The NHANES database employs a complex, multistage probability sampling design to reflect the nutritional and health status of the U.S. population. This study was approved by the National Center for Health Statistics (NCHS) Institutional Review Board, and all participants provided written informed consent. Notably, the NHANES database is publicly accessible and does not require additional ethical or administrative approval for use.

Our study meticulously adheres to the STROBE guidelines (16). Sleep health is a key factor in assessing the LE8, and applicable data in NHANES are only available after 2005. Therefore, our study incorporated data from a total of seven survey cycles between 2005 and 2018. A total of 70,190 individuals participated in the survey. Among them, 30,441 were under the age of 20, 4,422 lacked information on dietary live microbe intake, 8,322 lacked information on the LE8, 14,514 lacked valid sample weights, and 3,983 lacked information on other covariates. As shown in Figure 1, after screening, a total of 10,531 participants were finally included in the study.

In the NHANES dataset, the dietary intake of participants during two distinct 24-h periods is thoroughly recorded through face-to-face interviews and telephone follow-ups. The NCHS uses dietary nutritional data from the United States Department of Agriculture (USDA) to accurately assess the energy and nutritional content of each food and beverage. The estimated quantity of live microbes (per gram) in 9,388 foods across 48 subgroups in the NHANES database was determined by a team of four experts, including Marco. During these assessments, the experts relied on reported values in professional literature and authoritative reviews, or made inferences based on the known effects of food processing on microbial vitality. Foods were categorized into low (Lo; < 10⁴ CFUs/g), medium (Med; 10⁴-10⁷ CFU/g), and high (Hi; >10⁷ CFU/g) classes based on their live microbe content. Any discordance encountered during the assessment of live microbe content was resolved through external consultations and discussions.

Each of the eight indicators is scored according to publicly available official calculation methodologies, with the specific calculation methods detailed in the Supplementary Table 1. Within the eight CVH indicators, each indicator is scored on a scale ranging from 0 to 100. The overall LE8 score is derived from the unweighted average of these indicators. Based on the final score, CVH is categorized into three groups: low (0–49 points), medium (50–79 points), and high (80–100 points) (17).

The dietary indicator was assessed using the 2015 Healthy Eating Index (HEI) (18), and our study included participants with 2 days of dietary data (those with only 1 day of dietary data were excluded). Subsequently, dietary information was combined with data from the USDA to calculate the HEI (19), details are provided in the Supplementary Table 2. Self-reported survey questionnaires were used to collect information on participants' PA, smoking, sleep, history of diabetes mellitus (DM), and medication use. Height, weight, and BP were measured in mobile examination centers, and BMI was calculated. Blood samples were collected and sent to a central laboratory for testing of blood lipids, blood glucose, and glycated hemoglobin. Non-high-density lipoprotein cholesterol was calculated as total cholesterol minus high-density lipoprotein cholesterol.

In light of prior studies and clinical experiences (12, 20), confounding factors that could influence the relationship between dietary microbes and CVH were taken into consideration. These factors included age, gender and race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, other Hispanic, and other race—including multi-racial). Educational level was also considered, categorized as less than high school, high school graduate/GED or equivalent, and college graduate or above. Marital status was classified into married/living with partner, never married, and widowed/divorced/separated. Economic status was assessed using the poverty income ratio (PIR) with categories of < 1.30, 1.3–3.49, and 3.50 or higher. Health insurance status was noted as either Yes or No. Alcohol consumption included never drinking, former drinking, mild drinking, moderate drinking and heavy drinking. Obesity status includes normal weight, overweight and obesity. Medical history encompassed a history of CVD (Yes, No), DM (Yes, No), hypertension (HTN; Yes, No), and hyperlipidemia (HLD; Yes, No). The details of these variable assessments can be found in Supplementary Table 3. Additionally, we assessed the participants' average daily intake of energy, protein, carbohydrates, dietary fiber, and fat.

