Belief Updating in Bipolar Disorder Predicts Relapse

Post by Flora Moujaes

What's the science?

Bipolar disorder affects around 3% of the population and is characterized by periods of mania and depression, interspersed with periods of wellbeing. A key question in bipolar disorder research is whether it is possible to predict when relapse will occur in order to treat patients earlier. One approach to addressing this question is to examine how bipolar disorder patients learn about self-relevant information. We know that healthy individuals update their beliefs more in response to new positive information than negative information. This pattern of learning is often distorted in individuals with mental health disorders. For example, individuals with depression update their beliefs more in response to negative information than positive information, leading to a more pessimistic outlook. This week in Elife, Ossola and colleagues investigate whether it is possible to predict when an individual with bipolar disorder will relapse by examining how they update their beliefs.

How did they do it?

In order to explore whether changes in belief updating could predict relapse in bipolar disorder, 36 individuals diagnosed with bipolar disorder performed a belief-updating task during a period of wellbeing. They were then monitored for symptoms of bipolar disorder every 2 months for the next 5 years. In the task, participants were presented with 40 adverse life events, such as a robbery, and asked to estimate how likely it was to happen to them in the future (first estimate). They were then presented with information about how likely the event was to happen in a demographically similar population. The information provided was either positive (e.g. robberies occur less frequently than the participant estimated in the demographic similar to them) or negative (e.g. robberies occur more frequently than the participant estimated). In a second session, participants were then asked to provide an estimate of how likely the same event was to happen to them (second estimate). By measuring the difference between the first and second estimate, the task is able to capture how participants update their beliefs based on new information. The researchers also controlled for confounding factors such as participants' memory, and their familiarity with the adverse event.

What did they find?

The researchers found that there was an association between belief updating and time to relapse. Participants who were more likely to update their beliefs in response to good news relative to bad news took longer to relapse. Interestingly, the reduction of a positivity bias in belief updating was predictive of both depressive and manic episodes, which is consistent with the clinical observation that stressors are equally likely to trigger episodes of depression and mania. 

The researchers also showed that change in belief updating was a better predictor of when participants would relapse than traditionally used clinical and demographic indicators such as age, education, gender, medication, duration of illness, and depression score. Finally, they used a machine learning method (leave-one-out validation) to show that including the update bias in their model was crucial in order for the model to predict the time to relapse.

What's the impact?

This is the first study to show that greater belief updating in response to positive information relative to negative information predicts a longer time to relapse in individuals with bipolar disorder. This indicates that biased processing of information in a manner that supports an optimistic outlook is linked to a more favorable course of bipolar disorder, while biased processing of information in a manner that supports a pessimistic outlook may provide a more fruitful environment for clinical symptoms to manifest. Not only is this finding important for understanding the relationship between valence-dependent learning and mood, but it may have wide-reaching clinical implications in terms of developing preventative treatments, the identification of high-risk patients, and developing tools for the early diagnosis of affective disorders.

Ossalo et al. Belief updating in bipolar disorder predicts time of recurrence. Elife (2020). Access the original scientific publication here.