Marine n-3 fatty acid treatment for carotid plaques in patients with type 2 diabetes

Abstract

Background

Carotid plaques result from a blockage or narrowing of carotid arteries associated with severe atherosclerotic cardiovascular diseases. We aimed to assess the effect of marine n-3 fatty acid supplementation on carotid plaques among patients with diabetes and whether the effect was modified by genetics.

Methods

In this 14 month double-blind, randomized controlled trial, 415 patients with type 2 diabetes (T2D) were randomly assigned to receive high-dose (3.0 g/day) or low-dose (1.5 g/day) marine n-3 fatty acids (fish oil) or placebo (refined olive oil). The primary outcome was the prevalence of carotid plaques. Secondary outcomes included changes in NMR-derived lipoprotein subclasses. Genetic interactions with the supplementation were also explored.

Results

383 participants (92.3%) completed the 14 month intervention, and 359 patients (86.5%) completed carotid ultrasound exams. Fish oil supplementation did not significantly reduce the carotid plaque prevalence (P trend = 0.111). Compared to the placebo, the odds ratios (95% CIs) of carotid plaque risk were 1.11 (0.56–2.22) and 0.54 (0.26–1.13) for low-dose and high-dose groups, respectively. Fish oil supplementation also showed no significant effect on the incidence of new carotid plaques (P for trend = 0.304) or the regression of existing plaques (P for trend = 0.390). High-dose intervention significantly reduced remnant cholesterol, LDL-1-triglycerides (TG), HDL-4-TG, and HDL-3-TG. In subgroups by genetic risk, the high-dose intervention significantly reduced carotid plaque risk specifically in patients with a low genetic risk for remnant cholesterol, with no significant benefit observed in those with medium or high genetic risk (P for interaction = 0.008; FDR = 0.056).

Conclusion

In patients with T2D, 14 month marine n-3 fatty acid supplementation does not reduce carotid plaque risk but improves lipoprotein profile. The anti-atherosclerotic effect appeared significant in patients with a low genetic risk of remnant cholesterol, warranting further investigation into personalized nutritional strategies.

Trial registration

Clinicaltrials.gov NCT03708887.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-026-03082-7.

Research insights

What is currently known about this topic?

• Patients with type 2 diabetes (T2D) have a 2–3 times higher risk of developing atherosclerotic cardiovascular disease (CVD) due to complications from atherosclerosis.

• It remains uncertain whether dietary supplementation with marine n-3 polyunsaturated fatty acids (PUFAs) improves carotid plaques and lipoprotein profiles among T2D patients.

• It is hypothesized that marine n-3 PUFAs could interact with genetic factors to affect cardiovascular health, yet evidence supporting genetic personalization of treatment is lacking.

What is the key research question?

Does dietary supplementation with marine n-3 PUFAs reduce carotid plaque risk in patients with T2D, and is this effect modified by genetics?

What is new?

• In this 14-month trial, high-dose n-3 PUFA supplementation did not reduce the risk of carotid plaques in T2D patients.

• High-dose n-3 PUFA significantly reduced remnant cholesterol, LDL-1 triglycerides (TG), HDL-4 TG, and HDL-3 TG.

• Genetic analyses revealed the anti-atherosclerotic effects were observed in patients with low genetic risk for elevated remnant cholesterol.

How might this study influence clinical practice?

• This study indicates that 14 month marine n-3 PUFA supplementation may not reduce the risk of carotid plaque, despite improving the lipoprotein profile in patients with T2D.

• The findings do not support the routine use of n-3 PUFA supplementation for atherosclerosis prevention in T2D patients but suggest the importance of assessing genetic background and support the development of personalized approaches targeting responsive patient subgroups.

Introduction

Type 2 diabetes patients are two to three times more likely to develop cardiovascular disease (CVD), largely due to atherosclerosis [1, 2]. Even with effective management of LDL cholesterol using statins, a significant residual cardiovascular risk persists, which is often associated with atherogenic dyslipidemia, including elevated triglyceride levels [3, 4]. Thus, additional intervention strategies, such as nutritional therapy, are necessary. Marine n-3 polyunsaturated fatty acids (PUFAs) may offer cardiometabolic benefits through anti-inflammatory properties and regulation of lipoprotein metabolism [5]. However, randomized controlled trials (RCTs) such as the STRENGTH [6], VITAL [7], and REDUCE-IT [8] trials have reported inconsistent results regarding CVD prevention, which could be attributable to heterogeneity in trial designs, including the dosage of n-3 PUFAs, the composition of the placebo, background statin therapy, and the cardiovascular risk profiles of the study populations. This uncertainty is further compounded in the diabetic subpopulation, where two RCTs have failed to confirm a definitive benefit on hard cardiovascular outcomes [9, 10], extending to conflicting results for cardiometabolic risk factors such as triglyceride levels [11, 12].

These inconsistencies in previous outcome trials underscore the necessity of investigating more direct and proximate measures of cardiovascular risk and highlight the need for more individualized assessment strategies. Carotid plaque, a validated morphological measure of subclinical atherosclerosis, may serve as a more sensitive intermediate endpoint than distant clinical events for detecting intervention effects [13]. We therefore hypothesized that the impact of n-3 PUFAs would be more readily detectable at this stage of atherosclerotic development than on final cardiovascular outcomes. Furthermore, evidence regarding the impact of n-3 PUFAs on carotid plaques in T2D or dysglycemia patients is sparse and inconsistent [14, 15], despite benefits shown in patients with CHD or coronary atherosclerosis [16, 17], creating a critical knowledge gap.

