We used data from the Multi-Ethnic Cohort Phase 1 (MEC1), a population-based study of adult residents aged 21 years and older in Singapore that includes 3 major ethnic groups (Chinese, Indian, and Malay). MEC1 baseline data (T1) were collected between 2004 and 2010. Individuals with a history of cancer, CVD, stroke, renal failure, or serious mental illness at recruitment were excluded from cohort enrollment. The MEC1 baseline dataset was formed by combining 2 existing population-based studies conducted between 2004 and 2007: the Singapore Prospective Study Program and the Singapore Cardiovascular Cohort Study, with additional recruitment of participants from 2007 to 2010 (Singapore Consortium of Cohort Studies–Multi-Ethnic Cohort [SCCS-MEC]) through household visits and community outreach to achieve oversampling of Malay and Indian individuals in Singapore. All participants provided written informed consent, including consent for linkage with national disease registries. Detailed information on the cohort has been published previously [22].

Of the initial 12,988 participants, 12,384 (95.4%) remained after excluding those not from the 3 major ethnic groups (n=33), with missing interview data (n=42), or with preexisting CVD (n=328) or cancer (n=201). After further exclusions of those with missing lifestyle data (n=2376) or covariate data (n=19) and implausible energy intake (≤500 or ≥6000 kcal per day; n=69), 9920 participants were included in the final analysis (Multimedia Appendix 1). The median follow-up duration ranged from 14.1 years for cancer to 14.3 years for all-cause mortality, with 14.2 years for both AMI and stroke. Mental well-being data were available for 4791 participants from a subset of participants (Singapore Prospective Study Program and Singapore Cardiovascular Cohort Study) who completed the 36-item Short Form Health Survey version 2 (SF-36v2) emotional well-being items at baseline; therefore, analyses incorporating mental well-being were restricted to this subset. In the subset with mental well-being data, the median follow-up ranged from 16.1 years (AMI, stroke, and cancer) to 16.2 years (all-cause mortality).

Six lifestyle factors were assessed at baseline: BMI, cigarette smoking, physical activity, diet quality, and screen time, with mental well-being available in a subset of the participants. BMI was calculated from measured height and weight, while the other factors were self-reported via an interviewer-administered questionnaire [22]. Each factor was categorized into 3 levels and assigned 0 to 2 points, with higher scores representing healthier behaviors. Three levels were chosen to balance the need to capture meaningful variation with the need to maintain sufficient sample size within each category.

BMI was classified using established Asian cutoffs, defined as less than 23.0 kg/m² (2 points), 23.0 to 24.9 kg/m² (1 point), and 25.0 kg/m² or higher (0 points) [23,24]. Smoking status was categorized as “never” (2 points), “former” (1 point), or “current” (0 points). Leisure-time physical activity was assessed using a physical activity questionnaire [25] and expressed in metabolic equivalent of task (MET) hours per week. Participants were categorized into 7.5 or more MET hours per week (2 points), 0 to less than 7.5 MET hours per week (1 point), and none (0 points), consistent with the World Health Organization recommendation of 150 minutes or more of moderate-intensity physical activity per week [26]. Dietary intake was assessed using the validated semiquantitative food frequency questionnaire (FFQ) [27], and diet quality was evaluated using the Alternative Healthy Eating Index (AHEI) 2010 [19], with higher scores indicating better diet quality. Participants were categorized into tertiles: highest (2 points), intermediate (1 point), and lowest (0 points). Screen time was assessed through the amount of time spent watching television or playing computer or handheld video games daily. Participants were categorized into less than 2 hours per day (2 points), 2 to 4 hours per day (1 point), and more than 4 hours per day (0 points). In a subset of the population, mental well-being was assessed using the SF-36v2 emotional well-being subscale derived from 5 items assessing nervousness, low mood (being downhearted or depressed), calmness, and happiness over the previous 4 weeks, with higher scores indicating better mental well-being [28]. Scores were transformed into a norm-based T-score (mean 50, SD 10) following the SF-36v2 scoring manual [28]. Participants were categorized into tertiles: highest (2 points), intermediate (1 point), and lowest (0 points). As mental well-being data were available only for a subset of participants (n=4791), the main analyses were conducted using a 5-factor score in the entire cohort, while mental well-being was incorporated in additional analyses in the smaller subset.

The primary lifestyle score was calculated by summing all 5 factors, with possible scores ranging from 0 to 10. Participants were further grouped into 4 categories: score of 0 to 3 (least favorable lifestyle profile), score of 4 or 5, score of 6 or 7, and score of 8 to 10 (most favorable lifestyle profile; Table 1). In the subset with mental well-being data, an extended 6-factor lifestyle score was calculated by additionally including mental well-being, contributing scores from 0 to 2 to the previous 5-factor lifestyle score. This extended 6-factor score had a wider possible range from 0 to 12 and a different observed distribution. Participants were grouped separately into 4 ordered categories: scores of 1 to 4, 5 and 6, 7 and 8, and 9 to 12.

