Physiological measures variability and risks of heart disease and stroke: evidence from three cohort studies Authors: Wei Liu, Lieyang Fan, Da Shi, Linling Yu, Jiahao Song, Ruyi Liang, Xuefeng Lai, Hao Wang, Yongfang Zhang, Shuhui Wan, Yueru Yang, Bin Wang Published: 2024 ABSTRACT AbstractBackgroundThe overall effect of long-term variability in physiological measures on cardiovascular health of older adults and the underlying mechanic pathway remain uncertain.MethodsWe constructed a composite score (0 ~ 3) of variability in physiological measures, including blood pressure, pulse rate, and body mass index, in older adults from the China Health and Retirement Longitudinal Study (CHARLS) 2011 ~ 2015, the Health and Retirement Study (HRS) 2006/2008 ~ 2014/2016, and the UK Biobank 2006 ~ 2019. The associations of the composite score with incident risks of heart disease and stroke were assessed. The mediation roles of several biomarkers were explored.ResultsA higher composite score was related to increased incident risk of heart disease in older adults from the US and the UK and increased incident risk of stroke in all three cohorts. Upon pooling the results, each 1-point increment in the composite score was associated with a 19% (hazard ratio: 1.19; 95% confidence interval: 1.14, 1.30) and a 23% (1.23; 1.12, 1.35) increments in incident risks of heart disease and stroke, respectively. The composite score also exhibited an inverse relationship with grip strength while displaying positive associations with C-reactive protein, glycosylated hemoglobin Alc (HbAlc), and cystatin C. Reduced grip strength, elevated HbAlc, and elevated cystatin C significantly mediated the composite score-associated elevated risks of heart disease and stroke.ConclusionsLong-term variability in physiological measures was associated with increased incident risks of heart disease and stroke, and the associations were partially mediated through deteriorated biomarkers of muscle strength, hyperglycemia, and kidney function.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-024-03805-1.Keywords: Blood pressure variability, Pulse rate variability, Body weight variability, Heart disease, Stroke Background Heart disease and stroke, two types of cardiovascular disease (CVD), are significant global health challenges [1, 2]. The aging population is particularly susceptible to heightened risks of heart disease and stroke, posing substantial threats to the health of older adults [3]. The progressive functional degradation of biological system is a hallmark of aging and heightens incidence of CVD in the old adults [4]. Specifically, excessive disruption of homeostasis emerges as a key facet of aging, playing a pivotal role in the development of CVD in older adults [5]. The inability to maintain homeostasis can be indicated by long-term variability in some physiological measures, notably blood pressure, pulse rate, and body mass index (BMI) [6]. Long-term variability in physiological measure indicates the body’s regulatory systems are struggling to maintain stability or homeostasis [7]. Higher variability often reflects greater dysregulation, as it implies increased strain on feedback mechanisms and higher risks of overshooting corrections, leading to further instability [7]. For example, higher long-term variability in blood pressure may reflect arterial stiffness, a typical lesion that results in impaired blood pressure homeostasis in the body [8, 9]. Previous studies have linked long-term variability in single physiological measure, such as blood pressure [10, 11] or BMI [5, 12] alone, to CVD events. For instance, a meta-analysis has revealed that each 5 mmHg increase in standard deviation (SD) of systolic blood pressure was associated with a 17% increase in incident risk of stroke, a 27% increase in incident risk of coronary heart disease, and a 12% increase in incident risk of CVD [11]. Another meta-analysis has suggested that people with the highest BMI variability was associated with a 27% increase in incident risk of any cardiovascular event compared to those had the least BMI variability [12]. Nevertheless, the overall effect of long-term variability in major physiological measures, especially blood pressure, pulse rate, and BMI together, on incident risk of CVD in older adults across nations remains poorly understood. The urgency to explore such overall effect is underscored by the fact that assessing homeostasis disruption and cardiovascular health risk often requires the simultaneous consideration of multiple physiological measures rather than just a single one. Furthermore, the mechanistic pathway underlying long-term variability in these physiological measures in relation to incidence risk of CVD is unclear. Long-term variability in physiological measures may cause many negative effects, such as systemic inflammation, muscle damage, impaired blood glucose regulation, and stress on the kidneys [13–16], and these adverse changes in physiological functions are recognized risk factors for CVD [17–20]. Thus, exploring