Lipoprotein particles' biological functions vary based on size [18]. Small dense LDL particles and remnant lipoproteins (VLDL, IDL, and chylomicron remnants) are highly atherogenic [19]. However, past trials on marine n-3 PUFAs reported inconsistent results concerning lipoprotein particle size and subfractions [20–25]. Thus, precise measurement through quantitative nuclear magnetic resonance (NMR) [26] is warranted to address the variations in each lipid subclass. Moreover, considerable heterogeneity in previous atherosclerosis studies may be attributed to genetic variations and gene-nutrient interactions [27]. Genome-wide association studies (GWASs) have identified single-nucleotide polymorphisms (SNPs) associated with carotid plaques and lipoproteins [28, 29]. A genetic risk score (GRS) can be constructed to quantitatively depict genetic risk, valuable for gene-nutrient interactions [30]. Our prior studies have shown that genetic risks modify the relationship between n-3 PUFAs and risk of cardiometabolic diseases [30, 31]. In addition, genetic variants in enzymes like FADS1/2 and ELOVL2 (e.g., rs2295602 CC vs CT/TT) [32], involved in n-3 PUFA metabolism, may alter the effect of n-3 PUFA supplementation.

To address these specific limitations of prior research, we designed the current trial with several key novel features. First, we focused specifically on a Chinese T2D population with characteristically low statin usage [33], a large and clinically relevant demographic that is underrepresented in previous RCTs. Second, we selected a low dose (1.5 g/day) to model the effect of achieving guideline-recommended intake levels and a high dose (3.0 g/day) to investigate the potential added benefit of a supplemental, pharmacologic-level intake and potential dose–response relationship. Third, we employed a 14-month intervention and used carotid plaque risk as our primary outcome to directly measure anti-atherosclerotic efficacy. For secondary analyses, we employed advanced NMR-based lipoprotein profiling to characterize changes in lipoprotein subclasses and conducted genetic analyses to identify patient subgroups most likely to benefit from the intervention.

Methods

Study design

The Precision Nutritional Management for Diabetes (PNMD) study is a placebo-controlled, double-blind, randomized clinical trial designed to assess the effects of low and high doses of marine n-3 PUFAs compared to a placebo (refined olive oil) on glucose and lipid levels and carotid plaques (primary outcomes) in individuals with T2D (NCT03708887). Secondary outcomes included T2D complications and comorbidities. A secondary outcome—sleep quality—from the PNMD trial has been reported, with full protocol details published online [34]. The protocol was approved by the ethics committee for human research (reference no. 20180301) at Lanxi Red Cross Hospital in Jinhua, China. Written informed consent was obtained from all participants. Royal DSM, the donor of the capsules used in the trial, played no further role in the study's execution. This study conformed to the principles outlines in the Declaration of Helsinki.

Participants

Participants eligible for the study were those with a confirmed T2D diagnosis (WHO criteria) [35], aged between 40 and 75, and with no history or recent use of using n-3 PUFA supplements. The age range was selected to focus on individuals where T2D is more likely to have stabilized, avoiding the variability associated with younger individuals who may have early onset or older individuals who may experience additional age-related health complications that could confound the study outcomes. Exclusion criteria comprised a previous diagnosis of type 1 diabetes, existing coronary heart disease, prior stroke, cancer diagnosis, liver or renal conditions, premenopausal status, current pregnancy or breastfeeding, fish allergies, or participation in another clinical study within the last three months. Premenopausal women were excluded to eliminate the interference of estrogen on glucose and lipid metabolism [36]. Initially, potential candidates were identified and pre-screened by local primary care physicians (PCPs), who then referred them for further evaluation by specialized research personnel at the hospital to determine their eligibility for the trial.

Randomization and masking

Participants were randomly assigned in equal numbers to one of three groups: high dose, low dose, or placebo. An external statistician generated random sequences using a computer-based system (SAS PROC PLAN). To preserve the study's integrity, the appearance of the fish oil and placebo capsules was similar and they were distributed in identical, coded containers, making them indistinguishable to both participants and researchers. The group assignments remained confidential to both the investigators and participants until after the data analysis was completed.

Procedures

During the treatment phase, all participants were instructed to take four capsules per day. The composition of these four capsules was designed to maintain blinding and equalize total oil intake across the groups. Participants in the high-dose group received four capsules filled entirely with fish oil, providing a total daily dose of 3.0 g of marine n-3 PUFAs (2149 mg EPA and 851 mg DHA). Those in the low-dose group received four capsules, each filled with a 1:1 mixture of fish oil and the refined olive oil placebo, delivering a total of 1.5 g of marine n-3 PUFAs (1075 mg EPA and 425 mg DHA). The placebo group received four capsules containing only refined olive oil. This design ensured that all participants consumed an identical number of capsules and volume of oil daily. Capsules were taken with meals (two with breakfast and two with dinner). The exact formulations of both fish oil and olive oil capsules are presented in the Figs. S1 and S2. Refined olive oil is rich in oleic acid (73.6%) which cannot be converted to long-chain n-3 fatty acids in humans. Refined olive oil, unlike virgin or extra-virgin olive oil, lacks plant polyphenols, which could exert cardiometabolic benefits [37]. Moreover, the daily dose administered in our study (4 g) was substantially lower than the therapeutic quantity used in dietary interventions such as the PREDIMED trial (≥ 50 mL/day) [38]. Our preliminary in vitro research (Figs. S3 and S4) further confirmed the minimal antioxidant and anti-inflammatory effects of refined olive oil compared to other vegetable oils, establishing it as an effective neutral control.

