Life’s essential 8 health profiles and risk of incident neuropsychiatric diseases: a nationwide prospective cohort study

Abstract

Background

Although previous studies have established associations between Life’s Essential 8 (LE8) and neuropsychiatric diseases, they predominantly operationalised the LE8 as a composite score, assuming equal weighting and potentially obscuring heterogeneity in health behaviour patterns. Therefore, this study applied latent profile analysis (LPA) to identify distinct LE8 profiles and examined their associations with the incidence of neuropsychiatric diseases.

Methods

The sample comprised 242,947 UK Biobank participants without prevalent neuropsychiatric diseases. LE8 comprises eight components: healthy diet, physical activity, nicotine exposure, sleep health, glucose, blood lipid, blood pressure, and body mass index. LPA was utilized to identify distinct health patterns. Cox proportional hazards models were used to examine the associations between LE8 profiles and disease incidence.

Results

LPA identified five distinct profiles: Low LE8 score; Elevated adiposity, blood pressure, and glucose group; Low physical activity with poor sleep group; Elevated blood pressure and dyslipidaemia group; and High LE8 score group. Over a median follow-up of 13.8 years, 11,419 cases of anxiety, 341 cases of bipolar disorder, and 40 cases of schizophrenia were identified. Using the Low LE8 score group as a reference, all four comparison profiles exhibited significant protective associations for both anxiety (Hazard Ratio [HR] range: 0.65–0.78, all p < 0.001) and bipolar disorder (HR range: 0.37–0.51, all p < 0.001), with risk reductions of up to 35% and 63%, respectively. No significant associations were observed for schizophrenia, likely due to the small number of incident cases (n= 40), and these findings should be considered exploratory.

Conclusions

Distinct LE8 profiles showed heterogeneous associations with neuropsychiatric disease risk, with healthier profiles linked to lower anxiety and bipolar disorder risk, while schizophrenia findings remained inconclusive. These results support person-centred, pattern-based approaches over composite scoring for neuropsychiatric risk stratification and hypothesis generation.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12888-026-08341-1.

Introduction

Neuropsychiatric disorders represent a group of syndromes exhibiting abnormalities in cognition, emotions and/or behaviours, encompassing conditions such as schizophrenia, anxiety disorders, bipolar disorder, and depression [1]. These conditions affect nearly 30% of all individuals and are among the largest contributors to the global burden of disease, accounting for 32% of all years lived with disability, and 13% of disability-adjusted life years [2, 3], profoundly impacting individuals’ quality of life, family functioning, and healthcare systems [1, 3]. The global cost of mental health conditions in 2010 was estimated at US$2.5 trillion, and these costs are expected to grow to $6.0 trillion by 2030 [4]. Consequently, the World Health Assembly adopted the Intersectoral Global Action Plan on Epilepsy and Other Neurological Disorders 2022–2031, which aims to reduce the stigma, impact and burden of neurological disorders [5].

Given the substantial disease burden, identifying modifiable risk factors for neuropsychiatric disorders has become a research priority. Although the aetiology involves complex genetic and environmental interactions, lifestyle and metabolic factors represent particularly promising intervention targets owing to their modifiable nature [6–10]. The American Heart Association’s (AHA) Life’s Essential 8 (LE8) provides a comprehensive framework for assessing multiple modifiable health factors through eight components: healthy diet, physical activity, nicotine avoidance, healthy sleep, healthy weight, and optimal blood lipids, glucose, and blood pressure [11]. Although the LE8 was developed for cardiovascular health assessment, it has shown promising associations with neuropsychiatric outcomes [12].

Recent studies elucidated associations between LE8 components and neuropsychiatric outcomes. Cross-sectional studies demonstrated that higher overall LE8 scores were associated with reduced anxiety risk, with 10-unit increases in health behaviour and health factor scores associated with 15% and 7% reduced anxiety risks, respectively [12]. Individual components exhibited differential associations, with Mediterranean-style dietary patterns associated with 35% reduced odds for major depression and 32% reduced odds for anxiety disorders [13], and physical activity demonstrating protective effects against depressive symptoms and bipolar disorder [14]. However, most previous studies examined LE8 components in isolation, failing to capture how these health behaviours and biomarkers synergistically interact within individuals. Latent profile analysis (LPA) offers substantial advantages in identifying naturally occurring patterns among LE8 profiles, which may confer differential neuropsychiatric disease risks. This approach successfully identified distinct psychological profiles associated with dementia risk and mental health outcomes [15, 16], demonstrating the value of person-centred approaches for understanding complex health phenotypes.

This study’s objectives were twofold: (1) to use LPA to examine whether individual LE8 health behaviours and biomarkers aggregate into distinct profiles within a large prospective cohort, and (2) to examine the associations between these profiles and incident neuropsychiatric diseases (schizophrenia, anxiety, and bipolar disorder).

