Testing a theory of avoidance learning & overgeneralization in chronic pain

当我们碰到热炉子或把脚撞在客厅的桌子上时，就会疼。虽然让人恼火

When we touch a hot stove or bang our foot against the living room table, it hurts. While irritating in the moment, acute pain serves an important purpose. First, it triggers an immediate defensive response to prevent or avoid further tissue damage. Second, it serves as a potent teaching signal to make sure we avoid making the same mistake in the future. This is an adaptive response if pain perception subsides when the tissue heals. However, when pain persists longer than 3 months, maladaptive learning processes in the brain that misinterpret these teaching signals may contribute to pain chronification. Analogous to a fire alarm that continues to ring even after the fire has been extinguished, the brain is thought to continue to predict that it is in a context of harm and engages in avoidance behavior even after the tissue has healed. Importantly these harm beliefs are thought to translate to other non-harm contexts, leading to an over-generalization of harm avoidance. This harm-avoidance model of pain ties in nicely with recent neuroimaging findings showing alterations in the fronto-striatal circuitry involved in the representation of context values and reinforcement learning in chronic pain Patients. However, the mechanism by which these interacting brain systems contribute to chronic pain or amelioration thereof is unknown. Here we expand a learning task and computational models in combination with EEG to differentiate between (PFC-driven) estimates of context values and (BG-driven) learning in chronic pain patients. Specifically, we test the hypothesis that avoidance learning in chronic pain patients may be driven by an overemphasis on learning from negative outcomes and an overgeneralization from one context of harm avoidance to other contexts. We hypothesize that this can be additionally observed in increased EEG frontal theta-band power in chronic patients during harm avoidance learning. Theta-band power has been previously shown to covary with the representation of context values and learning and thus provides an important read-out of biased learning throughout the task. This work will provide the basis for a future longitudinal research study where this task will be used to predict the risk of developing chronic pain after an acute pain state.