Associations of physical frailty, depression and their interaction with incident all-cause dementia among older adults: evidence from three prospective cohorts Authors: Yihong Ding, Mingrui Duan, Jie Shen, Lisha Xu, Yuehui Ma, Di He, Yimin Zhu Published: 2025 ABSTRACT AbstractBackgroundPhysical frailty and depression may share common pathophysiological pathways associated with dementia and thus interact with each other. However, previous studies have primarily focused on the individual impact of these factors on dementia.AimsTo examine the joint effect and interaction of physical frailty and depression on the risk of all-cause dementia.MethodsWe conducted prospective analyses among participants aged ≥60 years from three cohorts: the UK Biobank (UKB), the English Longitudinal Study of Ageing (ELSA) and the Health and Retirement Study (HRS). Physical frailty was assessed using modified versions of the Fried frailty phenotype. Depression was evaluated through mental health questionnaires or combined with hospital admission records. The primary outcome was incident all-cause dementia, identified via active follow-up and passive surveillance. Cox proportional hazards models were used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs).ResultsA total of 220 947 participants (mean age: 64.5 years; 53.3% female) were included. Over 2 832 696 person-years of follow-up, 9088 participants (7605 in UKB, 1207 in HRS and 276 in ELSA) developed incident all-cause dementia. Compared with robust individuals, frail participants faced a 155% increased risk of dementia (pooled HR: 2.55, 95% CI 2.36 to 2.76; I2=72.3%). Depression conferred a 1.59-fold excess risk for dementia (pooled HR: 1.59, 95% CI 1.50 to 1.69; I2=56.8%). Adding physical frailty and depression to a traditional dementia risk model significantly improved prediction accuracy (all p-Δarea under the curve<0.05). Jointly, participants with both physical frailty and depression exhibited the highest dementia risk (pooled HR: 3.23, 95% CI 2.86 to 3.65; I2=41.6%) compared with those without physical frailty and depression. Moreover, a significant additive interaction between physical frailty and depression was observed (pooled relative excess risk due to interaction: 0.38, 95% CI 0.13 to 0.63), with 17.1% (95% CI 6.0% to 28.3%) of dementia risk attributed to their interactive effects.ConclusionsIndividuals with both physical frailty and depression had the highest risk of dementia. More importantly, these two factors interact in an additive manner, further amplifying dementia risk.Keywords: Depression, Cognition, Cohort Studies, Epidemiologic Studies, Psychiatry WHAT IS ALREADY KNOWN ON THIS TOPIC WHAT THIS STUDY ADDS Individuals with both physical frailty and depression had the highest risk of dementia. Notably, the combined effect of physical frailty and depression on dementia risk exceeded the sum of their individual effects, with 17.1% of the dementia risk attributable to their interaction. Individuals with both physical frailty and depression had the highest risk of dementia. Notably, the combined effect of physical frailty and depression on dementia risk exceeded the sum of their individual effects, with 17.1% of the dementia risk attributable to their interaction. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY Our findings underscore the need to integrate assessments of physical frailty and depression into routine clinical screenings, enabling the early identification of high-risk individuals and the implementation of targeted interventions that integrate physical and mental health strategies. Our findings underscore the need to integrate assessments of physical frailty and depression into routine clinical screenings, enabling the early identification of high-risk individuals and the implementation of targeted interventions that integrate physical and mental health strategies. Introduction All-cause dementia, a major type of neurodegenerative disease, imposes substantial individual and societal burdens.1 2 Globally, approximately 57 million individuals are living with dementia, and the number is expected to triple by 2050.2 Given the irreversible progression of dementia and the absence of widely available and effective treatments, identifying modifiable risk factors and early predictors is critical for crafting effective prevention measures.1 3 As well-known risk factors for dementia,4,8 both physical frailty and depression are highly prevalent among older adults and often coexist. Frailty is a multidimensional syndrome characterised by multisystem functional decline and heightened vulnerability to stressors.9 10 Depression, marked by persistent low mood, anxiety, cognitive impairment and somatic symptoms, accounts for the