Methods

Study participants

This study used data from the UK Biobank, a large prospective cohort study that recruited more than 500,000 participants aged 40–70 years between 2006 and 2010 within United Kingdom. At baseline, participants completed comprehensive touchscreen questionnaires, underwent standardised physical examinations, and provided biological specimens. The UK Biobank project received ethical approval from the Northwest Multicenter Research Ethics Committee, and informed consent was obtained from all participants before enrolment. This study was conducted under the UK Biobank project application number 482,700.

Sequential exclusion criteria were applied to define the study population. Participants with missing LE8 component data (n = 194,953), missing covariate information (n = 52,028), or prevalent schizophrenia, anxiety disorders, or bipolar disorder at baseline (n = 12,201) were excluded. Consequently, 242,947 eligible participants were included in the analysis (Supplementary Fig. 1).

Measurements

Assessment of LE8

Based on the AHA’s definition [11], the LE8 comprises eight elements, including four health behaviours (healthy diet, physical activity, nicotine exposure, and sleep health) and four health factors (body mass index [BMI], blood lipid, blood glucose, and blood pressure). Each element was assigned a score from 0 to 100, with higher scores denoting better health status. Detailed definitions of dietary intake and LE8 scores are provided in Supplementary Tables 1 and 2, respectively.

Definition of incident neuropsychiatric diseases

This study focused on neuropsychiatric outcomes, including schizophrenia, anxiety disorders, and bipolar disorder. Neuropsychiatric diseases during follow-ups were identified using specific ICD-10 codes, according to the “First occurrence fields” of health-related outcomes in the UK Biobank (Data category: 1712). These fields captured the earliest recorded instances of each condition from multiple linked healthcare sources, including hospital inpatient data, primary care records, death registries, and self-reported medical histories. Schizophrenia was identified using ICD-10 codes F20 and F25 (UK Biobank field IDs 130874–130875 and 130884–130885), anxiety using codes F40–F41 (field IDs 130904–130907), and bipolar disorder using code F31 (field IDs 130892–130893). The follow-ups were computed from the recruitment date to the earliest neuropsychiatric disease diagnosis, death, or censoring date. Participants with prevalent neuropsychiatric diseases at baseline were excluded to ensure that only incident cases were examined.

Covariates

Potential confounders were adjusted based on previous studies [17–19]. These included age at baseline (years), sex (male, female), ethnicity (white, other ethnicities), education (college or university degree, other qualifications or none), household income (<£18,000, £18,000–£30,999, £31,000–£51,999, £52,000–£100,000, >£100,000), Townsend Deprivation Index (TDI), employment status (employed/self-employed, not in paid employment), long-standing illness (yes, no), alcohol intake (moderate, non-moderate), and family history of severe depression (yes, no). TDI is a composite indicator of unemployment, car ownership, house ownership, and household overcrowding, which represents the level of deprivation, with higher values indicating higher deprivation [20]. Moderate alcohol intake was defined, in accordance with UK dietary guidelines, as daily consumption of one drink or fewer for women and two drinks or fewer for men (one drink = 8 g of alcohol); both abstainers and individuals exceeding these limits were classified as having non-moderate intake [21]. Family history of severe depression was derived from UK Biobank fields 20,107, 20,110, and 20,111 (father’s, mother’s, and siblings’ illnesses, respectively), with a positive history assigned if any first-degree relative was reported to have severe depression.

Statistical analyses

Participant characteristics are presented as mean ± standard deviation (SD) for continuous variables and as frequencies and proportions for categorical variables.

Latent profile analyses (LPA)

LPA was used to identify participant groups based on similar patterns of baseline LE8 health behaviours and biomarkers. Various model-fit statistics were used to determine the optimal profile solutions. A bootstrap likelihood ratio test (BLRT) was performed to identify the best-fitting model. This was further supplemented by an evaluation of common fit indices, including Akaike information criterion (AIC), Bayesian information criterion (BIC), adjusted BIC, and entropy values. The BLRT compares k with k-1 profile solutions, with p < 0.05 indicating that the k-profile model fits significantly better than the k-1 profile model [22]. Entropy values range between 0 and 1, with values > 0.80 indicating excellent classification accuracy, values between 0.60 and 0.80 suggesting adequate classification, and values < 0.60 indicating poor separation between profiles [23]. Lower values of AIC, BIC, and sample size-adjusted BIC indicate a better model fit, with decreasing values suggesting improved model performance as the number of profiles increases [24]. However, the optimal solution balances statistical fit with model parsimony and interpretability [15].

Survival analyses

Kaplan-Meier survival curves were constructed to visualise the cumulative incidence of neuropsychiatric diseases across the five LE8 latent profile groups. Log-rank tests were used to assess differences in survival distributions between groups [25, 26]. Cox proportional hazards regression models were employed to quantify the associations between LE8 latent profiles and incident neuropsychiatric diseases (schizophrenia, anxiety, and bipolar disorder) [27]. Follow-up time was calculated from baseline recruitment to the earliest occurrence of neuropsychiatric disease diagnosis, death, or administrative censoring. The models were adjusted for potential confounders, including age, sex, education, ethnicity, TDI, household income, employment status, long-standing illness, alcohol intake, and family history of severe depression. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated with the ‘Low LE8 score group’ as reference. The proportional hazards assumption was assessed using Schoenfeld residuals for all Cox models.

