This protocol has been registered in the PROSPERO database for systematic reviews, a web-based international registry of systematic review protocols: PROSPERO #CRD42017079826.
Eligibility criteria
We will include studies of patients with BPSD residing in long term care facilities, community, or in specialized geriatric assessment and psychogeriatric units. We will exclude studies conducted in acute care hospitals other than psychogeriatric units as the picture in an acute hospital admission is usually confounded by other factors such as delirium. Further, as a majority of BPSD is managed in LTC and the community as a chronic problem, we think that these criteria will capture the studies needed to answer our question. As dementia can occur in younger adults, we did not limit the search to older adults.
The diagnosis of dementia can be arrived at by a clinical interview and exam using criteria specified by the Diagnostic and Statistical Manual of Mental Disorders fourth or fifth editions, or International Statistical Classification of Diseases and Related Health Problems tenth revision (ICD-10), or by using internationally recognized criteria such as National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA).
We will include clinical trials of orally administered anticonvulsants, such as valproic acid, gabapentin, pregabalin, carbamazepine, phenytoin, topiramate, levetiracetam, zonisamide, oxcarbazepine, lamotrigine, and phenobarbital without restrictions regarding the type of anticonvulsant, dose, or frequency. We will include studies with both flexible and fixed doses. We will also consider studies in which combinations of anticonvulsants or anticonvulsants plus another drug or non-pharmacologic strategy was the intervention. The interventions will include all non-benzodiazepine anticonvulsants for BPSD. As seizure disorders are a physiologically different condition and the population characteristics of these patients are different, these studies will be excluded.
The control condition for studies can be either placebo, no intervention, or other active treatments including pharmacologic or non-pharmacological interventions. In studies where there is co-administration of other drugs that can impact the outcome such as co-administration of a benzodiazepine, we will only consider trials where the number of patients who received the additional drug does not differ significantly between randomized populations.
We will consider interventions and comparators that allow us to isolate the effect of the anticonvulsant. For example, a comparison between a combination of an anticonvulsant + antidepressant versus an antipsychotic + the same antidepressant may be considered as anticonvulsant vs antipsychotic, whereas a comparison between an anticonvulsant + antidepressant versus just an anticonvulsant would be excluded. We will consider studies that are longer than 2 weeks in length as symptoms lasting less than 2 weeks are more likely to be secondary to delirium. We will include randomized control trials and crossover trials. For crossover studies, we will include pre-crossover, and consider post-cross-over if there is an adequate washout. We defined an adequate washout period as the time required for the intervention drug to reach steady state concentration or five elimination half-lives.
We will include English publications only, but the study may be conducted in any country.
We will include studies that report outcomes based on validated measures of BPSD such as the Neuropsychiatric Inventory [
19], Cohen-Mansfield Agitation Inventory [
20], Brief Psychiatric Rating Scale [
21], and Clinical Global Impression Scale, and these will be considered as primary outcomes.
Trials will be identified using the following sources: the Cochrane Central Register of Controlled Trials (CENTRAL, latest version), MEDLINE (OVID SP), EMBASE (OVID SP, 1980 to present), PsycInfo (OVID SP), CINAHL (EBSCO),
ClinicalTrials.gov, and the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP
http://apps.who.int/trialsearch/Default.aspx); ISRCTN will be searched for on-going registered trials. The studies will be limited to those published in English but the study location can be worldwide. The search strategies will be developed by the Clinical Services Librarian Health Sciences Library at McMaster University. Strategies will consist of controlled vocabulary terms and keywords to describe the condition and the intervention and a filter to identify randomized controlled trials. The Boolean operator “OR” will be used to combine terms within each concept and the operator “AND” will be used to combine the concepts together. See Additional file
1 for a draft MEDLINE strategy. The strategy will be replicated as closely as possible across the other databases.
Data collection and analysis
Data will be managed using the Covidence software (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia) throughout the study.
Screening and selection of studies
We will screen for duplicate citations. After pilot testing for title exclusion achieves acceptable interindividual agreement between reviewers, two sets of authors will independently screen the citations followed by full text review of trials identified by the literature search. We will resolve disagreements regarding eligibility by consulting with an additional author who is not part of the initial dyad of reviewers.
Data collection
Two sets of authors (SB and JT or JH and HA) will independently extract data after piloting the forms with at least two full-text articles to ensure the validity of the forms. Any discrepancies in the extracted data will be resolved by discussion. We will use a standard data extraction form to extract the following information: characteristics of the study (design, method of randomization), participants, setting, interventions, and outcomes (types of outcome measures, serious adverse events). We will then check for accuracy before entering the data into Covidence software.
Assessment of risk of bias in included studies
For the assessment of study quality, we will follow the guidance of the Cochrane Collaboration [
22]. Initially, we will copy information relevant for making a judgment on criteria from the original publication into an assessment table. If additional information is available from study authors, we will also enter this in the table along with an indication that this is unpublished information. Two review authors (JT and JH) will independently make a judgment as to whether the risk of bias for each criteria is considered to be “low,” “unclear,” or “high.” Consensus will be reached with a third author (SB). We will resolve disagreements by discussion. We will consider trials which are classified as low risk of bias in sequence generation, allocation concealment, blinding, incomplete data, and selective outcome reporting as overall low bias risk trials. Blinding of assessors will carry a greater weight in terms or risk of bias as individuals with cognitive impairment may not be as susceptible to the risk of being unblinded, especially in the case of pharmacological intervention.
