Methods for Synthesizing Findings on Moderation Effects Across Multiple Randomized Trials

Abstract

This paper presents new methods for synthesizing results from subgroup and moderation analyses across different randomized trials. We demonstrate that such a synthesis generally results in additional power to detect significant moderation findings above what one would find in a single trial. Three general methods for conducting synthesis analyses are discussed, with two methods, integrative data analysis, and parallel analyses, sharing a large advantage over traditional methods available in meta-analysis. We present a broad class of analytic models to examine moderation effects across trials that can be used to assess their overall effect and explain sources of heterogeneity, and present ways to disentangle differences across trials due to individual differences, contextual level differences, intervention, and trial design.

Could scale-up of parenting programmes improve child disruptive behaviour and reduce social inequalities? Using individual participant data meta-analysis to establish for whom programmes are effective and cost-effective

“…In addition, trial outcome data are rarely broken down by equity factors; when trials do test socioeconomic or other predictors of treatment effects, statistical models are specified in varying ways, 80 for example, some calculating interaction effects but others only within-group predictors, which renders synthesis meaningless. 23,82 These problems apply no less to parenting intervention trials, 46,50 and can be overcome by use of pooled data.…”

“…23,82 These problems apply no less to parenting intervention trials, 46,50 and can be overcome by use of pooled data. Pooling individual-level data is an exciting new approach to data synthesis, 80,83 increasingly common in medicine in recent years, 84 but rarely used in public health or psychosocial fields. This study is the first to pool individual-level data from multiple independent trials of a parenting intervention (the IY Basic parenting series) in order to investigate moderator effects in a large, well-powered sample, and making use of individual rather than aggregate trial-level measurement of moderator variables.…”

“…80 Reporting bias is known to be a considerable problem in many areas of health care; systematic reviews find it to be linked to higher effect sizes. 111,112 It is especially problematic in this field, in which there are typically multiple secondary outcomes and multiple measures of the same construct within and between trials.…”

“…In this study we analyse data at an individual level rather than a trial level. Individual-level meta-analysis allows the separation of between-and within-trial effects, 80 meaning that individual-level effects can be distinguished from trial-level effects that are more likely to be confounded. Individual-level meta-analysis also allows for the analysis of a greater sample size (1799 individuals compared with 14 trials, if this had been a traditional meta-analysis), meaning that there is greater statistical power to detect effects.…”

Is Environment Really A Function?

“…I would not subscribe to the view that universal approaches "are generally more impactful on better off, lower risk population groups." Rather, it seems-when applying innovative statistical methodologies (Brown et al 2011;Muthén et al 2002)-that universal prevention can be more beneficial for the more vulnerable subgroups. Recent evaluations of universal prevention programs have shown such effects on the more vulnerable children in elementary school (Kellam et al 2008); some selective family based programs seem to be more effective the more vulnerable the targeted families are (Kumpfer et al 2008); a school-based program, evaluated in seven European countries under the EU-DAP trial, showed a more favorable effect on problematic drinking among students in underprivileged social environments (Caria et al 2011); and the evaluation of an alcohol prevention program in both school and family settings combined (Koning et al 2012) found that the more vulnerable subgroups within the universal target population benefited more from the intervention.…”