Sensitivity and subgroup analyses

Multiple supplementary analyses were performed to enhance the robustness of the findings. First, to minimise potential reverse causation bias, sensitivity analyses were performed, excluding participants diagnosed with neuropsychiatric diseases within the first 0.5, 1, 2, 3, and 5 years of follow-up. Second, missing data were addressed using multiple imputation by chained equations (“mice” package in R) for LE8 components and covariates. Third, to evaluate the incremental value of the latent profile approach, conventional LE8 total score categories (Low: 0–49, Moderate: 50–79, High: 80–100) were used as an alternative exposure in Cox regression models for comparison. Stratified analyses were conducted across demographic and socioeconomic strata, including age (< 60 vs. ≥ 60 years), sex (male vs. female), TDI (dichotomised at median), education (college or university degree vs. other qualifications or none), ethnicity (White vs. other ethnicities), employment status (employed/self-employed vs. not in paid employment), long-standing illness (yes vs. no), alcohol intake (moderate vs. non-moderate alcohol intake), and family history of severe depression (yes vs. no).

All statistical analyses were performed using the R statistical software version 4.5.0 (R Project for Statistical Computing). Statistical significance was set at p < 0.05 (two-sided test).

Results

Participant characteristics

The study population’s baseline demographic characteristics are presented in Supplementary Table 3. Among the 242,947 participants, the mean age was 55.83 ± 8.09 years, with 121,189 males (49.88%) and 121,758 females (50.12%). The cohort was predominantly white (233,745, 96.21%) with a high educational attainment (93,913 participants, 38.66% holding college or university degrees). The mean TDI of -1.54 ± 2.94 indicated a generally affluent population. Most participants were employed or self-employed (152,953, 62.96%). Approximately one-third reported long-standing illness (71,747, 29.53%), and 13.82% (33,568) had a family history of severe depression. Regarding alcohol consumption, 98,625 participants (40.60%) were classified as having moderate alcohol intake according to UK low-risk drinking guidelines. Based on UK low-risk drinking guidelines, 98,625 participants (40.60%) were classified as having moderate alcohol intake. The mean LE8 score was 66.78 ± 11.19, with substantial variation observed across individual health dimensions. Participants achieved the highest scores in sleep health (90.18 ± 17.70) and glucose control (90.23 ± 20.89), whereas the healthy diet scores were the lowest (31.92 ± 19.19). The remaining dimensions showed moderate scores: physical activity (74.57 ± 36.24), nicotine exposure (72.19 ± 30.12), BMI (70.19 ± 27.99), blood pressure (53.65 ± 26.34), and blood lipids (51.32 ± 30.89).

Latent profile identification and characteristics

The model fit statistics for the latent profile analysis are presented in Supplementary Table 4. The AIC, BIC, and adjusted BIC values decreased progressively from Profiles 1–7; the 5-profile solution provided an optimal balance between statistical fit and interpretability. This solution achieved adequate entropy (0.95), reasonable class proportions (smallest class: 9.73%), and significant BLRT results (p < 0.001). The 6-profile solution produced a negative BLRT statistic and lower entropy (0.85), indicating convergence to a suboptimal solution, further supporting the selection of the 5-profile model. The 7-profile solution, despite improved entropy (0.97), produced a small class comprising only 7.68% of participants, reflecting potential overextraction.

As presented in Fig. 1, five distinct profiles that demonstrated a comparable aggregation of LE8 characteristics were identified. Profile 1 (‘Low LE8 score group’, mean score = 55.96, n = 23,634) was characterised by the lowest mean LE8 score, with particularly poor nicotine exposure scores (0.06 ± 0.80), indicating near-universal current smoking, alongside uniformly low scores across most other health dimensions. Profile 2 (‘Elevated adiposity, blood pressure, and glucose group’, mean score = 59.94, n = 40,359) was characterised by markedly negative standardised scores for BMI, blood pressure, and glucose, consistent with elevated adiposity and metabolic dysfunction, while maintaining relatively higher nicotine exposure scores. Profile 3 (‘Low physical activity with poor sleep group’, mean score = 64.36, n = 75,118) was characterised by the lowest physical activity scores (50.06 ± 40.59) and poor sleep health (79.30 ± 20.90), while maintaining near-average scores in most metabolic indicators. Profile 4 (‘Elevated blood pressure and dyslipidaemia group’, mean score = 70.93, n = 74,509) was characterised by elevated blood pressure and dyslipidaemia as the primary compromised dimensions, with blood pressure scores (40.86 ± 15.17) and blood lipid scores (48.15 ± 29.59) being notably low, while other health behaviours remained relatively favourable). Profile 5 (‘High LE8 score group’, mean score = 80.59, n = 29,327) exhibited consistently positive standardised scores across nearly all LE8 dimensions, indicating superior performance relative to the overall sample mean. The standardised z-score distribution illustrated the distinct health behaviour and biomarker patterns across the five latent profile groups. Descriptive statistics for each profile are presented in Supplementary Table 5.

