16th EUSPR Conference and Members’ Meeting

Symposium 6B: Joys and challenges: Using pooled Individual Participant Data Meta-Analysis (IPDMA) to advance our understanding of what works for whom in preventive interventions

26 Sept 2025, 08:30

1h 30m

Innere Medizin/2-403 (Virchowweg 9)

Symposium Child and Youth Wellbeing Symposium 6B: Joys and challenges: Using pooled Individual Participant Data Meta-Analysis (IPDMA) to advance our understanding of what works for whom in preventive interventions

Speakers

FRANCES Gardner (OXFORD UNIVERSITY) Francisco Calderon (University of Oxford, UK) Liina Björg Laas Sigurðardóttir (University of Oxford)Prof. Nina Heinrichs (Dept of Psychology, Universität Bielefeld, Germany)Dr PATTY Leijten (UNIVERSITY of AMSTERDAM) Sophia Backhaus

Description

Authors: Frances Gardner (Oxford University), Francisco Calderon (University of Oxford, UK), Liina Björg Laas Sigurðardóttir (University of Oxford), Nina Heinrichs (Dept of Psychology, Universität Bielefeld, Germany), Patty Leijten (University of Amsterdam), Sophia Backhaus, Ankie Menting, Bram O. De Castro, Constantina Psyllou, European Parenting Program Research Consortium, GJ Melendez Torres, Ignacia Arruabarrena, Jamie Lachman, Judy Hutchings, Maria Filomena Gaspar, Qing Han, Sophia Backhaus, Vashti Berry
Co-chair: Frances Gardner
Discussant: Nina Heinrichs

Symposium
Background: For many preventive interventions there is substantial knowledge about their effectiveness from RCTs. However, evidence is more limited when it comes to understanding for whom they work, partly because of limitations in traditional methods for analysing moderators. There are compelling reasons to investigate moderators: to inform appropriate targeting of interventions, to identify where programmes may need to be adapted for subgroups, and to advance personalisation. Traditionally there are two main approaches to evaluating moderators, i) analysing individual RCT data; ii) synthesising data at trial-aggregate level in meta-analyses, but each has distinct drawbacks; typically both have limited power. Our symposium focuses on a solution that solves many problems of each approach, Individual Participant Data Meta-Analysis (IPDMA). We first introduce principles and advantages of IPDMA, then show 4 examples of how IPDMA can yield more powerful, precise and transparent estimates of moderator effects, addressing questions such as how moderators operate longitudinally across time; how IPDMA can help understand equity effects of interventions; and developmental change in prevention effects, plus discussant.
Methods: Presentations utilise two pooled datasets: parenting intervention RCTs conducted in i) Europe- 38 trials, 5500 families; ii) low-middle-income countries- 6 trials, 2000 families.
Paper 1: Sets the scene, providing a brief primer on principles and methods behind IPDMA.
Paper 2: Examines moderators of longitudinal trajectories of prevention effects on child maltreatment, using IPDMA from in low-middle-income countries in Europe, Asia, Africa.
Paper 3: Examines at what levels of baseline risk parenting programs are effective, using the large European IPDMA dataset.
Paper 4: Investigates whether earlier interventions are more effective than those delivered later in childhood, using the large European IPDMA.
Paper 5: Shows how IPDMA and traditional meta-analysis can together assess equity effects of parenting interventions for preventing maltreatment, using i) large European IPDMA and ii) traditional aggregate-level, global meta-analysis.

ABSTRACT 1
A brief primer on principles, methods and resources for IPDMA.

This brief paper sets the scene for the symposium by defining Individual Participant Data Meta-Analysis (IPDMA), summarising the principles behind its methods, and sharing our enthusiasm for its unique advantages. These include greatly enhanced power and precision, especially for analysing moderators; reduced confounding of moderators; ability to detect effects on rarer outcomes; and good fit with the culture of data sharing and open, reproducible science. IPDMA also reduces ‘research waste’ (Ioannidis et al, 2014), by maximising the utility of data that we already have. Links to open-source learning materials will be provided, including guidance for conducting and reporting IPDMA studies and video materials that members may find useful.