The intervention trial lasted for 14 months. Initially, participant baseline characteristics, including sociodemographic factors, lifestyle habits, and clinical details, were collected via questionnaires by trained graduate students and clinical staff. Dietary intake was evaluated using a validated food-frequency questionnaire [39] including 55 food items (covering 15 food groups), drinks and cooking oils [40]. Intakes of nutrients and total energy as calculated using the China Food Composition Tables (2018). A diet quality score was established to represent the overall diet quality based on dietary recommendations for cardiometabolic health [40, 41]. Briefly, points were awarded for the following factors: seafood, fruit, vegetable, milk, tea, and wholegrain intakes at or above the median; and for preserved vegetable, unprocessed red meat, processed red meat, refined grain, and sodium intakes below the median. Data on physical activity were collected using a questionnaire that assessed tasks involving light, moderate, and vigorous intensity activities. The metabolic equivalent of task hours per week (MET-h/wk) was calculated based on the Compendium of Physical Activities [42]. These questionnaires were administered by trained interviewers both at baseline and the end of the intervention. Additionally, participants underwent physical assessments and provided biological samples for analysis. Anthropometric measurements, including weight, height, and waist circumference, were performed by trained staff from hospitals. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured three times by well-trained physicians. Hypertension was defined as a SBP of ≥ 140 mm Hg, a DBP of ≥ 90 mm Hg, a prior diagnosis of hypertension by a physician, or the current use of antihypertensive medication.

After randomization, each participant received a month's supply of capsules in a labeled bottle and was required to exchange the bottle monthly, returning any unused capsules, throughout the treatment phase. Adherence to the regimen was monitored, evaluated by 25 PCP teams to ensure that more than 95% of the capsules were consumed by patients. Plasma levels of EPA and DHA were quantified at baseline and post-intervention for all participants who completed the trial. Analysis was performed using gas chromatography to confirm compliance and assess changes in n-3 PUFA status. During the study, participants were advised to maintain their usual diets, lifestyle habits, and medication routines. Adverse events were assessed through regular monitoring by PCP teams during the trial. Participants were asked to report any adverse events related to the fish oil supplementation. In this RCT, 56 of 415 patients (13.5%) were receiving lipid-lowering therapy (LLT). Based on survey data and PCP records, > 95% of this therapy was statin-based (with no ezetimibe use recorded). While specific statin types and doses were not collected, the frequency of use (categorized as ≥ 1 time/day or < 1 time/day) was documented. PCPs confirmed that all LLT regimens remained unchanged for enrolled participants throughout the study.

Genotyping

Genotyping was conducted using the Illumina Infinium Asian Screening Array (ASA-750 K). Detailed descriptions of DNA extraction, data handling, imputation methods, and quality control measures have been previously reported [43]. For the analysis, we focused on SNPs associated with circulating marine n-3 PUFA levels as identified in prior GWASs (Table S1). Additionally, we constructed GRSs [44] for T2D and carotid plaques using 46 SNPs (Table S2) and 9 SNPs (Table S3), respectively, in prior GWASs. We also used allele variations related to LDL, remnant cholesterol, and specific HDL-TG to construct corresponding GRSs (Tables S4–S6).

Outcomes

At baseline and the conclusion of the 14 month intervention, participants were asked to go to the study hospital for comprehensive assessments and to provide biological blood samples in the morning after an overnight fasting. Serum levels of glucose, triglycerides, HDL cholesterol, LDL cholesterol, total cholesterol and hs-CRP were measured using the AU5800 clinical chemistry analyzer series. HbA1c levels were analyzed by immunoturbidimetry using an automated chemistry analyzer.

A FUJIFILM SonoSite ultrasound system (Washington, USA) equipped with a high frequency (7–11 MHz) probe was used to assess carotid artery plaque lesions, including carotid intima-media thickness (CIMT). The exam was performed by four blinded ultrasonography doctors adhering to the Mannheim Consensus (2011 update) [45]. A plaque was defined as a focal structure encroaching into the arterial lumen by at least 0.5 mm or at least 50% of the surrounding CIMT value, or being > 1.5 mm thick from the intima-lumen interface to the media-adventitia interface [45]. Details of the carotid ultrasonography are provided in the Supplemental material.

A high-throughput NMR metabolomics platform (a 310 K and 600.13 MHz proton Larmor frequency NMR spectrometer) at Zhejiang NUTRIEASE Science and Technology Corporation was used to measure plasma lipoprotein subclasses. Details of the NMR platform, laboratory measurements, and quality control are provided in the Supplemental material. Blood creatinine levels were also measured by the NMR platform and the estimated glomerular filtration rate (eGFR) was then calculated using the CKD-EPI equation [46].

Statistical analysis

Assuming a dropout rate of approximately 15%, we calculated that enrolling 415 participants (about 138 per group) would provide the trial with over 80% statistical power to detect a 33% relative risk reduction (RRR) of carotid artery plaque in the fish oil-treated group at a significance level of 0.05 (two-sided test). The calculation was based on an anticipated plaque prevalence of 59.0% in the placebo group [47]. The 33% RRR was selected as a clinically meaningful effect size, informed by the magnitude of benefit observed in prior high-dose n-3 PUFA outcome trials (e.g., the 25% RRR in cardiovascular events in REDUCE-IT [8]), and was based on the hypothesis that the effect on a direct pathological marker of atherosclerosis might be more pronounced than on clinical events. For the primary analysis, Intention-to-Treat (ITT) analysis was performed with missing data handled by multiple imputation via Markov Chain Monte Carlo methods. Data are presented as means ± standard deviations (SDs) or medians (interquartile ranges). Baseline group differences were assessed using one-way ANOVA or the Kruskal–Wallis H test for continuous variables and the chi-square test for categorical variables. Changes in blood indicators from baseline to the end of the intervention were compared among groups using one-way ANOVA or the Kruskal–Wallis H test, followed by Dunnett’s or Dunn’s post hoc tests for pairwise comparisons. P values for differences in lipoproteins (114 types) were further adjusted by the false discovery rate (FDR). We used logistic regression models to assess the effects of low and high doses of marine n-3 PUFAs on carotid plaque prevalence at the end of the trial. Model 1 was unadjusted, while Model 2 was adjusted for age, sex, and baseline plaque status. Effects of low and high doses of marine n-3 PUFAs on the incidence of new plaques in patients free of plaque at baseline and the regression of existing plaques were also analyzed using logistic regression models. The P value for trend was calculated by modeling the continuous variable of fish oil dose (1, 1.5, or 3 g). Changes in CIMT and plaque height baseline to the end of the intervention were also compared among groups using the Kruskal–Wallis H test.

