This study was approved by the University of Alberta Human Research Ethics Committee (Pro00073641) to obtain corresponding health data from Alberta Health administrative health record registries. Additional details of the methodology have been described elsewhere (31). Briefly, from 2000–2015, ATP recruited almost 55,000 adults aged 35–69 years who had no history of cancer other than non-melanoma skin cancer, could complete questionnaires in English, and planned to reside in Alberta, Canada, for at least 1 year. This study includes participants who were recruited from 2000–2008 and completed Health and Lifestyle Questionnaires (HLQ) as well as diet and physical activity questionnaires (described below) in that same time period, who also consented to linkage to administrative health records, and subsequently provided a blood sample in 2009–2015. The demographic characteristics of participants have been described previously (15, 31).

Participants (n = 29,878) were originally recruited using random digit dialing sample selection and were mostly female (64%), Caucasian (91%), living in urban areas (77%), and had an average age of 50 ± 9 years (31). For this study, we further selected participants who did not have prevalent CVD at the time of enrollment. The main exclusion criteria included leaving Alberta, because the incidence of CVD was ascertained using linked provincial administrative health records (described below). Our final sample size was n = 23,248 participants; Figure 1 shows the flowchart of sample selection.

At the time of recruitment to ATP study (during 2000–2008), following self-administered ATP questionnaires were mailed to participants who consented to participate in ATP study: HLQ, for obtaining personal and family health history, reproductive health, and demographic characteristics, Canadian Diet History Questionnaire I (CDHQ-I) to obtain past 12 months dietary intake, and Past-Year Total Physical Activity Questionnaire (PYTPAQ) to obtain past 12 months physical activity. The PYTPAQ assesses the type, duration, frequency, and intensity of activities during the past year (31). The reliability and validity of PYTPAQ in the Canadian population have been assessed previously (32). CDHQ-I is a 124-item food frequency questionnaire modified from the DHQ from the U. S. National Cancer Institute, with 145 questions about food items, beverages, and dietary supplements intake. Details of the CDHQ and the validity of the questionnaire for use in Canadian population have been previously published (33).

CDHQ-I data was analyzed using Diet*Calc (version 1.4.2, National Cancer Institute, Bethesda, MD, USA) to assess the mean daily intake of energy, macronutrients, and micronutrients. To assess the n-3 FA intake, we summated the value of alpha linolenic acid [ALA (C18:3n3)], EPA (C20:5n3), docosahexaenoic acid [DHA(C22:6n3)], and docosapentaenoic acid [DPA(C22:5n3)]. For assessing diet quality, the American 2005 HEI adapted for Canada was used (34). The details of HEI score and the validity to use in Canadian population have been previously published (34). The HEI score ranges between 0–100, with higher scores indicating higher diet quality via adherence to dietary recommendations. It accounts for two primary aspects of diet quality, which are adequacy (adequate intake of healthy foods and nutrients) and moderation (moderate intake of unhealthy foods and nutrients). HEI scores <50 represent poor diet quality, 50–80 represent moderate diet quality that needs improvement, and higher than 80 represent high diet quality (34).

The information for CVD and comorbidity prevalence and/or incidence were obtained via data linkage to Alberta Health registry databases (including emergency department, physician claims, and Discharge Abstract Database) using the personal health numbers provided by participants (31).

Cases of CVD were identified using codes from the International Statistical Classification of Diseases and Related Health Problems (ICD) (31). Specifically, in physician claim data, ICD-9 is used all the time, whereas in inpatient (hospitalization) data, ICD-9 was used before 2003, and after 2003, ICD-10 was implemented. Therefore, in ATP cohort both ICD-9 and 10 have been used. Procedures were identified with the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures codes, following the definitions provided by the Alberta Diabetes Surveillance System (15). The primary outcome of this study was the CVD composite incidence, which is a cluster of CVD or procedures including ischemic heart disease (IHD), myocardial infarction (MI), angina, heart failure (HF), transient ischemic attack (TIA), acute ischemic stroke (AIS), percutaneous coronary intervention (PCI) and coronary artery bypass graft (CABG) defined by Clair et al. (35) in those without previous CVD, before or within 6 months of enrolment to ATP or within 1 year of data linkage to the Alberta Health databases. The Elixhauser index score (score ranges 0–30) was generated previously as a continuous variable (with Quan coding algorithm) (36) to indicate the presence of 30 different comorbidities that do not overlap with CVD (15, 37).

