📄 medicine

Treatment Response in a Randomized Controlled Trial on the Prevention of Mental Illness in the Offspring of Parents With Depression

This reanalysis of the PRODO randomized controlled trial reveals that baseline youth internalizing symptoms, rather than demographic or parental factors, are the most informative predictors of treatment response in offspring of depressed parents, with higher initial symptoms associated with more favorable outcomes in the intervention group.

Original authors: Alexandra Pelz, Johanna Löchner, Kornelija Starman-Wöhrle, Gerd Schulte-Körne, Belinda Platt

Published 2026-07-28

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Alexandra Pelz, Johanna Löchner, Kornelija Starman-Wöhrle, Gerd Schulte-Körne, Belinda Platt

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the human mind as a vast, bustling garden. Sometimes, the weather gets stormy, and the plants—our thoughts and feelings—start to wilt. This is what happens with depression and anxiety. For a long time, scientists have known that if a parent's garden is struggling with these storms, their children's gardens are at higher risk of facing similar weather. It's like inheriting a seed that makes the soil more sensitive to rain. Because of this, researchers have been trying to build "greenhouses"—special programs designed to protect these at-risk kids before the storm even hits. These programs teach kids how to cope with stress and how to talk about their feelings, hoping to keep their mental gardens healthy. But here's the tricky part: not every plant reacts to a greenhouse the same way. Some thrive immediately, while others seem to barely notice the extra protection. Scientists have been asking: Who benefits most from these greenhouses? Is it the older kids? The girls? The ones whose parents are having a really bad week? Or is it something else entirely?

This study dives into that question by looking at a specific greenhouse program called "GuG-Auf" (which stands for "Growing-up-happy-and-healthy"). The researchers took a fresh look at data from a big experiment where 100 families were split into two groups: one group got the special greenhouse program, and the other group didn't get any extra help (they were the control group). Instead of just asking, "Did the program work for everyone on average?", the team used a super-powered statistical microscope (called Bayesian mixed models) to zoom in on the individual kids. They wanted to see if things like a child's age, gender, or how sad they were before the program started could predict how their symptoms would change over the next 15 months. They also tried to group the kids into "teams" based on how their moods moved over time, like sorting runners into those who sped up and those who slowed down.

Here is what they found. The biggest surprise was that the kids who were already feeling a bit down or anxious before the program started were the ones who saw the most improvement. It's as if the greenhouse was most effective for the plants that were already starting to wilt; the program gave them the exact tools they needed to perk back up. In the group that got the program, kids with higher starting symptoms showed a clear drop in their negative feelings over time. On the flip side, the kids who started with very low symptoms (feeling great) didn't get much of a boost from the program; their symptoms stayed the same or even crept up slightly, but they didn't get worse because of the program.

In the group that didn't get the program (the control group), the pattern was a bit fuzzy but suggested that boys and kids who started with very low symptoms might have been more likely to see their feelings get a little worse over time. However, the researchers were very careful to say that they aren't 100% sure about this part because the data was a bit wobbly.

Crucially, the study ruled out some ideas that people often guess are true. The researchers found no clear evidence that a child's age or gender made a difference in how well the program worked. Whether a kid was 8 or 17, or a boy or a girl, the program seemed to work (or not work) in the same way for them. They also found that how depressed the parent was feeling didn't seem to change how the child responded to the program. It turns out that the child's own starting mood was a much better crystal ball for predicting their future than their age, gender, or their parent's mood.

The researchers also tried to sort the kids into different "mood teams." They found two main groups in the program: one team where symptoms went up (the "worsening" group) and another where symptoms stayed steady or went down (the "improving" group). The kids in the "improving" group were the ones who had higher symptoms to begin with. But the researchers warned that because the number of kids was small and some data was missing, these "teams" might just be a pattern in this specific group of people and not a permanent rule for everyone.

So, the main takeaway is that for kids whose parents have depression, the best way to guess who will benefit most from a prevention program isn't to look at how old they are or if they are a boy or a girl. Instead, it's to look at how they are feeling right now. The program seems to be a powerful tool for those who are already starting to struggle, helping them find their footing before the storm gets too heavy. For those who are already doing great, the program didn't hurt them, but it didn't seem to add much extra magic either. The study suggests that we might need to be more like gardeners who check each plant individually, rather than assuming all plants need the same amount of water.

Technical Summary: Treatment Response in the PRODO Trial

Problem Statement
While preventive interventions for offspring of parents with depression (e.g., the "GuG-Auf" program) have demonstrated efficacy in reducing the incidence of depression and internalizing symptoms, significant heterogeneity exists in individual treatment responses. Prior research has largely focused on between-group moderators (comparing intervention vs. control) at single time points, offering limited insight into within-arm variability. Furthermore, existing moderator analyses have yielded inconsistent findings regarding the influence of baseline child characteristics (age, gender, symptom severity) and parental factors (depressive symptoms) on intervention outcomes. This study addresses the gap in understanding which specific baseline factors predict the rate of symptom change within the intervention and control arms of the PRODO randomized controlled trial (RCT).