the roles of physiological functions such as muscle function, inflammation response, glucose regulation, and renal function in the association between long-term variability in physiological measures and CVD risk may provide important epidemiological clues to the underlying mechanisms. The aims of this study were to explore the associations between long-term variability in multiple physiological measures and incident risks of heart disease and stroke and provide clue for the underlying mechanistic pathways. We drew upon data from tree nationally representative prospective studies, the China Health and Retirement Longitudinal Study (CHARLS), the Health and Retirement Study (HRS) in the USA, and the UK Biobank. The associations of a composite score of variability, constructed by blood pressure variability (BPV), pulse rate variability (PRV), and body weight variability (BWV), with incident risks of heart disease and stroke were assessed, and the mediation roles of several biomarkers of muscle strength (grip strength), hyperglycemia (glycosylated hemoglobin Alc [HbAlc]), system inflammation (C-reactive protein [CRP]), and kidney function (cystatin C), for instance, were evaluated to identify the underlying mechanistic pathways. Methods Details concerning the study population are outlined in the supplementary material (Additional file 1: Study population). As depicted in Fig. 1, three times measurements of physiological measures were conducted in 2011 ~ 2015 for the CHARLS [21], 2006/2008 ~ 2014/2016 for the HRS [22], and 2006 ~ 2019 for the UK Biobank [23]. The time at which the 3rd physiological measurement was completed was defined as baseline. Outcome was ascertained in 2018 for the CHARLS, 2018 for the HRS, and up to January, 2021 for the UK Biobank. Study design. Abbreviations: BPV, blood pressure variability; PRV, pulse rate variability; BWV, body weight variability; HbAlc, glycosylated hemoglobin Alc; CRP, C-reactive protein; TC, total cholesterol; TG, triglycerides; HDL, high-density lipoprotein; LDL, low-density lipoprotein The measurements of physiological measures are described in the supplementary material (Additional file 1: Physiological measures). Blood pressure, pulse rate, and BMI were selected physiological measures due to their non-invasive nature and easy availability for monitoring the health of older adults. Visit-to-visit BPV, PRV, or BWV was calculated as the coefficient of variation (CV), which is the quotient of SD and mean, of blood pressure, pulse rate, or BMI, respectively. Consistent with previous studies [6, 24], systolic blood pressure variability was used to represent BPV due to its stronger association with CVD events when compared with diastolic blood pressure [10]. Following the approach of a previous study [25], the highest tertile of BPV/PRV/BWV was assigned 1 point, while the middle or lowest tertile was assigned 0 point. In the CHARLS and the HRS, self-reported stroke or heart disease diagnosed by a doctor or self-reported medication use for treating stroke or heart disease from the 3rd physiological measurement to the end of follow-up was considered incidence of stroke or heart disease. In the UK Biobank, outcome was ascertained by linking to electronic medical records up to January, 2021, via the International Classification of Diseases, 10th revision: heart disease was defined as I01, I02.0, I11, I13, I20 ~ I25, I26 ~ I28, or I30 ~ I52; stroke was defined as I60 ~ I69. Measurements of mediating markers were performed at the 1st and 3rd assessments of physiological measures. Detailed methodologies for mediating biomarker measurements have been documented elsewhere [26–28]. CRP, HbAlc, total cholesterol (TC), high-density lipoprotein (HDL), and cystatin C were measured in the CHARLS and the HRS. Triglycerides (TG) and low-density lipoprotein (LDL) were measured in the CHARLS. Grip strength was measured in the CHARLS, the HRS, and the UK Biobank. Covariate adjustment was performed according to previous literatures [6, 29, 30], the results of preliminary analysis, and biological and statistical considerations with the change-in-estimate method [31]. Demographic information, lifestyle, and histories of medication use and disease were collected via questionnaires at baseline, including age (years), sex (male/female), education (high school education or higher/below high school education), current smoking status (smokers/nonsmokers), current drinking status (drinkers/nondrinkers), and physical activity (active: ≥ 5 times per week moderate activity for at least 10 min or ≥ 3 times per week vigorous activity for at least 10 min/inactive: did not meet the active criteria) as well as histories of diabetes, stroke, heart disease, dyslipidemia, and hypertension and corresponding medication usage histories. Besides, systolic blood pressure, BMI, and pulse rate measured at baseline were also included as covariates. We employed Cox proportional hazard model to explore the associations of variability in single physiological measure (BPV/PRV/BWV) and the composite score with incident risks of heart disease and stroke. Meta-analysis with fixed-effect model was conducted to estimate pooled hazard ratios (HRs) and pooled 95% confidence intervals (CIs) of heart disease and stroke associated with the composite score. Participants with a history of heart disease/stroke were excluded a priori if the outcome was heart disease/stroke. Restrict cubic spline plots with knots at 5th, 50th, and 95th percentiles were used to explore whether there existed nonlinear relationships between BPV/PRV/BWV and outcomes or not. All models were adjusted for confounders/covariates including age, sex, education, smoking status, drinking status, physical activity, histories of diabetes, dyslipidemia, stroke, and heart disease, systolic blood pressure, BMI, and pulse rate. Stratified analyses were conducted according to sex (male/female), age (< 65/ ≥ 65 years), BMI (< 25/ ≥ 25 kg/m2), physical activity (active/inactive), smoking status (smoker/nonsmoker), and drinking status (drinker/nondrinker), and the interaction effects between the composite score and the stratified variables were assessed by adding product terms in the models. Sensitivity analyses of the associations of the composite score with outcomes were performed by additionally adjusting for residence (rural/urban) in the CHARLS, race (White/Black/others) in the HRS, race (White/other) in the UK Biobank; excluding participants with diabetes; excluding participants with hypertension; excluding participants with dyslipidemia; using SD or using variation independent of the mean (VIM) to present variability [6]; and using competing risk model with setting death as a competing event. The associations between the composite score and mediating biomarkers were assessed via linear regression models, where the mediating biomarkers measured at the 3rd assessment of physiological measure were included as the response variables and the mediating biomarkers measured at the 1st assessment of physiological measure were included as covariates in addition to the aforesaid covariates for adjustment. Because mediation analysis yields results only when there is a significant association between exposure and mediator [32], we further conducted mediation analyses by using R package regmedint (version 1.0.0) according to the results of linear regression models [33]. Details for the method of mediation analysis are described in the supplementary material (Additional file 1: Mediation analyses). All analyses were performed via R software version 4.1.0 (R Foundation for Statistical Computing), with a significance threshold of P < 0.05 in two-sided test. Results Screening process for the study population is illustrated in Additional file 1: Fig. S1. As shown in Table 1, the mean age of participants in the CHARLS, the HRS, and the UK Biobank were 62.93, 72.58, and 65.10 years, respectively. The proportion of female exceeded that of male in all three cohorts, with the highest female proportion observed in the HRS (60.80%). The HRS also recorded the highest proportion of participants with high school education or higher (85.31%), along with the highest mean level of BMI (29.12 kg/m2). The CHARLS had the highest proportion of smokers (38.17%), while the UK Biobank had the highest proportions of drinkers (51.08%) and participants with active physical activity (53.77%). In the CHARLS, the HRS, and the UK Biobank, the mean levels of BPV were 8.58%, 8.62%, and 8.04%, respectively; the mean levels of PRV were 8.51%, 9.17%, and 8.53%, respectively; and the mean levels of BWV were 4.34%, 4.03%, and 3.85%, respectively. Table 1Basic characteristics of the study participants at baselineCharacteristicsCHARLSHRSUK BiobankNumber736167997209Age, years62.93 ± 9.1172.58 ± 8.0165.10 ± 7.82Female3974 (54.04%)4133 (60.80%)3623 (50.26%)Body mass index, kg/m223.10 ± 3.7229.12 ± 5.4626.54 ± 4.31High school education or higher663 (9.01%)5800 (85.31%)5784 (80.23%)Smokers2808 (38.17%)678 (9.97%)2429 (33.69%)Drinkers1778 (24.17%)2688 (39.54%)3682 (51.08%)Active physical activity1924 (26.14%)2378 (34.98%)3876 (53.77%)Stroke284 (3.86%)640 (9.41%)234 (3.25%)Heart disease1346 (18.29%)2121 (31.20%)595 (8.25%)Diabetes744 (10.1%)1714 (25.21%)229 (3.18%)Dyslipidemia670 (9.1%)4109 (60.44%)1332 (18.48%)Systolic blood pressure, mmHg129.0 ± 20.05129.4 ± 20.65140.30 ± 19.88Pulse rate, bpm72.48 ± 10.6568.84 ± 10.7868.91 ± 12.04BPV, %8.58 ± 4.728.62 ± 4.498.04 ± 4.34PRV, %8.51 ± 4.549.17 ± 4.738.53 ± 4.54BWV, %4.34 ± 3.034.03 ± 3.433.85 ± 2.91Data are presented as mean ± standard deviation or case (percent)Abbreviations: CHARLS China Health and Retirement Longitudinal Study, HRS Health and Retirement Study, BPV Blood pressure variability, PRV Pulse rate variability, BWV Body weight variability Basic characteristics of the study participants at baseline Data are presented as mean ± standard deviation or case (percent) Abbreviations: CHARLS China Health and Retirement Longitudinal Study, HRS Health and Retirement Study, BPV Blood pressure