Fig. 1.

Z-score distribution of LE8 indicators across five latent profile groups. Bar plot indicates the standardized mean (z-score) for each LE8 indicator across five latent profiles. Positive and negative standardized values indicate scores above or below the overall sample mean for each dimension

Baseline characteristics across LE8 profiles

The baseline demographic characteristics differed significantly across the five latent profile groups (Table 1). The ‘Elevated adiposity, blood pressure, and glucose group’ comprised the oldest participants (mean age 59.31 years) and exhibited the highest prevalence of long-standing illness (47.69%), along with markedly low BMI (56.19), blood pressure (48.11), and glucose scores (48.83). The ‘Low LE8 score group’ demonstrated the lowest socioeconomic position (28.64% with higher education) and consistently poor LE8 scores, most notably a near-zero nicotine exposure score (0.06) indicative of pervasive smoking. The two largest subgroups—the ‘Low physical activity with poor sleep group’ (30.92%) and the ‘Elevated blood pressure and dyslipidaemia group’ (30.67%)—were characterised by domain-specific deficits in physical activity and sleep, and in blood pressure and lipids, respectively, while maintaining relatively favourable scores in other dimensions. In contrast, the ‘High LE8 score group’ comprised the youngest participants (mean age 52.45 years) and demonstrated the most favourable socioeconomic indicators alongside consistently high scores across all LE8 components. All variables differed significantly between profiles (all p < 0.001).

Table 1.

Baseline characteristics by five latent profile analysis groups

Characteristic	Total sample	Low LE8 score group	Elevated adiposity, blood pressure, and glucose group	Low physical activity with poor sleep group	Elevated blood pressure and dyslipidaemia group	High LE8 score group	Differences between profiles
N (%)	242,947 (100.00)	23,634 (9.73)	40,359 (16.61)	75,118 (30.92)	74,509 (30.67)	29,327 (12.07)	—
Age at baseline, mean ± SD	55.83 ± 8.09	54.04 ± 8.11	59.31 ± 7.08	55.10 ± 7.85	56.59 ± 8.03	52.45 ± 8.03	F = 4001.81, P < 0.001
Sex, N (%)							χ²=5038.40, P < 0.001
Male	121,189 (49.88)	13,533 (57.26)	22,145 (54.87)	36,041 (47.98)	39,972 (53.65)	9498 (32.39)
Female	121,758 (50.12)	10,101 (42.74)	18,214 (45.13)	39,077 (52.02)	34,537 (46.35)	19,829 (67.61)
Ethnicity, N (%)							χ²=2164.81, P < 0.001
White	233,745 (96.21)	22,525 (95.31)	37,368 (92.59)	72,443 (96.44)	72,959 (97.92)	28,450 (97.01)
Other ethnicity	9202 (3.79)	1109 (4.69)	2991 (7.41)	2675 (3.56)	1550 (2.08)	877 (2.99)
Education, N (%)							χ²=3206.45, P < 0.001
College or university degree	93,913 (38.66)	6768 (28.64)	13,092 (32.44)	29,837 (39.72)	29,684 (39.84)	14,532 (49.55)
Other qualifications or none	149,034 (61.34)	16,866 (71.36)	27,267 (67.56)	45,281 (60.28)	44,825 (60.16)	14,795 (50.45)
TDI, mean ± SD	-1.54 ± 2.94	-0.29 ± 3.40	-1.33 ± 3.05	-1.62 ± 2.88	-1.92 ± 2.70	-1.70 ± 2.81	F = 1501.47, P < 0.001
Household income, N (%)							χ²=7276.98, P < 0.001
<£18,000	45,497 (18.73)	6552 (27.72)	10,775 (26.7)	12,462 (16.59)	12,199 (16.37)	3509 (11.97)
£18,000 to £30,999	59,261 (24.39)	5983 (25.32)	11,471 (28.42)	16,927 (22.53)	18,798 (25.23)	6082 (20.74)
£31,000 to £51,999	66,407 (27.33)	5893 (24.93)	9996 (24.77)	21,266 (28.31)	20,949 (28.12)	8303 (28.31)
£52,000 to £100,000	55,864 (22.99)	4170 (17.64)	6496 (16.1)	19,113 (25.44)	17,692 (23.74)	8393 (28.62)
> £100,000	15,918 (6.55)	1036 (4.38)	1621 (4.02)	5350 (7.12)	4871 (6.54)	3040 (10.37)
Employment status, N (%)							χ²=6019.17, P < 0.001
Employed/self-employed	152,953 (62.96)	15,429 (65.28)	20,133 (49.88)	52,246 (69.55)	43,940 (58.97)	21,205 (72.31)
Not in paid employment	89,994 (37.04)	8205 (34.72)	20,226 (50.12)	22,872 (30.45)	30,569 (41.03)	8122 (27.69)
Long-standing illness, N (%)							χ²=9377.83, P < 0.001
Yes	71,747 (29.53)	7775 (32.90)	19,249 (47.69)	21,655 (28.83)	17,398 (23.35)	5670 (19.33)
No	171,200 (70.47)	15,859 (67.10)	21,110 (52.31)	53,463 (71.17)	57,111 (76.65)	23,657 (80.67)
Alcohol intake, N (%)							χ²=2476.69, P < 0.001
Moderate alcohol intake	98,625 (40.60)	8276 (35.02)	20,166 (49.97)	31,466 (41.89)	26,830 (36.01)	11,887 (40.53)
Non-moderate alcohol intake	144,322 (59.40)	15,358 (64.98)	20,193 (50.03)	43,652 (58.11)	47,679 (63.99)	17,440 (59.47)
Family history of severe depression							χ²=146.52, P < 0.001
Yes	33,568 (13.82)	3478 (14.72)	5274 (13.07)	10,803 (14.38)	9601 (12.89)	4412 (15.04)
No	209,379 (86.18)	20,156 (85.28)	35,085 (86.93)	64,315 (85.62)	64,908 (87.11)	24,915 (84.96)
LE8 score, mean ± SD
Healthy diet score	31.92 ± 19.19	26.62 ± 19.10	32.85 ± 19.65	31.22 ± 19.14	33.04 ± 19.13	33.90 ± 18.07	F = 647.40, P < 0.001
Physical activity score	74.57 ± 36.24	70.53 ± 38.82	71.25 ± 38.20	50.06 ± 40.59	94.30 ± 11.77	95.01 ± 10.72	F = 22842.53, P < 0.001
Nicotine exposure score	72.19 ± 30.12	0.06 ± 0.80	78.79 ± 20.01	80.33 ± 19.44	80.27 ± 19.36	79.81 ± 20.00	F = 98286.39, P < 0.001
Sleep health score	90.18 ± 17.70	87.79 ± 19.74	88.11 ± 19.59	79.30 ± 20.90	99.35 ± 2.46	99.52 ± 2.13	F = 18854.88, P < 0.001
BMI score	70.19 ± 27.99	71.01 ± 27.71	56.19 ± 30.71	69.50 ± 27.97	71.44 ± 25.92	87.37 ± 17.02	F = 5855.51, P < 0.001
Blood lipid score	51.32 ± 30.89	49.24 ± 31.29	55.38 ± 33.50	49.85 ± 30.25	48.15 ± 29.59	59.25 ± 29.71	F = 937.61, P < 0.001
Glucose score	90.23 ± 20.89	86.99 ± 22.99	48.83 ± 13.25	100 ± 0	100 ± 0	100 ± 0	F = 268437.58, P < 0.001
Blood pressure score	53.65 ± 26.34	55.47 ± 26.88	48.11 ± 24.36	54.60 ± 26.80	40.86 ± 15.17	89.89 ± 12.27	F = 27141.55, P < 0.001

