June 1, 2010 (New Orleans, Louisiana) — Placebo response in clinical trials of antipsychotic medications is highly heterogeneous and has implications for the design of placebo-controlled trials, according to a meta-analysis presented at the American Psychiatric Association 2010 Annual Meeting.
"Placebo response is often observed in psychiatric randomized clinical trials, and it presents a challenge for psychopharmacologic drug development," said principal study author Ofer Agid, MD, University of Toronto, Ontario, Canada, and the Centre for Addiction and Mental Health. "We found there are differences in the placebo response in studies. We tried to understand if there are predictors."
|Dr. Ofer Agid|
Data were extracted from 63 randomized, placebo-controlled studies conducted from 1970 to 2009 in patients with schizophrenia: demographics, baseline characteristics, diagnosis method, treatment, study design factors, treatment exposure duration, funding source, and publication year. In the meta-analysis, placebo response in short-term treatment (2-12 weeks) was assessed using standardized mean difference in the Positive and Negative Syndrome Scale (PANSS) or Brief Psychiatric Rating Scale (BPRS) total score based on a random-effects model. A meta-regression analysis was performed to identify influential moderators of placebo response.
A total of 3794 placebo-treated patients had valid PANSS or BPRS total changes scores, with a median placebo group size of 47.
"In the meta-analysis we found a greater mean decrease in BPRS score from baseline in those subjects with higher mean baseline score (P < .05), which indicates regression to the mean bias when examining mean change in placebo arms of the studies," Dr Agid said. "In other words, interpretation of mean changes from baseline can be confounded by regression to the mean."
Greater placebo response was associated with more recently published studies (P < .01), shorter trials (P < .01), shorter duration of illness (P < .05), higher baseline severity score (P < .01), and community hospital (or mixed) treatment settings (P = .01).
"Knowledge of placebo response patterns and predictor models of their effect sizes have important implications in the design of studies that opt to use a placebo control," he said.
They also conducted a patient-level analysis based on 2 identically designed 6-week, short-term, double-blind trials and a 1-year, long-term, double-blind, placebo-controlled study in the ziprasidone clinical trial dataset. This was done to verify meta-analysis results from the aggregate data, placebo response in special subpopulations, and the pattern of placebo response over time using a Growth Mixture Model.
In the short-term trials, they observed gradual symptom improvement in mean PANSS total score for 85% of patients, whereas the remaining 15% had symptom worsening.
In the 1-year trial, which was the longest placebo-controlled antipsychotic trial ever conducted, several "trajectory patterns" were observed, Dr. Agid reported.
The model showed no improvement in PANSS total score. Instead, the model identified 2 subgroups with gradual worsening (19%) and immediate worsening patterns (15%), who had a higher level of baseline severity symptoms. The other 2 subgroups identified patients who had no change (35%) or delayed worsening (32%) and lower level of baseline severity.
The pattern for the PANSS negative symptom score in the 1-year trial was different, with a gradual improvement observed in about 50% of patients.
"We found four distinct placebo response patterns, which shows us that schizophrenia may not be 1 disease in terms of response to antipsychotic medications," he said. "One group may really need medication, as they deteriorate immediately without them. Another group may not get better but may not really deteriorate, so perhaps medication for this group may not be so necessary."
"In many ways, this is like a natural history study," Dr. Agid noted. "It shows what happens when patients are not given medications—except that they think they are on drugs. We would like to be able to predict who will deteriorate quickly, or less quickly, or not at all. We need to look for predictors. The study also suggests, given this heterogeneity, there may be something wrong with how we design clinical trials."
Douglas L. Noorsdsy, MD, associate professor of psychiatry and an investigator with the Psychopharmacology Research Group at Dartmouth Medical School, Lebanon, New Hampshire, commented on the finding that higher placebo responses were observed in the more recent trials. "The same has been shown in antidepressant trials," he noted. "One factor may be that the placebo effect is related to patient expectation. Long ago, there was less evidence that medications help psychotic disorders, but that has changed recently and patient expectations are probably higher now."
He also commented on the placebo patients' gradual worsening. He suggested that because competition has increased for clinical trial subjects, some are included who are not as ill as the inclusion criteria stipulate. "There may be a tendency to err upward in rating patients at baseline in order to qualify them for the trial," he said. "If the control patients stretched upward to qualify, over time they will tend to rate downward — that is they will graduate to their more natural level of illness."
He also pointed out that the standard of care for usual treatment has been markedly elevated in recent years, so today's patients will respond better in the placebo arm. "Over time, the expectation as to standard treatment for the control arm has changed," he said.
Dr. Agid receives consulting fees from Pfizer and Eli Lilly. Dr. Noordsy has disclosed no relevant financial relationships.
American Psychiatric Association (APA) 2010 Annual Meeting: Abstract NR3-73. Presented May 24, 2010.
Medscape Medical News © 2010 Medscape, LLC
Send press releases and comments to email@example.com.
Cite this: Placebo Response Highly Variable in Antipsychotic Medication Trials - Medscape - Jun 01, 2010.