Incident cases of AMI, stroke, all-site cancer, and deaths were ascertained through linkage with national disease registries maintained by the National Registry of Diseases Office, including the Singapore Myocardial Infarction Registry, Singapore Stroke Registry, Singapore Cancer Registry, and Singapore Death Registry.

Follow-up time was calculated from the date of baseline interview to the date of first occurrence of the outcome of interest, death, or end of follow-up (December 31, 2021), whichever came first.

Covariates included age, sex, ethnicity, educational level, income, total energy intake, family history of cancer or heart disease, and menopausal status in women. Age was categorized as 35 years or younger, 36 to 45 years, 46 to 55 years, and 56 years or older. Sex was defined as male or female. Ethnicity was classified as Chinese, Malay, or Indian. Educational attainment was grouped into 4 categories: primary or lower, secondary, postsecondary, and university or higher. Household income was categorized into predefined levels, with a separate category created for participants who refused to answer, did not know, or had missing information.

Total energy intake (kcal per day) was derived from the FFQ administered at baseline and modeled as a continuous variable. Family history of cancer or heart disease was coded as “yes” or “no” based on self-report. Menopausal status was classified as premenopausal or postmenopausal among women, and male participants were coded as a separate category to retain all participants in the model.

Ethics approval was obtained from the Institutional Review Board of National University of Singapore (NUS-IRB-2022-27). All participants provided written informed consent, including consent for linkage with national disease registries. Research data used for this secondary analysis were deidentified and key coded, with access restricted to authorized members of the research team.

Descriptive statistics were presented according to the categories of the lifestyle score. For continuous variables, P values were estimated by modeling the median value of each lifestyle score category as a continuous variable in linear regression. For categorical variables, P values were calculated using the Cochran-Armitage trend test. To assess potential selection bias related to missing lifestyle data, baseline characteristics were compared between participants included in the final analytic sample and those excluded because of missing lifestyle data.

Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% CIs for associations between the lifestyle score and each outcome, including incident AMI, incident stroke, incident cancer, and all-cause mortality. The proportionality assumption was assessed using Schoenfeld residuals; no violations were detected. Tests for linear trend were performed by modeling lifestyle score categories as a continuous variable in the regression model. All models were adjusted for age, sex, ethnicity, educational level, income, total energy intake, family history of cancer or heart disease, and menopausal status.

Additional analyses were conducted in the subset of participants with available mental well-being data by incorporating mental well-being as a sixth component of the lifestyle score. To facilitate comparison between the primary and extended lifestyle scores, both scores were also modeled as continuous variables, with HRs estimated per 1-point increase. Ethnicity-stratified analyses were performed, and multiplicative interaction terms between the lifestyle score and ethnicity were included in the Cox models to assess potential effect modification by ethnicity. To assess whether the associations were driven by any single component of the lifestyle score, leave-one-out analyses were performed by recalculating the lifestyle score after omitting one component at a time.

In this multiethnic Asian cohort with over 14 years of follow-up, adherence to a healthy lifestyle as measured using a composite score was associated with substantially lower risks of all-cause mortality, AMI, and stroke. Participants with the least favorable lifestyle profiles had twice the risk of AMI and all-cause mortality and over 3 times the risk of stroke compared with those with the most favorable profiles. Results were consistent when mental well-being was incorporated in the lifestyle score. No significant associations were observed for overall cancer incidence. These findings highlight the importance of healthy lifestyle behaviors for cardiovascular health and longevity in Asian populations.

Our findings are consistent with those of previous cohort studies reporting that composite healthy lifestyle scores are inversely associated with mortality and CVD outcomes in both Western and Asian populations [29]. Large cohorts from Europe and the United States have demonstrated that combinations of nonsmoking, healthy body weight, regular physical activity, light-to-moderate alcohol consumption, and high-quality diet are associated with reductions in all-cause and cardiovascular mortality [30,31]. Comparable associations have been documented in East Asian cohorts, including China, Taiwan, South Korea, and Japan, where higher lifestyle scores incorporating BMI, smoking, physical activity, diet, and alcohol consumption were observed [32-35]. In our study, participants with the least favorable lifestyle profiles had approximately 2.0 times the risk of all-cause mortality, 2.1 times the risk of AMI, and 3.2 times the risk of stroke, broadly consistent with meta-analyses of prospective cohort studies reporting 2.2 times higher all-cause mortality and 2.2 to 3.4 times higher risk of various cardiovascular outcomes [4]. Our study extends this literature to a multiethnic Southeast Asian population and incorporates screen-based sedentary behavior, assessed as television viewing and computer or handheld video game use, into the composite lifestyle score.

