Impact of repetitive negative thinking on subjective cognitive decline: insights into cognition and brain structure

The global rise in dementia prevalence poses a concerning public health challenge, underscoring the urgent need for comprehensive strategies to address its impact on healthcare systems worldwide. Approximately 40% of global dementia, the leading cause of which is Alzheimer’s disease (AD), could potentially be prevented or delayed by addressing modifiable factors encompassing lifestyles across the lifespan. The impact of lifestyle factors on inter-individual differences in cognitive aging and susceptibility to clinical dementia has been described by the theory of cognitive reserve and resilience-related processes. Complementarily, the cognitive debt hypothesis suggests that cumulative cognitive deficits resulting from various psychological risk factors may deplete cognitive reserve (e.g., depression and anxiety symptoms, distress, neuroticism, sleep disorders, and life stressors), thereby increasing vulnerability to AD and leading to diminished cognitive functioning over time. Repetitive negative thinking (RNT) is a modifiable and central transdiagnostic construct within the cognitive debt hypothesis due to its applicability across various conditions. It is defined by self-relevant persistent negative thoughts that activate cognitive representations of previous (past-directed rumination) or anticipated (future-directed worry) stressful events. RNT prolongs stress physiological responses and has been independently related to amyloid and tau deposition, as well as cognitive decline in healthy adults. RNT better differentiated subjective cognitive decline (SCD) classification in older adults than other psychological factors. SCD refers to individuals experiencing a worsening in cognitive function compared to a previously normal status, without objective cognitive impairment in a neuropsychological assessment. Individuals with SCD have an increased risk of developing dementia compared with those without SCD, leading to the acknowledgment of SCD as a preclinical stage in the AD continuum. Indicators of cerebral status in SCD include cortical thickness (CTh) and white matter lesion (WML) volume assessed by structural MRI. SCD is associated with thinner temporal and parietal cortices and greater global WML burden. RNT has shown both positive and negative associations with CTh in regions involved in cognitive control and emotional processing, particularly in prefrontal and cingulate regions, and has been negatively associated with white matter microstructure alterations. Given these associations, it becomes relevant to explore how RNT may impact brain integrity in individuals with SCD. Aims: (1) characterize differences in global cognition, RNT, and brain integrity (CTh and WML) between individuals with SCD and matched controls; (2) examine the association between global cognition and brain integrity in the groups; and (3) study the influence of RNT levels on global cognition, brain integrity, and their relationships by group. Hypotheses: SCD individuals will exhibit higher levels of RNT, lower brain integrity, and lower global cognition compared to controls; higher RNT may strengthen negative associations between cognition and brain integrity in SCD.

Design: Cross-sectional study within the community-based Barcelona Brain Health Initiative (BBHI) cohort. Participants: 616 volunteers completed SCD screening statements and underwent in-person medical, cognitive, and MRI assessments. SCD classification: “Yes” to both statements (“In the last 2 or 3 years my memory has worsened considerably” and “I am worried about the state of my memory”) defined SCD; “No” to both defined controls. After inclusion criteria, 89 SCD participants remained; 89 matched controls (from 318 eligible controls) were selected by age, sex, and education using MatchIt (nearest method). Final sample: n = 178 (mean age 56.13 ± 7.72; 68.54% female; 69.10% tertiary education). Measures: RNT assessed via three questionnaires—Perseverative Thinking Questionnaire (PTQ; 15 items, 5-point Likert), Penn State Worry Questionnaire abbreviated version (PSWQ; 8 items, 5-point), Rumination Responses Scale (RRS; 22 items, 4-point, Spanish version). RNT composite: sum of z-scored PTQ, PSWQ, and RRS (higher scores = higher RNT). Global cognition: Preclinical Alzheimer’s Cognitive Composite 5 (PACC5; abridged)—MMSE (dementia screening), RAVLT delayed recall (episodic memory), SDMT total (processing speed/attention), category fluency (animals). Each test z-scored; unweighted average computed. MRI acquisition: 3T Siemens MAGNETOM Prisma, 32-channel head coil; structural T1-weighted images (0.8-mm isotropic). CTh maps generated with FreeSurfer v6.0; surface smoothing with 2D Gaussian kernel, FWHM 15 mm. WML burden: total volume (mm³) of voxels identified as white matter hypointensities in standard space, adjusted for estimated total intracranial volume. Statistical analyses: Normality assessed with Shapiro-Wilk; RNT, PACC5, WML non-normal (p < 0.05), group comparisons with non-parametric Wilcoxon/Kruskal-Wallis. Multivariate GLMs for pairwise relationships among PACC5, RNT, CTh, WML with group interaction terms; adjusted for age, sex, education. Model assumptions checked (Q-Q plots, residuals), residual normality via Shapiro-Wilk, outliers via Cook’s distance. Exploratory subgroup analyses within SCD: lowRNT-SCD vs highRNT-SCD for CTh and WML correlates.