highest global burden of disease in terms of disability-adjusted life years (DALYs) among mental disorders.11 Notably, Buigues et al reported that 16%–35% of frail individuals also experience coexisting depression, with the prevalence of depression among frail older adults reaching as high as 47%.12 In addition, physical frailty and depression may share the same pathophysiological pathways associated with dementia, such as chronic inflammation, oxidative stress and mitochondrial dysfunction, and thus constitute interactions.13 14 Therefore, physical frailty and depression should not be considered in isolation. However, the possible joint and interactive effects of physical frailty and depression on the risk of dementia remain largely unknown. An in-depth understanding of these relationships could advance our understanding of dementia’s underlying mechanisms and inform targeted prevention and intervention strategies. Therefore, this study aimed to evaluate the individual and joint associations of physical frailty and depression with incident all-cause dementia across three large-scale cohorts: the UK Biobank (UKB), the English Longitudinal Study of Ageing (ELSA) and the Health and Retirement Study (HRS). Moreover, we investigated the interactions between physical frailty and depression in relation to dementia risk by applying multiplicative and additive interaction measures. Methods Study participants This study included individual-level data from three prospective cohorts. The UKB is derived from a population-based cohort study that recruited over 500 000 participants aged 37–73 years across the UK between 2006 and 2010.15 Physical frailty and depression assessments were conducted by trained research nurses at recruitment.16 The ELSA and HRS are two large, ongoing, nationally representative, longitudinal surveys of community-dwelling adults aged ≥50 years in the UK and the US, respectively.17 18 In ELSA, participants began undergoing physical frailty assessments in wave 2 (2004) during nurse visits. In HRS, physical frailty assessments were introduced starting from waves 8 (2006) and 9 (2008). A randomly selected half of the participants underwent frailty measurements during the in-home survey visit in wave 8, while the other half were assessed in wave 9. These assessments in ELSA and HRS have been repeated every 4 years. The UKB received ethical approval from the National Health Service North West Multicentre Research Ethics Committee, the ELSA obtained ethical approval from the London Multicentre Research Ethics Committee, and the HRS was approved by the University of Michigan Institutional Review Board. All participants provided written informed consent before data collection. The study time frames and follow-up periods for each cohort are shown in online supplemental figure S1. In this study, baseline was defined as the time point of each participant’s first assessment of physical frailty and depression. For UKB, the baseline corresponded to the recruitment wave (2006–2010). For ELSA and HRS participants, the baseline was determined as their initial assessment wave during the study period: wave 2 (2004), wave 4 (2008) or wave 6 (2012) for ELSA and waves 8–11 (2006–2012) for HRS. These waves were selected to maximise sample size and ensure longer follow-up periods, as well as align baseline periods across cohorts as closely as possible within similar timeframes. For participants with multiple assessments, the earliest available wave was designated as baseline. The follow-up periods ended in 2022 for UKB, 2020 for HRS and 2018 for ELSA. We included participants who were aged ≥60 years with valid measurement data for physical frailty and depression. We excluded individuals with pre-existing stroke or dementia at baseline, those who developed dementia within the first 2 years of follow-up, and those without dementia diagnosis data during follow-up (figure 1). The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.19 Participant inclusion flowchart. ELSA, English Longitudinal Study of Ageing; HRS, Health and Retirement Study; UKB, UK Biobank. Assessment of physical frailty Physical frailty was assessed using the five-component frailty phenotype proposed by Fried and colleagues, comprising unintentional weight loss, self-reported exhaustion, low physical activity, slowness and weakness.20 Specific components were adapted to the protocols specific to each cohort21 22 and the methodologies for evaluating these components across cohorts are outlined in online supplemental table 1. Participants were classified as “frail” if three or more of the components were present, “pre-frail” if one or two of the components were present, and “robust” if none of the components were present. Assessment