ABSTRACT 2
The hidden effects of parenting interventions: Trajectories and equity-related moderators of intervention effects on child maltreatment using Individual Participant Data (IPD) meta-analysis.

Co-authors: Jamie Lachman, Qing Han, Frances Gardner, GJ Melendez Torres

Introduction: About one in two children aged 2-17 globally have experienced violence in the past year, with the majority of children worldwide being at risk. Parenting interventions have been developed to provide caregivers with tools that help reduce harsh parenting and contribute to the wellbeing and development of children.
Methods: This study pools individual participant data from six randomised controlled trials of Parenting for Lifelong Health group-based intervention collected at three different timepoints. Participants will be caregivers (n = 2,072) of children aged 2-17 from studies conducted in middle-income countries (Thailand, Philippines, South Africa, North Macedonia, Romania, Moldova). The effects of the parenting intervention on child maltreatment will be analysed using growth mixture modelling. Posterior analyses will be conducted on the resulting trajectories to examine class membership. The differential effects of the trajectories will be analysed through an equity lens, exploring variations in effects based on caregiver's age, gender, marital status, education level, relationship to child, experience of abuse, disability status, employment status, socioeconomic status, and children's age, sex, and education level.
Results: We anticipate distinct trajectories within the control and intervention groups. These trajectories will provide evidence of varying effects of the intervention, such as a high effect trajectory, a no-effect trajectory, among other possibilities. A post hoc analysis will examine trajectory membership to assess whether more disadvantaged groups benefit differently from the intervention.
Conclusions: The study's results will provide evidence of any differential effects of parenting interventions through a data-driven trajectory analysis. Further analyses will focus on the equity of outcomes, helping to assess how parenting interventions contribute to reducing inequalities in outcomes. The findings may be used to promote the intervention among high-effect groups and inform the need for tailored interventions for groups that show no effect or may even be negatively affected.

ABSTRACT 3
What is the optimal level of prevention in parenting support? Using individual participant data meta-analysis to identify when parenting support is (not) effective.

Co-authors: G.J. Melendez-Torres, Liina Björg Laas Sigurðardóttir, Constantina Psyllou, Sophia Backhaus, Frances Gardner, & European Parenting Program Research Consortium

Introduction: Parenting programs are disseminated across the world to enhance the quality of children’s upbringing and prevent parental and child mental health problems. Understanding the full promise of parenting programs as a prevention strategy requires understanding the magnitude of intervention effects that can be expected for individuals coming with different pre-intervention profiles. Generally, families with more difficulties at baseline benefit more from parenting programs. But at what level of baseline difficulties do intervention start yielding significant effects? And is there an upper limit to families with more severe problems benefiting more from parenting programs?
Methods: Our individual participant data meta-analysis includes 38 European randomized controlled trials (5,500 families) of parenting programs based on social learning theory principles for families with children aged 2 to 10 years old (PROSPERO preregistration CRD42019141844). We harmonized families’ pre- and post-intervention scores on child mental health (conduct problems, ADHD symptoms, and emotional problems), parental depression, and parenting behaviour (corporal punishment, harsh verbal discipline, not following through with discipline, praise, and tangible rewards). For each outcome, we tested baseline levels of the outcome as a moderator of intervention effects, using robust standard errors to account for the nested nature of the data. We tested for each moderator the best fitting functional form in a random intercept and random slope model using likelihood ratio tests determining whether additional polynomial terms improved model fit.
Preliminary results and discussion: Preliminary findings suggest that consistently across outcomes, there is a linear trend that families with more severe problems at baseline benefit more from parenting support. Preliminary findings suggest no consistent upper limits to this trend. Findings will be discussed in light of how they can help professionals and parents to have realistic expectations of parenting program benefits, and policy makers to implement programs in ways that maximize their effects.

