The mediating role of biological age in the impact of mood instability symptoms on neurodegenerative disease and mortality

Abstract

Background

Accumulating evidence links mood instability to an increased risk of adverse outcomes, yet its relationship with neurodegenerative diseases remains underexplored. This study examines the associations between mood instability symptoms and the risk of neurodegenerative diseases, exploring the mediating role of biological aging.

Methods

The participants in the UK Biobank without a diagnosed neurological condition at baseline were included. Mood instability symptoms were assessed at two time points: baseline (2006–2010, T1) and the third visit (2014 + , T2), using 13 items of mental symptoms (e.g., “mood goes up and down,” “miserable for no reason”). Biological aging was generated by two previously described measures of biological age based on 14 routinely measured clinical biomarkers (PhenoAge, Klemera-Doubal method age). Latent class analysis (LCA) was employed to classify individuals into different psychological clusters. Cox proportional hazard models and mediation analyses were used to examine the associations between mood instability symptoms, biological aging, and adverse health outcomes. The latent transition analysis (LTA) identified latent classes of transition, and multifactor logistic regression was used to examine the sex difference in the probability of transitioning between these latent classes.

Results

Of 185,818 included participants, 51.86% were female. During a median follow-up of 14.6 years, we identified three clusters with distinct mood patterns. Compared to individuals with a pattern of “stable,” individuals with severe mood instability symptoms at baseline was significantly associated with an increased risk of all-cause dementia (hazard ratio [HR] = 1.17, 95% CI: 1.02–1.35), Parkinson’s disease (HR = 1.47, 95% CI: 1.21–1.77), and all-cause mortality (HR = 1.07, 95% CI: 1.01–1.14). Accelerated biological aging was also linked to a higher risk of both dementia and Parkinson’s disease. Mediation analyses revealed that biological aging partially mediated the association between severe mood instability symptoms and neurodegenerative diseases, as well as mortality (ranging from 0.78% to 13.6%). Notably, females had a higher risk of transitioning to more severe groups over time.

Conclusions

These findings underscore the critical need to focus on severe mood instability symptoms in early intervention strategies to reduce neurodegenerative disease risk, particularly by focusing on emotional well-being in females.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04389-0.

Background

Neurodegenerative diseases, including Alzheimer’s disease (AD) and Parkinson’s disease (PD), are leading causes of mortality and morbidity worldwide, affecting more than 60 million individuals, particularly within the elderly population [1, 2]. Recent research suggests that the regulation of negative emotions may provide neuroprotective benefits in older adults, highlighting the potential for mental state modulation as a promising and innovative target for clinical intervention [3]. Despite this, the majority of the existing research has predominantly concentrated on pharmacological treatments [4], modifiable lifestyle factors [5, 6], and mental health disorders [7], with emotional factors receiving comparatively limited attention. Consequently, it is imperative to investigate further the associations between unstable mood and the increased risk of developing neurodegenerative diseases.

Mood instability, defined as ‘‘extreme and frequent fluctuations of mood over time’’ [8], has been linked to various adverse health outcomes [9–11], though its role in neurodegenerative diseases is still poorly understood. While mood instability often lacks uniform measurement, neuroticism is frequently equated with emotional instability in numerous studies [12–15]. Neuroticism is characterized by heightened propensity to experience adverse emotions, self-criticism, sensitivity to the criticism of others, and feelings of personal inadequacy. These attributes emanate from exaggerated responses to environmental stimuli [16]. Given the definitional similarities between the two concepts, we incorporated items from neuroticism assessments to draw on their established psychometric rigor for accurate risk identification. Furthermore, individuals with significant mood fluctuations are more prone to risk-taking behaviors [17–19], as mood swings can trigger impulsive actions [15, 20]. Several studies have used latent class analysis (LCA) to identify distinct psychological patterns, offering valuable insights into uncovering heterogeneous risk profiles that may be obscured by traditional, scale-based continuous scores [21, 22]. Building on this evidence, we utilized LCA to categorize individuals based on 13 items capturing core symptoms of mood instability (e.g., negative emotions, shame, rumination, and emotional extremity), intending to identify high-risk clusters. Additionally, there is an increasing body of literature that pays attention to sex differences in emotion, with different findings [23–25]. Given that, compared to a baseline assessment, examining changes in mood instability symptoms may provide a more effective indication of the target demographic for preventive interventions. Therefore, we further employed latent transition analysis (LTA) to estimate the sex difference in probabilities of transitions among mood instability symptoms over time.

Accelerated aging has emerged as a significant risk factor for the development of chronic diseases and mortality [26, 27]. Given recent findings suggesting that poor mental health is associated with more advanced and accelerated biological aging [28], exploring the impact of mood instability symptoms on accelerated aging is valuable. Studies have identified a correlation between premature aging and the onset of both dementia and PD [29]. Several methods for measuring biological age (BA) have been established, including omics-derived markers [30–32], clinically based measures (e.g., phenotypic age [33] and the Klemera-Doubal method (KDM) [34]), and phenotypic evaluations [35–37]. However, the lack of a gold standard remains a challenge. Among these, the KDM stands out for its strong predictive power for aging-related outcomes [34, 38]. In contrast, PhenoAge focuses on age-related physiological dysregulation, helping identify individuals at higher risk for chronic diseases. It also provides insights into genetic and environmental factors influencing aging in observational studies [39]. Thus, we employed both the KDM and PhenoAge algorithms in this research, as they have been extensively validated for predicting health outcomes in multiethnic cohorts of older adults [40, 41]. We hypothesize that the aging process may serve as a mediating factor in the associations between mood instability symptoms and neurodegenerative diseases, as well as mortality. This hypothesis, however, has received scant attention in the extant literature.