of depression Depression was defined using study-specific criteria.23,27 In UKB, depression was assessed through a multi-source approach: linked hospital admission records and the two-item Patient Health Questionnaire (PHQ-2), where a score of ≥3 (out of 6) indicated depression.23 In this way, the UKB provides a more detailed conceptualisation of depression, ranging from mild depressive symptoms to clinical depression.24 25 For ELSA and HRS, depression was evaluated using the eight-item Center for Epidemiological Studies Depression Scale (CES-D), with a score of ≥3 (out of 8) considered positive for depression.26 27 Ascertainment of dementia The primary outcome was all-cause dementia, identified via active follow-up and passive surveillance. Briefly, the UKB ascertained dementia using linkage to healthcare systems.28 In ELSA, dementia diagnosis was ascertained at each study wave using the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE).29 30 In HRS, dementia was identified according to the Langa–Weir criteria using data from objectively measured cognitive function and proxy interviews.31 Details are described in the online supplemental appendix. Covariates We included multiple covariates to adjust for confounding.4,8 Sociodemographic characteristics included age, sex (male or female), race and ethnicity (White or non-White) and education level (below college or college and above). Lifestyle factors included smoking status (current, former or never smokers) and drinking status (ever or never drinkers). Health-related factors included body mass index (BMI), type 2 diabetes (yes or no), hypertension (yes or no), hypercholesterolaemia (yes or no, not available in HRS) and heart diseases (yes or no). The assessments of these health conditions are shown in the online supplemental appendix. Statistical analysis For descriptive statistics, continuous variables were presented as mean (standard deviation (SD)) and categorical variables as number (percentage). Missing data on covariates were imputed using multiple imputations with a chained equation,32 following established protocols as previously described.33 The rates of missing data for covariates are summarised in online supplemental table 2. Cox proportional hazards models were used to evaluate the individual associations of physical frailty and depression with incident all-cause dementia. Survival time for each participant was calculated from the baseline of each cohort to the date of incident dementia, loss to follow-up or the end of follow-up, whichever occurred first. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated, adjusting for age, sex, race, education, smoking, drinking, BMI, type 2 diabetes, hypertension, hypercholesterolaemia (not included in HRS) and heart disease. The proportional hazards assumption was tested and verified using Schoenfeld residual methods, and the results showed no violation in the analyses (all p>0.05). We analysed data from each cohort separately and pooled their estimates using random-effects models.34 The incremental value of physical frailty and depression in predicting incident dementia was tested by adding these two factors to a clinical risk model (including the covariates mentioned earlier).35 36 Predictive accuracy was evaluated using a 10-year time-dependent area under the curve (AUC).37 38 We also assessed the relative importance of physical frailty, depression and other traditional risk factors in predicting all-cause dementia by calculating the R2 values derived from Cox models.39 In addition, participants were classified by joint categories of physical frailty and depression, with robust individuals without depression serving as the reference group. With the aforementioned multivariable model, we calculated the HRs with 95% CIs for the levels of the combined variables. Interaction terms were also introduced to evaluate whether the combined effects of physical frailty and depression exceeded the cumulative effects of the two on the additive and multiplicative scales. Multiplicative interaction was assessed by adding a multiplicative interaction term to the original Cox models. To assess additive interaction between physical frailty and depression on the risk of dementia, we considered physical frailty and depression as two binary variables (robust or pre-frail/frail for physical frailty and yes or no for depression). The relative excess risk due to interaction (RERI) and the attributable proportion (AP) were calculated as the indicators of additive interaction, as described in previous studies.40 41 Stratified analyses were performed by sex (male or female) and interactions were tested using likelihood ratio tests. We further conducted several sensitivity analyses to test