The menopause status for females was obtained via self-reported HLQ and clinical outcomes in routine evaluation (CORE) questionnaires.

In 2008, ATP joined the Canadian Partnership for Tomorrow’s Health (CanPath). As part of joining CanPath, ATP invited existing participants to complete further questionnaires and attend a study center to provide blood and urine samples. Non-fasting blood (~50 mL) sample was drawn at one time point, aliquoted to serum and plasma, and stored at −80°C for further analysis, with the majority of samples frozen within 2 h of blood draw (31). In 2017–2020 participants’ serum (0.5 mL) was used for assessing lipid biomarkers. HDL-C, TG, and total cholesterol (TC) were measured directly by Calgary Laboratory Services, an authorized clinical laboratory in Alberta (31). Non-HDL-C, RC, and LDL-C (by Friedewald formula) were calculated by the following equations:

Of the participants included in this study, a subset (n = 8,747) had lipid biomarkers data available.

Data was analyzed using STATA SE version 16.1. Descriptive statistics are shown as mean±standard deviation (SD) for continuous variables and percentage (number) for categorical variables. To compare the means at baseline, an independent t-test was used. Logistic regression was used for assessing the odds of occurring binomial outcome variable (CVD incidence), Cox proportional hazard model was used for assessing the relative risk of binomial outcome variable (CVD incidence), linear regression was used for assessing the linear relationship between exposure variables and continuous outcome variables (lipid biomarkers). Unadjusted and adjusted logistic regression [odds ratio (OR) and 95% confidence interval (95%CI)] was used for assessing the association of n-3 FA intake and HEI score with CVD incidence. The adjusted Cox proportional hazard model was used to calculate the hazard ratio (HR) and 95% CI. The adjusted linear regression [Coefficient (coef) and 95%CI] was used for the association of HEI score and n-3 FA intake with lipid biomarkers. The models were stratified by sex. For adjusting the confounders, we used two models; Model a: age, BMI (kg/m²), total energy(kcal), carbohydrate (%), protein (%), and fat (%) intake, Elixhauser index, as well as HEI score were adjusted to assess the association of n-3 FA intake (g/d), energy from n-3 FA (%), EPA intake (g/d), and energy from EPA (%) with CVD incidence and lipid biomarkers. Model b: age, BMI (kg/m²), total energy(kcal), carbohydrate (%), protein (%), and fat (%) intake, and Elixhauser index and were adjusted to assess the association of HEI score with CVD incidence and lipid biomarkers. p-value<0.05 (alpha error 5%) was considered as significant difference.

Table 1 shows the characteristics of participants with and without incident CVD. Those with CVD incidence (compared to those without) were significantly older (54 ± 9 vs. 49 ± 8.7 years, p-v < 0.001), had higher BMI (29.8 ± 6 vs. 28.2 ± 5.6 kg/m², p-v < 0.001), slightly lower HEI score (52.6 ± 9.6 vs. 53.8 ± 9.4, p-v < 0.001), and higher Elixhauser score means (2.1 ± 1.9 vs. 1.6 ± 1.6, p-v < 0.001). Comparing the baseline characteristics in participants with vs. without CVD incidence, no significant difference was seen in the means of total n-3 FA intake (g/d) (1.43 ± 0.75 vs. 1.43 ± 0.73), n-3 FA as proportion of total energy (%) (0.70 ± 0.21 for both), EPA intake (g/d) (0.02 ± 0.03 for both), EPA as proportion of total energy (%) (0.01 ± 0.01 for both), n-3 to n-6 PUFA ratio (0.11 ± 0.02 for both), and physical activity level (1.91 ± 0.42 vs. 1.98 ± 7.54).