variability, PRV Pulse rate variability, BWV Body weight variability There documented 428 heart disease cases and 372 stroke cases in the CHARLS 2018, 2195 heart disease cases and 566 stroke cases in the HRS 2018, and 720 heart disease cases and 118 stroke cases in the UK Biobank up to January, 2021. In the CHARLS, significant association between BPV and incident risk of stroke (HR: 1.12; 95% CI: 1.01, 1.24) was observed (Additional file 1: Table S1), while nonsignificant associations among variability in other physiological measures and outcomes were revealed (all P > 0.05). In the HRS, all physiological measures displayed significant associations with the outcomes (all P < 0.05), except for a marginal association between PRV and incident risk of stroke (HR: 1.08; 95% CI: 0.98, 1.19). In the UK Biobank, BPV was associated with increased incident risks of heart disease (HR: 1.12; 95% CI: 1.04, 1.19) and stroke (HR: 1.22; 95% CI: 1.03, 1.44), and similar results were observed between PRV and incident risks of heart disease (HR: 1.25; 95% CI: 1.18, 1.34) and stroke (HR: 1.19; 95% CI: 1.01, 1.39). Restricted cubic spline plots suggested no significant nonlinear relationship existed between any single physiological measure and outcome (all P for nonlinear > 0.05, Additional file 1: Fig. S2 and S3). As shown in Fig. 2, each 1-point increment in the composite score was associated with a 17% (HR: 1.17; 95% CI: 1.01, 1.38) elevation in incident risk of stroke in the CHARLS, but no significant association between the composite score and incident risk of heart disease was observed (all P > 0.05). In the HRS, each 1-point increment in the composite score was associated with an 18% (HR: 1.18; 95% CI: 1.12, 1.25) and a 20% (HR: 1.20; 95% CI: 1.07, 1.34) elevations in incident risks of heart disease and stroke, respectively. In the UK Biobank, each 1-point increment in the composite score was associated with a 27% (HR: 1.27; 95% CI: 1.16, 1.38) and a 31% (HR: 1.31; 95% CI: 1.07, 1.62) elevations in incident risks of heart disease and stroke, respectively. When pooling the results, each 1-point increment in the composite score was associated with a 19% (HR: 1.19; 95% CI: 1.14, 1.30) and a 23% (HR: 1.23; 95% CI: 1.12, 1.35) elevations in incident risks of heart disease and stroke, respectively, with no significant heterogeneity detected (both P for heterogeneity > 0.05). Associations of the composite score of variability with incident risks of heart disease and stroke. Abbreviation: CHARLS, China Health and Retirement Longitudinal Study; HRS, Health and Retirement Study; BPV, blood pressure variability; PRV, pulse rate variability; BWV, body weight variability; HR, hazard ratio; CI, confidence interval. Models were adjusted for age, sex, education, smoking status, drinking status, physical activity, histories of diabetes, dyslipidemia, stroke, and heart disease, systolic blood pressure, body mass index, and pulse rate. Participants with a history of heart disease were excluded a priori if the outcome was heart disease, and participants with a history of stroke were excluded a priori if the outcome was stroke. Significant effects are in bold In stratified analyses (Additional file 1: Table S2 and S3), only physical activity demonstrated a significant interaction with the composite score on incident risk of stroke (P for interaction < 0.001) in the CHARLS, with a more pronounced association observed in participants with inactive physical activity. The pooled results of the associations between the composite score and outcomes were persistent in all sensitivity analyses (Additional file 1: Table S4). Additional file 1: Table S5 outlines the associations between the composite score and mediating biomarkers. The composite score was significantly and inversely associated with grip strength, while significantly and positively associated with CRP, HbAlc, and cystatin C (all P < 0.05). As shown in Table 2, reduced grip strength (proportion mediated: 20.21%; 95% CI: 15.64%, 30.27%), elevated HbAlc (proportion mediated: 3.31%; 9%% CI: 1.05%, 6.79%), and elevated cystatin C (proportion mediated: 47.55%; 95% CI: 39.35%, 63.18%) significantly mediated the association between the composite score and incident risk of heart disease in the HRS. The association between the composite score and incident risk of stroke was significantly mediated by reduced grip strength (proportion mediated: 12.83%; 95% CI: 7.14%, 16.80%), elevated HbAlc (proportion mediated: 8.58%; 95%CI: 3.47%, 16.94%), and elevated cystatin C (proportion mediated: 14.32%; 95% CI: 7.07%, 26.53%) in the CHARLS (Table 3). Similar mediation patterns for incident risk of stroke were observed in the HRS, with the proportions reaching 41.60% (95% CI: 33.47%, 58.48%), 4.29% (95% CI: 2.34%, 9.75%), and 39.81% (95% CI: 27.66%, 54.13%) for reduced grip strength, elevated HbAlc, and elevated cystatin C, respectively. In the UK Biobank, reduced grip strength also significantly mediated the association between the composite score and incident risk of stroke by 