical Global Impression Scale (CGI). Initial CGI data reflected illness severity at the start of the current episode and the final CGI reflected change by October, 2006 (when this review was conducted) or resolution of episode and symptom remission, whichever came first. Resolution and symptom remission were determined by either empiric data (ie. rating scales) or by the use of terms “in remission” found in progress notes. All CGI ratings were done by the first author. Information was gathered regarding past medication trials and side-effects experienced. The Antidepressant Treatment History Form (Oquendo et al., 2003) was used to document whether a medication trial was complete (ie. fluoxetine at 40 mg for 6 wks) or incomplete (ie. fluoxetine at any dose for less than 4 weeks). The ATHF takes into account both duration and dose. Medications were classified as antidepressants (ie. SSRIs, TCAs, mirtazepine, buproprion) or general psychotropics (antipsychotics, benzodiazepines, sleep aids). Finally, the number of current psychiatric diagnoses, substance abuse history, and a history of physical or sexual abuse were all noted if that information was available in the charts. The addition of psychotherapy to treatment regimens was not included in the review. Statistics were performed using Excel and SPSS software. Pearson correlations and linear regression were conducted to determine correlations among various ordinal and continuous factors. A statistical significance level of 0.05 was used. 2. Results 2.1. Patient descriptives Patients included in this study were on average 45.6 years old (SD 13.9), with slightly more women than men (73 vs. 53). The population was predominantly Caucasian (N = 104, 83%).
20
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
2.2. Illness descriptives The average duration of the current or most recent Major Depressive Episode was 1.90 years (SD 2.10), ranging 0.08– 16.0. The most common age of illness onset was 20–40 yrs (40%, N = 48), followed closely by diagnosis before age 20 (38%, N = 46). Nineteen percent (N = 27) were diagnosed in middle age (40–60 yrs.) and rarely (3%, N = 4) were subjects diagnosed with depression after age 60. The majority of patients had diagnoses of unipolar depression without psychosis; 16 had MDD with psychosis. Fifty percent (N = 62) had current or past histories of substance abuse and more than half of those for whom information was available regarding physical or sexual abuse disclosed having been abused (54%, N = 55). The majority of patients had at least one comorbid diagnosis – commonly an anxiety disorder. The mean CGI score at baseline was 4.26 ± 0.96, which corresponds to a “moderate” severity of illness. 2.3. Treatment impact Overall, there was an improvement in depressive symptoms with treatment. The average final CGI score was 2.56 ± 1.16, which is in the “improved” range, with 76% (N = 96) of patients reporting improvement in depression symptomatology. More specifically, 20% (N = 25) achieved symptom remission, defined by a CGI score of 1, 33% (N = 42) demonstrated marked improvement (CGI = 2), and 23% (N = 29) showed minor improvement (CGI = 3). Twenty percent (N = 25) of all patients did not show any change in CGI score. Three percent (4 patients) did show minor worsening (CGI = 5), and 0.8% (1 patient) had significantly worse symptoms (CGI = 6). The amount of improvement, or final CGI score, was independent of the initial severity, the starting CGI score. 2.4. Impact of illness severity T-tests determined no statistically significant difference in the final CGI score between patients with and without substance abuse (p = 0.09). There was also no difference between those who reported and those who denied a history of abuse (p = 0.26). Mean final CGI scores did not significantly differ between those with MDD with psychosis and those without psychosis (p = 0.06). The majority of subjects had one or more comorbid diagnoses (78%). The mean final CGI score comparing those
Fig. 1. Statistically significant positive Pearson correlation (0.28) between the number of total psychotropic medications a patient is taking and the final CGI score p = 0.036, implying that greater number of medications is associated with worse outcomes. ANOVA not statistically significant p = 0.181.
Fig. 2. Statistically significant positive Pearson correlation (p = 0.009) between the number of antidepressant medications a patient is taking and the final CGI score. ANOVA significant at p = 0.013. Post-hoc analysis did not demonstrate significant differences between none, one, and two medications.
with and without comorbid diagnoses did differ (one-way ANOVA p = 0.035). Patients with two or more comorbid diagnoses had significantly higher mean final CGI scores (2.92 ± 1.14) when compared with patients without any comorbid diagnosis (2.31 ± 1.15) (p = 0.04). There was no significant correlation between the age of illness onset and the starting CGI, the marker of illness severity (p = 0.50). No statistically significant relationship was noted between baseline CGI scores and final CGI scores (p = 0.24), implying that the severity at the start of the depressive episode did not impact the final outcomes. 2.5. Medication impact Medication regimens were assessed based on total number of psychotropic medications a patient was taking, the number of antidepressant medications, and the number of antidepressant medications at full dose, as defined by the ATHF. Patients were prescribed a mean of 1.98 (SD 0.94) psychotropic medications, 1.15 (SD 0.52) antidepressants, and 0.64 (SD 0.59) full dosed antidepressants. Fifty-six percent (N = 70) of patients were on one full dose of an antidepressant medication. Forty percent (N = 51) did not have a full dose of any antidepressant medication.
