Background <p>Current evidence suggests an association between late-life depression (LLD), obesity, and cardiovascular metabolic diseases (CMD). However, no studies have yet investigated the correlation between the ratio of non-high-density lipoprotein cholesterol to high-density lipoprotein cholesterol(NHHR), residual cholesterol (RC), and LLD in China, nor the mediating function of CMD.</p> Methods <p>Participants in the China Health and Retirement Longitudinal Study (CHARLS) from 2011 to 2020 were included in the study. Multivariable logistic regression models were employed to assess the relationship between NHHR, RC, and LLD. Restricted cubic splines (RCS) analysis, in conjunction with threshold effect models, was employed to discern potential nonlinear associations and estimate the corresponding inflection points. Subsequently, subgroup analysis and interaction tests were carried out to bolster the robustness of the research findings. Furthermore, structural equation modeling (SEM) was utilized to corroborate the mediating effect of CMD.</p> Results <p>In the fully adjusted model, ln(NHHR) (OR 0.78, 95% CI: 0.70–0.88) and ln(RC) (OR 0.90, 95% CI: 0.85–0.95) showed significant negative links to the risk of LLD. RCS analysis revealed a nonlinear relationship between ln(NHHR) and ln(RC) with the risk of LLD (P for nonlinear &lt; 0.001), with threshold values of 0.993 and 1.994, respectively. Through threshold effect analysis, it was found that ln(NHHR) had a significantly negative correlation with LLD risk when below the inflection point of 0.993 (OR 0.55, 95% CI: 0.448–0.675). In contrast, ln(RC) was significantly negatively associated with LLD risk when above the inflection point of 1.994 (OR 0.811, 95% CI: 0.751–0.876). Mediation analysis confirmed a significant negative association between ln(NHHR) and LLD in the longitudinal time series (β -0.070, <i>P</i> = 0.036). While CMD was positively associated with LLD (β 0.290, <i>P</i> = 0.036) and exhibited an suppression effect, partially counteracting the direct negative association between ln(NHHR) and LLD (proportion of mediation = − 102%).</p> Conclusion <p>This research indicates a significant negative correlation between NHHR and LLD at both fixed time points and within the longitudinal time series, with CMD partially counteracting this relationship. In contrast, RC was only significantly negatively associated with LLD at the fixed time point. Therefore, controlling NHHR and RC levels, as well as managing CMD, may help prevent LLD.</p>