ABSTRACT 4
“The earlier the better” - is it really? Using IPDMA to test age effects of parenting interventions across Europe for reducing child behavior problems and harsh parenting

Co-authors: Liina Björg Laas Sigurðardóttir, Patty Leijten, G.J. Melendez-Torres, Sophia Backhaus, & European Parenting Program Research Consortium

Background: Early intervention is often thought to be more effective than later intervention - an idea that is embedded in policy and supported by evidence from numerous indirect sources, including economic evaluations, neuroscience, and studies of recovery from extreme early environments. However, the ‘earlier is better’ hypothesis has been questioned by some developmental and prevention scientists and economists. We use IPDMA to test the hypothesis directly, using larger and more generalisable samples than earlier studies, examining whether findings apply to harsh, maltreating parenting outcomes and parent depression, and disruptive child behaviour. We focus on parenting interventions whose goals are to prevent any of these outcomes.
Methods: We pooled data from 38 European randomised trials (N=5500) of social learning theory-based parenting programs for children aged 2-10 years (PROSPERO preregistration CRD42019141844). We harmonised pre-and post-intervention scores on child disruptive behaviour, harsh/maltreating parenting, and parental depression. Primary analyses will test baseline child age (treated as continuous variable) as moderator of intervention effects on each outcome. We will conduct a two-stage IPDMA, first estimating intervention-by-age interaction effects within each trial separately before pooling them in a random-effects meta-analysis, ensuring that only within-trial information informs moderation effects. Additional analyses will test: (i) non-linear effects across the age range using polynomial terms, consistent with a ‘critical periods’ perspective; (ii) whether baseline disruptive behaviour severity further moderates the intervention-by-age interaction effect via a three-way interaction term.
Results and Discussion: Interpretation will focus on the extent to which findings concur that ‘earlier is better’, extending prior work based on a single outcome and narrower range of interventions. It will discuss findings in the light of work on critical periods for intervention, theories about early-onset disruptive behaviour, economic analyses of early intervention effects, and current policy discourse and initiatives.

ABSTRACT 5
“Equity effects of parenting interventions to reduce violent parenting – Using traditional and IPD meta-analysis

Co-authors: Sophia Backhaus, Frances Gardner, G.J. Melendez-Torres, Ignacia Arruabarrena, Vashti Berry, Bram O. De Castro, Maria Filomena Gaspar, Judy Hutchings, Ankie Menting, Patty Leijten

Background: Violence against children is a global phenomenon with over one billion children affected. Parenting interventions are an effective strategy to reduce violence at home and are implemented rapidly across the globe. The recent publication of WHO Guidelines on parenting interventions further advanced global policy interest in parenting interventions. Yet, it remains unknown whether these interventions are equally effective in reducing violence against children in families from advantaged and disadvantaged backgrounds. This knowledge becomes urgent given the potential risk of social interventions to widen, rather than reduce, social inequalities.
Methods: This study harnessed the complementary strengths of individual participant data meta-analysis (IPDMA) and traditional aggregate-level meta-analysis (MA) of randomized trials to investigate whether socio-economic disadvantage (income, education, employment) and belonging to an ethnic minority moderate the effects of parenting interventions on violence against children: Study 1: IPDMA of European trials of the Incredible Years parenting intervention (k = 15, n = 1,999, mean age = 2−10); Study 2: global aggregate-level MA of trials of 32 different social learning theory-based parenting interventions (k = 56, n = 8,155, mean age = 2−10).
Results & Discussion: Both studies found parenting interventions to reduce violence against their children by parents (Study 1: physical violence: = –·21, 95% CI [–·33, –·19]; Study 2: physical and emotional violence: d = −0·37; 95% CI [–0·47, –0·27]). There was no evidence that parenting interventions benefitted families of disadvantage (income level, education, employment, or ethnic group status) less (or more) than other families. Parenting interventions are unlikely to widen social inequalities in violence against children. This knowledge is timely given the recent World Health Organization Guidelines to make parenting interventions based on social learning theory globally available to parents.