10.19% (95% CI: 5.38%, 17.12%). Directed acyclic graphs of the significant mediation effects are shown in Additional file 1: Fig. S4. No significant interaction exists between exposure and mediator. Table 2Mediation roles of biomarkers in the association between the composite score of variability and incident risk of heart diseaseMediatorsEstimated effectHRSUK BiobankGrip strength, each one-unit decreaseDirect effect1.21 (1.13, 1.30)1.12 (1.01, 1.24)Indirect effect1.03 (1.02, 1.05)0.99 (0.98, 1.01)Total effect1.25 (1.17, 1.33)1.12 (1.01, 1.24)Proportion mediated20.21% (15.64%, 30.27%)-HbAlc, each one-unit increaseDirect effect1.22 (1.15, 1.29)-Indirect effect1.01 (1.001, 1.02)-Total effect1.23 (1.16, 1.31)-Proportion mediated3.31% (1.05%, 6.79%)-CRP, each one-unit increaseDirect effect1.22 (1.15, 1.29)-Indirect effect1.01 (0.99, 1.02)-Total effect1.23 (1.16, 1.31)-Proportion mediated--Cystatin C, each one-unit increaseDirect effect1.08 (1.01, 1.13)-Indirect effect1.07 (1.06, 1.10)-Total effect1.20 (1.13, 1.28)-Proportion mediated47.55% (39.35%, 63.18%)-Models were adjusted for age, sex, education, smoking status, drinking status, physical activity, histories of diabetes, dyslipidemia, and stroke, systolic blood pressure, body mass index, pulse rate, and mediating biomarkersParticipants with a history of heart disease were excludedDirect, indirect, and total effects are presented as hazard ratio with 95% confidence interval, and significant effects are in boldProportion mediated was not calculated if the indirect effect was not significantData of mediation analysis for heart disease in the CHARLS was not shown given the association between the composite score and heart disease (i.e., the total effect) was not significantAbbreviation: CHARLS China Health and Retirement Longitudinal Study, HRS Health and Retirement Study, HbAlc Glycosylated hemoglobin Alc, CRP C-reactive protein Mediation roles of biomarkers in the association between the composite score of variability and incident risk of heart disease Models were adjusted for age, sex, education, smoking status, drinking status, physical activity, histories of diabetes, dyslipidemia, and stroke, systolic blood pressure, body mass index, pulse rate, and mediating biomarkers Participants with a history of heart disease were excluded Direct, indirect, and total effects are presented as hazard ratio with 95% confidence interval, and significant effects are in bold Proportion mediated was not calculated if the indirect effect was not significant Data of mediation analysis for heart disease in the CHARLS was not shown given the association between the composite score and heart disease (i.e., the total effect) was not significant Abbreviation: CHARLS China Health and Retirement Longitudinal Study, HRS Health and Retirement Study, HbAlc Glycosylated hemoglobin Alc, CRP C-reactive protein Table 3Mediation roles of biomarkers in the association between the composite score of variability and incident risk of strokeMediatorsEstimated effectCHARLSHRSUK BiobankGrip strength, each one-unit decreaseDirect effect1.16 (1.02, 1.31)1.15 (1.001, 1.33)1.21 (1.01, 1.55)Indirect effect1.02 (1.01, 1.04)1.09 (1.07, 1.12)1.02 (1.01, 1.04)Total effect1.19 (1.04, 1.35)1.26 (1.11, 1.43)1.24 (1.01, 1.57)Proportion mediated12.83% (7.14%, 16.80%)41.60% (33.47%, 58.48%)10.19% (5.38%, 17.12%)HbAlc, each one-unit increaseDirect effect1.13 (0.99, 1.28)1.27 (1.13, 1.43)-Indirect effect1.01 (1.003, 1.02)1.01 (1.001, 1.02)-Total effect1.14 (1.01, 1.30)1.28 (1.14, 1.44)-Proportion mediated8.58% (3.47%, 16.94%)4.29% (2.34%, 9.75%)-CRP, each one-unit increaseDirect effect1.13 (1.01, 1.28)1.26 (1.12, 1.41)-Indirect effect1.003 (0.99, 1.01)1.01 (0.99, 1.02)-Total effect1.14 (1.01, 1.29)1.27 (1.13, 1.43)-Proportion mediated---Cystatin C, each one-unit increaseDirect effect1.11 (0.96, 1.28)1.12 (0.99, 1.25)-Indirect effect1.03 (1.01, 1.04)1.09 (1.06, 1.11)-Total effect1.14 (1.01, 1.31)1.23 (1.11, 1.40)-Proportion mediated14.32% (7.07%, 26.53%)39.81% (27.66%, 54.13%)-Models were adjusted for age, sex, education, smoking status, drinking status, physical activity, histories of diabetes, dyslipidemia, and heart disease, systolic blood pressure, body mass index, pulse rate, and mediating biomarkersParticipants with a history of stroke were excludedDirect, indirect, and total effects are presented as hazard ratio with 95% confidence interval, and significant effects are in boldProportion mediated was not calculated if the indirect effect was not significantAbbreviation: CHARLS China Health and Retirement Longitudinal Study, HRS Health and Retirement Study, HbAlc Glycosylated hemoglobin Alc, CRP C-reactive protein Mediation roles of biomarkers in the association between the composite score of variability and incident risk of stroke Models were adjusted for age, sex, education, smoking status, drinking status, physical activity, histories of diabetes, dyslipidemia, and heart disease, systolic blood pressure, body mass index, pulse rate, and mediating biomarkers Participants