Cognitive Flexibility and Decision-Making in Anxiety and Depression: Meta-Analytic Evidence Facilitated by Machine-Learning Screening

This meta-analysis, utilizing machine learning for efficient screening, reveals that both anxiety and depression are associated with reduced cognitive flexibility and decision-making, with no significant differences between the two disorders, supporting a transdiagnostic perspective on these cognitive impairments.

The Gist

The Big Picture: A "Brain Report Card" for Anxiety and Depression

Imagine your brain is a high-tech command center. It has two main jobs that keep your life running smoothly:

1. Decision-Making: Choosing the best path forward when you have options (like picking a route to work or deciding what to eat).
2. Cognitive Flexibility: The ability to pivot quickly when the rules change (like switching lanes when traffic stops or changing your plan when it starts raining).

This paper is a massive "report card" that looks at how Anxiety and Depression affect these two jobs. The researchers didn't just look at one study; they gathered data from 92 different studies involving thousands of people. They used a special computer program (machine learning) to help them read through over 23,000 research papers to find the right ones, kind of like using a super-smart librarian to find the best books in a giant library.

What They Found: The "Traffic Jam" Effect

The main finding is simple: Both anxiety and depression make these brain jobs harder.

• The Result: People with anxiety or depression showed "lower scores" on tasks requiring decision-making and flexibility compared to healthy people.
• The Analogy: Imagine driving a car.
  • Healthy Brain: You see a detour sign, and you smoothly switch lanes and find a new route. You make decisions quickly.
  • Anxious/Depressed Brain: It's like driving in a heavy fog or a traffic jam. The car (your brain) is still there, but it's harder to see the road, harder to change lanes, and harder to make a quick turn. You might get stuck or take a longer, less efficient path.

The Big Surprise: They Are More Alike Than Different

Before this study, many experts thought anxiety and depression might mess with your brain in totally different ways.

• Old Idea: "Anxiety makes you too scared to decide, while depression makes you too sad to try."
• This Study's Finding: Nope. When it comes to these specific brain tasks, anxiety and depression act almost exactly the same.

The Analogy: Think of anxiety and depression as two different types of storms. One is a thunderstorm (anxiety), and the other is a heavy, gray fog (depression). They look different and feel different, but when it comes to knocking over a tree (impairing decision-making and flexibility), both storms knock the tree down with the same amount of force.

The researchers compared the two groups side-by-side and found no significant difference. Whether you have anxiety or depression, your ability to switch gears or make a choice is impaired to a similar degree.

Why Does This Happen? (The "Shared Engine" Theory)

The paper suggests that because the problems are so similar, they probably share a common "engine" or cause. It's not that anxiety and depression are totally separate diseases with separate brain problems; they seem to share a "transdiagnostic" (across-diagnosis) factor.

The Analogy: Imagine two different car models (anxiety and depression) that both have the same faulty spark plug. Even though the cars look different on the outside, the reason they both sputter and stall is the same broken part inside. The paper suggests that factors like stress, negative thinking loops, or how the brain handles uncertainty are the "faulty spark plugs" affecting both groups.

How They Did It (The "Smart Search")

One of the cool parts of this paper is how they did the work.

• The Problem: There are so many research papers (23,000+) that reading them all by hand would take a human team years.
• The Solution: They used a machine-learning tool called asReview.
• The Analogy: Imagine you need to find 100 specific needles in a haystack. Instead of a human digging through the whole haystack, you give a robot a picture of a needle. The robot scans the haystack, learns what a needle looks like, and starts pulling out the most likely candidates first. The humans then double-check the robot's picks. This made the process much faster and ensured they didn't miss any important studies.

The Bottom Line

This study tells us that when it comes to the mechanics of making choices and adapting to change, anxiety and depression are twins in their impact. They both create a "cognitive fog" that slows you down and makes it harder to switch tracks.

Because they affect the brain in such similar ways, the paper hints that we might need to stop treating them as totally separate problems when it comes to these specific brain functions. Instead of asking "Is this an anxiety problem or a depression problem?", we might need to ask, "How do we fix the shared engine that is causing the traffic jam for both?"

(Note: The paper explicitly states these are research findings and should not be used to guide clinical treatment on its own, as it is a preprint awaiting final peer review.)

Technical Summary

Technical Summary: Cognitive Flexibility and Decision-Making in Anxiety and Depression

Problem Statement
Anxiety and depressive disorders are highly prevalent conditions that significantly disrupt cognitive mechanisms, particularly attention, working memory, and higher-order processes like cognitive flexibility and decision-making. While existing literature documents impairments in these domains, prior meta-analytic efforts have relied heavily on self-report measures, which are distinct from task-based behavioral data and susceptible to social desirability bias. Furthermore, significant heterogeneity in task designs (e.g., Iowa Gambling Task vs. Balloon Analogue Risk Task for decision-making; Wisconsin Card Sorting Test vs. task-switching for flexibility) has complicated the interpretation of whether these disorders exert shared or distinct effects. There is a critical need to synthesize task-based evidence to clarify disorder-specific versus transdiagnostic impacts on these cognitive outcomes.