To ensure national representation, this study considered sample weights in all analyses. Participants were grouped into Low, Medium, and High dietary live microbe categories, where each group corresponds to the consumption level of foods within microbial content ranges, with the Low group consuming mostly low microbe-content foods, the Medium group consuming a balance excluding high microbe-content foods, and the High group predominantly eating foods rich in live microbes. We conducted one-way ANOVA (for continuous variables) and Chi-square tests (for categorical variables) to assess the baseline characteristics of participants among different groups. Continuous variables are represented by mean ± standard deviation, while categorical variables are expressed as the number of cases (n) and weighted percentages (%).

Weighted univariate and multivariate linear regression models are used to explore the relationship between dietary live microbes and LE8. Additionally, health behaviors and health factors were explored separately. To delve deeper into the connection between them, we consider a LE8 score below 50 as an indicator of high cardiovascular health risk (HCVHR). Weighted univariate and multivariate logistic regression models are used to explore the relationship between live microbes intake and HCVHR. In addition to the univariate model, we also developed three multivariate regression models. Model 1 was adjusted for age, gender, race/ethnicity, and education level. Model 2, building upon model 1, further incorporated adjustments for PIR, health insurance, marital status, and alcohol consumption. Finally, Model 3, expanding upon Model 2, additionally adjusted for dietary components including energy, protein, carbohydrate, fat, and fiber intake.

We also conducted subgroup analyses and interaction analyses based on age, gender, race/ethnicity, education level, PIR, health insurance, and marital status. Moreover, in each group with low, medium, and high live microbes intake, we employed the restricted cubic spline (RCS) analysis method to investigate the relationship between the quantity of food consumed (grams) and LE8. Finally, for the purpose of sensitivity analysis, we excluded individuals with a history of CVD, DM, HLD, and HTN. Statistical analyses were conducted using R software, version 4.3.1. A two-tailed P-value of < 0.05 was considered to indicate statistical significance.

Participants were categorized into three groups based on their intake levels of dietary live microbes: low, medium, and high. As shown in Table 1, the average age of the population included in the study is 47.59 years, with a slightly higher proportion of females than males. The predominant ethnicity is non-Hispanic White, and the majority of participants have a college education or higher. Most participants have health insurance, a habit of drinking alcohol, and are either married or cohabiting with a partner. In terms of weight status, the majority are obese, with 8.81% of participants suffering from CVD, 13.87% having DM, 37.44% experiencing HTN, and over 70% suffering from HLD. The assessment of CVH shows that 66.34% of the participants are at a moderate level. Among the three groups of participants, there were significant differences in all aspects except for daily carbohydrate intake, prevalence of HTN, and HLD. We also compiled the baseline characteristics of participants categorized by low, medium, and high levels of CVH, as shown in Supplementary Table 4.

Table 2 shows the association between different dietary live microbes intake and LE8. In the crude model, compared to the low group (reference group), the medium and high groups showed significantly higher LE8 scores, with estimated values (β) of 4.77 (95% CI: 3.89–5.65) and 7.08 (95% CI: 6.05–8.11), respectively, both reaching statistical significance (P < 0.0001). In the multivariate-adjusted models, the association between dietary live microbe groups and LE8 remained significant. In Model 1, the β coefficients for the medium and high groups were estimated at 4.54 (95% CI: 3.76–5.31) and 5.66 (95% CI: 4.65–6.67), respectively. The estimates slightly decreased in Model 2, with the medium group at 3.98 (95% CI: 3.23–4.74) and the high group at 4.94 (95% CI: 3.95–5.92). In the final Model 3, the β coefficients for the medium and high dietary groups further reduced to 2.75 (95% CI: 2.01–3.50) and 3.89 (95% CI: 2.97–4.81), while the associations remained significant (P < 0.0001).

As previously mentioned, LE8 primarily comprises two components: health factors and health behaviors. We further explored the relationship between these components and the intake of live microbes. As shown in Supplementary Table 4, across all analysis models, groups with higher intake of live microbes were generally associated with higher scores in health factors and health behaviors, and this association remained significant in the multivariate-adjusted models. It is noteworthy that in the crude model, there was no significant difference between the medium and low dietary live microbe groups.