Subgroup analyses were conducted to evaluate the effect of fish oil supplementation on carotid plaque risk stratified by baseline characteristics, including age, sex, BMI (above or below the median), HbA1c, total energy intake, hsCRP, triglycerides, hypertension, eGFR and baseline plasma marine n-3 PUFA levels. Sensitivity analyses were performed by further adjusting for hypoglycemic drugs, lipid-lowering drugs, fish consumption, plasma marine n-3 PUFA levels, total energy intake, and other dietary supplements. We tested potential interactions between high-dose fish oil supplementation and categories of T2D/plaque GRS, lipoprotein-related GRS, or SNPs related to n-3 PUFA metabolism on carotid plaque risk using logistic regression models that included a multiplicative interaction term. We applied a rigorous FDR correction to all genetic interaction analyses, considering an FDR < 0.1 as significant for these exploratory tests.

Results

Patients

Among the 2,198 T2D patients initially screened, 740 underwent baseline assessment to determine eligibility from September 2020 to November 2020. After exclusions, 415 patients were enrolled and randomly assigned to the low-dose fish oil (140 participants), high-dose fish oil (137 participants), or control (138 participants) group. The final follow-up was conducted in January 2022. Of the patients who underwent randomization, 383 (92.3%) completed the 14 month intervention, and 359 (86.5%) completed the carotid ultrasound exams (Fig. 1). Capsule counts indicated that more than 98% of the capsules were consumed by patients. No harms or unintended effects related to fish oil supplementation were observed. The characteristics of patients included and excluded from the per-protocol analysis are presented in Table S7. The mean age of the included patients was 63.2 years, and 45.3% were male. Baseline characteristics of the T2D patients were similar across the three groups (Table 1), indicating unbiased randomization. Changes in plasma EPA and DHA levels following the intervention are shown in Fig. S5 and Table S8, confirming the efficacy of the treatment.

Fig. 1.

Enrollment and completeness of data. T2D, type 2 diabetes

Table 1.

Characteristics of the patients at baseline

Characteristics	Control group	Low dose group	High dose group
	(n = 138)	(n = 140)	(n = 137)
Age—yr	62.3 ± 6.9*	63.2 ± 7.3	63.7 ± 7.1
Sex—no. (%)
Male	68 (49.3)	54 (38.6)	70 (51.1)
Female	70 (50.7)	86 (61.4)	67 (48.9)
BMI†—kg/m2	24.2 ± 3.3	23.7 ± 2.9	24.0 ± 3.0
Smoking status—no. (%)
Never	92 (66.7)	98 (70.0)	92 (67.2)
Former	25 (18.1)	16 (11.4)	22 (16.1)
Current	18 (25.8)	11 (40.3)	16 (33.9)
Alcohol drinking—no. (%)
Non-drinker	79 (57.7)	89 (64.5)	84 (61.3)
Drinker	58 (42.3)	49 (35.5)	53 (38.7)
Education—no. (%)
Illiteracy	40 (30.1)	35 (25.2)	26 (19.3)
Elementary school	52 (39.1)	41 (29.5)	52 (38.5)
Middle school	32 (24.1)	41 (29.5)	36 (26.7)
High school	9 (6.8)	15 (10.8)	17 (12.6)
Professional education	0 (0.0)	5 (3.6)	3 (2.2)
College or higher	0 (0.0)	2 (1.4)	1 (0.7)
Physical activity—MET-h/wk	130.0 ± 100.0	115.2 ± 98.4	104.2 ± 76.4
Blood pressure—mmHg
Systolic pressure	140.6 ± 21.7	138.1 ± 20.4	137.7 ± 20.0
Diastolic pressure	84.9 ± 10.4	83.9 ± 9.6	83.2 ± 9.2
Blood indicators
Fasting glucose—mmol/L	9.2 ± 2.9	8.8 ± 2.4	8.8 ± 2.3
HbA1c—%	7.3 ± 1.5	7.1 ± 1.4	7.2 ± 1.3
Insulin—pmol/L	48.6 ± 53.1	62.9 ± 146.9	60.6 ± 78.0
Total cholesterol—mmol/L	4.8 ± 1.1	4.9 ± 1.0	4.9 ± 1.1
Triglyceride—mmol/L	2.0 ± 2.4	1.9 ± 1.6	1.8 ± 1.8
Low-density lipoprotein cholesterol—mmol/L	2.6 ± 0.8	2.8 ± 0.8	2.7 ± 0.7
High-density lipoprotein cholesterol—mmol/L	1.5 ± 0.3	1.6 ± 0.4	1.5 ± 0.3
Medical history
Duration of diabetes—yr	7.2 ± 5.2	6.6 ± 5.1	7.0 ± 5.7
Hypertension—no. (%)	59 (42.8)	62 (44.3)	62 (45.3)
Lipid-lowering drug use—no. (%)‡	22 (15.9)	17 (12.1)	17 (12.4)
≥ 1 time/day—no. (%)	11 (8.0)	6 (4.3)	9 (6.6)
< 1 time/day—no. (%)	11 (8.0)	11 (7.9)	8 (5.8)
Antidiabetic drugs
Insulin—no. (%)	13 (9.4)	20 (14.3)	15 (11.0)
Metformin—no. (%)	54 (43.6)	62 (47.7)	66 (55.0)
Sulfonylurea—no. (%)	67 (54.0)	60 (46.2)	58 (48.3)
Glinide—no. (%)	5 (4.0)	7 (5.4)	8 (6.7)
TZDs—no. (%)	6 (4.8)	9 (6.9)	3 (2.5)
Glucosidase inhibitor—no. (%)	18 (14.5)	20 (15.4)	26 (21.7)
SGLT-2 inhibitors—no. (%)	0 (0.0)	1 (0.8)	2 (1.7)
GLP-1 receptor agonist—no. (%)	0 (0.0)	2 (1.5)	0 (0.0)
Pramlintide—no. (%)	0 (0.0)	0 (0.0)	0 (0.0)
DPP-4 inhibitor—no. (%)	5 (4.0)	13 (10.0)	8 (6.7)
Other glucose-lowering drugs—no. (%)	1 (0.8)	0 (0.0)	1 (0.8)
Dietary supplements
Mineral—no. (%)§	26 (18.8)	31 (22.1)	32 (23.4)
Vitamin—no. (%)	4 (2.9)	8 (5.7)	7 (5.1)
Lecithin—no. (%)	1 (0.7)	8 (5.7)	6 (4.4)
Royal jelly—no. (%)	6 (4.4)	10 (7.1)	11 (8.0)
Total energy intake—kcal/d	2204.5 ± 844.3	2286.4 ± 964.7	2380.4 ± 1190.3
EPA + DHA intake—mg/2000 kcal	47.6 ± 98.5	54.8 ± 91.9	47.0 ± 62.9
Diet quality score	4.5 ± 1.3	4.7 ± 1.4	4.8 ± 1.3
Carotid plaque—no. (%)	78 (56.5)	69 (49.3)	82 (59.9)