Methodology
This study is a secondary analysis of data from the PRODO trial, a parallel-group RCT conducted in Germany involving 100 families (135 children) with a history of parental depression. Families were randomized to either the "GuG-Auf" (Growing-up-happy-and-healthy) intervention ( $N=50$ families) or a no-intervention control condition ( $N=50$ families). Assessments occurred at baseline (T1), 6 months (T2), 9 months (T3), and 15 months (T4).

Participants: Children aged 8–17 years with at least one parent meeting DSM-IV criteria for a depressive disorder. Exclusion criteria included severe parental psychiatric comorbidities (e.g., bipolar disorder, substance dependence) and current child psychiatric disorders.
Measures:
- Outcome: Youth-reported internalizing symptoms measured via the Youth Self-Report (YSR) T-scores.
- Predictors: Baseline child age, gender, baseline internalizing symptoms (YSR), and parental depressive symptoms (Beck Depression Inventory-II).
Statistical Approach:
- Data Handling: Missing data were addressed using Multiple Imputation by Chained Equations (MICE) with $m=20$ imputations. Children with only one time point were excluded.
- Modeling: Bayesian linear mixed models were fitted separately for the intervention and control arms using the brms package in R. The models included fixed effects for predictors, time, and their interactions, with a random intercept for the child. Weakly informative priors were used.
- Clustering: Exploratory hierarchical clustering (Ward's method, Euclidean distance) was applied to the intervention arm's imputed data to identify distinct symptom trajectory subgroups. Scores were centered on baseline to emphasize trajectory shape.

Key Results

Intervention Arm (RQ1):
- Baseline Symptoms: A significant negative interaction was found between baseline internalizing symptoms and time ( $\beta = -0.49$ , 95% CrI $[-0.69, -0.28]$ ). Children with higher baseline symptoms showed a decline in symptoms over 15 months, whereas those with lower baseline symptoms showed a slight increase.
- Other Predictors: No significant associations were found between the rate of symptom change and child age, gender, or parental depressive symptoms. The credible intervals for these interactions included zero, indicating inconclusive evidence rather than a definitive absence of effect.
Control Arm (RQ2):
- Baseline Symptoms & Gender: Children with lower baseline symptoms and male gender were associated with steeper symptom increases over time, though these effects were characterized by substantial uncertainty (wide credible intervals).
- Other Predictors: Age and parental depressive symptoms showed no clear association with symptom change rates.
Exploratory Clustering (RQ3):
- Hierarchical clustering in the intervention arm identified two primary trajectory groups: a "Symptom-Increasing" cluster ( $n=17$ ) and a "Stable/Decreasing" cluster ( $n=29$ ).
- The "Symptom-Increasing" cluster was characterized by lower baseline internalizing symptoms.
- Robustness: The cluster solution varied across imputed datasets. In the complete-case analysis, the "Symptom-Increasing" cluster also appeared to have higher parental depressive symptoms and a higher proportion of females, but these differences were not observed in the imputed data, suggesting they may be artifacts of missingness rather than stable subgroup effects.

Key Contributions

Within-Arm Analysis: The study shifts the focus from between-group efficacy to within-arm heterogeneity, utilizing Bayesian mixed models to estimate uncertainty in small samples.
Baseline Symptom Severity as a Predictor: The analysis identifies baseline internalizing symptom severity as a more informative predictor of treatment response than demographic factors (age, gender) or parental symptom levels.
Trajectory Identification: The study provides an initial, exploratory data-driven identification of distinct symptom trajectory subgroups within a preventive intervention, distinguishing between those who improve/stabilize and those whose symptoms worsen.

Significance and Claims
The authors claim that for children of parents with depression, baseline symptom severity is a critical factor in understanding treatment response to the GuG-Auf program. The findings suggest that the intervention is particularly beneficial for children who already exhibit subclinical internalizing symptoms, aligning with the hypothesis that indicated prevention programs are more effective than universal ones. Conversely, children with low baseline symptoms may derive little additional benefit, though the program does not appear to cause harm to this subgroup.

The paper modestly concludes that while demographic tailoring (by age or gender) does not appear necessary based on current evidence, screening for baseline symptom severity could help prioritize children for preventive interventions. The authors emphasize that the observed heterogeneity and the specific trajectory patterns require validation in larger samples with reduced missing data and more advanced modeling (e.g., latent growth mixture modeling) to confirm whether these represent stable subgroups or sample-specific variability.

Technical Summary: Treatment Response in the PRODO Trial

More like this