Data synthesis
Assessment of heterogeneity
We will look for clinical heterogeneity by examining the study details to determine the appropriateness of combining studies quantitatively and, when summary estimates are computed, test for statistical heterogeneity between trial results using the
χ2 test and the
I2 tests statistics when they are combined (see Chapter 9 of
The Cochrane Handbook of Systematic Reviews of Interventions) [
22]. We will classify heterogeneity using the following
I2 values: 0 to 40%, might not be important; 30 to 60%, may represent moderate heterogeneity; 50 to 90%, may represent substantial heterogeneity; and 75 to 100%, considerable heterogeneity. If substantial heterogeneity exists, we will explore reasons for this through sensitivity and subgroup analyses on factors related to risk of bias, study design, and patient and intervention characteristics.
Measures of treatment effect
The outcome measures from the individual trials will be combined through meta-analysis when appropriate (based on the clinical comparability of population, intervention, and outcomes between trials) using a random-effects model. A P value of less than 0.05, using the χ2 test, indicates significant statistical heterogeneity. For dichotomous data, we will use relative risk (RR) and absolute risk reduction (ARR) as the effect measures with 95% confidence intervals (CI). For continuous data, we will present the results as mean differences (MD) with 95% confidence intervals (CI). When pooling data across studies, we will estimate the mean difference if the outcomes are measured in the same way between trials. If papers present odds ratios (ORs), then we will try to extract the data, and if not available, then we will contact the authors for datasets.
Unit of analysis
The unit of analysis will be the study.
Dealing with missing data
We will analyze studies on an intention-to-treat (ITT) basis, i.e., we will analyze patients according to the intervention they were allocated, whether they received the intervention or not. We will impute a poor outcome for a drop-out rate of > 5%. We will also perform a sensitivity analysis imputing a favorable outcome for patients who dropped out from the studies. For each trial, we will report whether or not the investigators stated if the analysis was performed according to the ITT principle.
Narrative summary
If a meta-analysis is not possible or appropriate, the results from clinically comparable trials will be described qualitatively in the text. We will narratively describe the major findings and conclusions from individual RCTs. We will identify evidence gaps by documenting conflicting evidence across identified studies and comments when the existing evidence base is insufficient to reach firm conclusions.
Subgroup analysis and investigation of heterogeneity
In the case of excessive statistical heterogeneity (I2 > 50%), we will use subgroup analysis to evaluate for potential sources of heterogeneity. Subgroup analyses are secondary analyses in which the participants are divided into groups according to shared characteristics, and outcome analyses are conducted to determine if any significant treatment effect occurs according to that characteristic.
If data permit, we will carry out the following subgroup analyses: (1) effect of intervention by setting such as LTC vs community-dwelling seniors, (2) different types of drug as the associated intervention (e.g., antidepressants versus antipsychotics), (3) types of dementia such as anticonvulsants in Lewy body dementia or Alzheimer’s dementia.
Sensitivity analysis
If there are an adequate number of studies, we will perform the following sensitivity analyses:
Apart from assessing the risk of selective outcome reporting, considered under assessment of risk of bias in included studies, we will assess the likelihood of potential publication bias using funnel plots, provided that there are at least ten trials [
23]. Although small sample effects in a funnel plot can be a marker of publication bias, other causes will be considered including selection biases, poor methodological quality, and heterogeneity, artefactual, and chance. Furthermore, we will contact drug companies and authors also as a strategy to assess reporting bias.
Confidence in cumulative estimates
We will use the principles of the GRADE system [
24], as recommended in
The Cochrane Handbook for Systematic Reviews of Interventions to assess the quality of the body of evidence associated with specific outcomes (such as decrease in agitation or aggression, global improvement indices, serious adverse events) in our review and construct a SoF table using the GRADE software GRADEPRO [
25]. GRADE specifies an approach to framing questions, choosing outcomes of interest and rating their importance, evaluating the evidence, and incorporating evidence with considerations of values and preferences of patients and society to arrive at recommendations [
26]. The main comparison will be anticonvulsants compared to other active treatments including both non-pharmacologic and pharmacologic treatments.
Factors that may decrease the quality of the evidence include risk of bias, imprecision, indirectness, and inconsistency such as when the decision is made on the basis of the variability of results across the included studies. It also takes into account whether all the research evidence has been taken to account or if there is publication bias.
The quality of the evidence for a specific outcome will be reduced by a level, according to the performance of the studies against these five factors. There are five levels of evidence: High-quality evidence: there are consistent findings among at least 75% of RCTs with low risk of bias, consistent, direct, and precise data and no known or suspected publication biases. Further research is unlikely to change either the estimate or our confidence in the results. Moderate quality evidence: one of the domains is not met. Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low-quality evidence: two of the domains are not met. Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low-quality evidence: three of the domains are not met. We are very uncertain about the results. No evidence: no RCTs were identified that addressed this outcome.