the robustness of the primary findings. First, physical frailty and depression were mutually adjusted in the same model to assess their independent associations with dementia. Second, we used different definitions to assess depression across the three cohorts. Specifically, in UKB, we restricted depression diagnoses from hospitalisation records to those that occurred within 2 years prior to baseline, as these are more likely to represent current or recent depressive symptoms. In HRS and ELSA, we removed two items from the CES-D scale that overlapped with frailty measurements to avoid inflated associations resulting from the repeated measurement of these items. For the modified six-item CES-D, a score of ≥2 was used to detect clinical depression, as recommended in previous studies.26 27 Third, considering the competing risk between mortality and incident dementia, we repeated the main analyses using the competing risk model.33 Fourth, we excluded participants who developed dementia in the first 5 years of follow-up to further reduce reverse causation. Fifth, we excluded participants with heart disease at the study baseline. Finally, we repeated the main analyses by excluding participants with missing data on covariates to explore the influence of multiple imputations on these variables.33 Statistical analyses were performed using R 4.3.0. We reported two-sided p-values throughout, and p<0.05 indicated statistical significance. Results Baseline characteristics of the study population According to the inclusion and exclusion criteria, a total of 220 947 participants were included. The mean (SD) age at baseline was 64.5 (3.7) years, and 53.3% of participants were female (table 1). Specifically, the study comprised 206 969 participants from UKB, with a mean (SD) age of 64.1 (2.9) years and 53.1% female; 7460 participants from HRS, with a mean (SD) age of 72.8 (6.4) years and 58.1% female; and 6518 participants from ELSA, with a mean (SD) age of 67.9 (7.1) years and 55.0% female. Across all cohorts, frail individuals were older, more likely to be female, with a lower prevalence of high education level, and had a higher level of BMI and a higher prevalence of chronic conditions, as compared with robust individuals (online supplemental table 3). Baseline characteristics stratified by depression status are presented in online supplemental table 4. Table 1Baseline characteristics of study participantsVariableOverallUKBHRSELSANumber, n220 947206 96974606518Age, years, mean (SD)64.5 (3.7)64.1 (2.9)72.8 (6.4)67.9 (7.1)Female, n (%)117 770 (53.3)109 852 (53.1)4335 (58.1)3583 (55.0)White ethnicity, n (%)213 248 (96.5)200 377 (96.8)6460 (86.6)6411 (98.4)Education level, n (%) Below college160 916 (72.8)152 730 (73.8)4134 (55.4)4052 (62.2) College and above60 031 (27.2)54 239 (26.2)3326 (44.6)2466 (37.8)Smoking status, n (%) Never smokers22 500 (10.2)16 791 (8.1)3264 (43.8)2445 (37.5) Former smokers92 976 (42.1)86 141 (41.6)3518 (47.2)3317 (50.9) Current smokers105 471 (47.7)104 037 (50.3)678 (9.1)756 (11.6)Drinking status, n (%) Never drinkers13 825 (6.3)9677 (4.7)3484 (46.7)664 (10.2) Ever drinkers207 122 (93.7)197 292 (95.3)3976 (53.3)5854 (89.8)BMI, kg/m2, mean (SD)27.5 (4.5)27.5 (4.5)27.7 (5.1)28.0 (4.9)Type 2 diabetes, n (%)16 218 (7.3)14 220 (6.9)1459 (19.6)539 (8.3)Hypertension, n (%)147 361 (66.7)140 254 (67.8)4457 (59.7)2650 (40.7)Hypercholesterolaemia, n (%)NA118 675 (57.3)NA1751 (26.9)Heart disease, n (%)20 998 (9.5)18 089 (8.7)1848 (24.8)1061 (16.3)BMI, body mass index; ELSA, English Longitudinal Study of Ageing; HRS, Health and Retirement Study; NA, not available; SD, standard deviation; UKB, UK Biobank. Baseline characteristics of study participants BMI, body mass index; ELSA, English Longitudinal Study of Ageing; HRS, Health and Retirement Study; NA, not available; SD, standard deviation; UKB, UK Biobank. Individual associations of physical frailty and depression with dementia Over 2 832 696 person-years of follow-up, 9088 participants (7605 of 2 711 802 person-years in UKB, 1207 of 65 403 person-years in HRS, and 276 of 55 491 person-years in ELSA) developed incident all-cause dementia. As shown in online supplemental table 5, frail participants had a significantly higher risk of incident all-cause dementia than robust participants (pooled HR: 2.55, 95% CI 2.36 to 2.76; I2=72.3%). The significantly elevated risk of dementia was also found in pre-frail participants as compared with robust participants (pooled HR: 1.45, 95% CI 1.39 to 1.52; I2=80.1%). In addition, individuals with depression had a 59% increased risk of dementia (pooled HR: 1.59, 95% CI 1.50 to 1.69; I2=56.8%) when compared with those without depression. Predictive value and relative importance of physical frailty