with a history of stroke were excluded Direct, indirect, and total effects are presented as hazard ratio with 95% confidence interval, and significant effects are in bold Proportion mediated was not calculated if the indirect effect was not significant Abbreviation: CHARLS China Health and Retirement Longitudinal Study, HRS Health and Retirement Study, HbAlc Glycosylated hemoglobin Alc, CRP C-reactive protein Discussion In this prospective study encompassing three cohorts, the composite score constructed by BPV, PRV, and BWV was significantly associated with incident risks of heart disease and stroke. Moreover, reduced grip strength, elevated HbAlc, and elevated cystatin C partially mediated the associations of the composite score with incident risks of heart disease and stroke. These findings underscore the importance of maintaining physiological stability for cardiovascular health in the older adults. Although BPV [34], PRV[35], and BWV [36] have been separately found to be associated with CVD events of older adults in previous studies, there exists limited evidence regarding the overall effect of long-term variability in physiological measures on cardiovascular health. Consistent with our findings, a cohort study of the general adults from Korean suggested that a composite score, which was constructed by 13-year variability in fasting blood glucose, TC, systolic blood pressure, and BMI, was significantly associated with elevated incident risks of stroke and myocardial infarctions in the subsequent 3 years, with the associations being stronger among participants older than 65 years [25]. The results of the present study revealed the adverse effects of variability in non-invasive physiological measures on CVD events by integrating the three national populations of older adults, suggesting a commonality in the detrimental impact of long-term variability in physiological measures on cardiovascular health across nations. The pooled results across populations were independent of the baseline levels of these physiological measures and demonstrated consistency in sensitivity analyses. Nevertheless, it should be noted that the follow-up durations for the outcomes in studies regarding long-term variability of physiological measures were relatively short [11, 12]. Further investigation with longer follow-up time is warranted to validate the causal relationship, especially given a previous population-based cohort study from the Netherlands observed that the HR of stroke incidence associated with BPV may fluctuate with increasing follow-up time [34]. We found that the association between the composite score and incident risk of stroke was significantly weakened in participants with active physical activity, indicating the potential protective effect of physical activity on long-term variability-associated risk of stroke. In addition, the associations were mostly absent in CHARLS, which may be resulted by multiple reasons. On the one hand, participants from the CHARLS had the youngest average age and the healthiest average BMI of the three cohort participants, making the CHARLS participants tend to have stronger compensatory capacity for health damage and thus partly obscure the association between physiological measures variability and CVD. On the other hand, the lower educational attainment of CHARLS participants and the self-reported approach used for outcome determination may lead to higher degree of outcome misclassification, which may furtherly dilute the association. We found that long-term variability in physiological measures was related to reduced grip strength as well as elevated CRP, HbAlc, and cystatin C, and these associations were independent of the levels of these biomarkers at recruitment. Previous studies have provided either mechanistic or observational support for these findings. In older adults from Japan [37] as well as France and Monaco [13], BPV was associated with increscent risks of reduced grip strength and even frailty. A plausible explanation is that greater variability in physiological measures signifies a state of physiological dysregulation that is also a definition of frailty (manifested as reduced grip strength) [13]. Previous animal studies have found that variability in physiological measures, like BPV, can lead to the increases of plasma thromboxane B2, interleukin-1, and tumor necrosis factor alpha in rats [38]. These findings indicate that physiological measurement variation may lead to the initiation of inflammation that can also manifest as an increase in CRP. Variability in physiological measures has been reported to be associated with glucose metabolism disorders and impaired renal function [16, 39], which align with the associations of variability with HbAlc and cystatin C observed in the present study. It is also mechanically plausible because aging-related dysregulation of homeostasis is often accompanied by impaired function of various organs, such as islets and kidneys. Moreover, fluctuating physiological measures are also typical risk