In this work, our objectives are to examine the following: (1) the relationship between mood instability symptoms and aging, (2) the association between aging and neurodegenerative diseases, (3) the mediating role of aging in the link between mood instability symptoms and adverse health outcomes, and (4) the sex disparities in the transition of mood instability symptoms.

Methods

Study population and design

Information on the design and methodology of the UK Biobank study has been previously documented [42]. The UK Biobank enrolled over 500,000 participants aged 37–73 Years from 2006 to 2010. Data on sociodemographic, lifestyle, and mental symptoms were collected via questionnaires, physical measures, and sample assays. Health outcomes were tracked through linkages with national data sets, including primary care, hospital inpatient, and death registries [43]. A recruitment flowchart is shown in Fig. 1.

Fig. 1.

Flow chart of inclusion and exclusion criteria

Identification of clusters of participants with distinct mood instability symptom patterns

Mood instability symptoms were assessed at baseline (2006–2010; T1) and follow-up (2014 + , T2) using a 13-item instrument capturing core symptoms (e.g., “mood goes up and down,” “miserable for no reason”; full items in the Additional file 1: Table S1). At recruitment, participants reported their current psychiatric symptoms through self-report questionnaires, where they could choose “Yes,” “No,” “Do not know,” or “Prefer not to answer.” Participants with non-response (“Do not know”/ “Prefer not to answer”) were excluded. The transitions of mood instability symptoms patterns were assessed in participants who completed psychological assessments at both T1 (2006–2010) and T2 (2014 +).

Biological aging assessment

In this study, we applied two distinct biological age metrics—KDM-BA and the PhenoAge, to assess biological aging more comprehensively. These validated biological age algorithms based on clinical parameters capture different facets of the biological aging process [33, 34, 44]. The details for the field codes of clinical biomarkers are available in the Additional file 1: Table S2.

We calculated KDM-BA using a regression model with nine clinical biomarkers [28, 34]. PhenoAge was calculated using data from the NHANES dataset, applying nine biomarkers and chronological age [38, 45]. We evaluated biological aging rates by calculating two forms of biological age accelerations, which are the residuals of KDM-BA and PhenoAge, adjusted for chronological age using natural spline regression with three degrees of freedom [46]. The calculation of these two biological age metrics was performed using the R package “BioAge” [47]. We evaluated the predictive ability of the two biological age accelerations for all-cause mortality risk and found significantly positive results (see Additional file 1: Table S3).

Ascertainment of outcome

Outcomes were determined using an “algorithm defining results,” which identifies the earliest recorded date of specific health events by integrating baseline data, hospital admissions, and death records. All-cause dementia was determined based on the International Classification of Diseases, 10th Revision (ICD-10) codes F00, F01, F02, F03, G30, G31.0, G31.1, G31.8, I67.3, A81.0, and F10.6. All-cause Parkinsonism was identified using ICD-10 codes G20–G26 and G90. The follow-up duration for each participant was calculated from the date of baseline assessment to the date of disease diagnosis, death, or end of follow-up, whichever came first. The details are available in the Additional file 1: Tables S4 and S5.

Covariates

At baseline, data on age, sex, ethnicity, education, drinking and smoking status, physical activity (PA), sleep duration, and family history of dementia/Parkinson’s were collected via a touchscreen questionnaire. Sleep duration was recorded as hours of sleep in 24 h, including naps. PA was assessed using the International Physical Activity Questionnaire (IPAQ), and participants were categorized by total PA (less than 500 or 500 + MET-min/week). The Townsend Deprivation Index (TDI) was calculated from residential postcode data. Body mass index (BMI) was determined from weight and height, while data on diabetes, depression, and hypertension were collected through self-reports and hospital records. For field code details, see Additional file 1: Table S6.

Statistical analyses

Demographic characteristics for each class at baseline were assessed. Descriptive statistics comprised the mean (SD) for continuous variables and percentages for categorical variables. Continuous variables were assessed through analysis of variance, and the χ² test was applied to categorical variables.

First, LCA was used to identify the potential clusters of participants with distinct mood status; participants with responses of “Prefer not to answer” and “Do not know” were treated as missing data and were not included in the analysis. The final model was chosen based on a combination of statistical fit indices and the interpretability of each class. Individuals were classified based on their posterior probabilities before the analysis. Additional file 1 (page 7) provides detailed descriptions of the LCA model fitting process. Second, LTA modeled the transition probabilities of these classes over a 4 + years follow-up from T1 to T2 [12] [48]. Measurement invariance was assumed that there is no difference in the way latent classes were constructed across the two waves [49]. The assumption of measurement invariance allows the transitions to be based on the changes in latent classes, instead of their compositions [50]. Third, a multifactor logistic regression analysis was performed to calculate odds ratios (ORs) for class transitions. Full methodological details are provided in the Additional file 1 (page 11).

