Table 2.

Important Methodological Problems in Current Antipsychotic Drug Trials and Some Suggestions to Resolve Them

Problem	What may be done about it?
Study population not representative for the majority of patients in daily clinical practice	Broader inclusion criteria; more pragmatic trials; corroborate the findings in RCTs by naturalistic studies
Lack of universal definitions for “acute exacerbation,” “relapse,” “response,” or “remission”	Consensus statements on commonly accepted definitions for the respective terms
Varying thresholds in PANSS/BPRS reduction for the outcome “response”	Use of 50% PANSS/BPRS reduction as threshold for response in acutely ill patients and of 25% in refractory patient populations; ideally display the distribution of patients meeting staged levels of PANSS/BPRS reduction and remission criteria (see table 1)
Limited evidence on specific aspects of symptomatology (eg, negative symptoms, aggression) and diagnostic subtypes	Careful selection of patients with target symptoms; conduct studies in specific diagnostic subtypes
Small sample sizes	Multicenter trials
High dropout rates	Ways of reducing dropout rates must be found; participants should be followed up even after dropping out of the study
Poor quality of ratings	Assessments by independent raters (not the therapists), eg, using telemedicine
Carryover effects or withdrawal effects from pretrial medication	Extension of wash-out periods, special rules for preceding depot treatment
Insufficient blinding	Use of prophylactic anti-parkinson medication; probing questions at study end to detect insufficient blinding
Dosing of trial drugs	Precursory obtaining of recommendations on dosing by the manufacturer of the study drugs; justify dosage in method section
Intervention groups can be switched in some long-term trials	Separate analysis of patients not switching the assigned treatment group
Lack of hard outcomes; difficult to interpret rating scales	Use intuitive, easy-to-use scales like the improved CGI62; development of further scales may be warranted
Little uniformity on the use of rating scales on aspects beyond symptoms	Consensus on relevant domains and measurement instruments must be achieved
Analysis of the time component of the new remission criteria in the context of high dropout rates	Alternatives to LOCF (inadequate approach, likely to overestimate remission rates) are “worst case” approaches (likely to underestimate remission rates) in addition to completer analyses, but better solutions are needed
High dropout rates have the potential to distort outcomes	Mixed-effect models appear to be statistically superior to LOCF but are not intuitive; alternative strategies such as the combination of the results of dropout rates and scale-derived efficacy results should be invested; analyses based on the study completers are useful sensitivity analyses
Quality of reporting of results is heterogeneous	More effective control of the reporting quality by reviewers and editors of scientific journals; special focus on abstract content as most commonly used source of information on a trial; completeness of the reporting of outcomes must be assured; strict application of the CONSORT statement; a description of the randomization method, blinding, and dropouts should be essential components of any publication
Incomplete side effect reporting	Common standard for the reporting of side effects must be defined; showing only those adverse events that occurred in at least 5% of the cases should be abandoned, all important adverse events should be reported instead, eg, as supplemental material on the web; “dropouts due to adverse event” should be restricted to side effect related dropouts and exclude efficacy-related events. Only then would it be a good measure of overall tolerability

Note: RCT, randomized controlled trial; PANSS, Positive and Negative Syndrome Scale; BPRS, Brief Psychiatric Rating Scale; CGI, Clinical Global Impression; LOCF, last observation carried forward.