Fenofibrate Use Is Associated With Lower Mortality and Fewer Cardiovascular Events in Patients With Diabetes: Results of 10,114 Patients From the Korean National Health Insurance Service Cohort

Patients with diabetes are at higher risk of cardiovascular disease (CVD) (1). Even with statin therapy to lower LDL cholesterol in the contemporary era, the residual risk of CVD remains substantial in these patients. Triglycerides (TG) are another target for the prevention and treatment of CVD, in addition to LDL cholesterol, for residual risk reduction. Major epidemiological (2,3) and genetic linkage studies using Mendelian randomization methods (4) have suggested that a high TG level is associated with CVD and even has a causal relationship with CVD (5).

Therefore, in an effort to reduce the remaining risk, the strategy of lowering TG levels has been investigated, but the results have been contradictory, although it has shown some potential in specific subset of patients. Large randomized clinical trials, such as the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) (6) and Action to Control Cardiovascular Risk in Diabetes (ACCORD) (7) studies, which tested the efficacy of fenofibrate alone or as an add-on to statin in patients with diabetes, demonstrated overall negative outcomes, but some signs of improved clinical outcomes were detected in subgroup of patients with diabetes with low HDL cholesterol and high TG levels (6,7). Moreover, a recent study, the Reduction of Cardiovascular Events With Icosapent Ethyl–Intervention Trial (REDUCE-IT), aiming to reassess the efficacy of a specific formula of omega-3 fatty acid provided new insight into the lowering of TG levels and the novel concept of introducing a specific form of omega-3 fatty acid, an icosapent ethyl acid, for the secondary and primary prevention of CVD, especially in patients with diabetes (8,9).

Despite these promising data, there has been no large-scale long-term follow-up study to assess TG-lowering therapy with fenofibrate in patients with diabetes in real-world practice. We investigated the true efficacy of the TG-lowering drug fenofibrate on clinical outcomes in patients with diabetes in a real-world setting using the South Korean National Health Insurance Service (NHIS) database.

We used the data from the South Korean NHIS cohort. A detailed description has been published and validated previously (10,11). Briefly, the NHIS provides obligatory health care for all South Korean citizens, with an enrollment rate of 97% (11). The NHIS cohort had information on socioeconomic status, admission and discharge information, hospital visit data, drug prescription data, and hospital information. A health examination is performed by the NHIS annually or biannually for all people aged ≥20 years, which is composed of a survey on health status and behaviors, body weight, height, blood pressure, and laboratory tests for blood and urine.

The NHIS provides the aforementioned data as well as death records, including cause and date, which are merged from Statistics Korea for research purposes. We used the NHIS Health Examination Database 2.0 sample cohort combined with death data including people aged 40–79 years; this group comprised 10% (n = ∼510,000) of the original NHIS full cohort of 5,150,000 people who underwent the national heath examination test in 2002–2003.

This study was approved by the Hallym University Sacred Heart Hospital Institutional Review Board (no. 2019-01-021-005). Informed consent was waived, because the database was made to be anonymous under the strict confidentiality guidelines of the NHIS (11).

We used the NHIS Health Examination Database 2.0 sample cohort, which included ∼510,000 people aged 40–79 years who underwent a health examination test in index years of 2002–2003. They were followed until 2015. Within this cohort, we enrolled 212,969 patients with type 2 diabetes first diagnosed between 2003 and 2014. Diagnosis of diabetes was made according to ICD-10 codes E11–E14. After excluding 149,242 people who had a combined type 1 diabetes code, history of fenofibrate or omega-3 fatty acid prescription before diagnosis of diabetes, and no data on laboratory tests, 63,727 patients were candidates for analysis (Supplementary Fig. 1).

To compare fenofibrate users versus nonusers, we divided these patients into two groups: patients receiving fenofibrate only and nonusers of fenofibrate and/or omega-3 fatty acid (controls). We excluded those receiving both omega-3 fatty acid and fenofibrate in the control group to investigate strictly the efficacy of fenofibrate. Patients who were prescribed fenofibrate at least once from January 2003 to December 2014 after diagnosis of diabetes were considered the study drug group, and those who were never prescribed either drug were considered the control group. Patients were matched 1:1 with propensity scoring, and 5,057 were included in each group (Supplementary Fig. 1).

The primary end points were a composite of the first occurrence of myocardial infarction (MI), stroke, percutaneous coronary intervention (PCI), and cardiovascular death. Occurrence of any event among the four end points of MI, stroke, PCI, and cardiovascular death was considered to be a primary end point occurrence. Secondary end points were the first occurrence of each individual event (MI, stroke, PCI, or cardiac death) and all-cause death.