We first fitted general multiple linear regression models to examine the associations between baseline mood status and aging. Model included age, sex, ethnicity, education, TDI, smoking status, drinking status, physical activity, BMI, sleep duration, family history, depression, diabetes, and hypertension. Secondly, Cox proportional hazards regression models were performed to investigate the associations of aging with neurodegenerative diseases and mortality and to estimate the hazard ratio (HR) and 95% confidence interval (CI) of mood instability symptoms and aging associated with neurodegenerative diseases. The Schoenfeld residuals method was employed to test the proportional hazards assumption, and the results indicated that the variables did not violate this assumption. Three models were developed: Model 1 was adjusted for age, sex, ethnicity, education, and TDI. Model 2 included additional adjustments for smoking status, drinking status, physical activity, BMI, sleep duration, and family history. Model 3 adjusted for the same variables as in Model 2, with further adjustments for depression, diabetes, and hypertension.

Formal mediation analyses examined whether the relationship between mood status, neurodegenerative disease, and mortality was mediated by aging, calculating mediation proportions and their 95% CIs. Mediation analyses were undertaken using the “CMAverse” package in R using the regression-based approach [51]. This counterfactual framework employed two interlinked regression models: a linear regression model for the continuous mediators (biological age acceleration) and a Cox proportional hazards model for the time-to-event outcome, both adjusting uniformly for age, sex, ethnicity, education, TDI, smoking status, drinking status, physical activity, BMI, sleep duration, family history, depression, diabetes, and hypertension. Using direct counterfactual imputation with 1000 nonparametric bootstrap replicates, we estimated the following: (1) the total effect (TE) of mood status on outcomes; (2) the total natural direct effect (TNDE), representing effects unmediated by biological aging; and (3) the total natural indirect effect (TNIE), quantifying mediation through biological aging pathways. We adjusted p-values for multiple testing using the Benjamini-Hochberg (BH) method, with a false discovery rate (FDR) threshold of < 0.05 for statistical significance.

We performed several sensitivity analyses. First, we conducted a stratified analysis to investigate the associations of mood status with neurodegenerative diseases in different subgroups. Second, using the HRs derived from the adjusted model, we employed Levin’s method to calculate the population-attributable fractions (PAFs) for estimating the share of adverse outcomes that would theoretically be avoided if the levels of severe mood instability symptoms were adjusted to the lowest levels. Third, to reduce the potential reverse causation, we repeated the primary analyses after excluding outcomes that occurred within the first 2 years of follow-up. Fourth, we further conducted a score based on the sum of 13 items of mental symptoms and repeated the primary analyses. Fifth, we used a multistate competing risks model to analyze the impact of PhenoAge acceleration and KDM acceleration on the risk from baseline to onset. Sixth, imputation of missing data on covariates for 51,869 participants using multiple imputation by chained equations (MICE) with 10 imputation templates generated, pooled using Rubin’s rules. Seventh, using available follow-up mood assessments (T1–T2, 4 years), we tested whether baseline biological age predicted subsequent mood instability symptoms transitions to evaluate the risk of reverse causality.

All data analyses were completed using R software version 4.3.1 and Mplus version 7. A two-tailed test with p < 0.05 was considered statistically significant.

Results

Baseline characteristics

According to the inclusion and exclusion criteria, 185,818 participants were included in the baseline. The three-class model was chosen as the final model based on statistical fit indices, clinical significance, and the interpretability of each class (Table 1, Figs. 2 and S1). For mean posterior probabilities and individual item loadings, see Additional file 1: Tables S7 and S8. Class 3 exhibited the highest likelihood of experiencing mood instability symptoms (e.g., mood swings, miserableness, irritability, sensitivity/hurt feelings), while class 3 showed the lowest likelihood. Cluster 1 was designated as the “stable group” (n = 75,314), cluster 2 as the “mild group” (n = 85,577), and cluster 3 as the “severe group” (n = 24,927). During a median follow-up of 14.6 years, 2343 cases of all-cause dementia, 1110 cases of Parkinson’s disease, and 1561 all-cause mortality were documented. Baseline characteristics of these participants are presented in Table 2. Compared to those in the stable group, participants with severe mood instability symptoms were younger, primarily female, biologically older, and had less education and physical activity, as well as higher BMI.

Table 1.

The model fitting index with identified clusters using latent class analysis (LCA)

Model	AIC	BIC	aBIC	Entropy	p-value		prob
					LMR	BLRT
1	3,679,144.511	3,679,279.357	3,679,238.042	-	-	-	-
2	3,218,670.197	3,218,950.261	3,218,864.454	0.821	< 0.001	< 0.001	0.41/0.59
3	3,144,241.905	3,144,667.188	3,144,536.888	0.770	0.3333	< 0.001	0.16/0.39/0.45
4	3,083,662.926	3,084,233.428	3,084,058.635	0.761	< 0.001	< 0.001	0.21/0.15/0.23/0.39
5	3,069,801.464	3,070,517.184	3,070,297.899	0.740	< 0.001	< 0.001	0.33/0.10/0.24/0.13/0.20

AIC, Akaike information criterion. BIC, Bayesian information criterion. aBIC, adjusted Bayesian information criterion. LMR, Lo-Mendell-Rubin adjusted likelihood ratio test. BLRT, bootstrap likelihood ratio test

Fig. 2.

3-class model of mood instability patterns

Table 2.