Fig. 3. Positive Pearson correlation coefficient (0.15) between the number of antidepressant medications at full trial dosing a patient is taking and the final CGI score. ANOVA significant at p = 0.019. Post-hoc analysis did not demonstrate significant differences between none, one, and two medications.
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
There was a statistically significant positive correlation (p = 0.028) between the number of total medications a patient was taking and the final CGI, and a statistically significant positive correlation (p = 0.009) between the number of antidepressant medications and a higher final CGI score. There was no statistically significant correlation between number of antidepressants at full dose a patient was prescribed and the final CGI, but the correlation trended positively (correlation coefficient = 0.15). No relationship existed between total number of psychotropic medications and final CGI score (ANOVA p = 0.181) Fig. 1. However, there was a statistically significant association between the final CGI score and total number of antidepressant therapies (ANOVA p = 0.013) (Fig. 2). There also was a statistically significant association between the final CGI score and the total number of full-trial antidepressant therapies (ANOVA = 0.019) (Fig. 3). The relationship between illness severity, as measured by starting CGI score, and the number of current and past medication trials was analyzed. While there was no significant relationship between starting CGI and the number of current antidepressant medications prescribed (p = 0.48), there was a highly significant correlation between the starting CGI and both the number of full dose antidepressant medications a patient was currently prescribed (p = 0.004) and the total number of psychotropic medications a patient was currently prescribed (p b 0.001). There were no relationships between starting CGI scores and number of past medication trials of full dose or incomplete dose antidepressants. Furthermore, the presence of psychosis in the diagnosis did significantly correlate with an increased total number of active psychotropic medications (p = 0.024), explained the by the addition of an antipsychotic medication to the regimen. An increased number of comorbid diagnoses was also significantly correlated with a higher number of past antidepressant complete trials (p = 0.041). The number of comorbid diagnoses did not impact the final CGI score. When controlling for the number of comorbid diagnoses, there remained a significant positive correlation between the final CGI score and both the number of prescribed antidepressants (p = 0.009) and the total number of psychotropic medications (p = 0.036). This implies that the prior finding of a worse outcome with greater numbers of medications is not explained by the presumption that patients with more comorbid diagnoses, a marker of severity, are prescribed more medications. Lastly, medication prescription patterns themselves were assessed, such as number of complete or incomplete trials, the use of SSRIs versus other antidepressants, and the impact of side effects on medication use. The mean number of incomplete or unknown trials reported was 1.71 (SD 1.65), ranging from 0–7, with 74% (N = 88) of patients having one or more incomplete trials of antidepressant medications or trials without known outcomes. However, 33% of patients (N = 44) had had at least one complete past medication trial (adequate dose and duration). Nineteen percent of patients (N = 24) in this study had no previous medication trials. Additionally, patients had an average of 2.17 (SD 1.16) trials (complete or incomplete) of an SSRI/SNRI medication and 1.21 (SD 1.29) trials of other antidepressant medications. Only 2 patients did not have any past or active trials of an SSRI/SNRI medication.