Table 2 shows the characteristics of male and female participants. The incidence of CVD in females (n = 14,729) in comparison to males (n = 8,519) was significantly lower (21.93 vs. 28.24%, p-v < 0.001). The mean intake of n-3 FA (g/d) and EPA (g/d) in females compared to males were also significantly lower (1.3 ± 0.6 vs. 1.6 ± 0.8, and 0.024 ± 0.029 vs. 0.029 ± 0.034 p = v < 0.001). The ratio of n-3 FA to total energy and EPA to total energy (%) was slightly higher in females vs. males (0.7 ± 0.2 vs. 0.6 ± 0.2, and 0.013 ± 0.015 vs. 0.012 ± 0.013, p-v < 0.001). The mean HEI and Elixhauser scores were significantly higher in females (55.2 ± 9.4 vs. 50.7 ± 8.8, and 1.9 ± 1.8 vs. 1.4 ± 1.5, p-v < 0.001).

Table 3 shows the characteristics of participants with and without CVD stratified by sex. In both sexes, those with CVD incidence were older and had higher BMI, lower HEI score, and higher Elixhauser score (p-v < 0.001). In females (with vs. without CVD), there was no significant difference in either n-3 FA intake (g/d) or n-3 FA as proportion of total energy (%). In males with CVD (versus without CVD), the mean of n-3 FA intake (g/d) was slightly lower (1.6 ± 0.8 vs. 1.64 ± 0.8, p-v = 0.03) but, the n-3 FA as proportion of total energy (%) was slightly higher (0.66 ± 0.2 vs. 0.65 ± 0.2, p-v = 0.006).

The unadjusted and adjusted logistic regression model was used to assess the association of HEI score and n-3 FA intake with CVD incidence (Table 4). After adjusting for confounders (age, BMI, total energy, carbohydrate, fat, and protein intake (%), and Elixhauser score), in either women, men, or the combined total cohort, higher HEI score was associated with lower odds of developing CVD [OR: 0.97 (95%CI 0.97–0.98), p-v < 0.001, OR: 0.98 (95%CI 0.98–0.99), p-v = 0.01, and OR: 0.97 (95%CI 0.97–0.97) p-v < 0.001 respectively]. After adjusting for the confounders (age, BMI, total energy, carbohydrate, fat, and protein intake (%), Elixhauser score, and HEI score) for assessing the association of n-3 FA intake (g/d) with CVD incidence, no association was found in women. However, for males and the total cohort, higher n-3 FA intake (g/d) was associated with higher odds of CVD incidence [OR: 1.13 (95%CI 1.01–1.27), p-v = 0.03, and OR: 1.09 (95%CI 1.00–1.18), p-v = 0.02, respectively]. Higher N-3 FA to total energy ratio (%) was associated with a higher odds of CVD incidence in males [OR: 1.62 (95%CI 1.19–2.21), p-v = 0.002], and no association was found in females and the total cohort.

Table 5 shows the adjusted and unadjusted HR for the association of HEI score, n-3 FA intake (g/d), and n-3 FA to total energy ratio (%) with CVD incidence. After adjusting for confounders (model a), no association was found between n-3 FA (g/d) and CVD incidence in women, men, and the total cohort. Higher n-3 FA as proportion of total energy (%) was associated with increased CVD risk in males (HR: 1.42, 95% CI 1.1–1.84, p-v = 0.006). After adjusting for confounders (model b) higher HEI score was associated with reduced risk of CVD incidence in women, men, and the total cohort (HR: 0.98, 95%CI 0.97–0.98, p-v < 0.001, HR: 0.99, 95%CI 0.98–0.99, p-v = 0.004, HR: 0.97, 95%CI 0.97–0.98, p-v < 0.001, respectively). No significant association was found between EPA (g/d) and CVD incidence in all groups.