Baseline characteristics of study participants

		Mood instability symptoms			Pa
Characteristics	Total (n = 185,818)	Stable (n = 75,314)	Mild (n = 85,577)	Severe (n = 24,927)
Age, years	55.58 ± 8.08	56.45 ± 8.03	55.37 ± 8.06	53.69 ± 7.9	< 0.001
Biological age at baseline (year)
KDM-BAge (Levine method)	55.51 ± 8.23	56.38 ± 8.16	55.27 ± 8.23	53.69 ± 8.06	< 0.001
PhenoAge-BAge (Levine method)	49.44 ± 9.10	50.34 ± 9.13	49.12 ± 9.04	47.8 ± 8.87	< 0.001
Accelerated age
KDM-AAge (Levine method)	− 0.08 ± 1.49	− 0.07 ± 1.37	− 0.11 ± 1.55	− 0.01 ± 1.61	< 0.001
PhenoAge-AAge (Levine method)	− 0.09 ± 4.12	− 0.05 ± 4.08	− 0.20 ± 4.08	0.17 ± 4.31	< 0.001
Sex					< 0.001
Female	96,371 (51.86)	32,502 (43.16)	48,935 (57.18)	14,934 (59.91)
Male	89,447 (48.14)	42,812 (56.84)	36,642 (42.82)	9993 (40.09)
College/university degree	70,292 (37.83)	30,380 (40.34)	31,777 (37.13)	8135 (32.64)	< 0.001
White British	178,908 (96.28)	72,458 (96.21)	82,689 (96.63)	23,761 (95.32)	< 0.001
Townsend Deprivation Index	− 1.6 ± 2.91	− 1.74 ± 2.83	− 1.61 ± 2.88	− 1.1 ± 3.16	< 0.001
BMI, kg/m2	26.99 ± 4.44	27.06 ± 4.25	26.88 ± 4.47	27.14 ± 4.89	< 0.001
Physical activity (MET-min/week)					< 0.001
< 500	159,145 (85.65)	65,685 (87.21)	73,150 (85.48)	20,310 (81.48)
≥ 500	26,673 (14.35)	9629 (12.79)	12,427 (14.52)	4617 (18.52)
Never smoked cigarettes	104,462 (56.22)	43,729 (58.06)	47,734 (55.78)	12,999 (52.15)	< 0.001
Never drink alcohol	6085 (3.27)	2519 (3.34)	2608 (3.05)	958 (3.84)	< 0.001
Hypertension, n (%)	38,528 (20.73)	15,250 (20.25)	17,742 (20.73)	5536 (22.21)	< 0.001
Diabetes, yes, n (%)	5975 (3.22)	2505 (3.33)	2610 (3.05)	860 (3.45)	< 0.001
Depressed, yes, n (%)	14,860 (8)	1935 (2.57)	6601 (7.71)	6324 (25.37)	< 0.001
Family history of dementia, yes	26,085 (14.04)	10,539 (13.99)	12,130 (14.17)	3416 (13.7)	0.154
Family history of Parkinson, yes	6826 (3.67)	2758 (3.66)	3155 (3.69)	913 (3.66)	0.961

^ap-values for comparison between groups, obtained from analyses of variance for continuous variables and χ²-test for categorical variables, reflect overall differences across three latent mood profile groups

Association of mood instability symptoms with incident neurodegenerative diseases and all-cause mortality

Table 3 presents the association between severe mood instability symptoms and risks of incident dementia, PD, and all-cause mortality. After adjusting for confounders, participants with severe mood instability symptoms exhibited a significantly higher risk compared to those with stable moods. Specifically, severe mood instability symptoms were associated with a 17% increased risk of incident dementia (HR = 1.17, 95% CI = 1.02–1.35), a 47% increased risk of Parkinson’s disease (HR = 1.47, 95% CI = 1.21–1.77), and a 7% increased risk of all-cause mortality (HR = 1.07, 95% CI = 1.01–1.14). Additionally, severe mood instability symptoms were linked to a 16% increased risk of late-onset dementia (HR = 1.16, 95% CI = 1.00–1.34) and late-onset Parkinson’s disease (HR = 1.44, 95% CI 1.19–1.75) (Additional file 1: Tables S9 and S10).

Table 3.

The association of mood instability symptoms and the adverse outcomes

		Mood instability symptoms				p trend	HR (95% CI)
	Stable	Mild		Severe
		HR (95% CI)	p-value	HR (95% CI)	p-value
All-cause dementia
No. of cases/person-years	1018/1,070,171	1037/1,220,969		288/354,930
Model 1	Ref	1.07 (0.98, 1.17)	0.113	1.34 (1.17, 1.52)	< 0.001	< 0.001	1.13 (1.06, 1.21)
Model 2	Ref	1.06 (0.97, 1.15)	0.179	1.30 (1.14, 1.49)	< 0.001	< 0.001	1.11 (1.04, 1.23)
Model 3	Ref	1.03 (0.95, 1.13)	0.445	1.17 (1.02, 1.35)	0.024	0.045	1.07 (1.00, 1.14)
Parkinson’s disease
No. of cases/person-years	484/1,070,987	460/1,222,041		166/355,095
Model 1	Ref	1.04 (0.92, 1.19)	0.633	1.65 (1.37, 1.97)	< 0.001	< 0.001	1.21 (1.11, 1.32)
Model 2	Ref	1.04 (0.92, 1.19)	0.523	1.65 (1.37, 1.97)	< 0.001	< 0.001	1.22 (1.11, 1.33)
Model 3	Ref	1.02 (0.89, 1.16)	0.811	1.47 (1.21, 1.77)	< 0.001	0.002	1.15 (1.05, 1.26)
All-cause mortality
No. of cases/person-years	5254/1,072,632	1539/355,815		5204/1,223,815
Model 1	Ref	1.01 (0.97, 1.05)	0.526	1.20 (1.13, 1.27)	< 0.001	< 0.001	1.07 (1.04, 1.10)
Model 2	Ref	0.99 (0.96, 1.03)	0.735	1.13 (1.07, 1.20)	< 0.001	0.003	1.04 (1.01, 1.07)
Model 3	Ref	0.98 (0.94, 1.01)	0.277	1.07 (1.01, 1.14)	0.033	0.214	1.02 (0.99, 1.04)