Categorical variables are presented as N (%), and continuous variables are presented as mean (SD). χ² test statistics are presented for categorical variables, and F-test are presented for continuous variables

Abbreviations: SD, Standard Deviation; TDI, Townsend Deprivation Index; LE8, Life’s Essential 8

Neuropsychiatric disease incidence by LE8 profiles

During a median follow-up of 13.8 years, the incidence patterns of the three neuropsychiatric conditions varied across the five LE8 profiles (Figs. 2, 3 and 4).

Fig. 2.

Kaplan–Meier curves showing cumulative incidence of anxiety across LE8 latent profile groups

Fig. 3.

Kaplan–Meier curves showing cumulative incidence of bipolar disorder across LE8 latent profile groups

Fig. 4.

Kaplan–Meier curves showing cumulative incidence of schizophrenia across LE8 latent profile groups

Anxiety disorders

Anxiety demonstrated the highest overall incidence and the greatest between-group separation among the three conditions (Fig. 2). Cumulative incidence reached approximately 0.07 by the end of follow-up, with divergence in risk trajectories emerging within the first two years and widening progressively thereafter. The ‘Low LE8 score group’ consistently exhibited the highest cumulative incidence throughout the observation period, whereas the ‘High LE8 score group’ and the ‘Elevated blood pressure and dyslipidaemia group’ tracked closely together at the lowest risk levels. The ‘Elevated adiposity, blood pressure, and glucose group’ and the ‘Low physical activity with poor sleep group’ displayed intermediate and largely overlapping trajectories.

Bipolar disorder

Bipolar disorder exhibited the clearest risk stratification relative to event frequency among the three conditions (Fig. 3). The ‘Low LE8 score group’ showed a markedly higher cumulative incidence (approximately 0.004), with early and sustained separation from all other profiles. The remaining four profiles clustered at substantially lower risk levels, with cumulative incidence ranging from approximately 0.001 and 0.0015 by the end of follow-up.

Schizophrenia

Schizophrenia had the lowest overall incidence across all profiles, with cumulative incidence remaining below 0.0015 throughout follow-up (Fig. 4). Between-group differences were minimal, with only slight divergence observed in the later years. The small number of incident cases (n = 40) limited the ability to visual differentiate risk trajectories across profiles.

