Psychopharmacology (2015) 232:3833–3840 DOI 10.1007/s00213-015-4047-2
ORIGINAL INVESTIGATION
Antidepressant treatment history and drug-placebo separation in a placebo-controlled trial in major depressive disorder Aimee M. Hunter 1,2 & Ian A. Cook 1,2,3 & Molly Tartter 4 & Simi K. Sharma 1 & Gregory D. Disse 1 & Andrew F. Leuchter 1,2
Received: 7 May 2015 / Accepted: 3 August 2015 / Published online: 29 August 2015 # Springer-Verlag Berlin Heidelberg 2015
Abstract Rationale A history of antidepressant treatment may predispose subjects toward placebo nonresponse in randomized controlled trials (RCTs) in major depressive disorder (MDD). Objective The objective of this study is to examine selfreported prior antidepressant treatment and response in relationship to clinical outcome in an 8-week randomized trial of reuptake inhibitor antidepressant medication (MED) versus placebo (PBO) administered along with limited supportive care. Methods Chi-square and MMRM analyses examined MED vs. PBO outcomes in antidepressant-naïve vs. antidepressantexperienced subjects. Linear regression models examined treatment history along with covariates as predictors of clinical improvement. Results Among completers (n=56), there was no significant difference in response rate between MED (53.3 %) and PBO (42.3 %) (χ2 =0.33, p=0.28, 1-tailed). The antidepressantexperienced subgroup (n=37), however, showed a significantly greater response rate to MED (52.4 %) than PBO (25.0 %)
(χ2 =2.82, p=0.047, 1-tailed). The full intent-to-treat (ITT) sample (n=69) did not show a significant difference between MED and PBO group improvement over time, but in the treatment-experienced subgroup (n=46), MED showed significantly greater improvement than PBO (coefficient=.39, SE=.23, p=.045, 1-tailed). A history of prior antidepressant treatment predicted poorer overall response independent of pretreatment symptom severity, number or length of previous episodes, subject expectations, or family history of MDD. Conclusions Treatment history appears to constitute a factor that is distinct from other commonly studied illness characteristics or expectancy measures, and that impacts overall response as well as drug-placebo separation in RCTs. Keywords Antidepressant treatment history . Tachyphylaxis . Tolerance . Placebo response . Antidepressant medication . Randomized clinical trial . Treatment-resistant depression . Expectations
Introduction * Aimee M. Hunter
[email protected] 1
Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine at UCLA, University of California Los Angeles, 760 Westwood Plaza, Rm. 57-455, Los Angeles, CA 90024-1759, USA
2
Neuromodulation Division, Semel Institute for Neuroscience and Human Behavior at UCLA, University of California Los Angeles, Los Angeles, CA, USA
3
Department of Bioengineering, University of California Los Angeles, Los Angeles, CA, USA
4
Department of Psychology, University of California Los Angeles, Los Angeles, CA, USA
The randomized controlled trial (RCT) represents the gold standard for assessing antidepressant treatment efficacy. Many factors have been shown to affect the ability to detect the Bsignal^ of antidepressant treatment efficacy, including baseline symptom severity (Fournier et al. 2010; Khan et al. 2002, 2010; Kirsch et al. 2008), nature of the outcome assessment, trial duration (Khan et al. 2010), and likelihood of receiving placebo (Rutherford et al. 2014). Signal detection is confounded by the highly variable benefits of placebo treatment (Walsh et al. 2002). The bulk of evidence suggests a link between greater baseline symptom severity and lower placebo response rates in antidepressant trials (Khan et al. 2002; Kirsch et al. 2008; Weimer et al. 2015). Additionally, a recent
3834
prospective RCT in major depressive disorder (MDD) showed that subjects’ expectation of response to antidepressant medication was a specific predictor of response to placebo, but not medication (Leuchter et al. 2014). The effects of medication treatment history are not routinely considered as a factor affecting signal detection in antidepressant RCTs. A number of studies have shown that prior antidepressant treatment is associated with a decreased likelihood of response to subsequent treatments, regardless of prior outcome (Amsterdam et al. 1994, 2009; Amsterdam and Shults 2005; Leykin et al. 2007). The Sequenced Treatment Alternatives to Relieve Depression Trial (STAR*D) demonstrated that each failed treatment is followed by a lower rate of remission in the successive open-label medication treatment (Rush et al. 2006). Past antidepressant use with failure to achieve response has also been associated with poorer clinical response to placebo (Brown et al. 1992). We previously demonstrated that prior antidepressant medication treatment may be associated with differentially poorer response to placebo versus medication in subsequent RCTs, even when controlling for illness severity (Hunter et al. 2010). Accounting for this effect enabled detection of significant drug-placebo separation in the treatment-experienced subgroup when the full sample did not demonstrate separation. It is unclear whether this poorer response to placebo is attributable to prior treatment, poor prior outcome, negative medication expectations, or other factors. The present study sought to replicate the effect of prior antidepressant treatment history as a predictor of drugplacebo separation (Hunter et al. 2010). We additionally assessed prior clinical response, as well as illness severity characteristics and subject expectations because these factors may contribute to any treatment history effect. Based on prior findings, we hypothesized that antidepressant-experienced (AE) subjects would be less likely to respond to placebo than antidepressant-naïve (AN) subjects, and that drug-placebo separation would therefore be greater in the AE subgroup.