Abbreviation: HR hazard ratio. No. of cases/person-years: number of incident cases over total person-years

Model 1 was adjusted for age, sex, ethnicity, education, and Townsend Deprivation Index

Model 2 was additionally adjusted for smoking status, drinking status, physical activity, body mass index, sleep duration, and family history

Model 3 was adjusted for the same variables as in model 2 and further for additionally adjusted for depression, diabetes, and hypertension

Association of biological aging with incident neurodegenerative diseases and all-cause mortality

After adjusting for the common covariates, KDM-BA acceleration (HR = 1.03, 95% CI = 1.01–1.07) and PhenoAge acceleration (HR = 1.02, 95% CI = 1.02–1.04) were all statistically significantly associated with an increased risk of all-cause dementia. Similarly, there was a statistically significant association between PhenoAge acceleration and increased risk of Parkinson’s disease (HR = 1.01, 95% CI = 1.01–1.02) (Additional file 1: Tables S11 and 12). PhenoAge acceleration (HR = 1.07, 95% CI = 1.06–1.07) and KDM-BA acceleration (HR = 1.12, 95% CI = 1.11–1.14) were all statistically significantly associated with an increased risk of all-cause mortality (Additional file 1: Table S3).

Association of mood instability symptoms with accelerated biological aging

The severe mood instability symptoms were associated with PhenoAge acceleration (β = 0.150 ± 0.030 SE, p < 0.001) and KDM-BA acceleration (β = 0.029 ± 0.011 SE, p = 0.006) (Additional file 1: Table S13).

The role of accelerated biological aging in the association of mood instability symptoms with incident neurodegenerative diseases and all-cause mortality.

Table 4 presents the mediation proportion of aging (measured by PhenoAge acceleration and KDM-BA acceleration, respectively) in adverse health outcomes attributed to severe mood instability symptoms. The mediation proportion of PhenoAge acceleration in associations of severe mood instability symptoms with risk of incident dementia, incident Parkinson’s disease, and all-cause mortality was 2.08%, 0.79%, and 13.63% (all p-values < 0.05), respectively. The mediation proportions of KDM-BA acceleration in the relationships between severe mood instability symptoms and the risk of all-cause mortality were 5.44%.

Table 4.

Associations of mood instability symptoms with outcomes and the mediation proportion of accelerated biological aging

Variables	Total effect	p- value	Natural direct effect	p-value	Natural indirect effect	p -value	Mediation proportion, %	p- value	PFDR- value
PhenoAge acceleration
All-cause dementia
Stable	Ref		Ref
Mild	1.00 (0.95, 1.06)	1.000	1.00 (0.96, 1.06)	0.800
Severe	1.17 (1.00, 1.36)	0.042	1.17 (1.01, 1.35)	0.050	1.00 (1.00, 1.01)	< 0.001	2.08 (0.41, 12.00)	0.042	0.044
Parkinson’s disease
Stable	Ref
Mild	0.93 (0.84, 0.99)	< 0.001	0.93 (0.84, 0.99)	< 0.001
Severe	1.47 (1.20, 1.69)	< 0.001	1.45 (1.20, 1.68)	< 0.001	1.00 (1.00, 1.01)	0.034	0.79 (0.10, 1.40)	0.034	0.044
All-cause mortality
Stable	Ref		Ref
Mild	0.97 (0.95, 0.99)	0.2	0.97 (0.95, 0.99)	0.200
Severe	1.08 (1.02, 1.14)	0.014	1.07 (1.01, 1.13)	0.020	1.01 (1.01, 1.01)	< 0.001	13.63 (6.88, 39.30)	0.014	0.044
KDM-BA acceleration
All-cause dementia
Stable	Ref		Ref
Mild	1.04 (0.97, 1.13)	0.364	1.04 (0.97, 1.14)	0.356
Severe	1.17 (1.01, 1.25)	< 0.001	1.16 (1.01, 1.35)	0.036	1.00 (0.99, 1.00)	0.088
Parkinson’s disease
Stable	Ref		Ref
Mild	1.02 (0.90, 1.16)	0.660	1.02 (0.90, 1.16)	0.664
Severe	1.48 (1.19, 1.77)	< 0.001	1.48 (1.20, 1.77)	< 0.001	0.99 (0.99, 1.00)	0.444
All-cause mortality
Stable	Ref		Ref
Mild	0.98 (0.94, 1.02)	0.336	0.98 (0.95, 1.02)	0.416
Severe	1.08 (1.02, 1.14)	0.012	1.08 (1.01, 1.14)	0.016	1.00 (1.00, 1.01)	< 0.001	5.44 (0.59, 38.40)	0.044	0.044