After adjusting the menopause status in models, a and b, no association was found between n-3 FA and EPA as proportion of total energy (%), and absolute n-3 FA and EPA intake (g/d) (Supplementary Table 5). Higher HEI score was associated with reduced CVD incidence [OR: 0.97 (95%CI 0.90–0.98), p-v < 0.001, and HR: 0.98 (95% CI 0.97–0.98), p-v < 0.00]. Supplementary Table 4 shows the comparison of baseline characteristics between pre- and postmenopausal females.

Table 6 shows the adjusted linear regression model to assess the association of HEI score and n-3 FA intake with lipid biomarkers (RC, TG, LDL-C, non-HDL-C, and HDL-C). No association was found between n-3 FA intake (g/d) and lipid biomarkers except for HDL-C in the total cohort and females [coef: 0.033 (95%CI 0.009–0.057), p-v = 0.007 and coef: 0.038 (95%CI 0.006–0.071), p-v = 0.02, respectively]. N-3 FA as proportion of total energy (%) was associated with only HDL-C in the total cohort [coef: 0.08 (95%CI), p-v = 0.002]. In women, a higher HEI score was associated with lower RC, TG, non-HDL-C, and higher HDL-C [coef: −0.006 (95%CI), −0.01 (95%CI), −0.008 (95%CI), +0.004 (95%CI), p-v < 0.001], but no association was found with LDL-C. The same trend was found in males between HEI score and RC, TG, non-HDL-C, and HDL-C [coef: −0.01 (95%CI −0.018–−0.005), p-v < 0.001, coef −0.01 (95%CI −0.02–−0.006), p-v < 0.001, coef −0.008 (95%CI −0.014–−0.001), p-v = 0.01, coef: +0.005 (95%CI 0.003–0.007), p-v < 0.001]. In the total cohort, higher HEI score was associated with lower RC, TG, non-HDL-C and higher LDL-C and HDL-C [coef: −0.01 (95%CI −0.016–−0.011), p-v < 0.001, coef −0.02 (95%CI −0.02–−0.01), p-v < 0.001, coef −0.009 (95%CI −0.012–−0.005), p-v < 0.001, coef: +0.004 (95%CI 0.001–0.008), p-v = 0.003, coef: +0.014 (95%CI 0.012–0.015), p-v < 0.001 respectively].

Our results are generally consistent with previous observational studies that reported a higher HEI score is associated with lower CVD incidence (38–42). These studies mainly compared the highest vs. the lowest quartile of HEI score and found that higher quartiles have lower CVD incidence than the lowest quartile. In our study, we found that both females and males had lower CVD incidence in the moderate (scores 51–80) vs. low HEI (scores ≤50) category (24.5% vs. 21 and 29.5% vs. 27%, respectively). Also, we found that with every one unit HEI score increase, the odds and risk of having CVD in the future decreased by 3%.

Studies on HEI score and lipid biomarkers have mainly focused on the relation with TG, TC, LDL-C, and HDL-C. To the best of our knowledge, this is the first study assessing the association of HEI scores with non-fasting RC. The results of previous studies on lipid biomarkers are controversial. While some studies have reported a significant inverse association between HEI score and TG, LDL-C, and TC and a direct association with HDL-C (43–45), others have found no association between HEI score and lipid biomarkers (46–48). In this study, one unit increase in HEI score was associated with reduced levels of non-fasting RC, TG, and non-HDL-C (−0.01, −0.02, and −0.0009 (mmol/L), respectively); as well as increased levels of HDL-C (+0.014 (mmol/L), respectively) in both females and males.

