Embracing Wellness in Later Years: A Shield Against Cognitive Decline for All Genetic Backgrounds

Research design and demographic

This research was conducted within CLHLS, an ongoing prospective population-based cohort study implemented in numerous regions across 23 provinces in China. Its intent is to explore the factors that contribute to healthy aging among older Chinese adults. Further insights regarding the research design have been previously detailed⁴². In summary, CLHLS was initiated in 1998, followed by subsequent surveys and recruitment of new participants in 2000, 2002, 2005, 2008–2009, 2011–2012, 2014, and 2018. The CLHLS obtained approval from the Ethics Committee of Peking University (IRB00001052-13074). Written informed consent was secured from all participants or their legal representatives during face-to-face interviews.

Out of the total number of participants recruited over 8 waves of CLHLS, 37,456 completed both baseline and follow-up assessments. We further excluded participants who had incomplete information on lifestyle scores at baseline (n = 671), those who were younger than 65 years old (n = 297), individuals exhibiting cognitive impairment at baseline (n = 9649), and those lacking MMSE data at baseline or in any follow-up (n = 8028). Consequently, 18,811 participants were included in the current analysis, and among these, 6301 participants with genotyping data were incorporated for the joint association examination concerning healthy lifestyle and genetic risk analysis (Supplementary Fig. 6).

Lifestyle components

Information regarding demographic characteristics, socioeconomic status, lifestyle components, physical health, and psychological wellness was collected by well-trained interviewers through a standardized questionnaire at baseline. A healthy lifestyle score was derived based on four alterable lifestyle components linked to cognitive function, as supported by prior research^{30,38,39,43,44}. These include current non-smoking status, never drinking alcohol, engaging in physical activity, and maintaining a healthy dietary regimen. Supplemental methods and Supplementary Table 6 provide additional specifics on the particular questions asked and how the healthy lifestyle scores were formulated.

Specifically, current non-smoking referred to never smoking or previous smoking as per prior studies¹⁵. Alcohol drinking status was categorized into non-drinker, current drinker, and former drinker. The classification of never drinking was considered a positive lifestyle factor³⁰. With respect to physical activity, participants reported the frequency of their involvement in regular exercises, household chores, personal outdoor activities, gardening, caring for domestic animals/pets, reading, playing games like cards/mahjong, watching television/listening to the radio, and participating in social events⁴⁵. The frequency of “almost every day”, “occasionally”, and “rarely or never” was designated scores of 2, 1, or 0 respectively. The total physical activity score was computed as the sum of 9 activities, ranging from 0 to 18; a higher score indicated a greater level of physical engagement. An ideal physical activity score was defined as being in the top 40% of the cohort’s distribution⁴⁶. Dietary intake was evaluated using a standardized food frequency questionnaire with acceptable reliability and validity⁷, encompassing 9 commonly consumed food categories in the Chinese diet: fresh vegetables, fresh fruit, legumes, meat, eggs, fish and seafood, salty vegetables, tea, and garlic^7,47. The aging process among older adults, particularly the very elderly, often leads to insufficient daily protein intake and an imbalance in protein synthesis and degradation. Therefore, the consumption of typical protein-rich foods such as legumes, meat, eggs, and fish, known for their significant beneficial impacts on mortality, was also regarded as a healthy lifestyle in this investigation^37,48. A total diet score was calculated similarly to physical activity scores. Dietary intake was considered ideal if it fell within the top 40% of the cohort distribution, as evidenced by previous research⁴⁶. All component scores were aggregated to derive a healthy lifestyle score that ranges from 0 to 4, with a higher score indicating a more favorable lifestyle and categorized further as “unfavorable” (fewer than three healthy lifestyle components), “intermediate” (three healthy lifestyle components), and “favorable” (all four healthy lifestyle components).

Control variables

Control variables were chosen based on earlier studies and available cohort measures^15,49. The Supplementary Information outlines the specifics of the covariates described in the questionnaires, including age, gender, educational background, residential area, marital status, employment, income sources, self-reported health status, optimism levels, and history of significant chronic illnesses. Optimism was assessed utilizing seven items (score range from 0–7, with higher scores indicating greater levels of optimism)⁵⁰. The presence of major chronic conditions, such as cardiovascular ailments, diabetes mellitus, hypertension, respiratory diseases, digestive disorders, or cancer, was self-reported.

Cognitive abilities

Cognitive abilities were evaluated through an adapted Chinese version of the MMSE at baseline and during each follow-up survey according to the standardized protocol²⁹. The MMSE comprised 24 items spanning six cognitive dimensions: orientation, attention and computation, visual construction, language skills, naming, and memory capabilities (Supplementary Table 7)²⁹. A score of zero was allotted for incorrect or unknown responses, and one point was awarded for correct answers. All items carried equal weights, resulting in a maximum total score of 30. Cognitive scores along with their individual dimensions were converted from raw scores to Z scores based on means and standard deviations at baseline¹⁵. A positive Z score signifies superior cognitive functioning compared to the average population score, while a negative score reflects poorer cognitive performance. Higher cognitive scores denote enhanced performance. Cognitive scores were computed for each cycle for all participants in the study, and based on the cognitive scores during follow-up, the rate of cognitive decline was established. Given the significant number of participants lacking formal education, cognitive impairment was defined as an overall cognitive score of <18 in line with previous findings²⁹.

