Part A: methods for patient and public involvement
This project brought together expertise from a range of fields to plan and deliver the review. The main part was review work. In order to assess if current research met the needs of people with experience of TD, a small consultation was planned, taking results from the reviews and exploring whether or not the assessed outcomes matched service user priorities for managing TD. The consultation was advertised by e-mail via the McPin Foundation’s large circulation list of people who are interested in being involved. It was also advertised on their website. Interested people were asked to contact the McPin Foundation to book a place to attend. Reimbursement for time and out-of-pocket expenses was offered.
A lay overview of the previously published version of a Cochrane review evaluating the effects of vitamin E in TD47 gave the foundation for the discussions. All of the researchers involved in the consultation were extremely experienced in involving patients and the public. The session was planned to provide time to reflect on current research on TD and to consider gaps in knowledge.
The discussion was audio-taped and the service users were invited to write comments on Post-it® (3M, Bracknell, UK) notes and paper tablecloths, which were then collected and reviewed. The researchers listened to the recordings after the session and noted any points relevant to the above-mentioned questions that would have an impact on the funded systematic review. Full transcription and formal analyses were not appropriate in this case, as the consultation was not a piece of empirical qualitative work. Furthermore, two of the consultation facilitators had extensive experience in involving patients and the public in research and expert knowledge in this paradigm, including hosting focus groups (or, in this case, a consultation).
Informed by the results of the consultation, we updated outcomes for the summary-of-findings table for the systematic reviews. See Appendix 1 for the full report.
Part B: methods for systematic review
Please see Appendix 2 for differences between the project protocol and the review.
Interventions being assessed
We aimed to evaluate any intervention used for treating or preventing deterioration of symptoms of antipsychotic-induced TD. There is a vast array of strategies to deal with TD – one review identified over 100.50 Based on our experience with Cochrane reviews in this research area, we grouped the interventions as follows:
vitamins
GABA agonists
benzodiazepines
anticholinergics
cholinergics
calcium channel blockers
non-antipsychotic dopaminergics and noradrenergics
specific antipsychotic drugs
antipsychotic reduction or cessation including intermittent therapy
other interventions, including botulin toxin, insulin or lithium, among others.
We compared interventions with other interventions used to treat or prevent deterioration of symptoms of antipsychotic-induced TD of relevance to people in the NHS, placebo or no intervention.
Prioritisation of interventions for the NHS
From the included studies we listed all interventions, regardless of the primary condition, in order to map research activity. From this mapping, we chose to target, for this report, the top 10 interventions that seem to have demonstrated some efficacy and that are relevant for clinical practice and the NHS.
Measurement of outcomes
The following outcomes were included in the overview:
clinical improvement of TD symptoms
deterioration of TD symptoms
adverse events – extrapyramidal symptoms
adverse events – all
mental state
acceptability of the treatment – leaving the study early
social confidence, social inclusion, social networks, or personalised quality-of-life measures [this outcome was designated as important to patients, informed by the results of the patient and public involvement (PPI) consultation].
The Cochrane reviews included several more outcomes.
Design and theoretical/conceptual framework
We included randomised or quasi-randomised controlled trials containing data related to antipsychotic-induced TD, irrespective of language or place of publication. We also considered observational studies for inclusion with the following designs: (1) non-randomised controlled trials, (2) prospective cohort studies with a control group and (3) case–control studies. The systematic reviews and the overview of reviews follow Cochrane design and methodology.53
Target population
We included studies of adults with a diagnosis of antipsychotic-induced TD (according to any criteria), regardless of the primary condition.
Inclusion/exclusion criteria
We excluded studies in which participants had used antipsychotic drugs for < 3 months or in which the antipsychotic doses had not been stable for at least 1 month4 (except in analyses of antipsychotic switch, withdrawal or reduction). In addition, we excluded studies evaluating children and adolescents, or studies evaluating interventions that are not relevant to the NHS.
We also excluded studies that were > 10 years old that otherwise qualified for inclusion, but reported no useable data and in which:
we contacted study authors requesting data, but received no reply
we were unable to contact any of the study authors.
Setting/context
Participants may be receiving treatment in any setting, any country or any health-care system.
Search strategy
We attempted to identify all relevant studies regardless of language or publication status (published, unpublished, in press and in progress).
We searched Cochrane Schizophrenia Group’s Study-Based Register of Trials on 16 July 2015 using the following string:
*Tardive Dyskinesia* in Healthcare Condition Field of Study.
