Can dementia risk be reduced by following the American Heart Association's Life's Simple 7? A systematic review and dose-response meta-analysis

This study aimed to quantify the relationships between the American Heart Association (AHA) Cardiovascular Health (CVH) metrics, namely AHA Life’s Simple 7, and cognitive outcomes. We searched PubMed and Embase (January 1, 2010—August 24, 2022) and finally included 14 longitudinal studies (311654 participants with 8006 incident dementia cases). Random-effects meta-analysis and one-stage linear mixed-effects models were performed. Increased CVH score seemed to associate with decreased risk of incident dementia in a linear manner, but this relationship varied by the measurement age of CVH metrics. That is, mid-life CVH tended to have a linear association with late-life dementia risk, whereas a J-shaped association was observed between the late-life CVH score and dementia. In addition, late-life dementia risk was reduced significantly if individuals maintained an ideal level of AHA’s CVH guidelines of physical activity, fasting plasma glucose, total cholesterol, and smoking. However, our meta-analysis did not show a significant association between CVH score and global cognitive decline rate. Following AHA’s CVH guidelines and maintaining CVH at an optimal level would substantially reduce the late-life dementia risk. More research is required to explore the link between a favorable CVH score and cognitive trajectories among cognitively asymptomatic older populations.


Introduction
Despite the number of people with dementia worldwide is expected to rise continuously, longitudinal cohort studies have shown evidence supporting a gradual decline in the agespecific dementia incidence over the previous decades. [1,2] Though the factors underlying J o u r n a l P r e -p r o o f the declining trends remain unclear, it has been suggested that improved cardiovascular health may partly contribute to this phenomenon. [1][2][3] Yet, it remains unclear the exposure timing and dose at which maintaining cardiovascular health generates optimal beneficial effects on dementia.
In 2010, the American Heart Association (AHA) defined ideal levels of seven modifiable cardiovascular health (CVH) factors, known as Life's Simple 7, that consist of smoking, physical activity, diet, body mass index, fasting blood glucose, total cholesterol, and blood pressure. [4] Maintaining ideal levels of AHA's CVH metrics has been recommended as a prevention strategy against not only cardiovascular diseases but also neurodegenerative disorders, e.g., cognitive decline and dementia. [5,6] According to previous novel evidence of systematic reviews and meta-analyses, cardiovascular disease can be indeed preventable through keeping at least an intermediate level of the AHA's CVH metrics. [7] However, studies exploring the beneficial effects of the AHA's CVH metrics on cognitive outcomes, especially among older populations, have been uncertain. [8,9] Solid evidence from systematic reviews and meta-analyses of longitudinal studies is lacking in this field.
In addition, cohort studies have displayed a U-or J-shaped relationship between certain CVH factors and dementia risk in old age. [10,11] For instance, increased dementia risk was evident in older adults both with high or low systolic blood pressure. [10,12] It is thus plausible that in older adults, the relationship between the AHA's CVH score and cognitive outcomes may tend to be non-linear and vary between age groups. Performing a doseresponse analysis in a systematic review, considering the dose exposures of the CVH metrics on cognitive outcomes, is essential for understanding the vascular mechanism in the development of dementia.

J o u r n a l P r e -p r o o f
In the current systematic review and meta-analysis, we aimed to quantify the relationship between the AHA's CVH metrics and cognitive outcomes. Specifically, we first estimated the pooled association of the CVH score with the risk of incident dementia and global cognitive decline and then analyzed the dose-response relationships of the CVH score and individual CVH components with cognitive outcomes.

Methods
We reported all results following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. [13] The PRISMA checklists were listed in Table A.1.
The present study was registered in PROSPERO (CRD42021242231).

Search strategy
Our searching strategy included all key terms that cover -cardiovascular health metrics‖, -dementia‖, and -cognitive decline‖. The detailed searching strategy is given in Table A.

