Placebo arms in acute schizophrenia trials improve, plateau and sometimes worsen
- Design and sample. Ren J, Zhang L and colleagues, led from the Center for Drug Clinical Research at Shanghai University of Traditional Chinese Medicine, pooled aggregate data from 48 placebo-controlled trials of oral medication in acute-phase schizophrenia: 4,909 participants in the placebo arms. The trials were published from 1993 to 2021, median year 2013 (Schizophrenia Bulletin 2026;52(1):sbaf013).
- At a baseline PANSS Total of 95, the modeled placebo reduction was 7.08 points at week 2, 9.56 at week 4 and 10.43 at week 6. On CGI-S the reduction was 0.31, 0.53 and 0.68 points at weeks 2, 4 and 6, with the plateau (90% of maximum) at about 12.1 weeks.
- Placebo groups whose score rose above baseline, which the paper calls a nocebo response, made up 12.5% of groups on PANSS Total (6 groups), 13.5% on Positive (5), 18.9% on Negative (7), 5.3% on General Psychopathology (1) and 8.1% on CGI-S (3). They were left out of the modeling.
- A higher baseline PANSS Total went with a larger placebo response: at week 6 the modeled PANSS Total reduction was 8.51, 10.43 and 12.33 points at baselines of 90, 95 and 100.
Placebo arms in acute schizophrenia trials do not stay flat. A model-based meta-analysis led from Shanghai pooled 48 placebo-controlled oral-medication trials and fitted how PANSS and CGI-S scores move over time on placebo alone.
Where the placebo curve flattens
The authors fitted time-course models to both scales. At a baseline PANSS Total of 95, the placebo reduction reached 7.08 points at week 2, 9.56 at week 4 and 10.43 at week 6, and the plateau came at about 4.3 weeks. CGI-S moved more slowly: 0.31, 0.53 and 0.68 points at weeks 2, 4 and 6, with a plateau near 12.1 weeks. A higher baseline PANSS Total went with a larger response. The median publication year of the trials was 2013.
The arms that got worse
The paper also counts placebo groups whose scores rose above baseline and calls this a nocebo response. Here the term means a worsening of the rating-scale score, not side effects reported by participants. It occurred in 12.5% of groups on PANSS Total (6 groups), 13.5% on Positive (5), 18.9% on Negative (7), 5.3% on General Psychopathology (1) and 8.1% on CGI-S (3). These groups were left out of the modeling, so the paper does not say what drove them.
In the consulting room, "nocebo" usually means the side effects a client expects and then notices. The meta-analysis measures something else, and the two should not be merged. What it shows is that on placebo alone a scale score can rise as well as fall. When a client's rating goes up in the first weeks, that rise is an outcome placebo arms also produce, so it does not by itself point to the drug. A concrete step follows: record the baseline score and ask what the client expects about side effects and about the first weeks, then read an early rise against both.
In 6 of the 48 placebo arms, the PANSS Total score ended above its baseline value.
The analysis uses aggregate data from published trials, so patient-level details are missing. It covers only oral medication, only short-term studies and only English-language publications. The nocebo groups were not modeled.