BMI, body mass index; MET, metabolic equivalent task

*Plus–minus values are means ± SD. There were no significant differences between the groups in any baseline characteristics (using one-way ANOVA or the Kruskal–Wallis H test for continuous variables and the chi-square test for categorical variables) at a two-sided alpha level of 0.05. The total number of participants in the subcategories does not equal the overall participant count due to missing values

^†Body-mass index is the weight in kilograms divided by the square of the height in meters

^‡The vast majority used statin lipid-lowering drugs (> 95%) and none used ezetimibe

^§One participant reported using both calcium and zinc supplements, while the others reported using only calcium supplements

Carotid plaques

A total of 359 patients completed the carotid ultrasound exams, with no differences in carotid plaque prevalence between the groups at baseline (Table 1). At the end of the intervention, 227 patients (54.7%) had carotid plaques: 51.8% in the high-dose group, 57.1% in the low-dose group, and 55.1% in the control group. In the unadjusted analysis, no significant trend was observed for the effect of fish oil supplementation on carotid plaque prevalence at the end of the intervention (P for trend = 0.111). Compared with the control group, the odds ratios were 1.11 (95% CI 0.56–2.22) for the low-dose group and 0.54 (95% CI 0.26–1.13) for the high-dose group. After adjustment for age, sex, and baseline carotid plaque status, the high-dose group showed a non-significant 51% reduction in odds (adjusted odds ratio, 0.49; 95% CI 0.23 to 1.04) compared with the control group (P for trend = 0.071; Model 2) (Table 2). In the analysis of new plaque incidence among patients free of plaque at baseline, fish oil supplementation did not significantly reduce the risk of developing new carotid plaques (odds ratio for low-dose group, 1.13; 95% CI 0.53–2.39; odds ratio for high-dose group, 0.62; 95% CI 0.26–1.48; P for trend = 0.304). Similarly, no significant differences were observed in plaque regression among the three groups (P for trend = 0.390) (Table 2). Adjusting for age and sex (model 2) yielded similar results for plaque incidence and regression.

Table 2.

Effects of marine n-3 PUFA supplementation on risk of carotid plaque

	OR (95% CI)*			P-trend
	Control	Low-dose fish oil	High-dose fish oil
Carotid plaque prevalence at the end of the intervention
N	138	140	137
Cases—no. (%)	76 (55.1)	80 (57.1)	71 (51.8)
Model 1†	1 (Ref.)	1.11 (0.56–2.22)	0.54 (0.26–1.13)	0.111
Model 2‡	1 (Ref.)	1.05 (0.51–2.14)	0.49 (0.23–1.04)	0.071
Carotid plaque incidence in patients free of plaque at baseline
N	60	71	55
Cases — no. (%)	17 (28.3)	20 (28.2)	11 (20.0)
Model 1†	1 (Ref.)	1.13 (0.53–2.39)	0.62 (0.26–1.48)	0.304
Model 2‡	1 (Ref.)	1.08 (0.50–2.34)	0.54 (0.22–1.32)	0.191
Carotid plaque regression
N	78	69	82
Cases regressed—no. (%)	4 (5.1)	4 (5.8)	8 (9.8)
Model 1†	1 (Ref.)	0.69 (0.20–2.34)	1.57 (0.53–4.71)	0.390
Model 2‡	1 (Ref.)	0.78 (0.22–2.75)	1.72 (0.56–5.30)	0.329

*Analyses were conducted under the intention-to-treat (ITT) principle. Missing values in covariates and outcomes were handled using multiple imputation. Odd ratios (95% CI) were analyzed using logistic regression models

^†Unadjusted

^‡Further adjusted for age, sex, and carotid plaque at baseline (only for carotid plaque prevalence)

Subgroup analyses revealed an inverse association between high-dose fish oil supplementation and carotid plaque prevalence among women (P for interaction = 0.018) and patients with lower energy intake (P for interaction = 0.044) (Fig. S6). Regarding IMT measurements, the changes in maximum IMT were 0.000 mm (IQR, −0.010 to 0.000) in the control group, −0.005 mm (IQR, −0.020 to 0.005) in the low-dose group, and −0.010 mm (IQR, −0.020 to 0.000) in the high-dose group (P = 0.075 for between-group differences) (Table S9). No significant differences were observed in changes in mean IMT (P = 0.553), mean plaque height (P = 0.695), or maximum plaque height (P = 0.677) between groups.