Methods Subjects and treatment interventions Data were collected as part of a larger study that examined predictors of response to placebo, medication, or supportive care treatments for depression (ClinicalTrials.gov Identifier: NCT00200902). Adult subjects with MDD, ages 18–65 years, were recruited for the primary study through community advertising as described in the primary report (Leuchter et al. 2014). Enrolled subjects met criteria for MDD as determined using the Mini-International Neuropsychiatric Interview (M.I.N.I.; Sheehan et al. 1998) and evidenced moderate or greater symptom severity based on a score ≥17 on the 17-
Psychopharmacology (2015) 232:3833–3840
item Hamilton Depression Rating Scale (Ham-D 17 ) (Hamilton 1960). Individuals with other primary Axis I disorders, or Axis II conditions that could interfere with their participation in the study, were excluded. Urine toxicology data were used to exclude illicit substance or psychotropic medication users. The UCLA IRB approved all procedures, and written consent was obtained from all subjects prior to participation. All subjects received supportive care in the form of weekly 30-min sessions of interpersonal clinical interaction including risk assessment and symptom monitoring with the research nurse. Subjects were randomly assigned to 8 weeks of treatment with either supportive care exclusively (provided at the randomization visit and at weeks 1, 2, 4, and 8; 28 % likelihood) or supportive care with the addition of either a placebo pill (28 % likelihood) or antidepressant medication (44 % likelihood, divided equally among venlafaxine XR 225 mg, duloxetine 90 mg, or escitalopram 10 mg). The present report focused only on the subjects who had been randomized to placebo (PBO) or antidepressant medication (MED) because these conditions are relevant to typical randomized controlled trials (RCTs). Subjects in either pill-taking condition received a 1-week single-blind placebo lead in prior to double-blind treatment with medication or placebo. Medication subjects were started on one tablet each morning of either venlafaxine XR 75 mg, duloxetine 30 mg, or escitalopram 10 mg with dosages increased by one pill every week until achieving the final dose. Escitalopram-treated subjects received 10 mg throughout the study but were given placebo pills after week 1 in addition to the escitalopram in order to maintain the blind. Placebo subjects also received placebo pills increased on the same schedule in order to maintain blinding. Clinical assessments Depression severity was evaluated weekly using the HamD17. Subjects’ pretreatment expectations of treatment outcome were assessed using two items from the Patient Attitudes and Expectations Form (PAEF) (Weiss et al. 1997) that rated the degree to which subjects believed that treatment in general, and medication in particular, would be helpful in relieving their depression. A Likert scale, with anchors ranging from 1=BI expect to feel completely better^ to 5=BI don’t expect to feel any different,^ was used to rate general expectations regarding treatment, while a Likert scale with anchors ranging from 1=Bwould not be helpful at all^ to 7=Bwould be extremely helpful^ was used to rate expectations concerning medication as a treatment. Subjects were interviewed at enrollment into the study regarding prior antidepressant treatment. Subjects who had never taken an antidepressant medication were categorized as antidepressant-naïve. Those who reported previous use and were able to name, or identify from a list, at least one
Psychopharmacology (2015) 232:3833–3840
antidepressant medication used were categorized as antidepressant-experienced. Subjects were asked to recall the beginning and ending date for each antidepressant trial and to rate the Bdegree of improvement in symptoms^ during each prior course of medication. Degree of improvement was rated on a nine-point Likert scale ranging from ‘1’ (no improvement) to ‘9’ (marked improvement). The midpoint rating of ‘5’ was anchored to indicate Bmoderate improvement.^ Data analysis Improvement was assessed as percent change in the total Ham-D17 score from baseline to week 8, with response defined as ≥50 % improvement. Treatment history was examined as a categorical predictor comparing AE and AN subjects. We used one-tailed directional tests of significance (p≤.05) to examine the hypothesis that treatment with MED would be associated with a higher response rate and greater symptom improvement than PBO. Analyses were performed using SPSS version 22 and STATA version 13.1. We examined drug-placebo separation in completers as previously reported (Hunter et al. 2010) in order to assess the impact of treatment delivered as intended (Ten Have et al. 2008). We also examined the intent-to-treat (ITT) sample because this approach is commonly used in RCTs. We used chi-square analysis comparing treatment (MED vs. PBO) with outcome (response vs. nonresponse) to examine response rates, first in the total sample and subsequently in the AE subgroup. In the ITT sample, we compared the trajectories of change in symptoms between MED and PBO using mixed-effects models with repeated measures (MMRM). Change from baseline in Ham-D17 score at each week was entered as the outcome, with time, treatment, and the timeby-treatment interaction as predictors, co-varying for baseline Ham-D17. Outcome analyses were performed first on all subjects and then separately on AE and AN subgroups. Linear regression models in the ITT sample were used to assess the role of treatment history as a predictor of symptom improvement along with covariates of age, baseline Ham-D17 score, episode length, and number of prior episodes. Exploration of relationships between prior improvement and baseline expectations We computed a mean rating of ‘degree of prior improvement’ averaged across all prior trials for each AE subject. Because the parent study found a relationship between baseline medication expectations and PBO outcome (Leuchter et al. 2014), we examined the association between prior improvement and baseline expectations. Because Shapiro-Wilk tests indicated that general- and medication-related expectation variables were not normally distributed (p
