Alpha-Linolenic Acid and Mortality Among Adults With Type 2 Diabetes: Findings From Two National Cohorts

2.1 Study Population

Our study was based on NHANES for the US population and CHNS for the Chinese population, respectively. NHANES is a large-scale survey designed to assess the health and nutritional status, utilizing representative samples of the US civilian population. Meanwhile, CHNS is an ongoing household-based longitudinal cohort study in China, employing a multistage random-cluster method to collect a sample of over 30 000 individuals from 9 provinces. The data of NHANES were reviewed and approved by the National Center for Health Statistics Research Ethics Review Board, and informed consent was obtained from all participants. Detailed methodological descriptions of both surveys have been published previously [16, 17].

In this study, individuals with T2D from continuous cycles of NHANES (1999–2018) and CHNS (1997–2011) were included. Individuals were diagnosed with T2D if they met at least one of the following criteria: (1) a self-reported diagnosis of diabetes by a specialist; (2) fasting blood glucose ≥ 7.0 mmol/L, hemoglobin A1c (HbA1c) ≥ 6.5%, or 2-h oral glucose tolerance test (2 h-OGTT) ≥ 11.1 mmol/L; (3) using hypoglycemic medicine or taking insulin. For the current analysis, participants with T2D who self-reported as pregnant, reported extreme energy intake (total energy < 500 or > 3500 kcal/day for women; < 800 or > 4200 kcal/day for men), had missing or duplicated dietary records, or had missing or illogical information about death were excluded. Overall, 9603 participants aged 20 years and above were included in analyses (n = 7953 in NHANES 1999–2018 and n = 1650 in CHNS 1997–2011).

2.2 Measurement of Dietary Alpha-Linolenic Acid Intake

The nutritional assessment in NHANES (1999–2018) involved a 24-h dietary recall interview conducted in person by trained interviewers in a private room in the Mobile Examination Center (MEC). In the first two cycles, a single dietary interview was conducted on a randomly chosen day. From 2003 onward, a second interview was conducted by telephone 3 to 10 days after the initial interview. To represent long-term dietary habits and minimize individual differences, the mean dietary information was calculated in the cycles from 2003 to 2018. For participants enrolled in 1999–2002 or lacking a second dietary interview, only one 24-h recall was used. Daily aggregates of food energy and macronutrient components (fats, protein, and carbohydrates) from all foods were calculated using the United States Department of Agriculture's (USDA) Food and Nutrient Database for Dietary Studies (FNDDS), based on cycle-specific versions [18]. In CHNS (1997–2011), dietary intake was assessed using a three-day consecutive 24-h dietary recall and household food checklist to estimate the typical consumption of participants. Dietary ALA and other nutrient values were calculated based on updated versions of the Chinese Food Composition Table (FCT) for each round [19].

2.3 Ascertainment of Mortality

The follow-up period was calculated from baseline to the date of death or until 31 December 2019 for NHANES, and until 31 December 2015 for CHNS, whichever occurred first. Information on all-cause mortality and specific mortality was ascertained by linking to the National Death Index (NDI) through 31 December 2019 for NHANES. The cause of death was determined using the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) guidelines. Death due to CVD was defined as ICD-10 codes I00–I09, I11, I13, I20–I51, or I60–I69. In CHNS, information on mortality was ascertained according to the report collected from household members in each survey. If the death record of a participant was reported multiple times, we initially used the data from the first report. As cause-specific death information was only reported in the 1991 survey, there was no information available on cause-specific mortality in CHNS.

2.4 Assessment of Covariates

Several possible confounding variables related to demographics, social economics, health-related behavior, and history of diseases previously associated with mortality [6, 14, 20], were collected by standardized questionnaires or physical examinations at baseline, including age, sex, race/ethnicity, education levels, economic status, drinking and smoking habits, leisure-time physical activity, duration and medication for diabetes, as well as a history of hypertension, hypercholesterolemia, CVD, and cancer. Additionally, the family history of diabetes and CVD was also documented.