2.
Since the AHA guideline of the ideal CVH metrics was released in early 2010, we performed a literature search in PubMed and Embase from January 1, 2010, until August 24, 2022 (the last date searched). The forward and backward citation tracking of the included studies was performed to identify the relevant references lists and citations.

Eligibility and study selection
Two authors (J.W. and Y.X.) reviewed all titles and abstracts of all obtained articles independently. The discrepancies were resolved by reviewing full texts or by discussing with another author (R.W.).
The inclusion criteria of studies were based on PECO (population, exposure, comparator, and outcomes) recommendations, [14] which included 1) the longitudinally designed studies including cohort, nested case-control, or case-crossover studies with assessment of the ideal CVH metrics using the AHA definition; 2) studies that reported at least one outcome of interest using standard assessments, with i) following the standard J o u r n a l P r e -p r o o f diagnosis procedures on dementia, Alzheimer's disease, or cognitive dysfunction, ii) maintaining standard and consistent assessments on cognitive performance over time; and 3) the study population was aged >18 years and without cognitive impairment at study entry.
We excluded studies that reported less than five cardiovascular health metrics, non-English or Chinese publications, or studies with insufficient information in the abstracts and without full texts available.

Data extraction and assessment for study quality
Two authors (Y.X. and J.W.) extracted data from each eligible study using prespecified and standard and data extraction forms and X.X. checked the data extracted from the study to ensure the accuracy of data extraction. We extracted data from each eligible study on 1) article information: name of the study, authors, publication year; 2) sample information: country of the study, number of participants, percentage of females, mean age; 3) study design: baseline year of assessment, follow-up period, number of follow-up waves, response rate, drop-outs; 4) measurements of exposures: implementation of CVH definitions; 5) measurements of outcomes: assessment tool, outcome definition; 6) analysis details: statistical methods, handling missing data, model adjustments; 7) reported estimates: number of cases, risk estimates. If the same study population was applied in different publications, we included the one with a longer follow-up period.

Risk of bias assessment
Using the Newcastle-Ottawa Scale that assesses the quality of cohort or case-control studies in meta-analyses, [15] two authors (Y.X and X.X) reviewed the risk of bias (ROB) of all eligible studies. Specifically, a total score of the ROB was calculated by summarizing the scores of corresponding items in three domains: 1) four items for selection domain (four scores), 2) one item for comparability domain (two scores), and 3) three items for the outcome domain (three scores). Studies with a total score over 6 were considered high quality. J o u r n a l P r e -p r o o f