Stability of class membership over time and sex difference: a transition analysis

Transitions among different mood instability symptom groups over time are illustrated (Additional file 1: Tables S14 and S15 and Fig. 3). The vast majority of individuals in the stable group at wave 1 (96.1%) remained stable at wave 2, with a small proportion transitioning to the mild group (3.8%) and a very few moving to the severe group (0.1%). Most individuals in the mild group at wave 1 remained in the mild group at wave 2 (80.8%), with a smaller fraction transitioning to the stable group (16.1%) and a few to the severe group (3.1%). Among those in the severe group at wave 1, most remained in the severe group at wave 2 (73.4%), while a smaller proportion shifted to the mild group (25.4%), and very few moved to the stable group (1.2%). These findings demonstrate a high degree of stability in mood states over time. Compared to males, females initially in a severe state demonstrated significantly higher status maintenance (75.2% vs. 71.0%). In contrast, males exhibited greater transition probabilities from severe to mild (27.4% vs. 23.8%) and from mild to stable (18.5% vs. 15.8%), reflecting distinct gender-specific transition patterns.

Fig. 3.

Classification plot of two waves

Logistic regressions were conducted to examine whether sex influenced class membership transitions. Specifically, females in the stable group at wave 1 had 1.41 times the odds of transitioning to the mild group at wave 2 compared to males (OR = 1.41, 95% CI (1.19, 1.68), p < 0.001). Females in the mild group at wave 1 had 1.34 times the odds of remaining in the mild group and 1.48 times the odds of transitioning to the severe group at wave 2 (OR = 1.48, 95% CI (1.24, 1.78), p < 0.001). Females in the severe group at wave 1 had 1.96 times the odds of staying in the severe group at wave 2 (OR = 1.96, 95% CI (1.22, 3.15), p = 0.006) (Table 5).

Table 5.

Logistic regression coefficients and odds ratio for LTA model examining the effect of sex on mood instability symptoms across waves

	T1–T2	Coefficient β	SE	z	p-value	Odds ratio (95% CI)
Sex	Effect for “stable” of Time 1
	Mild at Time 2	0.35 (0.17, 0.57)	0.09	− 3.94	< 0.001	1.41 (1.19, 1.67)
	Severe at Time 2	0.44 (− 0.70, 1.57)	0.58	− 0.75	0.452	1.55 (0.50, 4.79)
	Effect for “mild” of Time 1
	Mild at Time 2	0.30 (0.22, 0.37)	0.04	7.84	< 0.001	1.34 (1.25, 1.45)
	Severe at Time 2	0.39 (0.21, 0.57)	0.09	− 1.10	< 0.001	1.48 (1.24, 1.78)
	Effect for “Severe” of Time 1
	Mild at Time 2	0.44 (− 0.04, 0.92)	0.25	1.80	0.072	1.55 (0.96, 2.51)
	Severe at Time 2	0.67 (0.20, 1.15)	0.24	2.77	0.006	1.96 (1.22, 3.15)

The stable class at Time 2 is not present in the table because it is used as comparison group for the logistic regression. SE standard error, Z-value of the coefficient β divided by the standard error

Sensitivity analyses

To assess the robustness of our findings, we conducted comprehensive sensitivity analyses: (1) stratified analyses examining associations between mood instability symptoms and adverse health outcomes (Additional file 1: Tables S16, S17), (2) calculation of population attributable fractions (PAFs) to estimate the proportion of adverse outcomes attributable to mood instability symptoms (Additional file 1: Table S18), (3) exclusion of outcomes occurring within the first 2 years of follow-up to address potential reverse causation (Additional file 1: Table S19), (4) alternative operationalization of mood instability symptoms using a composite score based on 13 mental symptom items (Additional file 1: Table S20), (5) competing risk analyses accounting for neurodegenerative disease mortality (Additional file 1: Table S21), (6) multiple imputation of missing covariate data using multiple imputation by chained equations (Additional file 1: Tables S22, S23, Fig. S2), and (7) longitudinal analyses testing whether baseline biological age predicted mood instability symptoms transitions during follow-up (T1–T2, 4 + years) (Additional file 1: Table S24). These analyses consistently demonstrated the stability of our primary findings.

Discussion

In this cohort study of 185,818 adults from the UKB, we found that severe mood instability symptoms at baseline were associated with an elevated risk of all-cause dementia, Parkinson’s disease, and all-cause mortality. The results demonstrated that severe mood instability symptoms increased the risk of incidents of late-onset dementia and late-onset Parkinson’s disease and increased the risk of incidents of all-cause dementia. Both young-onset and late-onset dementia are significantly associated with accelerated aging. Furthermore, we demonstrated that accelerated aging partially mediated the associations of severe mood instability symptoms and neurodegenerative diseases, as well as all-cause mortality. Notably, we observed that females demonstrated a higher likelihood of transitioning to more severe groups compared to males during the follow-up, which contributed valuable insights to the formulation of public health interventions for neurodegenerative diseases.