factors for end-organ damage [38]. The associations of long-term variability in physiological measures with incident risks of heart disease and stroke may be partly explained by reduced muscle strength, hyperglycemia, and impaired kidney function. These mediating effects are mechanistically plausible. Reduced muscle strength characterized by reduced grip strength can decrease glycolysis, impair glucose regulation, and increase lipids levels, which are well-known risk factors for CVD [40]. Higher HbAlc and cystatin C are widely recognized risk factors for CVD, especially in older adults [41, 42]. The mediation roles of HbAlc and cystatin C observed in our study also corroborate prior findings from the observational studies that dysglycemia and renal impairment often precede cardiovascular damage associated with variability in physiological measures [43, 44]. Although inflammation is a fundamental pathway in the onset and progression of CVD [38, 45], we did not observe CRP act as a mediator. In fact, at present, the relationship between CRP, as one of the indicators of inflammation, and CVD is still controversial [46, 47]. To comprehensively illustrate the role of inflammation, future studies may need to evaluate the mediating effect of other inflammatory factors. Overall, these findings of mediation analyses may provide inspiration for clinical interventions. Specifically, improving muscle strength through resistance training could mitigate cardiovascular risks, while managing hyperglycemia and monitoring kidney function may enhance overall patient care, ultimately reducing the incidences of heart disease and stroke and improving long-term health outcomes. Leveraging a large sample size, multicenter approach, and extensive data collection, our study offers an epidemiologic understanding of associations and underlying mechanic pathways. Nevertheless, the results should be interpreted in the light of potential limitations of the study. Firstly, outcomes were ascertained through self-reported information in the CHARLS and the HRS, while medical records were used in the UK Biobank. Such distinction might introduce variability in results across populations, although no significant heterogeneity was found in the meta-analysis. Secondly, the follow-up time for outcome was relatively short, and future studies are warranted to examine the relationship between variability in physiological measures and long-term cardiovascular events. Thirdly, unhealthy people generally pay more attention to their health status that tend to undergo more physical examinations; thus, participants included in our study (those underwent three physical examinations) may be unhealthier than those excluded, possibly introducing selection bias. Moreover, such conditional selection may restrict the generalizability of the findings, particularly regarding earlier events or individuals who did not participate in the full assessment process. Fourthly, potential for type I error may exist due to the numerous hypothesis tests conducted. The increased risk of false positive arising from multiple comparisons may compromise the validity of the results. Therefore, findings should be interpreted with caution. Fifthly, three points of measurements of physiological measures over several years may not fully capture the real underlying variability during this long period of time, especially for indicators, including blood pressure and heart rate, that may change greatly in the short term. Finally, residual or unknown confounding, such as dietary factors and genetic factors, cannot be ruled out given the observational study design, which need to be further explored in future studies. Conclusions This prospective cohort study of older adults from three populations in China, the US, and the UK suggested that higher long-term variability in physiological measures was associated with increased incident risks of heart disease and stroke. Reduced muscle strength, hyperglycemia, and impaired kidney function may play mediation roles in the associations of variability in physiological measures with incident risks of heart disease and stroke. Our study suggests that avoiding excessive disruption of homeostasis may be important for cardiovascular health in older adults. Supplementary Information Additional File 1: Table S1-S5. Table S1. Associations of variability in physiological measures with incident risks of heart disease and stroke. Table S2. Stratified analyses for the association of the composite score of variability in physiological measures with incident risk of heart disease. Table S3. Stratified analyses for the association of the composite score of variability in physiological measures with incident risk of stroke. Table S4. Sensitivity analyses for the associations of the composite score of variability in physiological measures with incident risks of heart disease and stroke. Table S5. Associations of the composite score of variability (1-point increment) with mediating biomarkers. Fig. S1-S4. Fig. S1. Participant inclusion