• SCD prevalence among 616 screened: 16.55%; analysis sample: 89 SCD and 89 matched controls (n = 178; mean age 56.13; 68.54% female). • Group differences: SCD showed higher RNT than controls (Wilcoxon W = 2003, p < 0.001). Trend toward lower PACC5 in SCD (W = 4569.5, p = 0.055). No group difference in WML burden (W = 435, p = 0.326). • MRI vertex-wise: SCD had thinner cortex in right middle and inferior temporal regions vs controls (Monte Carlo-corrected cluster-wise p, CWP = 0.009). • PACC5–CTh: No whole-sample association; significant group interactions in four clusters (frontal/parietal). In SCD, greater right parietal CTh was positively associated with better PACC5 (CWP details in Figure 3). • PACC5–WML: No overall association (β = -0.000, SE = 0.000, t = -0.161, p = 0.872). Trend toward group interaction (β = -0.000, SE = 0.000, t = -1.965, p = 0.051), suggesting a negative, non-significant PACC5–WML relation only in SCD. • RNT–PACC5: Trend toward negative association in full sample (β = -0.029, SE = 0.016, t = -1.789, p = 0.075); no group interaction. • RNT–CTh: No full-sample effect; in SCD, higher RNT correlated positively with higher CTh in right supramarginal gyrus and superior temporal sulcus (CWP = 0.010). • RNT–WML: No full-sample effect (β = 0.000, SE = 0.000, t = 1.913, p = 0.058). Significant group interaction (β = 0.000, SE = 0.000, t = 2.305, p = 0.022); SCD showed a positive association (β = 0.000, SE = 0.000, t = 2.806, p = 0.006) contrasting with controls. • Exploratory SCD subgroup (highRNT-SCD vs lowRNT-SCD): Only highRNT-SCD showed positive PACC5–CTh associations—left hemisphere clusters including PCC, lingual, lateral occipital (CWP = 0.000) and superior parietal, precuneus, cuneus, isthmus cingulate (CWP = 0.007); right hemisphere cluster including superior parietal and precuneus (CWP = 0.035). HighRNT-SCD also showed positive RNT–CTh association in right insula and superior temporal regions (CWP = 0.041) and positive RNT–WML association (β = 0.000, SE = 0.000, t = 2.425, p = 0.020).

Community-based SCD individuals exhibited higher RNT and thinner temporal cortical thickness compared to matched controls, without significant differences in global cognition (PACC5) or WML burden. These findings point to early grey matter changes in SCD over white matter alterations and underscore the relevance of psychological factors such as RNT in SCD. Group interactions indicated that the PACC5—a preclinical dementia marker—was sensitive to early CTh variations in SCD across PCC, precuneus, cuneus, parietal, frontal, and ACC regions previously linked to early age-related cognitive changes and to amyloid-β burden in SCD with higher worry. While prior studies often found WML burden related to poorer cognition and SCD status, in this middle-aged to older adult sample WML burden did not differ by group and showed only a trend toward a negative association with PACC5 in SCD, suggesting WML levels observed may be insufficient to impact cognition. RNT demonstrated positive associations with CTh in parietotemporal areas (e.g., right supramarginal gyrus, superior/middle temporal gyri) only in SCD, consistent with literature suggesting that moderate levels of RNT in non-clinical populations may correspond to functional engagement of regions supporting emotion regulation and cognitive processes. Importantly, RNT was positively associated with WML burden in SCD, aligning with evidence linking RNT to white matter microstructural abnormalities and with findings that higher WML may relate to amyloid positivity and progression risk in SCD. Exploratory analyses in highRNT-SCD highlighted positive associations of cognition with CTh in PCC, parietal and precuneus regions, and of RNT with CTh in insula/superior temporal areas, suggesting potential RNT-related influences on salience network structures implicated in responding to salient stimuli and emotional demands. Overall, results support RNT as a modifiable psychological factor potentially impacting brain integrity in SCD and warrant longitudinal investigation of its role in the interplay between cognition and brain structure.

Individuals with SCD had higher repetitive negative thinking and reduced cortical thickness in the right temporal cortex compared to controls. Only in SCD, global cognitive status was associated with cortical thickness, and RNT was positively related to both grey and white matter integrity measures. These findings emphasize incorporating modifiable psychological factors such as RNT alongside neurobiological markers when assessing neurocognitive status in SCD. Further longitudinal studies are needed to determine whether and how RNT influences the relationship between cognition and brain integrity and to evaluate interventions targeting RNT to mitigate cognitive decline and preserve brain health in at-risk populations.

• Community-based SCD sample may limit generalizability to clinical populations. • Variability in SCD criteria (e.g., SCD vs SCD-plus) complicates replication and cross-study comparisons. • Cross-sectional design precludes causal and temporal inferences; longitudinal studies are needed to clarify sequences of cognitive decline and brain changes and to test RNT-targeted interventions.