Each outcome was identified by ICD-10 code, in accordance with the American Heart Association statistics guidelines (12); death was categorized as all-cause death if a death record was present and as cardiac death if both a death record and ICD-10 code of I (ischemia code) were present.

We also examined the covariates of age; sex; BMI; comorbidities; history of statin intake; history of PCI; laboratory data such as TG, HDL cholesterol, and LDL cholesterol; history of smoking; history of alcohol intake; and household income. MI was confirmed if the ICD-10 code was I21–I23, and stroke was confirmed if the ICD-10 code was I60–I64. Diagnosis of comorbidities such as angina (I20), other coronary artery disease (CAD) (I24–I25), chronic kidney disease (N18), cancer (C**), hypertension (I10), atrial fibrillation (I48), and heart failure (I50) was confirmed if ICD-10 code for each disease was present. The ICD-10 codes for use of the study drugs (fenofibrate, omega-3 fatty acid, and statins) and receipt of PCI are listed in Supplementary Table 1. The median follow-up period was 3.06 years (interquartile range [IQR] 1.94–4.45): 2.86 years (IQR 1.90–4.19) in the fenofibrate group, and 3.30 years (IQR 2.01–4.66) in the nonuser group.

We performed a subgroup analysis under the categories of BMI, with a cutoff value of 25 kg/m²; TG, 200 mg/dL; HDL cholesterol, 35 mg/dL; and LDL cholesterol, 130 or 100 mg/dL, as well as HDL cholesterol <35 mg/dL, TG ≥200 mg/dL, sex, age, income, and some medical and social histories.

We tried to overcome bias and the imbalance of baseline characteristics and control for confounders by performing a 1:1 propensity score matching (PSM) analysis with greedy nearest-neighbor matching between the fenofibrate and control groups. We entered age, sex, BMI, household income stratified by high and low levels (low level included 0–3, and high level 4–10 in decile classification of original data), LDL cholesterol, TG, HDL cholesterol, alcohol habits, smoking habits, statin intake, and aforementioned comorbidities including history of MI, angina, other CAD, stroke, cancer, chronic kidney disease, hypertension, atrial fibrillation, heart failure, and PCI.

Categorical variables were presented as numbers and percentages and compared using the χ² test or Fisher exact test. Continuous variables are presented as mean ± standard deviation and were compared using the independent sample Student t test.

After PSM, we used the paired Student t test for continuous variables in comparing the two groups and the Cochran-Mantel-Haenszel test for categorical variables. We categorized age with a cutoff value of 65 years; BMI, 25 kg/m²; TG, 200 mg/dL; HDL cholesterol, 35 mg/dL; and LDL cholesterol, 100 or 130 mg/dL for analysis. We also calculated the absolute standardized difference to examine the differences in comorbidities between the fenofibrate and control groups, regarding it as negligible if the value was <0.1.

The main outcomes were identified after PSM analysis, and we performed logistic regression analysis. We present the absolute event rate and the event rate per 1,000 person-years for primary end points, individual events, and all-cause mortality.

In addition, a stratified Cox proportional hazards regression analysis was used for risk analysis. The hazard ratio (HR) and 95% CI are presented. The incidence of the end points is shown with Kaplan-Meier curves, and differences were assessed using the stratified log-rank test.

A subgroup analysis was conducted to compare the incidence of the primary outcomes across different clinical conditions that can influence prognosis, and an interaction analysis was conducted to examine heterogeneity. A P value of <0.05 (two sided) or absolute standardized difference of ≥0.1 was regarded as statistically significant. Statistical analyses were performed with SAS Enterprise Guide 7.1. (SAS Institute).

The fenofibrate or omega-3 fatty acid group included 8,383 patients, and the nonuser group included 55,344. After excluding 3,244 patients using omega-3 fatty acid, 5,139 were fenofibrate only users (Supplementary Fig. 1). After 1:1 PSM, 10,114 patients were allocated, 5,057 to the fenofibrate only group and 5,057 to the nonuser group. Baseline characteristics of patients are presented in Table 1. The overall characteristics were well balanced after PSM, except for TG ≥200 mg/dL and HDL cholesterol <60 mg/dL (Table 1). The median follow-up was 3.1 years.

The primary composite outcome was significantly lower in the fenofibrate group compared with the control group (13.4 vs. 15.5 per 1,000 person-years; HR 0.76; 95% CI 0.62–0.94; P = 0.010) (Table 2 and Fig. 1A).