Methodology
The authors conducted a comprehensive meta-analysis examining the effects of anxiety and depression on decision-making and cognitive flexibility using behavioral task data.

• Search and Screening: A systematic search across PubMed, APA PsycInfo, ERIC, and Academic Search Ultimate yielded 23,587 records. The study employed a hybrid screening strategy combining manual review with machine-learning-assisted screening using asReviewLab (version 1.6.3). The machine learning model utilized a Naive Bayes classifier with TF-IDF feature extraction, trained on a precursor set of 4 relevant and 5 irrelevant articles. Screening followed a mixed stopping rule: a time-based limit (20 hours of AI-assisted screening) and a data-driven heuristic (50 consecutive irrelevant records).
• Inclusion Criteria: Studies were included if they: (a) sampled individuals with clinical anxiety/depressive diagnoses or high self-reported traits (excluding state-based or induced measures); (b) utilized behavioral tasks measuring decision-making or cognitive flexibility (excluding self-report measures); (c) included a healthy control or low-trait comparison group; and (d) focused on participants aged 18–44.
• Data Extraction and Conversion: 92 unique studies (yielding 100 effect sizes) met the criteria. Reported statistics (t-tests, ANOVAs, means) were converted to Hedges' g using the R package esc. For prospect-theory models, the loss-aversion parameter ($\lambda$) was extracted. Outliers were identified based on non-overlapping confidence intervals relative to the pooled effect.
• Statistical Analysis: A random-effects model was employed to account for anticipated between-subject heterogeneity, using the restricted maximum likelihood estimator for $\tau^2$. Pooled effects were calculated via the inverse-variance method with Knapp-Hartung adjustments for confidence intervals. Subgroup analyses compared anxiety versus depression, and sensitivity analyses (Egger's test) assessed publication bias.

Key Results

• Overall Effects: Both anxiety and depression were associated with negative effect sizes (indicating impairment) across all four outcome domains.
  • Anxiety & Decision-Making: Moderate negative effect ($g = -0.45$, 95% CI [-0.60, -0.30]) after outlier adjustment.
  • Anxiety & Flexibility: Small to moderate negative effect ($g = -0.25$, 95% CI [-0.48, -0.03]) after outlier adjustment.
  • Depression & Decision-Making: Moderate negative effect ($g = -0.39$, 95% CI [-0.53, -0.25]) after outlier adjustment.
  • Depression & Flexibility: Moderate negative effect ($g = -0.43$, 95% CI [-0.58, -0.28]) after outlier adjustment.
• Subgroup Comparisons: Crucially, subgroup analyses revealed no significant differences between anxiety and depression for either decision-making ($Q(1) = 0.31, p = 0.577$) or flexibility outcomes ($Q(1) = 2.15, p = 0.142$). The effect sizes for both disorders overlapped significantly.
• Heterogeneity: High heterogeneity was observed in the full models (e.g., $I^2$ ranging from 64.7% to 81.0%), which was reduced but not eliminated after outlier removal.
• Prediction Intervals: While subgroup differences were non-significant, prediction intervals for depression in decision-making were wider and spanned zero, suggesting future studies might show improved outcomes, whereas anxiety intervals remained consistently negative.

Key Contributions

1. Task-Based Synthesis: This study provides the first large-scale meta-analytic synthesis of task-based behavioral data for anxiety and depression regarding decision-making and flexibility, moving beyond the limitations of self-report meta-analyses.
2. Transdiagnostic Evidence: The findings challenge disorder-specific hypotheses (e.g., that anxiety causes larger decision-making deficits or depression causes larger flexibility deficits). Instead, the comparable effect sizes support a transdiagnostic perspective, suggesting that anxiety and depression share underlying mechanisms that impair higher-order cognition.
3. Methodological Innovation: The paper demonstrates the efficacy of integrating machine-learning-assisted screening (asReview) with manual verification in large-scale meta-analyses, successfully managing a dataset of nearly 24,000 records while maintaining rigorous inclusion criteria.

Significance and Claims
The authors claim that their results align with dimensional models of psychopathology (such as HiTOP and RDoC), which posit that anxiety and depression share a latent "internalizing" factor (specifically a "distress" subfactor) that drives cognitive impairment. The comparable deficits suggest that the mechanisms disrupting decision-making and flexibility are not unique to specific diagnostic categories but are likely driven by transdiagnostic factors such as working memory capacity, perseverative negative thought (rumination/worry), and neurobiological substrates like the amygdala and prefrontal cortex.

The paper modestly concludes that while distinct behavioral phenotypes exist, the underlying cognitive disruptions in decision-making and flexibility are broadly similar across anxiety and depression. The authors note limitations, including the exclusion of comorbid samples (which may limit generalizability to real-world presentations), the underrepresentation of social decision-making tasks in depression studies, and the potential for small-study effects in anxiety decision-making data. They suggest future research should employ computational modeling to distinguish between transdiagnostic and disorder-specific factors (e.g., learning rates, reward sensitivity) and investigate whether these findings extend to state-based anxiety.