Severe mood instability symptoms manifested across several dimensions, and participants with severe mood instability symptoms were more likely to experience frequent negative emotions (“feelings of miserable,” “fed up,” “lonely,” and “isolated”), shame, ruminating and worrying feelings (“tense”/ “highly strung,” “nervous feelings/suffering from nervousness,” “guilty,” “worrying too long after embarrassment,” “worried feelings,” and “sensitivity”), extreme emotion (“mood swings,” “irritability,” “risk behavior”). Among these symptoms, the severe group was particularly characterized by notably high probabilities of experiencing mood swings, feelings of being fed up, and feeling worried after embarrassment. These findings align with previous research that defines emotional lability as frequent, abrupt, self-critical, and unpredictable changes in irritability, arousal/activation, and anxiety/depression [16, 52].

Several studies have utilized Mendelian randomization techniques to investigate the potential causal relationship between mood swings and other adverse outcomes [9, 53]. While direct evidence linking mood instability to dementia risk remains limited, this study supports existing research on the connection between emotional factors and dementia. Earlier observational studies have also highlighted associations between mental health disorders and neurodegenerative conditions [22, 54, 55]. Moreover, repetitive negative thinking, which consists of worry (negative thoughts regarding future issues) and rumination (negative thoughts regarding past issues), has been acknowledged as a psychological risk factor for dementia [56]. In addition, Chapman et al. [57] and Wang et al. [58] discovered that increased neuroticism correlated with poorer average cognitive performance and a more rapid decline during follow-up.

Our findings indicated that severe mood instability symptoms increased 47% risk of PD, specifically in late-onset Parkinson’s disease. To our knowledge, there is no longitudinal evidence on whether individuals who report mood instability are at greater risk of developing PD. But studies have shown that more depressive symptoms and neuroticism were independent associations with incident Parkinsonism [59, 60]. PD is progressive; current studies have focused on feelings of shame, embarrassment, and nervousness in people with PD [61]; and Few have investigated whether these personality traits contribute to the development of PD. The result in this study showed that by adjusting severe mood instability symptoms to stable levels, the incidence of neurodegenerative diseases and death could potentially decrease by 0.29% to 5.91%. We provided new evidence of the link between mood instability symptoms and neurodegenerative diseases. However, reverse causality may be a concern, as the disease could already have progressed due to the often-delayed diagnosis of neurodegenerative diseases. Although we excluded patients with neurodegenerative diseases and implemented sensitivity analyses, neuropsychiatric symptoms may represent early manifestations of conditions like Parkinson’s disease or cognitive decline [62, 63]. Future studies incorporating neurodegeneration biomarkers assessed alongside mood instability symptoms would provide more substantial evidence.

Notably, mood instability symptoms exerted more potent effects on Parkinson’s disease than dementia, possibly reflecting distinct neurobiological mechanisms. In PD, sustained cognitive load from excessive worry or rumination triggers glucocorticoid release [64], leading to oxidative stress [65], mitochondrial dysfunction [66], and dopaminergic neuron degeneration—all of which are critical to PD progression [67]. Consequently, mood instability symptoms directly exacerbate these processes and may exert a more pronounced effect on PD progression. In contrast, dementia progression is shaped by multifactorial interactions involving genetic predispositions (e.g., APOE ε4), vascular health issues (e.g., ischemia), amyloid pathology (e.g., plaques), tauopathy (e.g., tangles) [68], and age-related changes. Unlike PD, mood instability symptoms are not considered a central driver in dementia pathogenesis.

Despite an increasing body of evidence supporting the association between mental health and the aging process [69–71], the existing research on the physiological mechanisms that explain how mood instability symptoms affect the acceleration of biological aging is still not well understood. Despite this, earlier research still offers valuable insights. Firstly, stress induced by psychological symptoms such as anxiety, worry, paranoia, and shame can weaken social support [72]. Psychological stress plays a key role in linking psychopathology to age-related diseases by elevating levels of inflammatory and stress hormones [73]. Immune function is changed by psychological stress through the release of hypothalamic–pituitary–adrenal (HPA) and sympathetic-adrenal-medullary (SAM) hormones, which attach to and modify immune cells [74]. Prolonged exposure to stress hormones can impair memory and cause hippocampal shrinkage and atrophy. Additionally, repeated exposure to stressful life events can overwhelm the body’s compensatory mechanisms, a phenomenon known as “toxic stress,” which accelerates aging [75, 76]. Moreover, psychological disorders have been linked to an increased production of reactive oxygen species (ROS), which can damage DNA and elevate cancer risk [77]. Blood biomarkers associated with PhenoAge residual, such as C-reactive protein (an inflammation marker), and immune-related markers reflect individuals’ inflammatory and immune status. Consequently, PhenoAge acceleration provides a preclinical perspective on the risks associated with these physiological conditions [78].

Various mechanisms supported that accelerated biological aging was implicated in the incidence of neurodegenerative diseases. Increasing levels of tumor necrosis factor, C-reactive protein, interleukin-6, and other inflammatory molecules lead to an increase in dementia risk [79]. Notably, similar to findings from previous studies, accelerated biological aging (measured by KDM) does not appear to increase PD risk in this study and may be protective [80]. The apparent protective effect could stem from the inclusion of systolic blood pressure in the KDM measures, which is negatively correlated with PD risk but positively correlated with other neurological outcomes [80, 81]. Although we did not explore the incident risks of CVD in this study, research has found that the development of various diseases, including CVD and cancer, is influenced by inflammation and impaired immune function, ultimately leading to mortality [78].