Incidence of the individual outcome of stroke was lower in the fenofibrate group compared with the control group (6.5 vs. 8.6 per 1,000 person-years, respectively; HR 0.621; 95% CI 0.463–0.833; P = 0.0015) (Table 2 and Fig. 1C). Rates of cardiac death and all-cause death were also significantly lower in the fenofibrate group, at 1.8 vs. 3.1 per 1,000 person-years (HR 0.59; 95% CI 0.352–0.987; P = 0.0446) and 7.6 vs. 15.3 per 1,000 person-years (HR 0.437; 95% CI 0.340–0.562; P < 0.0001), respectively (Table 2 and Fig. 1E and F).

However, MI and PCI were not different between the fenofibrate and control groups, at 1.6 vs. 1.6 per 1,000 person-years (HR 1.158; 95% CI 0.627–2.139; P = 0.640) and 5.7 vs. 4.7 per 1,000 person-years (HR 1.159; 95% CI 0.827–1.623; P = 0.392), respectively (Table 2 and Fig. 1B and D). Kaplan-Meier analysis showed statistically significantly favorable efficacy with fenofibrate only use versus nonuse in the primary outcome, stroke, cardiac death, and all-cause death (Fig. 1).

When we stratified the duration of fenofibrate use by quartile (Q), with Q1 as 1–59 days, Q2 as 60–193 days, Q3 as 194–485 days, and Q4 as ≥486 days, in increasing order, the adjusted HR of the composite outcome in the Q1 group increased to 1.52 (model 8; 95% CI 1.169–1.977; P = 0.0018) (Supplementary Table 2), with no change in the Q2 or Q3 group compared with the control group. However, the risk decreased in Q4, with an HR of 0.347 (95% CI 0.226–0.532; P < 0.0001) (adjustment model 8). (Supple mentary Table 2 and Supplementary Fig. 2).

In the subgroup analysis, the favorable effect of fenofibrate was sustained consistently across all subsets of patients (Fig. 2). Notably, the efficacy of fenofibrate persisted regardless of LDL cholesterol, HDL cholesterol, TG level, or the combination of both high TG and low HDL cholesterol levels. The efficacy was more pronounced in men, patients with a history of stroke, and patients with LDL cholesterol of <100 mg/dL, with significant P values for interaction.

We demonstrated that fenofibrate use is associated with a better long-term clinical outcome of the composite of MI, stroke, PCI, and cardiac death, as well as some individual outcomes, including stroke, cardiac death, and all-cause death, in patients with type 2 diabetes. The benefit was consistent across all subgroups and was pronounced in longer-duration users of fenofibrate.

Our study has value in its large scale, its reflection of a real-world practice setting, and its definite and consistent results after thorough multiple adjustments of covariables and subgroup analyses.

Patients with diabetes are prone to a more atherogenic environment, with small dense LDL cholesterol, high TG-rich lipoprotein (TRL) cholesterol, and low HDL cholesterol (13–15). An elevated TG level, aside from the number of particles carrying TG, has been reported to be associated with CVD, despite of some controversies, even when LDL cholesterol is under control with statin treatment (16–18). Furthermore, some genetic studies using Mendelian randomization methods have raised the possibility that TG or TRL cholesterol could cause CVD (4,19). Therefore, we can assume that lowering of TG levels can be a therapeutic target, and fenofibrate can be an alternative to statins to reduce residual risk, especially for patients with diabetes (15,20). Koreans and other Asian people are appropriate candidates for TG lowering, because they tend to have a lipid profile including high TG and low HDL cholesterol, mainly as a result of a sedentary lifestyle and high-carbohydrate diet (21). Thus, Korean patients with diabetes may be fit for fenofibrate treatment, and our study results support this. Moreover, a recent South Korean study corresponds well with our study in that the authors also showed positive results with fenofibrate in a population of patients with metabolic syndrome, a group at lower risk than our patients with diabetes (22). Considering that the cardiovascular risk of patients in our study was higher (e.g., 10-fold CAD history) as well as the fact that our entire study population had diabetes and the population was larger (2.34-fold), our study results firmly support the role of fenofibrate in Koreans.

Previous trials have been frustrating, mainly because sample sizes have been relatively inadequate to determine statistical significance or because patients with diabetes have already been using statins, as was the case in the ACCORD trial (7). However, in the FIELD study, the benefit of fenofibrate was seen more prominently than in the ACCORD trial in that incidence of nonfatal MI and total CVD was lower in the fenofibrate group, despite the overall failure to show a significant risk reduction (6). We believe this was mainly due to the nonuse of statins at baseline and the larger sample size compared with the ACCORD trial. In both trials, a specific subset of patients with high TG and low HDL cholesterol benefited from fenofibrate, even with the use of statins, because these patients were at a risk high enough to realize the effect of fenofibrate. Our study population was much larger at ∼10,000 (ACCORD n = ∼5,500; FIELDS n = ∼6,000), the baseline risk was higher, with a CAD history of 37% among our participants (ACCORD 36%; FIELDS 22%), and the rate of statin use was 65% compared with ACCORD (statin use 100% in the active drug group of simvastatin plus fenofibrate) and FILED (no statin use). These factors may have contributed to the definite positive results seen in the fenofibrate group in our study.