21
Data was collected regarding number of antidepressant medication trials stopped due to side effects. Almost 40% (N = 50) of subjects had a history of discontinuing one or more medications secondary to side-effects. Of those that did terminate a trial, 62% (N = 31) had one truncated medication trial due to side-effects, 16% (N = 8) had discontinued two trials, and 22% (N = 11) had stopped three or more medication trials. Overall, 215 past trials of medications were discontinued without a known complete trial, and 88 (41%) total trials were discontinued secondary to side-effects. While the final outcome as measured by CGI did not correlate with number of medications discontinued in the past due to sideeffects (p = 0.59), an increased number of trials terminated due to side-effects did significantly correlate with an increased number of past medication trials (p b 0.001). Clearly, an increased number of side-effects may lead to greater frequency of premature medication trial discontinuation, thereby increasing the numbers of incomplete medication trials. However, there was insufficient data to determine whether side-effects were more likely to be experienced as the number of concurrent psychotropic medications prescribed increased. 3. Discussion Rates of polypharmacy prescription in the treatment of depression have been increasing. This study was conducted to elucidate the pattern of transition from first line treatments to polypharmacy and to evaluate the relationship between the use of multiple concurrent psychotropic medications and treatment efficacy. The first main goal of the investigation – to address prescribing patterns – was attained. The hope was to determine if patients are undergoing trials of first-line therapies like SSRIs before progressing to more complex regimens. In this study sample, the majority of patients (98%) had undergone a trial of an SSRI/SNRI, which is encouraging as these medications are often recommended as first-line choices. On the other hand, almost three quarters of patients (74%) had experienced at least one incomplete trial or one without a known outcome. More startling was that almost half (40%) of patients did not receive a full dose of any medication. Coupled with these results was that patients were, on average, prescribed two (mean = 1.98) psychotropic medications and more than one (mean = 1.15) antidepressant. This suggests that patients had medications added to their regimen without optimizing the initial treatments, leading to polypharmacy. While possible that some patients were started on augmentation therapies to boost response, as reported to be efficacious in other clinical trials like STAR⁎D (Trivedi et al., 2006), these strategies are recommended after exhaustion of monotherapy. The experience of side-effects is a common reason for terminating a medication trial. However, less than half of all incomplete trials (41%) were those stopped secondary to side-effects. Other reasons for premature trial cessation included self-discontinuation by the patient due to cost, lack of desire to continue medication, or running out of refills. Oftentimes, the reason was unknown or the data unavailable. Additionally, the data suggest a relationship between treatment outcome and number of prescribed medications.
22
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23
There was a significant correlation between a worse treatment outcome (as defined by the CGI score) and both the total number of psychotropic medications and the number of antidepressants. This may be evidence that polypharmacy does not improve clinical outcomes. One question introduced with such investigations is how to distinguish between patient severity and polypharmacy as both affect treatment response. That is, patients prescribed more medications may inherently have a more complicated course of illness, thereby explaining the worse outcome. This study attempted to address this issue by isolating certain features of clinical illness that lead to a more severe picture. A history of substance abuse, physical or sexual abuse, and presence of psychosis were not associated with worse outcomes. Nor were these factors associated with greater numbers of prescribed medications – except that a diagnosis of MDD with psychosis correlated with greater number of total psychotropic medications. Controlling for the presence of comorbid diagnoses led to similar conclusions, namely that worse outcomes were associated with greater numbers of antidepressants and psychotropic medications in general. It has been suggested by prior investigations that patients suffering from treatment resistant depression benefit from polypharmacy prescription at the outset. However, this was not the targeted population in this study. One potential future direction for investigation is delving more deeply into treatment resistant depression and how other measures of severity impact treatment outcome and the use of multiple medications.
advantage to polypharmacy in the treatment of unipolar depression versus an adequate monotherapy trial of a different FDA-approved medication. Until those studies are conducted, the results of this investigation should prompt clinicians to assess their prescribing patterns and evaluate the medication regimens of patients on polypharmacy and those whose monotherapy regimens may be suboptimal. Role of funding sources There was no funding provided for this research as it was primarily done during the primary investigator's time in medical school. Conflicts of interest and author disclosures Dr. Anthony J. Rothschild has received research support from NIMH, Wyeth, Novartis, and Cyberonics; served as a consultant to Pfizer, Glaxo, Forest, and Lilly; and has received honoraria and served on the speakers bureaus of Bristol-Myers, Lilly, Pfizer, and Forest. Drs. Anna Glezer, Nancy Byatt, and Richard Cook do not have any disclosures.
Acknowledgements The authors of this research paper would like to acknowledge the support staff of the Center for Psychopharmacologic Research and Treatment at the University of Massachusetts Medical School as well as those involved in the Senior Scholars Research Program at the University of Massachusetts Medical School. We would also like to thank the residents of the Psychiatry Training Program who provided patient names and data for this project.
