| Method | Unused methods |
| Measures of treatment effect | Continuous data For continuous outcomes we will calculate the MD and corresponding 95% CI if studies use the same rating scales. We will calculate the SMD with 95% CIs if studies use different scales to measure the same outcomes. As recommended in section 9.4.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2017), we will focus on final values unless some of the studies use change scores. We will combine studies that report final values with studies that report only change scores in the same meta‐analysis, provided that the studies use the same rating scale. We will conduct the analysis according to age, as children and adults respond differently to medication. We will combine the data from all groups in studies that have trihexyphenidyl in more than one group (i.e. different frequencies) and then separate these when performing the subgroup analysis to see how the different frequencies influence the results (see item four in the Subgroup analysis and investigation of heterogeneity section). |
| Multiple outcomes If studies provide multiple, interchangeable measures of the same construct at the same point in time, we will calculate the average SMD across the outcomes and the average estimated variances, as recommended in section 16.1.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). |
| Unit of analysis issues | Cluster‐RCTs If included trials use cluster randomisation, we will extract an ICC and use this to reanalyse the data. Where no ICC is given and a unit of analysis error appears to exist, we will contact the trial authors and ask them to provide either an ICC or the raw data to enable calculation of an ICC. Where no ICC is made available, we will search for similar studies from which we can impute an ICC, or seek statistical advice to obtain an estimate of the ICC. |
| Dealing with missing data | We will contact trial investigators to request missing data. If the trialists provide missing data, we will conduct a meta‐analysis according to intention‐to‐treat principles using all data and keeping participants in the treatment group to which they were originally randomised, regardless of the treatment they actually received, as recommended in section 16.1.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). If missing data are not provided, we will analyse only the available data. If there is concern regarding a high level of missing data, such that data could not be included in a meta‐analysis, we will include a qualitative summary in the text of the review. We will document missing data and attrition in the 'Risk of bias' tables, and we will explore how missing data might affect the interpretation of the results by conducting a sensitivity analysis. |
| Assessment of heterogeneity | We will assess clinical heterogeneity by comparing the between‐trials distribution of participant characteristics (e.g. children versus adults) and intervention characteristics (e.g. treatment type and dose), and assess methodological heterogeneity by comparing trial characteristics (e.g. cross‐over versus parallel design). We will evaluate statistical heterogeneity using the I2 statistic and the Chi2 test of heterogeneity, with statistical significance set at P value < 0.10. As recommended in section 9.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2017), we will consider I2 values as follows. -
0% to 40% might not be important. -
30% to 60% may represent moderate heterogeneity. -
50% to 90% may represent substantial heterogeneity. -
75% to 100% represents considerable heterogeneity. We will report Tau2 as an estimate of the between‐study variance when reporting the results from the random‐effects model. |
| Assessment of reporting biases | If we identify 10 or more studies, we will use funnel plots to investigate the relationship between intervention effect and study size. We will explore possible reasons for any asymmetry found. We will analyse the funnel plot of the data to ascertain asymmetry. Asymmetry of a funnel plot may indicate, among other things, publication bias or poor methodological quality (Egger 1997). |
| Data synthesis | We will synthesise results in a meta‐analysis using a fixed‐effect model when studies are similar enough with regard to the intervention, population and methods, to assume that the same treatment effect is estimated. We will synthesise results in a meta‐analysis using a random‐effects model when statistical heterogeneity is found or when studies differ enough with regard to the intervention, population, and methods, to assume that different yet related treatment effects are estimated, and when it is deemed to be clinically relevant, as recommended in section 9.4.3.1 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2017). |
| Subgroup analysis and investigation of heterogeneity | We will conduct the subgroup analyses listed below. -
Classification of cerebral palsy according to motor ability using the Gross Motor Function Classification System.Movement disorder and whether or not there is a mixed pattern (spasticity, dystonia, ataxic, choreoathetoid, hypotonic, mixed). -
Participant age (e.g. adults versus children, preschool children versus school‐age children). -
Treatment dosage and frequency of medication (2 versus 3 times a day; and low dose versus high dose). -
Polypharmacy: whether sole agents or other agents are used (trihexyphenidyl alone versus trihexyphenidyl and another agent). -
Length of follow‐up (e.g. ≤ 3 months versus > 3 months). We will also look at the number of participants per study to determine if this is sufficient to perform a subgroup analysis. |
| Sensitivity analysis | We will conduct sensitivity analyses to investigate the effect on the overall results of excluding trials that meet the criteria described below. -
Inadequate allocation concealment or sequence generation (selection bias) -
No blinding (performance bias) -
Incomplete outcome data (attrition bias). To investigate the effect of incomplete outcome data we will re‐perform the meta‐analysis including only those studies with high rates of complete data and compare this result to the full meta‐analysis that contains studies with both high and low rates of complete outcome data (i.e. all included studies). In this way we will be able to determine if the results of the full meta‐analysis remain true and have not been influenced by incomplete outcome data. We will also conduct a sensitivity analysis for studies with very low risk of bias. In addition, we will conduct a sensitivity analysis using a range of ICCs to assess the impact on treatment effect. |