In such a study-based register, searching the major concept retrieves all the synonym keywords and relevant studies because all the studies have already been organised based on their interventions and linked to the relevant topics. The Cochrane Schizophrenia Group’s Register of Trials is compiled by systematic searches of major resources [including Allied and Complementary Medicine Database (AMED), Bioscience Information Service, Cumulative Index to Nursing and Allied Health Literature (CINAHL), EMBASE, MEDLINE, PsycINFO and PubMed, and registries of clinical trials including CT.Gov, International Standard Randomised Controlled Trial Number (ISRCTN) and the World Health Organization’s International Clinical Trials Registry Platform registries] and their monthly updates, hand-searches, grey literature and conference proceedings (see Group’s Module: http://onlinelibrary.wiley.com/o/cochrane/clabout/articles/SCHIZ/frame.html). There are no language, date, document type or publication status limitations for inclusion of records into the register.
We also searched the Cochrane Dementia and Cognitive Improvement Group’s Register of Trials via the Cochrane Register of Studies Online (CRSO; http://crso.cochrane.org/) on 21 July 2015 using the following string:
DEMENTIA:CC AND (*Tardive* OR *Dyskinesia*):TI,AB,KY.
For more information about this register, see the register’s page (www.medicine.ox.ac.uk/alois/content/about-alois).
Finally, we searched EMBASE, MEDLINE, and PsycINFO for observational studies on 9 January 2017, and details of the search strategy can be found in Appendix 3.
We inspected references of all identified studies for further relevant studies.
As some of the Cochrane reviews have not been updated during the past decade, and systematic reviews methods have changed considerably during this period of time, we also cross-checked all included, awaiting assessment, ongoing and excluded studies in the suite of nine Cochrane reviews on antipsychotic-induced TD.
Selection of studies
We uploaded search results into a web-based system (DistillerSR®, Evidence Partners, Ottawa, ON, Canada; www.systematic-review.ca). At least two reviewers (out of Antonio Grande, Rosie Asher, Hanna Bergman and Karla Soares-Weiser) independently screened all citations and abstracts identified by the search. Two reviewers (Hanna Bergman and Karla Soares-Weiser) inspected all studies from the nine Cochrane reviews on TD. We obtained full reports for potentially eligible studies and these were independently screened by two review authors (Antonio Grande and Rosie Asher). Disagreements were resolved through discussion with reviewers (Hanna Bergman and Karla Soares-Weiser). We documented justifications for excluding studies from the review.
Data extraction and management
Reviewer Rosie Asher extracted data from all included studies. These were cross-checked by Antonio Grande, and further validated by Hanna Bergman. Any disagreements about data extraction were documented and resolved by consensus. Any potential differences or data entry problems were discussed and decisions documented.
If more than one publication was identified reporting data from the same participants, the main publication was considered as the one with more information or with longer-term outcomes; all others were considered companion publications and data were only collected from these if they had not been provided in the main publication.
We attempted to contact authors in order to obtain missing information or for clarification whenever necessary.
We extracted data into tabular format, with an ‘address’ to each point in the document from which each data element had been taken. This allows future researchers to verify extraction and avoid duplication of effort. All data extracted in this way are fully available to researchers.54
We extracted data from graphs in GetData Graph Digitizer software version 2.26 (GetData Graph Digitizer, S Federov, Moscow, Russia).
Some specific outcomes
No clinically important improvement in tardive dyskinesia
‘No clinically important improvement’ was defined as < 50% improvement on any scale measuring TD, or as defined by triallists of the individual studies. For this outcome we assumed that participants with missing data did not improve.
We have shown details of the scales that provided usable data below.
Brief Psychiatric Rating Scale
The Brief Psychiatric Rating Scale (BPRS) is an 18-item scale measuring positive symptoms, general psychopathology and affective symptoms.55 The original scale has 16 items, although a revised 18-item scale is commonly used. Total scores can range from 0 to 126. Each item is rated on a seven-point scale, with high scores indicating more severe symptoms.
Extrapyramidal Symptom Rating Scale
The Extrapyramidal Symptom Rating Scale (ESRS) was developed to assess four types of drug-induced movement disorders: Parkinsonism, akathisia, dystonia and TD.56 The score for TD, ranging from 0 to 42, is based on the sum of all seven items in the TD objective examination.
Simpson–Angus Scale
The Simpson–Angus Scale (SAS)57 is a 10-item scale, with a scoring system of 0–4 for each item, measuring drug-induced Parkinsonism, a short-term drug-induced movement disorder. A low score indicates low levels of Parkinsonism.
Udvalg for Kliniske Undersøgelser Side-Effect Rating Scale
The Udvalg for Kliniske Undersøgelser (UKU) was developed to provide a comprehensive side-effect rating scale with well-defined and operationalised items to assess the side effects of psychopharmacological medications.58 The scoring sheet includes 48 items, with higher scores indicating more side effects.
Assessment of risk of bias of the included studies
Rosie Asher classified and Hanna Bergman cross-checked studies as being at low, unclear or high risk of bias, based on domain-specific assessments of risk of bias done using the Cochrane Collaboration’s existing risk-of-bias tool.53 If the raters disagreed, we made the final rating by consensus, with the involvement of another member of the review group. Where inadequate details of randomisation and other characteristics of trials were provided, we contacted authors of the studies in order to obtain further information.