Genetic testing and risk score formulation

In line with the findings of the CLHLS genome-wide association study on longevity, a replication study was undertaken involving 13,228 individuals utilizing a meticulously designed and specialized chip focused on 27,656 single nucleotide polymorphisms (SNPs) previously linked to longevity and related characteristics⁵¹. Among these selected SNPs, 3966 were associated with prevalent age-related diseases, including Alzheimer’s disease, cardiovascular conditions, type 2 diabetes, cancer, and immune-mediated disorders. Comprehensive details on SNP selection and the genotyping technique employed in the CLHLS study have been published earlier^51,52. In the current study, the polygenic risk score was derived based on the count of risk alleles from 34 genetic variants that elevate the risk for Alzheimer’s disease, as previously documented^51,53.

Information concerning the chosen SNPs is presented in Supplementary Table 8. Each SNP was designated as 0, 1, or 2 based on the count of risk alleles. The weight coefficient for every SNP was disclosed in earlier studies^51,52. The genetic risk score was constructed as the total of risk alleles at each location multiplied by their corresponding weight coefficient. This score was divided into two categories based on the median: low and high genetic risk groups.

Statistical evaluations

Participant baseline characteristics were reported as mean values (standard deviation) for continuous variables and counts (percentage) for categorical variables. The marginal mean estimates for cognitive scores and its individual components based on lifestyle categories were assessed through multiple linear regression models, adjusted for baseline covariates.

Linear mixed-effects models were utilized to examine the relationship between a healthy lifestyle and longitudinal changes in cognitive function as well as in each cognitive domain individually. To account for the within-individual associations across repeated measurements, all models incorporated time since the baseline along with random intercepts and random slopes of time. The analysis considered age (years, continuous), gender (female/male), entry time, educational level (years of schooling completed <1 year, 1–6 years, or >6 years), place of residence (urban/rural), current marital condition (married, not married), occupation (agriculture/forestry/husbandry/fishery, commercial, service, industrial worker/self-employed, professional/governmental/managerial roles, or homemaker/never worked/other), income source (independent, dependent), and baseline cognitive score (continuous). The same approach was employed to scrutinize the association between genetic risk and the rate of cognitive decline. Furthermore, we evaluated the relationship between a healthy lifestyle and cognitive decline stratified by genetic risk, alongside assessing the interaction between a healthy lifestyle and genetic risk. Additionally, we analyzed the joint influences of both healthy lifestyle and genetic risk on the rate of cognitive decline. Participants were sorted into six categories based on lifestyle group (unfavorable, intermediate, and favorable) and genetic risk (low, high), taking the combination of high genetic risk and an unfavorable lifestyle group as the reference.

Regarding the time-on-study as the timescale, person-years were calculated for each participant from baseline until the occurrence of cognitive impairment, death, or the conclusion of follow-up, whichever occurred first. Cox proportional hazard models were employed to evaluate the associations between lifestyle and genetic factors with the incidence of cognitive impairment, adjusting for age, gender, entry time, educational level, place of residence, current marital status, occupation, and income source. The proportional hazards assumption was assessed through the Schoenfeld residuals method, with no violations detected (P > 0.05). Additionally, we examined the impact of death as a competing risk for cognitive impairment through competing risk analyses.

To verify the robustness of our findings, we conducted several sensitivity analyses. Initially, in addition to primary adjustments, we further controlled for self-reported health status (yes/no), optimism level (continuous), and history of chronic disease status (yes/no) within the multivariate mixed model, as these factors were recognized as risk determinants for cognitive function¹⁵. Second, to investigate the likelihood of reverse causation resulting from impaired cognitive function, which could skew the accuracy of reported lifestyle behaviors, participants with baseline MMSE scores below the 10th percentile were excluded³⁴. Moreover, changes in cognitive scores were limited at the 0.5th and 99.5th percentiles to reduce the impact of outliers. Third, to mitigate the negative effects of past smoking, we formulated a new healthy lifestyle score considering “never smoking” as a healthy lifestyle attribute. Fourth, due to the frequent occurrence of missing values for specific items on the MMSE test, such as visual construction, the analysis was limited to participants who completed all MMSE items⁴¹. Fifth, considering that lifestyle scores might vary during follow-up, the latent class trajectory model was used to distinguish between different lifestyle score trajectories, and subsequently evaluate the impact of these lifestyle trajectory groups on cognitive decline. Lastly, to account for the non-linear relationship, quadratic terms of time were incorporated into the multivariate model¹⁵. Two-sided P < 0.05 was regarded as statistically significant, barring separate analyses for individual cognitive domains, where the Bonferroni correction was applied to adjust for multiple testing (P < 0.008 considered significant [=0.05/6]). All data analyses were conducted using SAS V.9.4 (SAS Institute, Cary, NC, USA), along with R V.4.3.1 (R Foundation for Statistical Computing, Vienna, Austria).

Reporting summary

Additional information on the research design can be found in the Nature Portfolio Reporting Summary linked to this article.