Schizophrenia Research Forum - A Catalyst for Creative Thinking

Separating Placebo and Drug Response In Antipsychotic Trials

August 13, 2013. A new meta-analysis examining 40 years' worth of trials of antipsychotic drugs for schizophrenia points to several factors likely to increase placebo response in trial participants. In their study published online July 30 in the American Journal of Psychiatry, Ofer Agid of the Centre for Addiction and Mental Health in Toronto, Canada, and colleagues confirm several factors associated with greater placebo response in previous studies (e.g., shorter trials and smaller placebo groups) but also point to variables in trial design and makeup of study populations that were not found in previous work, including younger age of patients, shorter duration and greater severity of illness, and greater number of trial sites, particularly if those sites were non-academic or not affiliated with the Veterans Administration.

"We think these are all tractable design and patient population factors that can be addressed in future trials," Agid and co-author Cynthia Siu wrote in an e-mail to SRF.

The trials of drug development—increasing placebo response
Strong placebo responses in clinical trials are a considerable obstacle in drug development in many fields of medicine (see Enck et al., 2013, for a review). But placebo responses are particularly vexing in trials of pain medications and of psychiatric drugs, a fact that has been recognized for some time. For example, a 1994 review found that placebo responses had occurred in 21 percent to 70 percent of patients in controlled trials of psychiatric drugs (Laporte and Figueras, 1994). However, such placebo responses appear to be increasing over time for reasons that are not fully understood (see SRF related conference story). This diminishes the significance of any meaningful drug benefit seen in treatment groups and leads to failed trials of new compounds to treat psychiatric disorders (see SRF related news story).

The new study is based on a literature review conducted over the past several years by Agid and his international collaborators. The group identified thousands of randomized controlled trials of antipsychotics conducted between 1970 and 2010, and by applying various quality and design criteria to this initial pool of records they winnowed down the dataset to 50 trials deemed suitable for meta-analysis.

Focusing just on patients receiving placebo, the researchers used a measure called standardized mean change (SMC), derived from reported scores on the Positive and Negative Symptom Scale (PANSS) or Brief Psychiatric Rating Scale (BPRS), to capture whether symptoms improved or worsened over the course of the trials under study, with a negative SMC indicating better outcomes. Meta-regression analyses were then performed with a number of patient and study characteristics.

A shrinking difference
The researchers found considerable heterogeneity in the SMC of patients given placebo across studies ranging from -1.4 to 0.9, with a mean of 0.33. But when the studies were taken as a whole, greater improvement in symptoms in the group treated with active drug was positively correlated with a greater response in those receiving placebo. For example, in trials in which drug response reached a mean SMC of 1.3, the mean placebo response was 0.75, but in trials in which active drug treatment resulted in a poorer SMC of 0.19, the placebo response was -0.75.

As these mean SMC values indicate, in addition to the general correlation between drug response and placebo response, the difference between the drug response and placebo response—a measure crucial to a given trial’s success—was smaller in studies with a greater placebo response: The respective difference between the mean SMCs for the trials just cited, for example, are 0.55 (higher placebo response) and 0.94 (lower placebo response).

Researchers have consistently found that the placebo response is increasing over time, and the current study is no exception, showing an SMC of -0.17 for studies conducted from 1970-1989; -0.28 from 1990-1999; and -0.39 from 2000-2010. The studies themselves were found to have improved in overall quality over the same time periods, but that improvement was reflected in more consistency in SMCs among studies rather than a lowering of mean SMCs.

The placebo is in the details
The critical question for future trials is determining what drives the placebo effect. In terms of individual patient factors, Agid and colleagues report that greater placebo response was seen in studies that enrolled younger patients and patients with a shorter duration of illness and/or more severe baseline symptoms. In the realm of study design, consistent with earlier analyses, the researchers found that the placebo response was higher in shorter studies and grew with the number of study sites. They also confirmed that this effect was more pronounced if trials included non-academic centers or research sites unaffiliated with the Veterans Administration. This is a significant corroboration of earlier work, say the authors, because the median number of sites involved in trials of antipsychotics has increased from two before 1990 to 38 in the period 2005-2010.

Based on these results, the authors recommend that investigators designing future trials attempt to minimize the placebo response by conducting studies no shorter than six weeks, exercising caution when expanding the number of study sites (especially if including non-academic sites), and modifying inclusion criteria to screen out younger patients, or patients with very short duration of illness or particularly severe symptoms.—Pete Farley.

Reference:
Agid O, Siu CO, Potkin SG, Kapur S, Watsky E, Vanderburg D, Zipursky RB, Remington G. Meta-regression analysis of placebo response in antipsychotic trials, 1970-2010. Am J Psychiatry. 2013 Jul 30. Abstract

Comments on Related News


Related News: ICOSR 2009—Unpleasing Placebos Cloud Antipsychotic Drug Trials

Comment by:  Paul Shepard
Submitted 23 April 2009
Posted 26 April 2009

When the 17 sites with high placebo responders were removed from the analysis, were only participants randomized to placebo removed or were all subjects who were recruited at these sites removed?

View all comments by Paul Shepard

Related News: ICOSR 2009—Unpleasing Placebos Cloud Antipsychotic Drug Trials

Comment by:  C. Anthony Altar
Submitted 28 April 2009
Posted 2 May 2009

Reply to P. Shepard
At ICOSR, Dr. Kinon presented the effects on PANSS positive values over 4 weeks for the placebo group, the groups receiving various LY2140023 doses, and those receiving olanzepine, but "without the 17 sites." I am reasonably sure, but not 100% positive, that this excluded all data from those sites, not just the placebo responders. Anything less would have introduced an unacceptable bias, even for a post-hoc analysis.