We incorporated these judgements in assessing limitations in study design for outcomes in the summary-of-findings table (see Table 2).
Risk-of-bias assessment for observational studies was performed by a senior systematic reviewer (Artemisia Kakourou) using a tool that is currently being tested by Cochrane.59 The following domains were assessed: (1) confounding and selection bias (including confounders measured and addressed, use of matching and methods of adjustment), (2) performance bias (including any considerations of co-intervention), (3) missing data, (4) detection (for cohort studies) or recall bias (for case–control studies) and (5) selective reporting bias.
Data analysis
Analyses of single studies
Dichotomous data
For each study, the risk ratio (RR) and 95% CI were derived for people receiving the intervention compared with the control.
Continuous data
We included continuous data from rating scales only if:
For each study, the mean difference (MD) between groups and 95% CIs were estimated.
We also produced descriptive tables summarising information about study design, risk of bias and results of all included studies. Data were presented by each specific intervention according to the main diagnosis (schizophrenia or dementia).
Crossover trials
A major concern of crossover trials is the carry-over effect. This occurs if an effect (e.g. pharmacological, physiological or psychological) of the treatment in the first phase is carried over to the second phase. As a consequence, on entry to the second phase the participants can differ systematically from their initial state, despite a washout phase. For the same reason, crossover trials are not appropriate if the condition of interest is unstable.61 As both effects are very likely in severe mental illness, we used only data of the first phase of crossover studies.
Meta-analyses
Where studies were considered substantively similar enough for meta-analysis to be appropriate, we carried out fixed-effects analyses using the RevMan software version 5.3.5 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark).
We understand that there is no closed argument for preference for use of fixed- or random-effects models. The random-effects method incorporates an assumption that the different studies are estimating different, yet related, intervention effects. This often seems to be true to us and the random-effects model takes into account differences between studies, even if there is no statistically significant heterogeneity. There is, however, a disadvantage to the random-effects model. It puts added weight onto small studies, which often are the most biased ones. Depending on the direction of effect, these studies can either inflate or deflate the effect size. We chose the fixed-effects model for all analyses.
Observational studies
We provided an overview of evidence from observational studies. Study characteristics, results and conclusions were tabulated and summarised.
Variation in efficacy according to characteristics of individuals and studies
Visual inspection of the forest plots was used to evaluate the potential statistical heterogeneity (differences between the true intervention effects in the different studies). Heterogeneity was quantified by estimating the between-study variance τ2- and the I2-statistics,62,63 which measures the percentage of observed variation that can be attributed to true differences between the studies.62 In forest plots and meta-analyses, τ2 was estimated using the restricted maximum likelihood estimator,64 whereas its 95% CIs were estimated by the Q-profile method.65
Summarising and interpreting results
We used the Grading of Recommendations, Assessment Development and Evaluation (GRADE) approach66–68 to assess the evidence of the various interventions. For all NHS-prioritised interventions and outcomes, we have presented a summary-of-findings table (see Table 2) based on the GRADE results.
Investigation of heterogeneity
We considered a degree of heterogeneity inevitable, and hence we planned to explore only important heterogeneity (I2 ≥ 75%) using metaregression or subgroup analyses for the effect modifiers: (1) risk of bias in the different study designs; (2) length of antipsychotic use; (3) underlying disease (dementia or schizophrenia); (4) sex/age; (4) type of treatment use, specifically first- or second-generation antipsychotics; and (5) whether or not other concomitant drug interventions were used. Analyses were homogeneous with no important heterogeneity (I2 ≥ 75%).
Sensitivity analyses
To ensure that our imputations did not bias our results, we planned to restrict the analyses to studies considered to be at low, and low or unclear risk of selection and detection bias. However, all studies were at unclear risk of selection and detection bias, and we did not carry out this restricted analysis.
Planning of future studies
To judge the sufficiency of the evidence for the comparison of switching to any FGAs versus any SGAs, we calculated the conditional power of an updated meta-analysis for the particular comparison as described in Sutton et al.69 We further investigated whether or not hypothetical future studies are likely to alter the meta-analysis results using extended funnel plots.70 Given the small number of studies available, a fixed-effect inverse-variance meta-analysis model was assumed for this analysis.
Power of an updated meta-analysis based on simulations of new studies
We estimated the power of an updated meta-analysis through the simulation of (sufficiently similar) hypothetical ‘new’ studies and calculating the proportion of times that the meta-analysis result would be statistically significant.69 The event rate was assumed to be equal to that observed, and the number of simulations on which we estimated power was 1000.
Extended funnel plots
We further assessed whether or not future studies are likely to alter the meta-analysis result via extended funnel plots.70 A colour code appended in conventional funnel plots illustrates where the result of an updated meta-analysis would lie, depending on the effect estimate and the standard error of a hypothetical new study to be added to the evidence base.