Given the evidence above, we propose that the effect of accelerated aging on neurodegenerative diseases and all-cause mortality may be through multiple pathways, including inflammation/immune pathways. Numerous studies of peripheral blood and cerebrospinal fluid from PD patients indicate that changes in inflammation markers and immune cell populations may initiate or worsen neuroinflammation. Furthermore, disruptions in adipokine expression with aging may lead to obesity, which is associated with PD and AD [82, 83]. Moreover, in contrast to PhenoAge, we did not find that KDM-BA acceleration mediated the relationship between severe mood instability symptoms and all-cause dementia. These inconsistent results may be due to the different algorithms used to generate the biological age measures. In contrast to KDM-BA, which mainly reflects chronological age, PhenoAge integrates additional mortality risk factors along with chronological age [47]. Future research should delve deeper into the complex roles of biological aging in the development of neurological outcomes.

The current study fills an essential gap in our knowledge of women’s health by providing prospective data on sex-specific rates of progression to mood symptoms. We found that compared to males, women are at a higher risk of transitioning to the severe mood instability symptoms group over time. This could be explained by the fact that the presence of gynecological conditions can affect mood stability through the physical and emotional distress they cause [84]. Additionally, the psychological stress, hormonal imbalances, and altered immune function linked to severe mood instability symptoms could potentially contribute to the onset of gynecological disorders [85]. Moreover, Civieri et al. [86] found that the most profound effect of anxiety and depression on neuro-immune stress pathways was observed in younger women, potentially contributing to their heightened relative risk of developing cardiovascular disease risk factors.

Therefore, supported by our findings, we emphasize the importance of integrating brief mood instability screening into routine healthcare for middle-aged and older adults, with a particular focus on women. The early identification of at-risk individuals, particularly those who exhibit core symptoms such as frequent mood swings or heightened sensitivity to criticism, is essential for implementing timely interventions before severe mood disorders develop. Recommended interventions include cognitive behavioral therapy [87], mindfulness-based training [88], or emotion-focused therapy [89], all of which can help older adults improve emotion regulation and may ultimately reduce the risk of neurodegenerative diseases.

The primary strength of our study is the utilization of the UK Biobank cohort, which has a substantial sample size, prospective and independent data collection, and an extensive follow-up period for the occurrence of neurodegenerative diseases. This enables us to conduct detailed subgroup and sensitivity analyses. Second, LTA findings highlighted mood status stability, minimizing recall bias and reflecting longitudinal emotional changes. There are several limitations to this study that need to be addressed. First, while posterior probabilities indicated robust classification, the three-step approach may yield biased estimates due to unaddressed classification error [90], and optimal methods for time-to-event outcomes remain under investigation [91]. Second, despite sensitivity analyses excluding dementia cases occurring within 2 years of baseline, we cannot definitively rule out reverse causation. Third, although baseline biological age showed mostly non-significant associations with mood transitions, a significant association was observed between KDM-BA and the transition from mild to stable. This supports the hypothesized mediation direction that biological age is a mediator of mood instability symptoms rather than the reverse pathway. However, due to a lack of longitudinal data for biological age, the causal directions cannot be confidently established. Furthermore, limited cases prevented examining the association between mental health transition patterns and neurodegenerative diseases. Finally, covariates were only assessed at baseline, lacking longitudinal data. Future research should also explore the effectiveness of different types of emotional interventions (e.g., psychotherapy, behavioral interventions, medication) in regulating severe mood instability symptoms in various populations, especially in women.

Conclusions

Our findings provide robust evidence that severe mood instability symptoms are associated with increased risks of dementia, PD, and mortality, with accelerated biological aging serving as a potential mediating pathway. These results underscore the importance of focusing on mood instability symptoms as part of early intervention strategies to mitigate neurodegenerative disease risk and improve long-term health outcomes.

Abbreviations

AD

Alzheimer’s disease

PD

Parkinson’s disease

LCA

Latent class analysis

LTA

Latent transition analysis

BA

Biological age

KDM

Klemera-Doubal method

PA

Physical activity

BMI

Body mass index

TDI

Townsend Deprivation Index

BH

Benjamini-Hochberg

FDR

False discovery rate

ORs

Odds ratios

HR

Hazard ratio

CI

Confidence interval

TE

Total effect

TNDE

Total natural direct effect

TNIE

Total natural indirect effect

PAFs

Population-attributable fractions

MICE

Multiple imputation by chained equations

Authors’ contributions

All authors have made a substantial intellectual contribution to the conception, design, or conduct of the study. JL conceived and designed the study. JW1 and JZ acquired the data. JW1, JZ, KL, XM, YW, JW2, and ZL analyzed and interpreted the data. SH, XL, and PC made critical revisions to the manuscript. JL supervised the study and improved the manuscript drafts. Graphical abstract was drawn by Figdraw. All authors read and approved the final manuscript.

Funding

This study is funded by the Natural Science Foundation of Ningxia Province (2022AAC03174).

Data availability

UK Biobank is an open-access resource (https://www.ukbiobank.ac.uk/). All bona fide researchers can apply to use its data for health-related research that is in the public interest (http://www.ukbiobank.ac.uk/register-apply). The analytical code has been uploaded to a GitHub repository https://github.com/sssccc-a11y/code.git.

Declarations

Ethics approval and consent to participate

The study was approved by both the National Health Service National Research Ethics Service (Ref.: 11/NW/0382) and the Institutional Review Board of Tulane University (2018–1872). All participants gave written informed consent to participate. This research conforms to the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.