Data synthesis and statistical analysis
We calculated the logarithm of hazard ratios (HRs) with 95% confidence intervals (CI) for dementia associated with a one-point increase in a CVH score. Four studies [9,[16][17][18] were with incomplete information on linear associations between the CVH score and dementia that only with estimations in the relationship between dichotomized CVH categories and dementia. Thus, we estimated the hazard ratio and corresponding 95% CI per one-point increase in CVH score using a linear regression model with inverse-variance weights. The pooled mean differences and related standard errors were extracted to estimate the cognitive change corresponding to a one-point increase in CVH score. We then reported the pooled reduction in the 10-year global cognitive decline rate with a one-point increase in CVH score.
Secondly, we extracted risk estimations between the CVH categories and dementia for doseresponse analysis. Specifically, we extracted the median point of CVH score in each CVH strata to link to the corresponding hazard ratio in each study. All results were extracted from multivariable models that adjusted for relevant confounding factors. We applied randomeffects meta-analyses to estimate (1) summary hazard ratio and 95% CI of dementia with a one-point increase in a CVH score; and (2) pooled β-coefficient with 95% CI of change in global cognitive decline rate per 10 years with a one-point increase in the CVH score. To investigate the shape of the dose-response curve between favourable CVH score and cognitive outcomes, we employed the one-stage dose-response meta-analysis [19] to estimate the hazard ratios of dementia across the reported CVH categories. We modelled the favourable CVH score with restricted cubic splines with 3 knots presented in the 10 th , 50 th , and 90 th percentiles. The non-linear relation was tested by hypothesizing that the coefficient of the second spline is zero. Subgroup analyses were conducted by linking late-life dementia risk to midlife CVH score (age of assessment: 30-50 years) and late-life CVH score (age of assessment: ≥65 years), respectively. The same dose-response and meta-analyses were J o u r n a l P r e -p r o o f performed for each CVH metric as well to estimate the specific contribution of involved CVH components to dementia.
We assessed the heterogeneity in all the random-effects meta-analyses with I 2 and τ 2 statistics reported for each model. I 2 values of >25%, >50%, >75% indicate a low, moderate, and high heterogeneity, respectively. Publication bias was visually assessed using a funnel plot, and asymmetry was tested using Egger's regression and Begg's rank correlation test at the P<0.10 level of significance. In case of possible publication bias was detected, the trim and fill method was applied to evaluate the number of missing studies in a meta-analysis and recalculated the pooled effect estimations with the imputed studies. We applied a leave-oneout procedure to remove outliers.
To investigate the potential causes of heterogeneity, additional analyses were carried out. Stratification analyses were conducted between groups with distinct profiles, namely, region (Europe vs. North America), the average age of the study population (<70 vs. ≥70 years), follow-up time (<10 years vs. ≥10 years), age of CVH assessment (midlife CVH score vs. late-life CVH score), APOE-ε4 status (carriers vs. non-carriers), subdomains of the CVH metrics (behavioural vs. biological metrics), and ROB subgroups. Furthermore, the same analyses were repeated by 1) excluding estimates from linear extrapolation; 2) excluding studies with incomplete CVH components; 3) excluding studies with high ROB.
All data analyses were performed using Stata SE software, version 17 (StataCorp, College Station, LP), using the meta routine and the -drmeta" command. [20] 3. Results  Table 1. Detailed information on study design, exposures, outcome measurements, and confounding of each selected study is presented in Tables A.3&A.4. Table A. 5 shows the qualities of studies evaluated following the Newcastle-Ottawa Scale.

Association of dementia and global cognitive decline with the CVH score
Using categorical estimations, we derived the hazard ratios of dementia with 95%CI in relation to the CVH score from 11 studies (see Figure 2, penal A). [9, 16-18, 21-24, 26, 27, 29, 30] Overall, a one-point increase in CVH score is associated with a 6% decrease in incident dementia (hazard ratio 0.94, 0.92 to 0.96, I 2 =94.98.04%). In addition, by pooling the estimation from three longitudinal studies (see Figure 2, penal B), data were compatible with the hypothesis of no linear association (β per 10-year 0.020 to -0.007 to 0.046, I 2 =86.57%).
Publication bias, potential small-study effect, asymmetry, and influential effect of individual studies are presented in Figures A.2

-A.7. No publication bias was detected using
Begg test nor Egger's linear regression test for studies with dementia as an outcome (see In comparison to other studies with global cognitive decline as outcome, omitting the study by Pase et al [8] shows the smallest effect size (β per 10-year 0.009, 0.001 to 0.017; see Figure   A.7).

Dose-response Meta-analysis between favourable CVH score and hazard of dementia
Because of a limited number of selected studies and insufficient data for synthesis, the doseresponse meta-analyses were only performed when dementia was treated as an outcome. We included ten studies that reported sufficient information for the non-linear dose-response analyses and applied the average CVH score (7-points) as a reference value. [9, 16-18, 22-24, 26, 27, 30] The dose-response curve (see Figure 3, panel A) shows a nearly J-shaped association between a favourable CVH score and incident dementia. Specifically, the proportion of hazard ratios of dementia drops 38.3% from CVH 0 (hazard ratio 1.49, 1.27 to 1.75) to CVH 9 (hazard ratio 0.92, 0.89 to 0.95), and 16.3% from CVH 9 to CVH 14 (hazard ratio 0.77, 0.64 to 0.92).
After conducting a stratification analysis by the age at CVH assessments, we found that the relationship between midlife favourable CVH score and reduction in the hazard of dementia tend to be linear (see We further carried out the dose-response meta-analysis to investigate the beneficial effect of individual CVH components on the hazard of dementia (see Figure 4). The pooled estimates of up to five studies [9,16,18,23,27] showed that the hazard of incident dementia Published data were compatible with the hypothesis of no association between CVH components in diet (hazard ratio ideal 0.84, 0.54 to 1.32, n=2), body mass index (hazard ratio ideal 1.14, 0.69 to 1.88, n=4), blood pressure (hazard ratio ideal 0.98, 0.72 to 1.33, n=5) and the dementia risk.

