BPSD is typically treated as a singular entity. Yet it is heterogeneous and challenges simple phenotyping by behavioral inventory. Some investigators recognize BPSD more as ‘obstreperous,’ disruptive behavior, or unwanted behavior. Others conceptualize it as a neuropsychiatric entity with an underlying pathobiology, or as the expression of an unmet need. Treatments for BPSD have been challenging since before the first clinical trials with chlorpromazine.

We systematically reviewed interventional studies to understand the successes, limitations, and knowledge gaps in terms of methodology that might misinform practice. Questions addressed included: What do these studies look like? How is BPRS operationalized, and does it vary between studies? What interventions have been tested? How are we measuring eligibility and outcomes? Are there methodological factors that influence the outcomes and validity of these trials? Are the trials methods fit for purpose and how can we better test interventions?

From a search yielding 6497 candidate studies, we included 474 of which 413 were randomized, 340 parallel group, 197 double-blinded, 51 unblinded. About 30% were in nursing homes only and 20% outpatient only. Most NH studies were drug studies; most outpatient studies were non-pharmacological. Over time, study durations consolidated to 6–12-week treatment periods and samples grew exceptionally large, involving 400 to 1200 participants.

Of studies that specified a target, 171 were for ‘agitation.’ 50 investigated sleep disturbance, 25 apathy, 25 depression, 21 psychosis. 150 described only ‘BPSD’ or ‘neuropsychiatric symptoms.’ Two-thirds of the agitation studies were single drug interventions; most used a scale score cut-off to define agitation. Important characteristics, secular trends in design, and quality of the BPSD studies will be detailed.

The important trends in methods for interventions and assessment of BPSD are not necessarily toward quality. Eligibility criteria have become designed for convenience, are misspecified relying on the same scales used for outcomes, although randomization is the rule, allocation concealment and treatment blinding is poor. There is marked autoregression of outcomes. Studies have become larger and designed to detect small effects even when clinical meaning is uncertain. BPSD studies need reconsideration and a few simple fixes to better discover effective treatments. Only a little care is needed to improve the quality and reliability of studies. This includes study management that is independent of patient selection and outcomes and from most procedures, and truly blinded assessments.