Associations between LE8 profiles and neuropsychiatric diseases

Cox proportional hazards regression analyses were conducted using the ‘Low LE8 score group’ as the reference category, and were adjusted for age, sex, ethnicity, education, household income, TDI, employment status, long-standing illness, alcohol intake, and family history of severe depression (Fig. 5).

Fig. 5.

HRs (95% CI) for schizophrenia, anxiety, and bipolar disorder according to LE8 latent profile groups. HRs were adjusted for age, sex, education, ethnicity, Townsend deprivation index, household income, employment status, long-standing illness, alcohol intake, and family history of severe depression.Abbreviations: CI, Confidence Interval; HR, Hazard Ratio; LE8, Life’s Essential 8

Anxiety disorders

During follow-up, 11,419 incident anxiety cases were identified. In the fully adjusted model, all four comparison profiles showed significant protective associations relative to the ‘Low LE8 score group’ in (all p < 0.001). The ‘High LE8 score group’ demonstrated the strongest protective association (HR 0.65, 95% CI: 0.60–0.70), followed by the ‘Elevated blood pressure and dyslipidaemia group’ (HR 0.68, 95% CI: 0.63–0.72), the ‘Elevated adiposity, blood pressure, and glucose group’ (HR 0.77, 95% CI: 0.72–0.82), and the ‘Low physical activity with poor sleep group’ (HR 0.78, 95% CI: 0.74–0.83). Overall, this graded pattern of risk reduction corresponded broadly to increasing mean LE8 scores across profiles.

Bipolar disorder

A total of 341 incident bipolar disorder cases were observed. All four comparison profiles reflected significant protective associations (all p < 0.001). The ‘Elevated blood pressure and dyslipidaemia group’ showed the strongest protective effect (HR 0.37, 95% CI: 0.27–0.52), followed by the ‘Low physical activity with poor sleep group’ (HR 0.41, 95% CI: 0.30–0.55), the ‘High LE8 score group’ (HR 0.41, 95% CI: 0.27–0.62), and the ‘Elevated adiposity, blood pressure, and glucose group’ (HR 0.51, 95% CI: 0.36–0.71). Notably, the effect sizes for bipolar disorder were larger than those observed for anxiety, with substantial risk reductions ranging from 49% to 63%.

Schizophrenia

A total of 40 incident schizophrenia cases were identified. No significant associations were observed between LE8 profiles and schizophrenia, with the exception of the ‘Elevated blood pressure and dyslipidaemia group’ (HR 0.18, 95% CI: 0.05–0.68, p = 0.011). All other profiles yielded non-significant HRs with wide confidence intervals, indicating the severely limited statistical power due to the small number of incident cases.

Sensitivity and subgroup analyses

Multiple sensitivity analyses were conducted to evaluate the robustness of the primary findings. First, to minimise potential reverse causation, participants who developed neuropsychiatric conditions within the first 0.5, 1, 2, 3, and 5 years of follow-up were sequentially excluded; the magnitude and direction of associations remained consistent across all exclusion thresholds (Supplementary Table 6). Second, multiple imputation was used to include participants excluded from the primary analysis due to missing LE8 indicators and covariate data, resulting in a substantially larger analytic sample; these findings were consistent with the complete-case analysis (Supplementary Table 7). Third, when conventional LE8 total score categories (Low, Moderate, High) were used as an alternative exposure, higher scores were associated with lower risk of anxiety and bipolar disorder in a monotonic pattern, but this approach yielded only a single gradient of risk reduction and could not distinguish the profile-specific patterns identified by LPA (Supplementary Table 8). The proportional hazards assumption was satisfied for bipolar disorder (global p = 0.419) and schizophrenia (global p = 0.637). For anxiety disorders, the global test reached statistical significance (p < 0.001) due to the large sample size (Supplementary Table 9); however, visual inspection of Schoenfeld residual plots confirmed no substantial time-dependent trend for the primary exposure variable (Supplementary Fig. 2).

Subgroup analyses stratified by sex, age, ethnicity, socioeconomic status, educational attainment, employment status, long-standing illness, alcohol intake, and family history of severe depression produced broadly consistent associations between LE8 profiles and incident neuropsychiatric diseases (Supplementary Tables 10–18). Across all strata, healthier LE8 profiles were generally linked to lower neuropsychiatric disease risk compared with the ‘Low LE8 score group’. These protective associations were particularly stable for anxiety, with significant risk reductions observed across virtually all demographic and health-related subgroups.

Discussion

This study applied LPA to examine LE8 patterns and their associations with neuropsychiatric diseases in a large prospective cohort from the UK Biobank. This person-centred approach identified five distinct LE8 profiles, suggesting that LE8 components cluster into data-driven statistical groupings with differential neuropsychiatric disease risk profiles. Fully adjusted Cox regression models revealed observational protective associations for anxiety disorders and bipolar disorder across multiple LE8 profiles, whereas no significant associations were observed for schizophrenia owing to the small number of incident cases. Accordingly, the schizophrenia findings should be regarded as exploratory.