Sensitivity analyses demonstrated consistent null associations of fish oil supplementation with carotid plaque prevalence, incidence, or regression after additional adjustment for hypoglycemic drugs, lipid-lowering drugs, fish consumption, plasma marine n-3 PUFA levels, total energy intake, and other dietary supplements (Table S10).

Blood lipids and lipoproteins

After the 14-month intervention, significant differences were observed in serum triglycerides (P < 0.001) and total cholesterol (P = 0.029) levels among the three groups (Table S11). The high-dose fish oil group had a significantly greater reduction in serum triglycerides (median change, −0.405; IQR, −0.800 to −0.130) compared with the control group (median change, −0.085; IQR, −0.540 to 0.305). The low-dose group showed a similar trend in reducing triglycerides but had significantly higher levels of total cholesterol (median change, 0.260; IQR, −0.250 to 0.650) compared with the high-dose group (median change, 0.010; IQR, −0.600 to 0.490).

For plasma lipoprotein subclasses, high-dose fish oil supplementation significantly reduced remnant cholesterol compared to the placebo (median change, −14.00 mg/dL vs. −0.50 mg/dL), including reductions in VLDL and IDL particle numbers (all FDR < 0.001) (Fig. 2A and Table S12). VLDL particle reductions were observed across VLDL-1, -2, -3, and -4. The largest reductions were seen in VLDL-TG and IDL-TG, while other lipid constituents showed smaller reductions. All lipid constituents within LDL-1 particles were significantly decreased in the high-dose fish oil group compared with the placebo, with LDL-1-TC showing the greatest reduction (median change, −4.44 mg/dL vs. 1.56 mg/dL). LDL-2-TG was also reduced by high-dose fish oil supplementation. Regarding HDL subclasses, high-dose fish oil supplementation significantly lowered concentrations of HDL-3 and HDL-4, with reductions consistent across all lipid constituents. Apo-A1 within HDL-3 and HDL-4 exhibited the greatest reductions. Low-dose fish oil intervention showed a trend towards lower remnant cholesterol, VLDL, and significantly decreased LDL-1 particles, whereas raised concentrations of LDL-5 and LDL-6 compared with the placebo (Fig. 2B and Table S12). Additionally, HDL-2-TC, HDL-2-CE, and HDL-1-FC were significantly elevated in the low-dose fish oil group.

Fig. 2.

Effects of low- and high-dose fish oil supplementation on plasma lipoprotein profiles in T2D Patients. Panel A shows the changes from baseline in plasma lipoprotein profiles in the high-dose fish oil group. Bars indicate medians (standardized). P values for the difference between the high dose group and control group were computed using Mann–Whitney U tests and adjusted for FDR. *indicates FDR < 0.05, **indicates FDR < 0.01 and ***indicates FDR < 0.001. Panel B shows changes from baseline in plasma lipoprotein profiles in the low-dose fish oil group. Bars indicate medians (standardized). *indicates FDR < 0.05, **indicates FDR < 0.01 and ***indicates FDR < 0.001 for the difference between the low dose group and control group using Mann–Whitney U tests

Interactions of fish oil supplementation with genetic profiles

Analysis of GRS interactions revealed that the effect of fish oil supplementation was significantly modified by the GRS for remnant cholesterol (P for interaction = 0.008; FDR-corrected P for interaction = 0.056) (Table S13) Specifically, a reduced risk of carotid plaques with high-dose supplementation, compared to the control group, was observed only among patients in the lowest tertile of the remnant cholesterol GRS, with no significant benefit seen in the middle or highest tertiles (Fig. 3). No significant interactions were found between fish oil supplementation and GRS for T2D (FDR-corrected P for interaction = 0.966), carotid plaque (FDR-corrected P for interaction = 0.940) (Table S13), or with genetic variants related to n-3 PUFA metabolism (Table S14).

Fig. 3.

Effects of fish oil supplementation on carotid plaque prevalence among T2D patients stratified by genetic risk score of remnant cholesterol. ORs (95% CIs) of carotid plaque risk associated with fish oil supplementations across tertiles of remnant cholesterol GRS using logistic regression models. Vertical bars indicate 95% confidence intervals. P values were computed by logistic regression models adjusting for age, sex, carotid plaque at baseline

Discussion

In this 14 month trial, patients with T2D who received low or high-dose marine n-3 PUFAs showed no significant reduction in carotid artery plaque risk, though serum triglyceride levels were significantly decreased compared to placebo. NMR analyses revealed a dose-dependent improvement in the atherogenic lipoprotein profile, characterized by reductions in remnant cholesterol and specific LDL-TG and HDL-TG subfractions. Furthermore, exploratory analyses suggested high-dose supplementation reduced plaque prevalence in patients with low genetic risk for remnant cholesterol.