and depression in dementia We tested the incremental predictive value of adding physical frailty and depression to a traditional clinical risk model that included baseline sociodemographic, lifestyle and clinical risk indicators. After the addition of physical frailty and depression, the AUC of the clinical risk model increased from 0.684 (95% CI 0.641 to 0.728) to 0.701 (95% CI 0.646 to 0.733) in UKB (ΔAUC: 0.017, p=0.038), from 0.783 (95% CI 0.764 to 0.802) to 0.800 (95% CI 0.780 to 0.816) in HRS (ΔAUC: 0.017, p<0.001), and from 0.802 (95% CI 0.763 to 0.841) to 0.841 (95% CI 0.796 to 0.863) in ELSA (ΔAUC: 0.039, p<0.001) at the 10-year follow-up (figure 2). Performance of multivariable models with physical frailty and depression for prediction of 10-year risk of all-cause dementia. Base model included sociodemographic (age, sex, race and education level), lifestyle factors (smoking, drinking and body mass index) and clinical risk factors (type 2 diabetes, hypertension, hypercholesterolaemia (not in HRS) and heart disease). AUCs were calculated at 10 years of follow-up in UKB (A), HRS (B) and ELSA (C). The 95% confidence interval is shown in parentheses. AUC, area under the curve; ELSA, English Longitudinal Study of Ageing; HRS, Health and Retirement Study; Ref, reference; UKB, UK Biobank. In the analysis of estimating the importance of physical frailty and depression in predicting incident all-cause dementia, we found these two factors ranked among the top predictors in relative strength compared with traditional risk factors (online supplemental figure 2). Specifically, physical frailty ranked first across all three cohorts. Depression, though less influential than frailty, still surpassed most lifestyle factors and chronic conditions (ranking fourth in UKB, and second in HRS and ELSA). Joint association of physical frailty and depression with dementia We further assessed the joint effect of physical frailty and depression on dementia risk (figure 3). Participants with coexistence of physical frailty and depression exhibited the highest dementia risk compared with robust individuals without depression (pooled HR: 3.23, 95% CI 2.86 to 3.65; I2=41.6%). The second-highest risk was observed in frail individuals without depression (pooled HR: 2.39, 95% CI 2.18 to 2.62; I2=0.0%). Joint association of physical frailty and depression with risk of all-cause dementia. HRs and 95% CIs were estimated using Cox proportional hazards models adjusted for age, sex, race and ethnicity, education level, smoking, drinking, body mass index, type 2 diabetes, hypertension, hypercholesterolaemia (not in HRS) and heart disease. Depression-free participants with robust physical status served as the reference group. The pooled estimates were from random-effects models. CI, confidence interval; ELSA, English Longitudinal Study of Ageing; HR, hazard ratio; HRS, Health and Retirement Study; UKB, UK Biobank. No significant multiplicative interactions between physical frailty and depression were observed (p-interaction=0.216; table 2). However, significant additive interactions were identified between these factors in all three cohorts (table 2). The pooled RERI was 0.38 (95% CI 0.13 to 0.63; p-interaction=0.003), with 17.1% (95% CI 6.0% to 28.3%) of dementia risk attributable to the interaction. Table 2Interaction effects between physical frailty and depression on dementia riskVariableUKBHRSELSAPooledMain effects Physical frailty (pre-frail/frail vs robust)1.45 (1.38 to 1.52)1.55 (1.35 to 1.77)1.94 (1.42 to 2.66)1.47 (1.40 to 1.54) Depression (yes vs no)1.40 (1.23 to 1.60)0.82 (0.42 to 1.59)0.77 (0.19 to 3.17)1.37 (1.20 to 1.55) Joint effect2.16 (1.99 to 2.35)2.26 (1.92 to 2.66)3.34 (2.37 to 4.70)2.22 (2.07 to 2.39)Multiplicative interaction P-value for multiplicative interaction0.4320.0710.2200.216Additive interaction RERI0.31 (0.07 to 0.56)0.89 (0.27 to 1.51)1.62 (0.27 to 2.98)0.38 (0.13 to 0.63) P-value for additive interaction0.0070.0020.0100.003 AP due to interaction, %14.4 (3.6 to 25.1)39.4 (13.5 to 65.3)48.6 (12.3 to 84.9)17.1 (6.0 to 28.3)Estimates were adjusted for age, sex, race and ethnicity, education level, smoking, drinking, body mass index, type 2 diabetes, hypertension, hypercholesterolaemia (not in HRS) and heart diseases. RERI>0 indicates additive interactions. The pooled estimates were from random-effects models.AP, attributable proportion; ELSA, English Longitudinal Study of Ageing; HRS, Health and Retirement Study; RERI, relative excess risk due to interaction; UKB, UK Biobank. Interaction effects between physical frailty and depression on dementia risk Estimates were adjusted for age, sex, race and ethnicity, education level, smoking, drinking, body mass index, type 2 