3.1. Study limitations References Several methodological shortcomings limit this investigation. First, this was a retrospective chart review with a relatively small N, making it less ideal for deciphering the roots behind the relationship between polypharmacy and treatment efficacy, although valid to describe medication prescription patterns. Additionally, only one rater was used to gauge CGI, which promotes consistency, but cannot speak to inter-rater reliability of the scores. Finally, several confounds were addressed to better elucidate the relationship between polypharmacy and treatment efficacy, such as the presence of psychosis, substance abuse, and trauma history. However, this information was not present on every patient, and other confounding factors may exist that could be addressed in future investigations, such as additional medical or psychiatric co-morbidities. The impact of psychotherapy was not dealt with in this study, and can be another avenue of future investigation. 4. Conclusions The increasingly common practice of polypharmacy for treatment of mental illness is concerning given the potential increased risk of side-effects, decreased compliance, and increased costs. This investigation demonstrated that patients often receive incomplete trials of medication before advancing to more complicated regimens. Additionally, greater numbers of antidepressants and psychotropic medications in general are not necessarily correlated with improved outcomes in the outpatient treatment of unipolar depression. What requires further investigation is whether there is any
Barbui, C., Ciuna, A., Nose, M., Levi, D., Andretta, M., Patten, S.B., Amaddeo, F., Tansella, M., 2005. Drug treatment modalities in psychiatric inpatient practice – a 20-year comparison. Eur. Arch. Psychiatry Clin. Neurosci. 255, 136–142. Centorrino, F., Eakin, M., Bahk, W., Kelleher, J., Goren, J., Salvatore, P., Egli, S., Baldessarini, R., 2002. Inpatient antipsychotic drug use in 1998, 1993, and 1989. Am. J. Psychiatry 159, 1932–1935. Frye, M.A., Ketter, T.A., Leverich, G.S., Huggins, T., Lantz, C., Denicoff, K.D., Post, R., 2000. The increasing use of polypharmacotherapy for refractory mood disorders: 22 years of study. J. Clin. Psychiatry 61, 9–15. Ganguly, R., Kotzan, J.A., Miller, L.S., Kennedy, K., Martin, B.C., 2004. Prevalence, trends, and factors associated with antipsychotic polypharmacy among medicaid-eligible schizophrenia patients, 1998–2000. J. Clin. Psychiatry 65 (10), 1377–1388. Haynes, R., 1976. A critical review of the ‘determinants’ of patient compliance with therapeutic regimens. In: Sackett, D., Haynes, R. (Eds.), Compliance with Therapeutic Regimens. Johns Hopkins University Press, Baltimore. Janseen, B., Weinmann, S., Berger, M., Gaebel, W., 2004. Validation of polypharmacy process measures in inpatient schizophrenia care. Schizophr. Bull. 30 (4), 1023–1033. Kendrick, T., Peveler, R., Longworth, L., Baldwin, D., Moore, M., Chatwin, J., Thornett, A., Goddard, J., Campbell, M., Smith, H., Buxton, M., Thompson, C., 2006. Cost-effectiveness and cost-utility of tricyclic antidepressants, selective serotonin reuptake inhibitors and lefepramine. Br. J. Psychiatry 188, 337–345. Nurnberg, H.G., Hensley, P.L., Thompson, P.M., Paine, S.S., 1999. Modeling the pharmacoeconomic cost of three selective serotonin reuptake inhibitors. Psychiatr. Serv. 50 (10), 1351–1353. Oquendo, M.A., Baca-Garcia, E., Kartachov, A., Khait, V., Campbell, C.E., Richards, M., Sackeim, H., Prudic, J., Mann, J., 2003. A computer algorithm for calculating the adequacy of antidepressant treatment in unipolar and bipolar depression. J. Clin. Psychiatry 64, 825–833. Osser, D.N., 1996–2003. Algorithm for the pharmacotherapy of depression. Mental Health Connections, Inc., Lexington, MA. http://www.mhc.com/ Algorithms/Depression. Rush, A.J., Trivedi, M.H., Wisniewski, S.R., Nierenberg, A.A., Stewart, J.W., Warden, D., Niederehe, G., Thase, M.E., Lavori, P.W., Lebowitz, B.D., McGrath,
A. Glezer et al. / Journal of Affective Disorders 117 (2009) 18–23 P.J., Rosenbaum, J.F., Sackeim, H.A., Kupfer, D.J., Luther, J., Fava, M., 2006. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR⁎D report. Am. J. Psychiatry 163 (11), 1905–1917. Simon, G., 2001. Choosing a first-line antidepressant. JAMA 286, 3003–3004. Texas Medication Algorithm Project, 2007. http://www.dshs.state.tx.us/ mhprograms/TMAP.htm.
23
Trivedi, M.H., Fava, M., Wisniewski, S.R., Thase, M.E., Quitkin, F., Warden, D., Ritz, L., Nierenberg, A.A., Lebowitz, B.D., Biggs, M.M., Luther, J.F., ShoresWilson, K., Rush, A.J., 2006. Medication augmentation after the failure of SSRIs for depression. N Engl J Med 23;354(12):1243–52.