Comparison with previous studies and possible explanations

Marine n-3 PUFAs have been considered anti-atherosclerotic, and a meta-analysis of six trials demonstrated that high-dose n-3 PUFAs significantly slowed atherosclerosis progression [48]. However, evidence specifically for diabetic patients, who are at high risk of atherosclerotic CVD, remains limited. Our findings of a non-significant effect on carotid plaque risk, together with the absence of significant changes in IMT or plaque height, consistently suggest no beneficial effect on atherosclerotic plaque progression. A trial involving 1184 individuals with dysglycemia and high CVD risk reported no effect of 1 g/day n-3 PUFAs on CIMT over 4.9 years [15], which was consistent with our findings showing no significant changes in CIMT and plaque height. Similarly, our observation of no significant effect on new plaque onset or plaque remission aligns with a large trial of 15,480 diabetic patients that found no difference in major cardiovascular events between 1-g fish oil and placebo groups [9]. This may suggest that hyperglycemia attenuates the anti-atherosclerotic effect of marine n-3 PUFAs. In contrast, a study of 81 Japanese T2D patients showed that 1,800 mg/day EPA improved CIMT [14], and the REDUCE-IT trial demonstrated that icosapent ethyl, a highly purified EPA formulation, reduced cardiovascular events by 25% in patients with prior CVD or diabetes [8]. These findings indicate that pure EPA may possess stronger anti-atherosclerotic properties than the mixed EPA and DHA formulation used in our study, potentially mediated through EPA's ability to reduce inflammation, oxidative stress, and plaque formation [49]. Further long-term RCTs are warranted to evaluate the effects of EPA versus DHA on carotid plaque progression or cardiovascular events in T2D patients.

Marine n-3 PUFAs are FDA-approved for severe hypertriglyceridemia, but previous RCTs among T2D patients have yielded inconsistent results on triglyceride levels, which may be due to low dose, short treatment duration, use of linoleic acid-rich placebos, and low EPA/DHA ratio [50–52]. Our study used two doses of fish oil with a high EPA/DHA ratio and refined olive oil as a placebo. With a large sample size (n = 415) and a long duration (14 months), the study found dose-dependent reductions in serum triglycerides in Chinese T2D patients, confirming the potent triglyceride-lowering effect in this population.

NMR analysis revealed that high-dose fish oil shifted lipoprotein particle distribution towards a cardioprotective profile with reduced remnant cholesterol, LDL-1-TG, HDL-4-TG, and HDL-3-TG [53]. High-dose icosapent ethyl has been reported to reduce atherogenic remnant particles by up to 30% in phase 3 trials [54]. Remnant particles may exhibit greater atherogenicity per particle because they contain up to 4 times more cholesterol per particle than LDL [53]. Here we revealed both FC and CE within remnant particles (VLDL and IDL) decreased in the n-3 PUFA treatment group, although a recent study suggests a similar association pattern of cholesterol and TG within lipoprotein particles with CVD mortality [55]. LDL-1, which carries most TG among LDL subfractions, was significantly reduced by n-3 PUFAs in our study, which was consistent with a prior trial [56]. Additionally, increased LDL-5, LDL-6, HDL-2-TC, HDL-2-CE, and HDL-1-FC in our low-dose group may be driven by DHA known to increase both LDL and HDL concentrations [57]. DHA's ability to enhance the activity of hepatic lipase and cholesteryl ester transfer protein (CETP) could promote the remodeling of larger, buoyant LDL particles into smaller, denser subspecies [58]. Supplementation with DHA has also been shown to significantly increase the concentration of large HDL particles, a finding attributed to the incorporation of DHA into the phospholipid membrane of HDL, which can increase particle size and cholesterol content [59]. Interestingly, we detected an unexpected reduction in HDL-TG in the middle subclasses (HDL-3 and HDL-4) in the high-dose group, possibly due to CETP activation by n-3 PUFAs [58], which facilitates the transfer of CE from HDL to VLDL and LDL in exchange for triglycerides. Our findings that n-3 PUFAs reduce atherogenic lipoproteins like IDL-Apo-B and VLDL-Apo-B align with the known mechanisms of n-3 PUFAs. It is plausible that the reduction in apo B-containing lipoproteins is linked to a downregulation of apolipoprotein C-III, a key inhibitor of lipoprotein lipase. This is supported by a meta-analysis by Sahebkar et al. [60], which confirmed that n-3 supplementation significantly lowers plasma apo C-III concentrations, providing a potential pathway for the triglyceride-lowering effects we observed.

Although the overall protective effect was not significant for the total study patients, a key question for clinical translation is identifying which patients are most likely to benefit. We found that the effect of high-dose fish oil on carotid plaque prevalence was modified by the genetic risk for elevated remnant cholesterol. Patients with a low genetic risk demonstrated a significant reduction in plaque risk, whereas no significant effect was observed in those with high genetic risk. This gene-treatment interaction is mechanistically plausible. Remnant cholesterol particles are established contributors to atherosclerosis [53], and in our trial, marine n-3 PUFA supplementation significantly reduced remnant cholesterol levels. Mechanistically, n-3 PUFAs modulate the expression of genes governing remnant cholesterol synthesis and clearance—for example, by suppressing HNF-4α, APOB, SREBP-1c, and ACAT2 to reduce VLDL assembly and secretion [61, 62], while upregulating CYP7A1 to promote bile acid synthesis and LDLR to enhance lipoprotein clearance [58]. In patients with a high genetic risk, the pathways driving remnant cholesterol synthesis may be constitutively activated, while clearance mechanisms could be genetically impaired. Consequently, the regulatory effects of n-3 PUFAs may be insufficient to overcome this "genetically blocked" physiology, resulting in an attenuated reduction of remnant cholesterol and no significant decrease in plaque risk. Conversely, individuals with a low genetic risk likely lack these strong, constitutive genetic variants, leaving their lipoprotein metabolism more normative and responsive to n-3 PUFA modulation. Future trials should investigate whether even higher dosages of marine n-3 PUFAs can overcome the attenuated treatment response observed in individuals with genetically determined elevations in remnant cholesterol.