diabetes, hypertension, hypercholesterolaemia (not in HRS) and heart diseases. RERI>0 indicates additive interactions. The pooled estimates were from random-effects models. AP, attributable proportion; ELSA, English Longitudinal Study of Ageing; HRS, Health and Retirement Study; RERI, relative excess risk due to interaction; UKB, UK Biobank. Subgroup and sensitivity analyses In sex-specific subgroup analyses, the individual associations between physical frailty, depression and the risk of incident all-cause dementia remained consistent in both males and females (p-difference>0.05; online supplemental table 6). Similarly, the joint effect of physical frailty and depression on dementia risk was consistent across sex groups (p-difference>0.05; online supplemental table 7). Multiple sensitivity analyses demonstrated the robustness of our main findings. When we mutually adjusted for physical frailty and depression, the individual associations remained robust (online supplemental table 8). In analyses using different definitions to assess depression, the associations were consistent with those using the original definitions (online supplemental tables 8–10). We observed similar associations with the primary analyses after conducting the competing risk analyses between mortality and incident dementia (online supplemental tables 8,9 and 11). Additional exclusion of incident dementia cases within the first 5 years of follow-up did not alter the observed associations (online supplemental tables 8,9 and 12). The findings were consistent when we restricted the analyses to participants without heart disease (online supplemental tables 8,9 and 13) or without missing data on covariates (online supplemental tables 8,9 and 14). Discussion Main findings In this investigation of three cohort studies, we extend findings in previous publications exploring the associations of physical frailty and depression with incident dementia. Our findings indicate that both physical frailty and depression are significantly associated with an increased risk of incident dementia. Furthermore, the inclusion of physical frailty and depression in a traditional risk model significantly improves dementia risk prediction, with both showing greater predictive power than lifestyle-related risk factors and other chronic conditions. Individuals with both physical frailty and depression exhibit the highest risk of dementia. Importantly, participants with coexisting physical frailty and depression face a dementia risk that exceeds the sum of their individual risks, with approximately 17.1% of the total risk attributable to their additive interaction. In general, our study aligns with previous evidence on the relationships of physical frailty and depression with dementia.4,8 For instance, a recent cohort study from the Korean National Health Insurance System (mean follow-up: 6.7 years) reported that frail individuals had a substantially higher risk of dementia when compared with robust individuals (HR: 3.59, 95% CI 3.42 to 3.77).4 Consistently, our analysis reveals elevated risk in pre-frail/frail participants. For depression, Diniz and colleagues’ meta-analysis of 23 community-based prospective cohort studies demonstrated a robust association between depression and increased dementia risk (HR: 1.85, 95% CI 1.67 to 2.04),5 which was echoed by our findings. Additionally, adding physical frailty and depression to the traditional risk prediction model significantly improves its prospective predictive performance for 10-year dementia risk. It should be emphasised that this enhancement reflects improved risk forecasting capability, not diagnostic accuracy for current or future dementia status. Notably, when comparing the relative importance of predictors, physical frailty and depression ranked higher than lifestyle-related risk factors and other chronic conditions in their association with incident dementia. These results suggest that physical frailty and depression may be among the most critical predictors of long-term dementia incidence. In addition to their individual effects, we evaluated the joint association of physical frailty and depression with dementia risk, which has been rarely examined previously. Jointly, participants with coexistence of physical frailty and depression were at the highest risk of dementia compared with robust and depression-free individuals. Notably, we observed that the dementia risk associated with a combination of physical frailty and depression was greater than the sum of the risks associated with each of these factors, indicating a significant interaction on an additive scale. Specifically, if both physical frailty and depressive symptoms were present, this would result in an additional 17.1% of cases of incident