Measuring and modifying information bias in depression

Background: Depression is a common illness which is believed to be caused, at least in part, by a tendency to preferentially process negative, at the expense of positive, events. Recent work in computational neuroscience indicates that individuals estimate how informative an event is and are more influenced by those events they judge to be more informative. This suggests a potential mechanism which may lead an individual to develop a negative bias—they may estimate that negative events are more informative than positive events. It also suggests a novel approach to the treatment of depression—interventions which reduce the estimated information content of negative relative to positive events should act to improve mood. A number of factors can influence how informative an event is, one of which is the changeability, or volatility, with which it is associated with a given outcome. The main aim of this study was to examine the role of the estimated information content of positive and negative outcomes, in depression. The study addresses two specific questions: First, do previously depressed patients, relative to never depressed controls, demonstrate a tendency to over-estimate the information content of negative outcomes? Second, can a simple training intervention alter an individual’s estimate of the information content of positive relative to negative outcomes, and does this influence other cognitive measures and symptoms of depression? The study described in this thesis is a large ongoing clinical study. The thesis describes an analysis of data from this study collected by myself during the first year of its recruitment, data collection for the larger study is ongoing. Methods: The study assessed the baseline cognitive biases using a volatility learning task and standard face based task, as well as depressive symptoms in two groups of participants: Previously depressed patients (N=16) and never depressed controls (N=18). After this, the participants were randomly assigned into either positive or neutral learning training for two weeks, and were tested using the same outcome measures immediately after training and then again four weeks later. Results: Across groups participants used a higher learning rate for informative relative to non-informative outcomes indicating that participants’ behaviour was influenced by the information content of the outcomes. No baseline difference in learning rate and inverse temperature between groups was found, although previously depressed patients did show a negative bias on the faces task. The study found that the learning training intervention led to the expected alteration in learning rates as well as having some impact on measures of anhedonia. Conclusion: Modification of learning rate for positive relative to negative outcomes with training was achieved although it did not lead to a convincing antidepressant effect. The facial based task is an efficient tool to measure previously depressed individuals’ negative cognitive bias. It will be important to assess the impact of learning training in a larger sample of participants.