TSPO PET brain inflammation imaging: A transdiagnostic systematic review and meta-analysis of 156 case-control studies

Introduction: The 18-kDa translocator protein (TSPO) is increasingly recognized as a molecular target for PET imaging of inflammatory responses in various central nervous system (CNS) disorders. However, the reported sensitivity and specificity of TSPO PET to identify brain inflammatory processes appears to vary greatly across disorders, disease stages, and applied quantification methods. To advance TSPO PET as a potential biomarker to evaluate brain inflammation and anti-inflammatory therapies, a better understanding of its applicability across disorders is needed. We conducted a transdiagnostic systematic review and meta-analysis of all in vivo human TSPO PET imaging case-control studies in the CNS. Specifically, we investigated the direction, strength, and heterogeneity associated with the TSPO PET signal across disorders in pre-specified brain regions, and explored the demographic and methodological sources of heterogeneity.Methods: We searched for English peer-reviewed articles that reported in vivo human case-control TSPO PET differences. We extracted the demographic details, TSPO PET outcomes, and technical variables of the PET procedure. A random-effects meta-analysis was applied to estimate case-control standardized mean differences (SMD) of the TSPO PET signal in the lobar/whole-brain cortical grey matter (cGM), thalamus, and cortico-limbic circuitry between different illness categories. Heterogeneity was evaluated with the I-2 statistic and explored using subgroup and meta-regression analyses for radioligand generation, PET quantification method, age, sex, and publication year. Significance was set at the False Discovery Rate (FDR)-corrected P < 0.05.Results: 156 individual case-control studies were included in the systematic review, incorporating data for 2381 healthy controls and 2626 patients. 139 studies documented meta-analysable data and were grouped into 11 illness categories. Across all the illness categories, we observed a significantly higher TSPO PET signal in cases compared to controls for the cGM (n symbolscript 121 studies, SMD symbolscript 0.358, PFDR < 0.001, I-2 symbolscript 68%), with a significant difference between the illness categories (P symbolscript 0.004). cGM increases were only significant for Alzheimer's disease (SMD symbolscript 0.693, PFDR < 0.001, I-2 symbolscript 64%) and other neurodegenerative disorders (SMD symbolscript 0.929, PFDR < 0.001, I-2 symbolscript 73%). Cortico-limbic increases (n symbolscript 97 studies, SMD symbolscript 0.541, P < 0.001, I-2 symbolscript 67%) were most prominent for Alzheimer's disease, mild cognitive impairment, other neurodegenerative disorders, mood disorders and multiple sclerosis. Thalamic involvement (n symbolscript 79 studies, SMD symbolscript 0.393, P < 0.001, I-2 symbolscript 71%) was observed for Alzheimer's disease, other neurodegenerative disorders, multiple sclerosis, and chronic pain and functional disorders (all PFDR < 0.05). Main outcomes for systemic immunological disorders, viral infections, substance use disorders, schizophrenia and traumatic brain injury were not significant. We identified multiple sources of between-study variance to the TSPO PET signal including a strong transdiagnostic effect of the quantification method (explaining 25% of between-study variance; V-T-based SMD symbolscript 0.000 versus reference tissue-based studies SMD symbolscript 0.630; F symbolscript 20.49, df symbolscript P < 0. 001), patient age (9% of variance), and radioligand generation (5% of variance).Conclusion: This study is the first overarching transdiagnostic meta-analysis of case-control TSPO PET findings in humans across several brain regions. We observed robust increases in the TSPO signal for specific types of disorders, which were widespread or focal depending on illness category. We also found a large and transdiagnostic horizontal (positive) shift of the effect estimates of reference tissue-based compared to V-T-based studies. Our results can support future studies to optimize experimental design and power calculations, by taking into account the type of disorder, brain region-of-interest, radioligand, and quantification method.