Additional analyses
To explore the possible explanations of heterogeneity, we performed additional analyses by comparing estimations between subgroups and excluding studies without sufficient information (see Table 2 When we divided favourable CVH scores into behavioural and biological scores, more heterogeneity was shown in the CVH behavioural score (I 2 =85.43, n=4) than in biological score (I 2 =23.21, n=4).

Principal findings
In this systematic review and meta-analysis, we pooled data from 14 population-based longitudinal studies to quantify the association of AHA's CVH metrics with dementia and cognitive decline in older adults. Our results show a considerable effect of favourable CVH score on reduced risk of incident dementia in older adults. A nearly J-shaped dose-response relationship is observed between an increased favourable CVH score and a reduced risk of incident dementia, especially when the CVH score is assessed in late life. Dementia risk can be reduced significantly if older adults achieved the recommended intermediate-to-ideal level J o u r n a l P r e -p r o o f of physical activity, blood glucose, or total cholesterol, while the reduction of dementia risk with non-current smoking seems to be borderline. The meta-analysis with three available cohort studies did not exhibit a substantial effect of CVH score on cognitive decline in older adults.

Mechanisms and Interpretations
Maintaining optimal cognitive function is a fundamental element of successful ageing. The pathophysiological changes in brain and cognition may start decades before the clinical manifestation of cognitive deficits and dementia,[31] highlighting identifying modifiable risk/protective factors of cognitive ageing and dementia at early stages are essential for disease prevention. [32] Studies have indeed demonstrated that dementia, including all-cause dementia and Alzheimer's disease, is a multifactorial disorder, whose aetiology is closely linked to vascular diseases. [33] Keeping an optimal profile of cardiovascular risk factors during the entire lifespan, especially in young adulthood and midlife, has been found to contribute to cognitive maintenance in ageing. [34] The protective effects of favourable CVH metrics on cognitive decline and dementia may be attributed to the shared pathogenesis processes (e.g., atherosclerosis and ischemia) between dementia and cardiovascular diseases. [35] Obtaining optimal levels of CVH metrics may decelerate the progression of atherosclerosis and thus benefit heart and brain vessels from ischemic and microvascular diseases. [36] Longitudinal cohort studies have proved that carotid atherosclerosis is associated with an increased risk for dementia. [37,38]  Furthermore, we observed that the patterns of relationship between CVH score, and dementia risk differed by the age of CVH assessment. That is, the dose-response relationship between a midlife CVH score and a reduction of dementia risk tends to be linear, whereas a Jshaped dose-response curve is seen in the association between late-life CVH score and risk of dementia. These patterns may be attributable to the age-varying associations of individual CVH components (e.g., blood pressure) with dementia. Specifically, the link between blood pressure and body mass index with dementia risk is complex, and it has been reported that older adults with high blood pressure and obesity tend to have a reduced dementia risk. [47] This can be interpreted by, at least partly, reverse causation, which refers to the damage of J o u r n a l P r e -p r o o f specific brain areas (e.g., brainstem, insula, amygdala, and hypothalamus) that can cause autonomic nervous system dysfunction and impair control of blood pressure and fat mass. [47] Although survival bias may provide certain levels of effect on the J-shaped association between late-life CVH metrics and dementia risk reduction, our dose-response analyses reinforced that the magnitude of the beneficial effect of AHA CVH metrics on dementia appeared to relay on age.
Our findings suggested that the association of optimal level of diet defined by CVH guidelines with dementia risk required further investigation. The uncertain result for diet effect on dementia may be largely attributable to the fact that only two available studies were included in the meta-analysis. [18,27] We observed that dementia risk was substantially reduced among individuals with intermediate-to-optimal levels of physical activity (certain moderate/vigorous activity per week), blood glucose (HbA1c≤6.4% or fasting glucose<126 mg/dL), total cholesterol (<240 mg/dL), and smoking (never or former smoking). This indicates that the suggested optimal CVH levels in smoking, physical activity, glucose, and lipid levels benefit not only cardiovascular health but also brain health. Yet, future studies need to explore whether lipid subcomponents are necessary to be included in the CVH metrics, e.g., including guidelines for low-density lipoprotein and high-density lipoprotein. [48]