BMI was determined by dividing the weight in kilograms by the square of the height in meters (kg/m²) and then categorized into 4 groups for NHANES (< 18.5, 18.5–24.9, 25.0–29.9, or ≥ 30.0) and for CHNS (< 18.5, 18.5–23.9, 24.0–27.9, or ≥ 28.0) [21]. Education levels were grouped as less than high school graduate, high school graduate or General Equivalent Diploma (GED), and college or above. Economic status was classified based on the ratio of family income to poverty level for NHANES or household per capita annual income for CHNS. Drinkers were categorized as non-drinker, ever-drinker, low and moderate drinker, and heavy drinker based on consumption levels and frequency. For smoking status, individuals were divided into never smoker, ever smoker, and current smoker with further subdivision based on the number of cigarettes smoked per day. Leisure-time physical activity was defined as inactive group, insufficiently active group, and active group. The duration of diabetes was categorized into < 3, 3 to 10, and > 10 years. Individuals with hypertension/hypercholesterolemia should have self-reported high blood pressure/cholesterol levels or be taking prescription or prescribed medicine for treatment. Suffering from CVDs was defined as self-reported suffering from at least one of the following diseases: congestive heart failure, coronary heart disease, angina/angina pectoris, heart attack, and stroke. Participants in NHANES were classified into two groups based on whether they reported taking ALA supplements or not. They were also asked about their ALA supplement usage in the past 30 days, including the amount and duration of supplementation.

In addition, cardiometabolic markers such as fasting glucose, fasting insulin, HbA1c, triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and C-reactive protein (CRP) were measured at recruitment. The insulin resistance (HOMA2-IR) was determined based on fasting blood glucose (mmol/L) and insulin (pmol/L), calculated using the HOMA calculator developed by the University of Oxford, UK [22].

2.5 Statistical Analysis

Participants were categorized into three groups based on the tertiles of dietary ALA intake. Baseline characteristics were presented as weighted means (standard errors, SE) or proportions. P values for differences across ALA tertiles were estimated using ANOVA for continuous variables or chi-squared tests for categorical variables. The generalized linear model was used to examine the associations of ALA intake with cardiometabolic biomarkers at baseline, including fasting glucose, insulin, HOMA2-IR, HbA1c, triglyceride, TC, HDL-C, LDL-C, and CRP.

Multivariable Cox proportional hazards regression models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) to examine the associations between dietary ALA intake and all-cause mortality or CVD mortality (only for NHANES) among adults with T2D. In model 1, we adjusted for age, sex, and race/ethnicity. In model 2, we further adjusted for BMI, education level, economic status, drinking status, smoking status, leisure-time physical activity, history of hypertension, history of hypercholesterolemia, history of cancer, history of CVDs, family history of CVDs (only for NHANES), family history of diabetes (only for NHANES), duration of diabetes, diabetes medication use, ALA supplement (only for NHANES), and total energy intake. Linear trend tests were conducted using the median of each tertile of ALA intake as a continuous variable. Missing values were imputed with missing indicators for categorical variables. Analyses were conducted separately for each cohort, followed by a fixed-effects meta-analysis using the inverse-variance weighted method to combine the results of both cohorts. Heterogeneity was evaluated using the I² statistic.

Restricted cubic spline (RCS) regression with three knots (at the 25th, 50th, and 75th percentiles) was used to explore the dose–response relationship between ALA intake and mortality. Nonlinearity was assessed using likelihood ratio tests. Stratified analyses were conducted by median age of individuals in two cohorts (< 60, ≥ 60 years), sex (male or female), race/ethnicity (NHANES: non-Hispanic white, other; CHNS: Han, the minority), education level (less than high school graduate, or high school graduate or above), BMI (< median or ≥ median, NHANES: < 30 or ≥ 30; CHNS: < 25 or ≥ 25), economic status (NHANES: < 3.0, or ≥ 3.0; CHNS: ¥, < 5000, or ≥ 5000), drinking status (nondrinker or drinker), smoking status (never, ever or current), leisure-time physical activity (inactive or active), medication for diabetes (none or pills and/or insulin), diabetes duration (< 3 or ≥ 3 years), hypertension and/or hypercholesterolemia (yes or no), and CVD and/or cancer (yes or no). Further stratified analyses were conducted by median intake of marine n-3 PUFA for all-cause mortality and CVD mortality. The interaction effects were estimated by likelihood ratio tests. To ensure the robustness of the results, several sensitivity analyses were performed. First, we excluded adults with a history of CVD or cancer to mitigate the potential influences of fatal diseases on our results. Second, participants who died within the first 4 years of the follow-up period were excluded to reduce the potential reverse causation bias. Third, participants from 1999 to 2002 in NHANES were excluded. Fourth, adults with extreme BMI (< 18.5 or > 40 kg/m²) were excluded. Fifth, dietary ALA units were converted to the percentage of energy intake. Sixth, multiple imputation was used to address missing values. Finally, the E-value was calculated to estimate the robustness of the analysis of associations between ALA intake and mortality due to the unmeasured confounders [23]. Higher E-value indicates stronger evidence supporting the observed associations. The analyses were conducted using SAS (version 9.4). All statistical tests were two-sided, and p < 0.05 was considered to indicate statistical significance.

