Abstract
Current evidence-based treatments for post-traumatic stress disorder (PTSD) are efficacious in only part of PTSD patients. Therefore, novel neurobiologically informed approaches are urgently needed. Clinical and translational neuroscience point to altered learning and memory processes as key in (models of) PTSD psychopathology. We extended this notion by clarifying at a meta-level (i) the role of information valence, i.e. neutral versus emotional/fearful, and (ii) comparability, as far as applicable, between clinical and preclinical phenotypes. We hypothesized that cross-species, neutral versus emotional/fearful information processing is, respectively, impaired and enhanced in PTSD. This preregistered meta-analysis involved a literature search on PTSD+Learning/Memory+Behavior, performed in PubMed. First, the effect of information valence was estimated with a random-effects meta-regression. The sources of variation were explored with a random forest-based analysis. The analyses included 92 clinical (N = 6732 humans) and 182 preclinical (N = 6834 animals) studies. A general impairment of learning, memory and extinction processes was observed in PTSD patients, regardless of information valence. Impaired neutral learning/memory and fear extinction were also present in animal models of PTSD. Yet, PTSD models enhanced fear/trauma memory in preclinical studies and PTSD impaired emotional memory in patients. Clinical data on fear/trauma memory was limited. Mnemonic phase and valence explained most variation in rodents but not humans. Impaired neutral learning/memory and fear extinction show stable cross-species PTSD phenotypes. These could be targeted for novel PTSD treatments, using information gained from neurobiological animal studies. We argue that apparent cross-species discrepancies in emotional/fearful memory deserve further in-depth study; until then, animal models targeting this phenotype should be applied with utmost care.