Comparison with other studies
This systematic review and meta-analysis bring new evidence on to what extent reaching the ideal CVH goal defined by AHA may protect older adults against age-related disorders such as dementia. Regarding the patterns of the relationship between CVH score and risk of dementia, the protective effect of a favourable CVH score on dementia tends to be J-shaped.
The aggregated effect of maintaining an ideal CVH level on dementia is somehow consistent with findings from the FINGER study, which investigated the effects of multidomain The results from Three-city reporting global cognitive scores were excluded from the metaanalysis because time was treated as a nonlinear function,[21] and we were unable to extract the estimated β-coefficients of change in 10-year cognitive decline in relation to a one-point increase in CVH score. Only one study reported the association between ideal CVH metrics and mild cognitive impairment, and we decided to remove it from our main result. [29] J o u r n a l P r e -p r o o f Since 2010, the AHA has established a 10-year goal to promote population CVH, and the ideal AHA's CVH metrics thus are recommended to adults aged 18 years and older.
Though the 10-year goal was not entirely achieved in 2020, a considerable 15% reduction in age-adjusted CVD mortality was observed. [6] Notably, the ideal CVH metrics have been recently committed as one of the main missions for achieving healthy ageing and longevity by 2030 and recommended as a tool to promote brain health for older adults. [5,6] The current timely systematic review and meta-analysis provides convincing evidence on the link between the AHA's CVH score and dementia, filling knowledge gaps in this field.
Specifically, our findings highlight the fact that there is an urgent need for more research that

Strengths and limitations of this study
To the best of our knowledge, the present study is the first meta-analysis that considered both linear and non-linear dose-response association between the AHA's CVH score and cognitive outcomes in longitudinal studies. We included studies with large sample sizes and long follow-up periods. As we only included the longitudinal studies, the potential influence of recall and selection bias could be minimized. The estimations were extracted from fully adjusted models to minimize the potential confounding bias. Also, the advanced doseresponse analysis can provide more quantified associations and test the shape of the possible associations between CVH scores and dementia risk. Moreover, the studies included in the present meta-analysis were with high qualities and low ROBs.
Some limitations of this study deserve mentioning. Firstly, even if we performed the random-effects models to take heterogeneity into account, the heterogeneity across studies was high in the main analysis. Therefore, we performed a series of subgroup and sensitivity analyses to examine the potential sources of heterogeneity and testify to the robustness of the results in the subgroups. We found the heterogeneities were reduced to low or moderate levels. Secondly, due to the small number of publications that reported the association between individual CVH components and dementia risk, our study did not have enough power to detect the possible mildly-to-moderately strong associations. Thirdly, in most of the included studies, the time-varying CVH levels were not taken into consideration. In addition, some studies did not include full CVH components due to lack of data sources, for instance, diet or physical activity. [9,16,23] However, the associations between CVH scores and risk of dementia did not differ significantly after the exclusion of studies without full metrics.

Conclusions
Our findings provide evidence that maintaining a favourable level of CVH score, either in mid-or late-life, would substantially reduce the risk of dementia among older adults.

Competing interest
The authors declare no conflict of interest. J o u r n a l P r e -p r o o f