and exclusion flowchart. Fig. S2. Restricted cubic spline plots of the association of variability in physiological measures with incident risk of heart disease. Fig. S3. Restricted cubic spline plots of the association of variability in physiological measures with incident risk of stroke. Fig. S4. Directed acyclic graphs of the significant mediation roles found in mediation analyses. Acknowledgements The China Health and Retirement Longitudinal Study was funded by the National Institute on Aging (R01-AG037031 and R01AG067625), the Natural Science Foundation of China (72061137005, 81903392, 71603013, 72173008, and 71873010), and the China Medical Board (20-364). The Health and Retirement Study was sponsored by the National Institute on Aging (NIA U01AG009740) and the Social Security Administration. The study director is Dr. David R. Weir of the Survey Research Center at the University of Michigan’s Institute for Social Research. For the UK Biobank, our analyses had gotten ethical approval and were conducted under the UK Biobank application no. 88159. The authors would like to thank the participants and workers of the CHARLS, the HRS, and the UK Biobank for their selfless contributions which made the current study possible. Abbreviations Blood pressure variability Body mass index Body weight variability Cardiovascular disease China Health and Retirement Longitudinal Study Coefficient of variation Confidence intervals C-reactive protein Glycosylated hemoglobin Alc Hazard ratios Health and Retirement Study High-density lipoprotein Low-density lipoprotein Pulse rate variability Standard deviation Total cholesterol Triglycerides Variation independent of the mean Authors’ contributions W.L. and B.W. conceived and designed this study. B.W. was responsible for obtaining funding. W.L. performed the statistical analyses, interpreted the data, and drafted the manuscript. W.L., L.F., D.S., L.Y., J.S., R.L., X.L., H.W., Y.Z., S.W., Y.Y., and B.W. contributed to manuscript revision, important intellectual content, and final version approval. B.W. is the guarantors of this work and, as such, have full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors read and approved the final manuscript. Funding This work was supported by the National Natural Science Foundation of China (82203996) and the China Postdoctoral Science Foundation (2022T150230, 2021M691131). The funders had no role in the study design, data collection, data analysis and interpretation, writing of the report, or the decision to submit the article for publication. Data availability Cohorts descriptions and questionnaires are available at http://charls.pku.edu.cn/, https://hrs.isr.umich.edu/, and https://biobank.ndph.ox.ac.uk/. Statistical code was available from the corresponding author. Declarations Ethics approval and consent to participate Ethical approval for all the CHARLS waves was granted from the Institutional Review Board at Peking University (approval number: IRB00001052, available at: http://charls.pku.edu.cn/), ethical approval for the HRS was granted from the University of Michigan Institutional Review Board (approval number: 16-19185, available at: https://hrs.isr.umich.edu/), and ethical approval for the UK Biobank was granted from the North West Multi-centre Research Ethics Committee (approval number: 299116, available at: https://biobank.ndph.ox.ac.uk/). Informed consent was obtained from all individual participants included in the study. Consent for publication All participants consented to the use of their de-identified data and access to their national health-related hospital and death records. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Additional File 1: Table S1-S5. Table S1. Associations of variability in physiological measures with incident risks of heart disease and stroke. Table S2. Stratified analyses for the association of the composite score of variability in physiological measures with incident risk of heart disease. Table S3. Stratified analyses for the association of the composite score of variability in physiological measures with incident risk of stroke. Table S4. Sensitivity analyses for the associations of the composite score of variability in physiological measures with incident risks of heart disease and stroke. Table S5. Associations of the composite score of variability (1-point increment) with mediating biomarkers. Fig. S1-S4. Fig. S1. Participant inclusion and exclusion flowchart. Fig. S2. Restricted cubic spline plots of the association of variability in physiological measures with incident risk of heart disease. Fig. S3. Restricted cubic spline plots of the association of variability in physiological measures with incident risk of stroke. Fig. S4. Directed acyclic graphs of the significant mediation roles found in mediation analyses. Data Availability Statement Cohorts descriptions and questionnaires are available at http://charls.pku.edu.cn/, https://hrs.isr.umich.edu/, and https://biobank.ndph.ox.ac.uk/. Statistical code was available from the corresponding author.