Alzheimer's disease (AD), the most prevalent form of dementia, is characterized by amyloid-β peptide aggregation and neurofibrillary tangles, leading to neuronal loss in specific brain regions like the hippocampus. Cortical atrophy, including changes in surface area and thickness, is frequently observed in AD. However, the causal relationship between cortical structure and AD remains unclear. Observational studies have shown associations, but these cannot reliably distinguish correlation from causation or reverse causation. Confounding factors like aging further complicate the issue. Mendelian randomization, using single-nucleotide polymorphisms (SNPs) as instrumental variables, offers a powerful approach to address this causal inference problem by leveraging naturally occurring genetic variations. This study employed bidirectional Mendelian randomization to investigate the causal effects of cortical structure on AD risk and vice-versa, aiming to clarify the etiology and progression of AD.

Existing literature extensively documents the association between cortical atrophy and AD, with studies highlighting atrophy in regions such as the hippocampus, entorhinal cortex, medial temporal lobe, precuneus, and orbitofrontal cortex. Different patterns of atrophy have even been used to classify AD subtypes. However, the selective vulnerability of neurons in AD pathology and the heterogeneity of atrophy patterns make it challenging to establish a definitive relationship between cortical structure and AD. The difficulty is further compounded by the fact that neurodegenerative changes begin before clinical symptoms appear, and the effects of aging on brain structure can confound observational analyses.

This study used a bidirectional two-sample Mendelian randomization analysis. Cortical structure data (surface area and thickness of the whole cortex and 34 brain regions) from 33,992 participants of European ancestry (from the ENIGMA Consortium and UK Biobank) were used as the exposure. Genetic variants associated with AD were obtained from a meta-analysis of GWAS data (71,880 cases, 383,378 controls). The inverse-variance weighted (IVW) method was the primary analysis method, with sensitivity analyses (MR-Egger, weighted median) conducted to assess robustness against potential violations of Mendelian randomization assumptions. Pleiotropy analysis (MR-PRESSO) was used to detect horizontal pleiotropy. Leave-one-out and single SNP analyses were performed to identify influential SNPs. Two sets of p-value thresholds were used for genetic variants associated with the exposure in the bidirectional analysis. A threshold of p<10^-6 was set for the forward Mendelian randomization analysis and p ≤ 5×10^-8 for the reverse analysis. The study aimed to determine the causal effect of cortical structure on AD risk, and the causal effect of AD vulnerability on cortical structure. This bidirectional approach allows for a more comprehensive understanding of the relationship.

The study revealed suggestive evidence of an association between decreased surface area of the temporal pole (OR (95% CI): 0.95 (0.90, 0.997), p = 0.04) and decreased thickness of the cuneus (OR (95% CI): 0.93 (0.89, 0.98), p = 0.006) with increased AD risk. In the reverse analysis, suggestive associations were found between AD vulnerability and decreased surface area of the precentral (β (SE): -43.4 (21.3), p = 0.042) and insular cingulate (β (SE): -18.5 (7.3), p = 0.011). Sensitivity analyses generally supported these findings. However, none of the Bonferroni-corrected p-values were significant. The analysis also identified specific SNPs (e.g., rs68552426 for temporal pole, rs442495 near ADAM10 gene for cuneus) that significantly influenced some causal estimates. Further exploration of other brain regions (lateral orbitofrontal, supramarginal, lingual, lateral occipital) yielded some suggestive, but not consistently robust, associations.

The findings suggest a potential causal link between atrophy in specific cortical regions (temporal pole, cuneus) and increased AD risk. The association between AD vulnerability and reduced surface area in precentral and insular cingulate regions aligns with known motor and cognitive impairments in AD. The unexpected finding of a positive association between genetically predicted increased volume in occipital lobes and AD may be related to amyloid plaque effects or neurogenesis. The study's limitations, including the relatively small number of SNPs used for some analyses and the focus solely on surface area and thickness as measures of cortical structure, need to be considered.

This study provides suggestive evidence for a causal link between specific cortical atrophy patterns and increased AD risk, and also points towards a causal relationship between AD vulnerability and cortical surface area reduction in certain brain regions. However, the limitations of the study necessitate further research with larger sample sizes and more comprehensive measures of cortical structure to confirm these findings and fully elucidate the complex relationship between brain structure and AD.

The study's limitations include the relatively small number of SNPs used as instrumental variables for some analyses, the focus on only surface area and thickness as measures of cortical structure, and the reliance on the assumptions of Mendelian randomization. The power of the analysis could have been increased by including more SNPs but this would risk violating the assumptions of Mendelian randomization and increase the likelihood of bias. The relatively low fraction of variance explained by the SNPs also presents a limitation. Future studies with larger sample sizes are needed to confirm these results.