Instrumental Activities of Daily Living as a Mediator Between Cognition and Anti-Saccade Uncorrected Error Rate in CSVD: An Exploratory Analysis with Education Moderation Testing

Background: Cerebral small vessel disease (CSVD) is one of the major causes of cognitive impairment and functional decline in older adults. Instrumental activities of daily living (IADL) may serve as a bridge between cognitive impairment and more basic executive control; however, whether IADL mediates the association between global cognition and anti-saccade inhibitory control in patients with CSVD remains unclear. This study examined the mediating role of IADL between the Montreal Cognitive Assessment (MoCA) and anti-saccade metrics, and exploratorily tested whether years of education moderated this pathway. Methods: A total of 171 patients with CSVD were enrolled. The anti-saccade uncorrected error rate, corrected error rate, and correct rate were collected. Using IADL as the mediator, MoCA as the independent variable, and relevant factors as covariates, mediation and moderated mediation analyses were conducted with PROCESS Model 4, Model 7, and Model 14. The same model specifications were applied to all three anti-saccade outcomes. Benjamini–Hochberg FDR correction was used to address multiplicity across outcome screening and interaction tests. Because no formal direct comparisons of indirect effects among the three outcomes were performed, differences in statistical significance across outcomes were not interpreted as evidence of outcome-specific mediation or distinct neural mechanisms. Results: (1) In Model 4: in the uncorrected analysis, the indirect effect of IADL between MoCA and the anti-saccade uncorrected error rate was significant (95% BootCI [−1.138, −0.003]); however, after Benjamini–Hochberg FDR correction, this indirect effect was no longer significant (Bootstrap p = 0.153, BH-corrected p = 0.212). It should be noted that the unadjusted bootstrap p value was already greater than 0.05; therefore, its non-significance was not solely attributable to FDR multiplicity correction. The indirect effects for the anti-saccade corrected error rate and correct rate remained non-significant after FDR correction (both BH-corrected p = 0.212); no direct comparisons among outcomes were performed in this study. (2) In Model 14: the IADL × years of education interaction term for the uncorrected error rate was significant before correction (p = 0.034), but was no longer significant after FDR correction (BH-corrected p = 0.051). More importantly, the moderated mediation indices for the uncorrected error rate, corrected error rate, and correct rate were all non-significant (uncorrected error rate: 95% BootCI [−0.021, 0.123]; corrected error rate: 95% BootCI [−0.186, 0.016]; correct rate: 95% BootCI [−0.052, 0.119]), suggesting that years of education did not significantly moderate the overall indirect effect. (3) HC3 robust standard error sensitivity analysis showed that in the uncorrected error rate model, path a in Model 4 was no longer significant (p = 0.077), and the Bootstrap 95% confidence interval for the indirect effect was [−1.132, −0.002], with its upper limit close to 0; in Model 14, the IADL × years of education interaction term was also no longer significant (p = 0.211), indicating limited robustness of the above indirect association. Conclusions: In this exploratory cross-sectional analysis, IADL may partially mediate the association between global cognition and the anti-saccade uncorrected error rate in patients with CSVD before correction; however, this indirect effect was not statistically supported after multiplicity correction. It should be emphasized that the unadjusted bootstrap p value was itself non-significant, and this p value was based on a normal approximation and serves as an exploratory reference only; after considering the robustness of the bootstrap confidence interval, this indirect effect did not reach statistical robustness. The findings of this study did not provide evidence for outcome-specific mediation, and the moderated mediation effect of years of education was not statistically supported. These results should be interpreted as exploratory associations and require further validation in independent longitudinal samples.