all-cause dementia. These results underscore the complex relationship between frailty, depression and cognitive function. Lower levels of frailty may allow the health system to partially offset the cognitive burden of depression and, similarly, lower levels of depression may enable the system to mitigate the burden of frailty. However, once both factors exceed a certain threshold, this compensatory ability may be compromised, leading to a sharp increase in the risk of dementia. Our study of three nationally representative cohorts provides valuable information, utilising rigorous study designs and a relatively long-term follow-up. However, the results should be interpreted within the context of several methodological challenges in exposure assessment and dementia ascertainment. Specifically, the definitions of physical frailty, depression and dementia varied across the three cohorts, which could have influenced the observed associations. For example, in UKB, hospital records were used to capture historical depression diagnoses, which may not accurately reflect current or recent depressive symptoms. Nevertheless, when we restricted UKB depression diagnoses to those recorded within 2 years before baseline, the associations remained consistent. Dementia ascertainment methods also differed. In UKB, reliance on electronic health records may have led to missed cases of undiagnosed dementia. In contrast, in HRS and ELSA, dementia was defined algorithmically using objective cognitive tests rather than clinical diagnoses, potentially misclassifying individuals with severe cognitive impairment who did not meet formal dementia criteria. However, these differences may also represent a key strength. The differences in exposures and outcome assessment could have introduced non-differential misclassification, likely biasing the results toward the null and making our observed associations conservative estimates. Variations in assessment protocols and cohort demographics created a natural experiment that allowed us to test our hypotheses across diverse settings. Crucially, consistent direction and significance of the associations and interactions across all cohorts demonstrated the robustness of the results. Heterogeneity in population characteristics further enhanced the generalisability of our findings. Our findings can potentially be explained by shared risk factors and overlapping pathophysiological pathways. For example, previous studies have confirmed the association between inflammatory markers and the incidence of physical frailty in non-frail older adults.42 43 Similarly, individuals with depression exhibit elevated levels of proinflammatory cytokines, indicating that inflammatory mechanisms play a crucial role in the pathophysiology of depression.44 Neuroinflammation is known to accelerate neurodegeneration while impairing neuroprotection and neuronal repair, both of which contribute to dementia onset.45 Furthermore, clinical and population-based studies have also demonstrated that physical frailty and depression are strongly associated with mitochondrial dysfunction and oxidative stress,46,49 both of which are hallmark features of dementia.50 51 Our findings have critical implications for clinical practice and public health. First, accurately predicting dementia risk is vital in clinical settings to identify high-risk populations and implement early, tailored interventions such as lifestyle modifications, nutritional support and pharmacological strategies. Therefore, it is imperative to incorporate the assessment of physical frailty and depression into standard clinical procedures. Our study demonstrates that capturing physical frailty status and depression not only reflects an individual’s physical and mental health but also serves as a valuable predictor of cognitive status. Second, our findings suggest that approximately 17.1% of dementia risk can be attributed to the additive effects of physical frailty and depression. Considering the high prevalence of physical frailty and depression comorbidity among older adults,12 this observed increase is noteworthy in the context of global ageing. Therefore, improving the contributing factors of physical frailty and depression comorbidity, along with effective physical frailty management and the maintenance of a healthy psychological state, is essential for the prevention of dementia. Finally, targeted interventions that simultaneously address physical frailty and depression may offer the most effective strategy for mitigating dementia risk in older adults. Policymakers and healthcare providers should prioritise integrated care models to support this vulnerable population. As suggested by previous studies, personalised lifestyle interventions and