We used Cohen d standardized mean differences (SMDs) and odds ratios (ORs) to summarize effect sizes for continuous and dichotomous variables. Standardized mean differences of 0.2, 0.5, and 0.8 corresponded to small, medium, and large effect sizes.31 Negative Cohen d SMDs or ORs less than 1 indicated that the treatment reduced the parameter of interest relative to the control condition (eg, signifying a beneficial effect for suicidal ideation).31,32

We followed the same analytic approaches used in previous NMAs of studies examining psychiatric disorders (eMethods in the Supplement).33,34,35,36,37,38 We used the RStudio netmeta package, version 3.5.1 (RStudio).39,40 Forest plots were graphed for each outcome measure (self-harm, retention in treatment, study withdrawals, suicidality, and depression), and treatment rankings were created to represent each therapy’s effect size compared with treatment as usual. To preserve randomization, we used frequentist random-effects models, which accommodate different measures for the same outcome (eg, alternative instruments measuring suicidal ideation).41 To maximize available data, outcomes presented as dichotomous were pooled with continuous data using an inverse variance method. We assumed a jointly randomizable network, in which participants were equally likely to be randomized to any of the treatments.21,41,42,43 To determine NMA goodness of fit, transitivity (the extent of network heterogeneity) and consistency (the extent of agreement between direct and indirect comparisons)44 were assessed. To quantify transitivity, τ2 (total variation) and I2 (percentage of τ2 not caused by random error) were measured, with higher values indicating more heterogeneity.45,46 The Cochrane Q statistic was used to evaluate consistency, with the assumption of a full design-by-treatment interaction random-effects model; P > .05 indicated that the model was consistent. Dual analyses were conducted by distinguishing outcomes at the end of treatment from outcomes at the end of follow-up.

Network-level subgroup or meta-regression analyses could not be performed owing to limitations in the currently available RStudio packages. Data were analyzed from October 15, 2020, to February 15, 2021.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.