3.1 Participant Characteristics

The final analysis included 9603 adults with T2D, comprising 7953 participants from NHANES and 1650 participants from CHNS. In NHANES, the mean (SE) age was 59.8 (0.3), with 4046 males weighted at 50.4%; in CHNS, the mean (SE) age was 59.9 (0.3), with 811 males weighted at 49.2%. The median (IQR) dietary ALA intake was 1.32 (0.87 to 1.94) g/day for American participants and 1.54 (0.54 to 2.84) g/day for Chinese participants. Table 1 shows the baseline characteristics of adults with T2D in NHANES and CHNS categorized by tertiles of ALA intake. At baseline, adults with T2D with higher ALA intake tended to be younger men, more likely to have higher education and better quality of living, and to consume higher calories, carbohydrates, and various fatty acids, including SFAs, monounsaturated fatty acids (MUFAs), and polyunsaturated fatty acids (PUFAs) in both cohorts. In NHANES, individuals with higher ALA intake were more likely to be drinkers, overweight/obese, and suffer from hypercholesterolemia, but less likely to have a family history or suffer from CVDs. In CHNS, participants with higher ALA intake tended to be smokers and were more likely to have a short duration of diabetes.

3.2 Main Associations Between ALA and Risk of Mortality

During 63 153 person-years of follow-up in NHANES, 2301 deaths were documented, including 780 attributed to CVD. In CHNS, during 12 382 person-years of follow-up, 167 deaths were recorded. As depicted in Table 2, ALA intake was associated with a significantly reduced risk of all-cause mortality after multivariate adjustment for demographics, health-related behavior, history of diseases, diabetes duration, and diabetes medication use. Hazard ratios (HRs) across categories were 1.00, 0.89 (95% CI, 0.77–1.03), and 0.80 (0.66–0.96) for NHANES (p_trend = 0.02). A linearly reverse association was observed (p_nonlinear > 0.05, Figure 1). For CHNS, the HRs for all-cause mortality were 1.00, 0.72 (0.49–1.06), and 0.75 (0.50–1.11) (p_trend = 0.19), showing no significant linear association. In the pooled analysis, the HRs were 1.00, 0.87 (0.76–0.99), and 0.79 (0.67–0.94) (p_trend = 0.01), with the linearly reverse association remaining significant (p_nonlinear > 0.05, Figure 1). For CVD mortality, compared with the lowest tertile, the HRs were 0.81 (0.65–1.01) for tertile 2 and 0.68 (0.50–0.91) for tertile 3 (p_trend = 0.01) in NHANES (Table S1), showing linear association (Figure S2). However, no further analysis was conducted for ALA and CVD mortality in CHNS because specific causes of death were not available.

In the subgroup analysis, consistent findings were observed and no significant interactions were found between ALA intake and the risk of all-cause mortality (all p_interaction > 0.05, Table 3). As well as for marine n-3 PUFA, although the result showed that dietary ALA had a stronger association with the risk of all-cause and CVD mortality in the population with insufficient intake of marine n-3 PUFA (< median), this interaction term was not statistically significant (p_interaction > 0.05, Tables 3 and S2). In the stratified analysis, the associations of ALA intake with all-cause mortality persisted, even after excluding participants with CVD or cancer, those who died within the first 4 years since baseline, participants from 1999 to 2002 in NHANES, and participants with extreme BMI values. Furthermore, the conversion of ALA unit from grams to percentage of total energy intake or using multiple imputation to address missing values did not alter the primary results (Table S3). The results of the relation between dietary ALA and CVD mortality were consistent with the primary analysis (Table S4). Additionally, the E-value for the point estimate and upper confidence bound for the association between ALA intake and all-cause mortality was 1.63 and 1.26, respectively, and for CVD mortality was 2.30 and 1.43.

3.3 Cardiometabolic Biomarkers

The least-squares means of cardiometabolic biomarkers based on ALA intake were shown in Table S5. Higher ALA intake was significantly associated with lower levels of TC in both cohorts at baseline. In addition, ALA intake demonstrated an inverse association with TG levels in NHANES and LDL-C in CHNS, while exhibiting a positive association with TG levels in CHNS (all p_trend < 0.05).