social support may help reverse frailty and alleviate depression.52,55 Specifically, it is advisable to focus on multi-domain lifestyle interventions, such as smoking cessation, maintaining a balanced diet and providing nutritional support. Additionally, engaging in at least 150 minutes of moderate to vigorous physical activity each week, combined with resistance exercises for >3 months, is recommended.52,55 Limitations Several limitations should be considered when interpreting our findings. First, we were unable to specifically assess the associations of physical frailty and depression with different subtypes of dementia due to inconsistent definitions across cohorts. Second, we could not account for temporal changes in physical frailty and depression, as only a limited number of participants had two or more measurements, particularly in UKB. Third, since the three cohorts included in this study primarily consisted of Western populations, the generalisability of our findings to other cultural and ethnic groups may be limited. Fourth, because only a small number of studies were included in our meta-analysis, the I2 statistics should be interpreted with caution. Fifth, although we adjusted our models for a range of demographic, lifestyle and health-related confounders, residual confounding and bias from unmeasured factors (such as dietary habits, epigenetic changes or polygenic risk) cannot be excluded. Finally, given the observational nature of this study, our findings do not necessarily imply causal relationships between the exposures and dementia. Implications In this analysis of three large, prospective and cross-national cohort studies, both physical frailty and depression were associated with higher dementia risk. These two factors offered significant improvement in the predictive value of the traditional risk model. They ranked higher in relative strength for predicting dementia than lifestyle-related risk factors and other chronic conditions. Individuals with both physical frailty and depression faced the highest risk of dementia. Furthermore, the joint effect of physical frailty and depression on dementia risk exceeded the sum of their individual effects. Given that physical frailty and depression are modifiable, concurrent interventions targeting these conditions could significantly reduce dementia risk. Supplementary material Acknowledgements We thank the participating individuals and staff of the cohorts for their continuing dedication and efforts. The results reported in this article are solely the responsibility of the authors. Biography Yihong Ding obtained his bachelor's degree of Medicine in Preventive Medicine from Zhejiang University, China in 2024. He has been pursuing a master’s degree at the Department of Epidemiology and Biostatistics, School of Public Health, Zhejiang University since September 2024. His main research focuses on identifying modifiable lifestyle and clinical risk factors for neurodegenerative and other age-related diseases through advanced statistical analysis of multinational cohort studies. He further develops evidence-based preventive interventions targeting these conditions, with particular emphasis on translational applications in ageing populations. online supplemental tabe 4 Footnotes Funding: This work was supported by the National Key Research and Development Program of China (2017YFC0907004 to YZ). The funding agencies had no role in study design, data collection, analysis, decision to publish or manuscript preparation. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: This study involves human participants. This research was conducted using the UK Biobank (UKB, Application No.102158), the English Longitudinal Study of Ageing (ELSA) and the Health and Retirement Study (HRS). The UKB received ethical approval from the National Health Service North West Multicentre Research Ethics Committee (21/NW/0157), the ELSA was approved by the Ethics Review Committee of London Multicentre Research (MREC/01/2/91) and the HRS was approved by the Ethics Review Committee of the University of Michigan (HUM00061128). All participants provided written informed consent to participate in the study before taking part. Data availability statement Data supporting the results of this study are available from official websites of three cohort studies: https://www.ukbiobank.ac.uk for UKB, https://ukdataservice.ac.uk for ELSA, and https://hrs.isr.umich.edu for HRS. References Associated Data Supplementary Materials Data Availability Statement Data supporting the results of this study are available from official websites of three cohort studies: https://www.ukbiobank.ac.uk for UKB, https://ukdataservice.ac.uk for ELSA, and https://hrs.isr.umich.edu for HRS.