View all comments by C. Anthony Altar

Related News: ICOSR 2009—Unpleasing Placebos Cloud Antipsychotic Drug Trials

Comment by:  Ralph Hoffman
Submitted 19 May 2009
Posted 20 May 2009

These placebo results are certainly irksome, but may be important in positive ways. I am thinking of two hypotheses to account for these results. First, perhaps second-generation antipsychotic drugs (that are now more widely in use than ever) have more sustained therapeutic effects after discontinuation, so when patients are taken off their prescribed drugs to participate in these trials, their vulnerability to symptomatic worsening is less.

Of course, this would not explain the greater improvements in placebo groups. But perhaps with growing expectations regarding patient safety and support during randomized clinical trials overall, participants are getting more contact with research staff, which may have non-specific positive effects. We have, for instance, solid data indicating that significant social isolation is a trigger for psychotic symptoms independent of neuropsychological impairment in vulnerable individuals (unpublished data). The combination of reduced social isolation, increased staff support, plus (perhaps) sustained protective effects of second-generation drugs might account for emergence of greater positive placebo response.

View all comments by Ralph Hoffman

Related News: Opinions Mixed on Future for Lilly’s mGluR2/3 Agonist for Schizophrenia

Comment by:  Philip Seeman (Disclosure)
Submitted 15 August 2012
Posted 22 August 2012

The Lilly results of 11 July 2012 are not surprising, considering that the main ingredient of LY2140023 is LY404039, which is both a glutamate agonist and a weak partial dopamine agonist with only one-hundredth the potency of aripiprazole (Seeman and Guan, 2009; Seeman, 2012a), and considering that closer inspection of the clinical data (Kinon et al., 2011) showed that olanzapine was effective in schizophrenia, while LY2140023 was not (Seeman, 2012b).

References:

Kinon BJ, Zhang L, Millen BA, Osuntokun OO, Williams JE, Kollack-Walker S, Jackson K, Kryzhanovskaya L, Jarkova N, . A multicenter, inpatient, phase 2, double-blind, placebo-controlled dose-ranging study of LY2140023 monohydrate in patients with DSM-IV schizophrenia. J Clin Psychopharmacol . 2011 Jun ; 31(3):349-55. Abstract

Seeman P, Guan HC. Glutamate agonist LY404,039 for treating schizophrenia has affinity for the dopamine D2(High) receptor. Synapse. 2009 Oct ; 63(10):935-9. Abstract

Seeman P. An agonist at glutamate and dopamine D2 receptors, LY404039. Neuropharmacology. 2012a Jul 4. Abstract

Seeman P. Comment on "A multicenter, inpatient, phase 2, double-blind, placebo-controlled dose-ranging study of LY2140023 monohydrate in patients with DSM-IV schizophrenia" by Kinon et al. J Clin Psychopharmacol. 2012b Apr ; 32(2):291-2; author reply 292-293. Abstract

View all comments by Philip Seeman

Related News: Opinions Mixed on Future for Lilly’s mGluR2/3 Agonist for Schizophrenia

Comment by:  Hugo Geerts
Submitted 15 August 2012
Posted 22 August 2012

This is indeed another setback for the schizophrenia patient community, and it underscores the difficulty of translating animal model outcomes to the clinical situation. We have to think about introducing a new technology in schizophrenia drug discovery and development that would combine the best of preclinical animal information, but transplanted into a humanized environment to reverse this string of clinical failures.

One such approach is Quantitative Systems or Network Pharmacology, a computer-based mechanistic disease model of biophysically realistic neuronal networks that combines preclinical neurophysiology with human pathology, and clinical and imaging data (the topic of a recent NIH White Paper). Such an approach can be calibrated with retrospective clinical data, and then used to predict and examine future clinical trials. Applying this quantitative paradigm to the (also much publicized) failure of Dimebon in AD, researchers found that there was a fundamental off-target effect that precluded Dimebon from having cognitive benefits. Further analyses suggested that an imbalance in a common dopaminergic phenotype could increase part of the clinical signal difference as observed in the first (successful) Phase 2 trial.

In the case of schizophrenia, we find that affecting glutamatergic (such as with the mGluR2/R3 agonist) or GABA neurotransmission almost always leads to an inverse U-shaped dose response, because of the intrinsic balance between excitation and inhibition in cortical networks. Using such an approach forces discovery scientists to look beyond the single target and think about the impact on networks and circuits that ultimately drive human behavior and pathology in CNS disorders.

Unlike the traditional, currently used "cartoon"-based qualitative drawings, this approach allows for a quantitative outcome that, in principle, can help define the optimal "sweet spot" of the dose response by looking at the outcome of endophenotypes such as BOLD fMRI.

References:

Athan Spiros, Hugo Geerts. 2012. A quantitative way to estimate clinical off-target effects for human membrane brain targets in CNS Research and Development. Exp Pharmacology, 4; 53-61.

Athan Spiros, Patrick Roberts, Hugo Geerts. (2012) A Quantitative Systems Pharmacology Computer Model for Schizophrenia Efficacy and Extrapyramidal Side Effects, Drug Dev. Res, 73(4): 1098-1109.

View all comments by Hugo Geerts