Our subgroup analysis also revealed that the anti-atherosclerotic effect was significant in women but not in men, which could be attributed to differences in fat distribution and metabolism [63]. Additionally, women generally exhibited healthier lifestyles than men, which may have contributed to the observed difference in effect. While our exploratory subgroup analyses suggested potential variations in the treatment effect, these findings should be interpreted with caution as they were derived from post-hoc analyses and were not adjusted for multiple comparisons.

Strengths and limitations

The strengths of our RCT include a large sample size, high adherence, use of both low and high doses, a neutral placebo, and focus on T2D patients in China, where n-3 PUFA deficiency is severe. We assessed effects on NMR-derived lipoprotein subclasses and explored genetic interactions, providing implications for personalized n-3 PUFA treatment. However, the trial has limitations. First, limitations for NMR include potential sensitivity issues with minor subclasses and the high cost and complexity of data interpretation. Despite these drawbacks, NMR remains a powerful tool for lipidomics, providing valuable insights into lipoprotein-related cardiovascular risk. Second, the observed dissociation between a significantly improved atherogenic lipoprotein profile and the absence of immediate plaque reduction underscores that the regression of established atherosclerosis is a slower process. This indicates that a longer follow-up period may be necessary for the biochemical benefits of n-3 PUFAs—specifically, the reduction of remnant cholesterol and atherogenic lipoprotein subfractions—to translate into measurable structural changes in the carotid artery wall. While our sample size provided adequate power (80%) to detect our pre-specified, clinically meaningful effect (33% relative risk reduction), the observed point estimate suggested a larger treatment effect (51% relative risk reduction) than anticipated. The resulting wide confidence intervals indicate that our study was underpowered to precisely estimate this larger effect size, which should be considered in the interpretation of the non-significant primary outcome. Furthermore, while our sample size is large for its specific focus (the largest trial of its kind in China), it remains smaller than that of major primary prevention RCTs, which may affect the precision of our estimates for specific patient subgroups. Third, plaque stability, defined by features such as echogenicity, surface characteristics, and composition, was not specifically assessed in our study, which could have provided more implications. Fourth, differences in lipoprotein subclasses between doses may be due to divergent EPA and DHA effects. Trials using purified EPA in T2D patients are needed. Fifth, medication use during the intervention may have influenced results, but findings remained consistent after further adjustment. The low prevalence of lipid-lowering therapy in our trial, while reflective of the broader Chinese population [33], allowed for the assessment of marine n-3 PUFAs' effect on atherosclerosis in a statin-naïve context. Although our trial findings were generalizable for many low-income and middle-income countries with a low rate of statin use [64], future studies are warranted to determine the effect in patients already receiving optimized statin therapy. Last, our findings regarding genetic interactions should be considered exploratory and require validation in independent RCTs. Nevertheless, their biological plausibility is supported by established literature detailing how n-3 PUFAs modulate the expression of genes central to remnant cholesterol synthesis and clearance.

This trial demonstrated that 14 month marine n-3 PUFA supplementation did not significantly reduce carotid plaque risk but did improve specific atherogenic lipoprotein profiles in Chinese patients with T2D. Our trial does not support a broad recommendation for marine n-3 PUFAs to prevent carotid atherosclerosis in T2D, it generates the hypothesis that patients with low genetic risk for remnant cholesterol might derive benefit. This observation warrants validation in studies specifically designed to test gene-supplement interactions. Ultimately, confirming cardiovascular benefits in diabetic populations and identifying responsive subgroups through precision medicine approaches are crucial next steps to inform clinical practice.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P.Z. drafted the manuscript and supervised data analysis. X.H.L. conducted data analysis, organized tables and figures, and revised the manuscript. P.Z., W.J., W.F.Z., X.Z.W, W.M.C., X.H.L., X.C.L., H.Y.L., W.Z.T., Y.L., Y.Q.W, Y.A., Y.R.L., H.Y., F.W., L.Z., L.G.Z., D.H.M., Y.M.T., A.L.W., S.Y.W., F.H.Z., Y.J.Z., H.B.S., X.M.Y., and F.Z. were responsible for population-based sample collection, with P.Z., W.J., W.F.Z., and W.C. also managing fieldwork and follow-up supervision. The PNMD-Trial Group was responsible for follow-up tracking and management. X.H.L., Y.L., Y.A., Y.Q.W., D.H.M., L.G.Z., W.J., and X.Z.W. performed plasma fatty acid profiling. J.P.G., X.M.W., X.H.L., Y.A., and Y.H. conducted NMR metabolomics analysis and contributed to data interpretation. Y.Z. was involved in the study design, the review of data, the manuscript editing, and review. J.J.J. conceived and designed the study, obtained funding, oversaw project administration, and critically revised the manuscript. All authors reviewed and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (grant no. 82473610), Fund of the National Natural Science Foundation of China (grant no. 32202057), Natural Science Foundation of Zhejiang Province (grant no. LY23C200007) and Young Elite Scientists Sponsorship Program by CAS (grant no. 2022QNRC001).

Data availability

To ensure patient confidentiality, individual de-identified participant data will not be made publicly available. Bona fide researchers can request access to these data by submitting a proposal to the PNMD Steering Committee (y_zhang@zju.edu.cn). Approval is subject to a review of the scientific merit and proposed use of the data.

Code availability

The code supporting the findings of this study is available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This trial was conducted in compliance with the Declaration of Helsinki and approved by the ethics committee for human research (reference no. 20180301) at Lanxi Red Cross Hospital in Jinhua, China. Written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare that there is no conflict of interest.