The post-acute effects of COVID-19 are not associated with robust long-term regional or global brain volumetric changes: A Bayesian meta-analytic evidence

Acute respiratory and systemic manifestations associated with SARS-CoV-2 infection have been extensively investigated, including their potential effects on brain structure. However, whether structural magnetic resonance imaging (MRI) studies converge towards reproducible long-term neuroanatomical alterations following COVID-19 remains unclear, mostly due to high variability in methods and reported results across neuroimaging investigations. The present study applied a Bayesian meta-analytic approach to quantify the strength of the evidence supporting the presence or absence of reproducible brain volumetric alterations following COVID-19 infection. The systematic literature search identified nine independent magnetic resonance imaging (MRI) experiments reporting whole-brain volumetric measures and 12 MRI experiments reporting regional volumetric measures (i.e., for specific regions of interest), including a total of 1107 subjects with a history of COVID-19 disease and 609 control participants. The results reported in the selected studies were quantitatively synthesized using the Rouder-Morey common-effect Bayesian meta-analytic framework. Bayes’ Factors (BF) were computed to quantify how strongly the available evidence supported the alternative hypothesis of volumetric differences relative to the null hypothesis of no differences. Global analyses provided limited evidence of volumetric differences, with BF₁₀ = 3.286 for whole-brain volume and BF₁₀ = 1.206 for total gray matter volume, corresponding to moderate and anecdotal evidence in favor of the alternative hypothesis, respectively. Regional analyses did not identify reproducible volumetric alterations in any of the 26 examined regions of interest. Overall, the present findings indicate that although several studies reported brain abnormalities detectable by MRI following COVID-19, the available volumetric evidence does not support a consistent pattern.