In this study, higher HEI scores were associated with lower CVD incidence risk and lipid biomarkers. Thus, for prevention of CVD, a dietary pattern with a higher intake of total fruits and vegetables, whole fruits, dark green and orange vegetables, legumes, total and whole grains, milk, meats, beans, and non-hydrogenated vegetable oil (or oil in fish, nuts, and seeds), and moderate intake of saturated fats, sodium, sugar, and alcohol (34), is recommended. The Canadian Cardiovascular Society, the American College of Cardiology/American Heart Association, and the European Society of Cardiology guidelines recommend this dietary pattern for preventing and managing CVD (16, 17, 49). This dietary pattern is associated with higher plant-based foods, fiber, vitamins including vitamin C and niacin, minerals, beneficial unsaturated fats intake, and lower saturated and trans FA intake (49). From a mechanistic point of view, higher intake of fiber, vitamin C, and niacin, can reduce CVD incidence by lowering TC, LDL-C, and inflammatory markers (50–52). In the intestine, fiber binds to bile acids and reduces cholesterol absorption. This leads to a lower hepatic cholesterol pool. As a compensatory response, the liver upregulates LDL receptor expression to increase uptake of circulating LDL-C, resulting in reduced TC, LDL-C, and CVD risk (53). Vitamin C is also associated with reduced LDL-C by promoting the conversion of cholesterol to bile acids, thereby depleting the liver cholesterol pool. This initiates a similar compensatory mechanism, leading to reduced circulating TC and LDL-C. Additionally, vitamin C acts as an antioxidant that prevents atherosclerosis by inhibiting LDL oxidation (54, 55). Niacin reduces CVD risk by decreasing inflammation, TG, and LDL-C, while increasing HDL-C by inhibiting the catabolism of HDL apolipoprotein A-I (ApoA-I). Niacin also decreases fatty acid mobilization from adipose tissue and inhibits hepatocyte diacylglycerol acyltransferase-2, a key enzyme in triglyceride synthesis, thereby reducing the secretion of VLDL and LDL lipoproteins (56, 57). Saturated and trans FAs activate the sterol regulatory element binding protein (SREBP)-2 pathway which upregulates hepatic cholesterol synthesis and increases the LDL-C levels (58, 59). Therefore, lower intakes of saturated and trans FAs are associated with reduced risk of CVD (60). In this study, dietary intake of niacin and vitamin C increased progressively across low, moderate, and high HEI score categories (Supplementary Table 1).

The absolute intake of n-3 FA (g/d) for females was lower than males (1.3 vs. 1.6 g/d) but the n-3 FA as proportion of total energy (%) was marginally higher (0.7 vs. 0.6%) due to lower total energy intake in females. However, this difference in proportion of energy from n-3 FA for both females and males is minimal and not clinically significant. In our cohort, we observed a direct association between total n-3 FA intake and CVD incidence in males, but no association was found either in females or the total cohort.

The dietary intake was measured only once at the beginning of the study and was not repeated during the study. Since individuals may alter their diet in the long term, we could not assess more recent dietary intake. Although dietary intake was assessed using a 12-month food frequency questionnaire, which may be subject to recall bias, the data were collected prospectively and not specifically focused on n-3 fatty acid intake. This prospective design, combined with subsequent ascertainment of CVD outcomes, strengthens the validity of the observed associations. Another key strength of this study is assessment of n-3 fatty acid intake in the context of overall diet quality, allowing for a more comprehensive evaluation of their combined impact on lipid biomarkers and CVD risk. This study provides understanding of how dietary patterns as a whole—rather than a single nutrient—may influence CVD risk. Also, this study is the first study that assessed the relation of diet quality and n-3 FA intake with non-fasting RC level. We recommend future studies assess the association of HEI scores using the newer version of the HEI score (2019). Also, we recommend RCT studies to assess the association of n-3 FA with CVD risk using a higher dosage (> 1 g/d of EPA + DHA).