This study identified five distinct LE8 profiles, representing a significant methodological and conceptual advancement over previous LE8 studies, which predominantly employed variable-centred approaches to examine individual LE8 components in isolation or calculating simple composite scores that assume equal weighting and a uniform health continuum. Traditional studies typically categorised participants into broad ‘high’, ‘moderate’, and ‘low’ LE8 groups based on total scores, failing to capture the complex heterogeneity of health patterns within populations [17]. The latent profiles identified here are data-driven statistical groupings derived from the analytic sample. Nonetheless, this person-centred approach revealed that LE8 components aggregate in distinct patterns, which would have been missed by conventional scoring methods [28]. When LE8 total scores were categorised into conventional Low, Moderate, and High groups, higher scores were associated with lower risk of anxiety and bipolar disorder in a monotonic gradient. However, this approach yielded only a single dimension of risk reduction and could not distinguish the profile-specific patterns identified by LPA. For example, the ‘Elevated adiposity, blood pressure, and glucose group’ was characterised by elevated metabolic dysfunction alongside relatively preserved lifestyle behaviours, whereas the ‘Low physical activity with poor sleep group’ exhibited the reverse pattern [29]. Such distinctions may carry important implications for targeted prevention strategies and would be entirely missed by summary-score approaches.

The observed differences in neuropsychiatric disease risks across LE8 profiles are biologically plausible and broadly consistent with established pathways linking cardiovascular health to brain function. The ‘Low LE8 score group’, which exhibited the highest neuropsychiatric disease burden, may reflect the convergence of multiple adverse health exposures. Chronic nicotine exposure pharmacologically has been shown to increase Hypothalamic-Pituitary-Adrenal (HPA) axis activity and sympathetic arousal, which may paradoxically heighten long-term vulnerability to anxiety and mood instability despite subjective anxiolytic effects [30]. Physical inactivity has been associated with reduced brain-derived neurotrophic factor (BDNF)-mediated hippocampal neuroplasticity [31], while poor dietary quality has been linked to elevated systemic inflammatory markers and oxidative stress [32]. These co-occurring behavioural and metabolic deficits may collectively contribute to a state of heightened neuropsychiatric vulnerability, which is consistent with the elevated risk observed in this profile for both anxiety and bipolar disorder.

In contrast, profiles characterised mainly by metabolic dysfunction, such as the ‘Elevated adiposity, blood pressure, and glucose group’, showed intermediate neuropsychiatric risk despite relatively preserved lifestyle behaviours. This pattern suggests that relatively preserved lifestyle behaviours may partially offset the neuropsychiatric consequences of metabolic dysfunction [33]. The ‘Low physical activity with poor sleep group’, the largest subgroup comprising approximately 31% of participants, may involve a distinct but overlapping pathway. Chronic sleep insufficiency has been associated with impaired prefrontal cortical regulation of amygdalar reactivity [34], while physical inactivity may limit exercise-mediated HPA axis downregulation and anti-inflammatory effects [33]. Despite relatively favourable metabolic indicators, the risk reductions identified in this group, relative to the ‘Low LE8 score group’, are consistent with the possibility that the behavioural pathway operates through mechanisms that are partially independent of metabolic dysfunction.

The ‘Elevated blood pressure and dyslipidaemia group’ and the ‘High LE8 score group’ exhibited the most favourable neuropsychiatric outcomes. In the former, preserved physical activity, adequate sleep, and nicotine abstinence may promote circadian rhythm stability, which is relevant because circadian disruption is a well-established precipitant for bipolar episodes [35]. Previous research has also implicated mitochondrial dysfunction in bipolar disorder pathology [36], and healthy lifestyle behaviours may help preserve mitochondrial function, although this remains speculative in the present context. The ‘High LE8 score group’ showed the strongest protective associations, consistent with the possibility that optimal scores across multiple health dimensions may confer additive or synergistic benefits for neuropsychiatric resilience [37, 38]. The graded dose-response pattern observed across progressively healthier profiles supports this interpretation, although the underlying mechanisms cannot be determined from the present observational data.

The effect sizes for bipolar disorder were notably larger than those for anxiety, but rest on only 341 incident cases. The consistent direction across all four comparison profiles lends some confidence, yet the precision of individual point estimates is limited, and these findings require replication in cohorts with greater event counts.

