Development of a patient-reported assessment to identify placebo responders in a generalized anxiety disorder trial

Abstract

Placebo response is thought to be a primary contributor to uninformative (failed) trials in clinical drug development. This study describes the development of a patient-reported assessment to detect likely placebo responders. A novel scale, the Placebo Response Screening Scale (PRSS), was developed to assess domains expected to be associated with placebo response. The scale was administered during the screening visit of a 4-week, placebo-controlled study of alprazolam and an investigational compound in 211 patients with generalized anxiety disorder (GAD). Items that predicted placebo response were identified. Sensitivity and specificity of the instrument were used to determine a threshold score for use in screening likely placebo response. The PRSS was then evaluated by comparing active treatment and placebo groups and subsetting the groups based on subject PRSS scores. Twenty items were selected for being predictive of patient global improvement rating, clinician global improvement rating, or improvement on Hamilton Rating Scale for Anxiety (HAM-A) scores in placebo-arm patients. Receiver operating characteristic concordance values ranged from 0.77 to 0.96 for the different definitions of placebo responder. A cut-score of 50 on a scale of 0–100 was chosen to maximize sensitivity (range 0.67–0.79) and specificity (range 0.78–1.00). Fifty-six patients with scores of 50 or higher were flagged as potential placebo responders. Excluding these 56 patients from the analysis resulted in a greater separation of active treatment from placebo. The PRSS is a promising tool for predicting placebo response in clinical trials and requires further use and validation.

Introduction

Placebo-controlled trials have been used to identify effective psychotropic medications for the past 50 years, and are necessary to demonstrate a specific treatment effect of drug candidates in indications where there is a variable degree of improvement over time. Unfortunately, a known active control drug (i.e., a marketed medicine with demonstrated efficacy in the study population) sometimes fails to demonstrate a treatment effect, indicating a lack of assay sensitivity. When the active drug fails to separate from placebo, the study has failed. The uninformative results from failed trials require costly repetition of studies, exposing additional subjects to potentially ineffective drug candidates. Failed trials have been a significant problem in studies of subjects with major depressive disorder, where approximately half of placebo-controlled trials have been reported to fail to distinguish a known active control from placebo (Khan et al., 2002a). Failed trials have also been a significant problem in studies of anxiety and other disorders. For example, Khan et al. (2002a), in a review of the available data in the FDA SBA database, found that only 45% of anxiolytics produced statistically significant results in GAD clinical trials. Trials with a high placebo response are thought to be less likely to distinguish known active controls from placebo.

A number of investigations have tried to identify predictors of placebo response in order to minimize this response and improve trial outcomes. Stein et al. (2006) found that placebo response in GAD and major depressive disorder (MDD) was not related to gender, age at onset of disorder, or duration of episode. Enhanced placebo response has generally been associated with lesser symptom severity in anxiety and depression studies, although not always (Khan et al., 2002b, Stein et al., 2006, Schweizer and Rickels, 1997). In concordance with this, greater treatment benefits have generally been shown in more severely ill patients in depression and anxiety disorder clinical trials (Khan et al., 2005, Stein et al., 2006, Pande et al., 2000, Kirsch et al., 2008), but data to support this are stronger for depression trials than for anxiety disorder trials. Efforts to increase symptom severity exclusion criteria raise concerns about the possible upward-biasing of subjective rating scale scores (Mundt et al., 2007) and decreased recruitment.

In addition to clinical variables associated with specific anxiety disorders, as well as demographic characteristics, if one looks across a variety of clinical disorders and experimental paradigms, a number of psychological and contextual factors have been suggested to contribute to the placebo response (Price et al., 2008). Psychological variables suggested to be associated with placebo response have included motivations (Geers et al., 2005), social acquiesence (traditionalism) (McNair et al., 1979), and openness to experience and suggestibility (De Pacalis et al., 2002). Substantial evidence indicates that a patient’s expectations of nonvolitional responses to a therapy (response expectancies) can have a substantial influence on the magnitude of the placebo response; and that manipulation of these response expectancies through verbal suggestion or other means can change the magnitude of the placebo response (Kirsch, 1985, Pollo et al., 2001). The therapeutic alliance can have a substantial effect on treatment response, in some instances including response to placebo (Horvath and Symonds, 1991, Schweizer and Rickels, 1997). Prior experience with treatment (Colloca and Benedetti, 2006) and classical conditioning (Goebel et al., 2002, Benedetti et al., 2003) are additional contextual factors that have been associated with the placebo response, and classical conditioning may be one mechanism that shapes expectancies (Stewart-Williams and Podd, 2004). Placebo effects may be most pronounced when psychological factors, such as motivation and suggestibility, combine with contextual factors (Geers et al., 2005, De Pacalis et al., 2002, Hyland et al., 2007).

Past efforts to identify patient characteristics that can be used to exclude placebo responders from clinical trials of new anxiolytics in development have met with limited success. This may not be surprising as, in general, in industry-sponsored clinical trials, relatively limited information is collected prior to randomization that addresses the psychological and contextual factors that are thought to contribute to the placebo response.

In order to further understand the placebo response and to begin the development of a measure that could be used to prospectively identify subjects as likely placebo responders, we developed a series of questions based on concepts hypothesized to be components of the placebo response and compiled them into a patient-reported assessment, the Placebo Response Screening Scale (PRSS), and then evaluated the PRSS as a predictor of placebo response in a GAD trial.

This study describes the development and evaluation of the PRSS and then examines the resulting impact of excluding potential placebo responders from the analysis of the Hamilton Rating Scale for Anxiety (HAM-A) (Hamilton, 1959) total score in a placebo controlled double-blind randomized clinical trial.