Adaptive coding of pain prediction error in the anterior insula

Abstract

BackgroundUnderstanding the mechanisms behind the influence of expectation and context on pain perception is crucial for improving analgesic treatments. Prediction error (PE) signals how much a noxious stimulus deviates from expectation and is therefore crucial for our understanding of pain perception. It is thought that the brain engages in adaptive coding of pain PE, such that sensitivity to unexpected outcomes is modulated by contextual information. While there is behavioural evidence that pain is coded adaptively, and evidence that reward PE signals are coded adaptively, controversy remains regarding the underlying neural mechanism of adaptively-coded pain PEs. MethodsA cued-pain task was performed by 19 healthy adults while undergoing FMRI scanning. BOLD responses to the task were tested using an axiomatic approach to identify areas that may code pain PE adaptively. ResultsThe left dorsal anterior insula demonstrated a pattern of response consistent with adaptively-coded pain PE. Signals from this area were sensitive to both predicted pain magnitudes on the instigation of expectation, and the unexpectedness of pain delivery. Crucially however, the response at pain delivery was consistent with the local context of the pain stimulation, rather than the absolute magnitude of delivered pain, a pattern suggestive of an adaptively-coded PE signal. ConclusionsThe study advances our understanding of the neural basis of pain prediction. Alongside existing evidence that the periaqueductal grey codes pain PE and the posterior insula codes pain magnitude, the results highlight a distinct contribution of the left dorsal anterior insula in the processing of pain. Significance statementAlthough there is behavioural evidence that pain is coded adaptively, the neural mechanisms serving this process are not well understood. This study used functional MRI to provide the first evidence that the left dorsal anterior insula, an area associated with aversive learning, responds to pain in a manner consistent with the adaptive coding of pain prediction error. This study aids our understanding of the neural basis of subjective pain representation, and thus can contribute to the advancement of analgesic treatments.