The absence of significant associations for schizophrenia should not be interpreted as evidence that LE8 profiles are unrelated to schizophrenia risk, but rather as a reflection of several methodological and clinical constraints. With only 40 incident cases, these analyses were severely underpowered and should be regarded as exploratory. First, the UK Biobank cohort largely excludes the typical age of onset for schizophrenia, and the cases captured predominantly represent late-onset presentations. Compared with early-onset schizophrenia (onset before age 18), late-onset cases are characterised by more prominent affective symptoms, predominantly executive cognitive impairment, and better premorbid functioning, and are thought to involve neurodegenerative processes rather than the neurodevelopmental abnormalities that predominate in early-onset cases [39–41]. Consistent with this, 90% of the 40 incident schizophrenia cases in the present cohort occurred after age 50, with 55% occurring after age 60, confirming that the observed cases almost exclusively represent late-onset presentations, and our findings should therefore be interpreted as specific to this phenotype rather than schizophrenia as a whole. Second, a ‘therapeutic paradox’ may obscure genuine associations: atypical antipsychotics induce metabolic side effects, including weight gain, sedation, and glycaemic elevation, that adversely affect LE8 indicators, thereby introducing the possibility of generating reverse causality in observational data [42]. Finally, evidence from identical twin studies indicates that schizophrenia is a complex trait shaped by both genetic and environmental influences, with substantial variance attributable to non-genetic factors such as trauma exposure, psychosocial stress, and lifestyle behaviours [43].

From a clinical perspective, the profile-specific risk patterns observed in this study generate hypotheses regarding LE8-based risk stratification for neuropsychiatric diseases that merit investigation in future interventional studies. For individuals in the ‘Low LE8 score group’, integrated interventions combining smoking cessation with comprehensive lifestyle modification could be prioritised in future trials. For the ‘Low physical activity with poor sleep group’, the largest subgroup, structured exercise programmes combined with sleep hygiene interventions represent a plausible intervention target warranting further evaluation. These observational findings provide a basis for designing profile-targeted interventions, although their clinical utility requires prospective validation before implementation in practice.

Several limitations warrant consideration. First, the UK Biobank cohort’s demographic characteristics (96.21% white, generally affluent) may limit generalisability to more diverse populations. Second, LE8 components were assessed only at baseline, and profiles may change over follow-up; this limits inference about how shifts in lifestyle behaviours might affect long-term neuropsychiatric risk. Third, the number of incident schizophrenia cases was extremely small (n = 40), severely limiting statistical power to detect associations between LE8 profiles and schizophrenia. These findings should accordingly be considered exploratory. Fourth, unmeasured confounders including genetic predisposition, social support networks, and environmental stressors cannot be entirely excluded despite comprehensive covariate adjustment. Fifth, although the prospective design offers stronger evidence than cross-sectional approaches, causal inference remains limited, and the mechanistic pathways discussed above should be understood as interpretive frameworks rather than conclusions drawn from the present data. Finally, neuropsychiatric disease identification relied on clinical diagnoses recorded in healthcare databases, which may potentially under-ascertain milder cases that do not receive clinical attention.

In conclusion, this prospective study demonstrated that LE8 components aggregate into distinct data-driven profiles with differential neuropsychiatric disease risk associations. Healthier LE8 profiles were associated with lower risk of both anxiety disorders and bipolar disorder, whereas schizophrenia findings remained inconclusive owing to insufficient statistical power. These findings support the utility of person-centred, pattern-based approaches over composite scoring methods, and generate hypotheses for future research aimed at evaluating whether profile-targeted interventions can reduce the burden of neuropsychiatric diseases.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

LE8

Life’s Essential 8

LPA

Latent profile analysis

HR

Hazard ratio

CI

Confidence interval

TDI

Townsend deprivation index

SD

Standard deviation

AIC

Akaike Information Criterion

BIC

Bayesian Information Criterion

BLRT

Bootstrap Likelihood Ratio Test

AHA

American Heart Association

Author contributions

Huihui Huangfu: Conceptualization, Formal analysis, Visualization, Writing – original draft. Haiyi Jia: Conceptualization, Formal analysis, Visualization, Writing – original draft. Chen Chen: Data curation, Formal analysis. Yunyi Xu: Data curation, Formal analysis. Xinke Zhou: Data curation, Formal analysis. Yilin Tang: Supervision, Validation, Writing – review and editing. Xiang Gao: Resources, Data curation, Funding acquisition, Supervision, Writing – review and editing. All authors contributed to manuscript review and editing, and approved the submitted version.

Funding

This work was supported by the Fundamental Research Funds for the Central Universities (grant number YG2025QNB01), the National Natural Science Foundation of China (grant number 72004032), Startup Fund for Young Faculty at Shanghai Jiao Tong University (grant number KJ3-0221-22-6333), Jiangsu Education Department Foundation (grant number 2025SJYB0231), China Association of Higher Education 2025 Annual Higher Education Scientific Research Planning Project (grant number 25BJ0312) and Interdisciplinary Programs of Chinese Medicine and Aging Science of Nanjing University of Chinese Medicine (grant number 2025YLFWYGL009).

Data availability

Data from the UK Biobank are publicly available and can be accessed at (https://www.ukbiobank.ac.uk/.

Declarations

Ethical approval and consent to participate

The study was approved by the National Health Service (NHS) National Research Ethics Service (REC reference: 11/NW/03820). The present study was conducted under UK Biobank Application Number (482700). All procedures performed in studies involving human participants were in accordance with the Declaration of Helsinki. All participants provided written informed consent to participate in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.