To further explore the relationship between different levels of live microbes intake and the risk of cardiovascular health, we employed logistic regression analysis. The study builds on previous linear regression analyses and defines individuals with an LE8 score below 50 as HCVHR. As shown in Table 3, compared to low group, moderate and high groups are associated with a significantly reduced risk of cardiovascular health. In the crude model, without adjusting for any confounding factors, moderate live microbes intake was associated with a reduced risk of cardiovascular health (OR = 0.58, 95% CI: 0.49–0.69), and the reduction was more significant for high live microbes intake (OR = 0.43, 95% CI: 0.34–0.55), with a P-value for the trend < 0.0001. This trend remained consistent after progressively adjusting for confounding factors. In the final model, taking into account a variety of potential confounders, moderate and high groups were still significantly associated with lower cardiovascular health risk. Specifically, the odds ratio for moderate live microbes intake was 0.73 (95% CI: 0.61–0.89), and for high live microbes intake, it was 0.65 (95% CI: 0.50–0.84).

To investigate the consistent link between dietary live microbes intake and CVH across various populations, subgroup analyses were conducted. The analysis covered various population subgroups, including age, gender, race, education level, household income ratio, health insurance status, and marital status. As shown in Figure 2, the results indicate a significant positive correlation between the intake of live microbes and LE8, which was validated in the vast majority of subgroups. However, there were some non-significant results in specific subgroups, such as non-Hispanic blacks, suggesting caution when generalizing these findings to a broader population. Trend tests showed that as the intake of live microbes increased (from moderate to high amounts), their positive impact on CVH tended to increase across all subgroups. Interaction analysis revealed that the relationship between dietary live microbe levels and LE8 was significantly influenced by gender, race and education level. The impact of gender was particularly significant (P = 0.001), suggesting that men and women might respond differently to the intake of live microbes. Additionally, subgroup analyses on the relationship between dietary live microbes levels and HCVHR were also conducted, with the results largely consistent (Supplementary Figure 1).

To verify the reliability of results in a healthy population, we also excluded individuals with a history of cardiovascular and related diseases. We conducted five sensitivity analyses, each excluding populations with a history of CVD, DM, HLD, HTN, and any combination of these conditions. Multivariable regression analyses were carried out under the fully adjusted model (Model 3), with the results presented in Table 4. Across the five different sets of exclusion criteria, the positive correlation between moderate to high intake of dietary live microbes and LE8 was consistently significant. Excluding individuals with CVD history (n = 9,368), both moderate (β = 2.64, 95% CI: 1.84–3.45, P < 0.0001) and high (β = 3.88, 95% CI: 2.90–4.86, P < 0.0001) intake groups showed significant positive associations. Excluding DM history (n = 8,551), results were similar (Moderate: β = 2.73, High: β = 3.66; both P < 0.0001). Excluding HLD history (n = 2,948), high intake remained significantly associated (β = 2.50, 95% CI: 1.06–3.95, P < 0.0001), while moderate intake showed no significant association. Excluding HTN history (n = 6,059), both intake groups showed significant associations (Moderate: β = 2.85, High: β = 3.81; both P < 0.0001). Excluding all above conditions (n = 2,050), significant associations persisted (Moderate: β = 2.04, High: β = 3.06; both P < 0.0001).

We employed RCS analysis to explore the dose-response relationship between food intake at three different levels of live microbes and CVH. The observations show that in the low live microbe group (Figure 3A), there is a significant negative correlation between the intake of food and LE8 (P < 0.001), and there is no significant non-linear trend (non-linear P < 0.36). In the medium live microbe group (Figure 3B), the analysis revealed a significant inverted “U” shaped relationship (P < 0.0001, non-linear P < 0.0001), meaning that the intake of food is positively correlated with LE8 before reaching 326.04 g, and the relationship turns negative after reaching this point. In the high live microbe intake group (Figure 3C), the intake is positively correlated with LE8 (P < 0.001), but this correlation is not curvilinear (non-linear P = 0.12).