Reward-Processing Behavior in Depressed Participants Relative to Healthy Volunteers: A Systematic Review and Meta-analysis

Abstract

Importance: Dysfunctional reward processing is a leading candidate mechanism for the development of certain depressive symptoms, such as anhedonia. However, to our knowledge, there has not yet been a systematic assessment of whether and to what extent depression is associated with impairments on behavioral reward-processing tasks.

Objective: To determine whether depression is associated with impairments in reward-processing behavior.

Data sources: The MEDLINE/PubMed, Embase, and PsycInfo databases were searched for studies that investigated reward processing using performance on behavioral tasks by individuals with depression and nondepressed control groups, published between January 1, 1946, and August 16, 2019.

Study selection: Studies that contained data regarding performance by depressed and healthy control groups on reward-processing tasks were included in the systematic review and meta-analysis.

Data extraction and synthesis: Summary statistics comparing performance between depressed and healthy groups on reward-processing tasks were converted to standardized mean difference (SMD) scores, from which summary effect sizes for overall impairment in reward processing and 4 subcomponent categories were calculated. Study quality, heterogeneity, replicability-index, and publication bias were also assessed.

Main outcome and measures: Performance on reward-processing tasks.

Results: The final data set comprised 48 case-control studies (1387 healthy control individuals and 1767 individuals with major depressive disorder). The mean age was 37.85 years and 58% of the participants were women. These studies used tasks assessing option valuation (n = 9), reward bias (n = 6), reward response vigor (n = 12), reinforcement learning (n = 20), and grip force (n = 1). Across all tasks, depression was associated with small to medium impairments in reward-processing behavior (SMD = 0.345; 95% CI, 0.209-0.480). When examining reward-processing subcomponent categories, impairment was associated with tasks assessing option valuation (SMD = 0.309; 95% CI, 0.147-0.471), reward bias (SMD = 0.644; 95% CI, 0.270-1.017), and reinforcement learning (SMD = 0.352; 95% CI, 0.115-0.588) but not reward response vigor (SMD = 0.083; 95% CI, -0.144 to 0.309). The medication status of the major depressive disorder sample did not explain any of the variance in the overall effect size. There was significant between-study heterogeneity overall and in all subcomponent categories other than option valuation. Significant publication bias was identified overall and in the reinforcement learning category.

Conclusions and relevance: Relative to healthy control individuals, individuals with depression exhibit reward-processing impairments, particularly for tests of reward bias, option valuation, and reinforcement learning. Understanding the neural mechanisms driving these associations may assist in designing novel interventions.

Figure 1.. Flow Diagram of Study Selection and Inclusion

Figure 2.. Forest Plot of Option Valuation, Reward Response Vigor, and Grip Force

Rectangles and horizontal lines represent, respectively, standardized mean difference (SMD) scores and 95% confidence intervals of individual studies (A, Option valuation studies. B, Reward response vigor studies. C, Grip force task). Diamonds represent the summary effects and 95% confidence intervals for the respective reward processing subcomponent category. HC indicates healthy control; MDD, major depressive disorder.

Figure 3.. Forest Plot of Reward Bias and Reinforcement Learning

Rectangles and horizontal lines represent, respectively, standardized mean difference (SMD) scores and 95% confidence intervals of individual studies (A, Reward bias studies. B, Reinforcement learning studies). Diamonds represent the summary effects and 95% confidence intervals for the respective reward processing subcomponent category. HC indicates healthy control; MDD, major depressive disorder.

Figure 4.. Contour-Enhanced Funnel Plot

The solid black vertical line represents the observed overall summary effect. The unfilled funnel represents 95% confidence intervals for individual effect estimates, assuming no bias. The dashed black vertical line represents the summary effect when including imputed studies (using the trim-and-fill method). The dashed orange line represents the Egger regression line. The light blue funnel represents the 95% confidence intervals and the dark blue funnel represents the 99% confidence intervals for individual effect estimates, including imputed studies. The colored shapes represent individual included effect estimates: red circles for option valuation, blue squares for reward response vigor, orange inverted triangle for the single grip force task, green diamonds for reward bias, and purple triangles for reinforcement learning. All dark gray shapes represent effect estimates imputed by trim-and-fill.