A notable contribution of this study is the integration of screen time into the composite lifestyle score. Sedentary behavior, including prolonged screen time, has been increasingly recognized as an independent risk factor for cardiometabolic diseases. Epidemiological evidence across multiple cohorts has shown that increased leisure screen time is associated with greater cardiometabolic risk in both adolescents and adults, as well as with elevated risks of coronary artery disease and mortality [21,36,37]. Consistent with these findings, our leave-one-out analysis showed that removing screen time from the score substantially attenuated the association with stroke, suggesting that screen time may be particularly relevant for cerebrovascular risk. Potential mechanisms include impaired circulation, central adiposity, dysregulated glucose metabolism, elevated blood pressure, and inflammation, all of which are closely linked to ischemic stroke [38-41].

Although there is accumulating evidence suggesting that psychological distress and low mental well-being are associated with higher risks of mortality and CVDs [42-44], mental well-being has rarely been incorporated into lifestyle indexes. While mental well-being is not a lifestyle behavior per se, it is largely shaped by them, and including them in the combined lifestyle score allows for a broader assessment of health. In our study, mental well-being was assessed using the SF-36v2 mental health subscale, which has been used to predict mortality and CVD outcomes in prior prospective cohorts [45]. In the subgroup with available data, adding mental well-being to the score did not substantially change the overall patterns of association. The estimates remained similar for all-cause mortality and slightly attenuated for AMI and stroke but were statistically significant despite the smaller sample size. These findings suggest that mental well-being represents a complementary domain of lifestyle assessment, capturing psychosocial aspects of health that may support overall disease prevention and well-being.

The composite lifestyle score was designed to capture the cumulative influence of multiple lifestyle-related risk factors that commonly co-occur in real-world settings as compared to considering each lifestyle component independently. Smoking, poor diet quality, lower physical activity, and sedentary behaviors may cluster within broader unhealthy lifestyle patterns, and the score provides a practical summary measure of this overall lifestyle risk. In sensitivity analyses excluding smoking, the associations with all-cause mortality and stroke were attenuated and no longer statistically significant, suggesting that smoking was a particularly influential component for these outcomes. This pattern aligns with extensive evidence showing that smoking promotes endothelial dysfunction, systemic inflammation, oxidative stress, and thrombotic pathways that accelerate atherosclerosis and increase risks of ischemic stroke and premature death [46-49]. These findings highlight the importance of smoking prevention and cessation for reducing the risks of premature mortality and stroke.

There was no significant evidence of interaction by ethnicity for the associations with all-cause mortality, AMI, and stroke. Although an interaction was observed for overall cancer incidence, the ethnicity-stratified analyses did not show clear dose-response trends. This finding should be interpreted cautiously. The null association between the lifestyle score and overall cancer incidence should also be interpreted with caution. Cancer encompasses multiple diseases with distinct etiologies and typically long latency periods, and aggregating all cancer types into a single outcome may obscure lifestyle-related associations that differ by cancer site. Furthermore, the relatively young baseline age of our cohort and comparatively short follow-up for cancer outcomes resulted in a limited number of cancer events during follow-up, reducing the power to detect associations.

This study has several limitations. First, the lifestyle score was constructed as an unweighted index summing all lifestyle components equally and assuming each factor had equal importance. Previous studies have examined both weighted and unweighted indexes, showing that both effectively predict CVD risk [50]. The unweighted index is outcome agnostic, which is more practical and can be applied to a broader range of health outcomes. Second, lifestyle behaviors were primarily self-reported, which may introduce recall error or reporting bias. However, such errors are likely nondifferential as outcomes were assessed prospectively. The physical activity questionnaire and FFQ have been validated in independent samples of the local population with similar characteristics [22,27,51]. In addition, differences in several baseline characteristics between included participants and those excluded because of missing lifestyle data may have introduced selection bias. Third, lifestyle behaviors and covariates were measured at baseline only, and changes in these behaviors during follow-up may have introduced nondifferential misclassification that biased the association toward the null. Fourth, alcohol consumption was not included as a separate lifestyle component as it was already incorporated into the AHEI 2010 score to avoid double counting. Fifth, although we adjusted for multiple confounders, our findings could still be influenced by residual confounding and unmeasured factors. While the median follow-up of approximately 14 years allowed for assessment of long-term outcomes, longer follow-up may be needed to further clarify associations, particularly for cancer outcomes with longer latency. Sixth, cancer was analyzed as a single combined end point. Although overall cancer incidence provides a broad assessment of cancer burden alongside other major chronic disease outcomes, cancer comprises a heterogeneous group of diseases with distinct etiologies, and combining all incident cancers may obscure site-specific associations with lifestyle factors. Therefore, findings for overall cancer incidence should be interpreted cautiously. Finally, as this study was conducted in a multiethnic Asian cohort, the findings may not be generalizable to other populations with different sociodemographic or cultural contexts.

Our findings highlight the importance of combined lifestyle behaviors, including screen-based sedentary behavior, for cardiovascular and overall health. Incorporating mental well-being yielded similar results, suggesting that a broader conceptualization of lifestyle may be relevant for chronic disease prevention. These results highlight opportunities for comprehensive lifestyle interventions and public health strategies to reduce CVD and premature mortality in Asian populations.