Notably, when we consider the moderate to high levels of baseline non-HDL and indirect TRL cholesterol (total cholesterol − HDL cholesterol − LDL cholesterol) along with the moderate LDL cholesterol (110 mg/dL) and TG levels (240 mg/dL), it is possible for fenofibrate to have had a beneficial effect in our study (18) (Table 1). In view of the importance of remnant lipoprotein particles in the setting of elevated TG in diabetes, the indirect calculation of TRL cholesterol has value and may help interpret the positive results of our study.

The ACCORD lipid extension study, which extended the follow-up period of the original ACCORD study by 5 years (total of 10 years), still demonstrated an overall neutral effect. However, it also showed a potential benefit of fenofibrate in a specific subset of patients with high TG and low HDL cholesterol (23). Another extended posttrial follow-up study (total of 10 years) of the ACCORD trial also demonstrated a legacy effect of fenofibrate, with reduced CVD death and all-cause death. These results complement our results (24).

Our study showed consistently favorable results with fenofibrate in all subsets of patients across all TG, HDL cholesterol, and LDL cholesterol levels, as well as in the group with a combination of high TG and low HDL cholesterol, irrespective of statin use. Moreover, the effect was more pronounced in longer-duration users (Q4) of fenofibrate, whereas users in Q1 experienced a high event rate in composite outcomes. This might have occurred by chance, but it might also be explained by the high probability of experiencing a clinical event among patients at risk of CVD in the initial stage of study drug use, when duration of use was not yet long enough for the drug to have efficacy. Consistent subgroup analysis and examination of the effect of duration on response are major strengths of our study.

To induce benefits in addition to lowering TG levels, we believe peroxisome proliferator–activated receptor α (PPAR-α) agonists, which have anti-inflammatory and antioxidative effects and improve endothelial function, could contribute to the beneficial effect of fenofibrate (25–27). More specifically, PPAR-α agonists modulate the expression of cytokines and adhesion molecules and have antiatherosclerotic effects (28). They reduce activation of nuclear factor-κB and prevent nuclear factor-κB translocation to the nucleus, terminating activation of the inflammatory pathway (28). In vascular smooth muscle cells, activation of PPAR-α inhibits interleukin-1–induced production of interleukin-6 and prostaglandins as well as the expression of cyclooxygenase II (28).

Our study demonstrated the definite and consistent beneficial effect of fenofibrate in a Korean population of patients with diabetes. For Koreans and presumably other Asian people with diabetes who have a sedentary lifestyle and carbohydrate-oriented food culture, lowering of TG with fenofibrate, with or without statins, could potentially improve cardiovascular outcomes, including reduction of cardiac death and all-cause death rates, regardless of TG and/or HDL cholesterol level.

Our study has some limitations. First, this was a population-based cohort study, which has inevitable biases, even though we tried to avoid expected confounders by PSM analysis. However, hidden biases may have remained. Therefore, our real-world results should be interpreted with caution, in contrast to the results of a randomized clinical trial. We recognize there may be limitations in drawing the conclusion of fenofibrate benefit in patients with type 2 diabetes based on our study results. However, we consider our results to be in accordance with previous registry data, and fenofibrate might be a treatment option for high-risk patients with diabetes, at least in the Korean population (22).

Second, the diagnosis codes for baseline comorbidities could have been missed or inaccurate as a result of, for example, government insurance reimbursement issues. Third, the criteria for inclusion in each group were dependent on prescription of the study drug at least once. The true drug intake and adherence are uncertain. To overcome this limitation, we analyzed the group according to drug use duration by Q and found efficacy of the study drug with longer-duration use. Fourth, because of a lack of data, we do not present information on drug name or dose or patient adherence, which could have made some difference in the study results. Data on blood and urine examinations were also sparse. Fifth, our study results were derived from a population-based cohort based on data from the South Korean NHIS. We acknowledge that the results indicating a fenofibrate benefit cannot necessarily be applied to other ethnicities or populations. Finally, we could not provide information on the level of LDL cholesterol achieved or other lipid data, which could have affected the outcomes of fenofibrate users, because we had so few consecutive data on these factors.

Fenofibrate use was associated with a reduced risk of cardiovascular events, cardiac death, and all-cause death during 3-year follow-up in patients with type 2 diabetes in a real-world setting.

Funding. This research was supported by grant HURF-2020-12 from the Hallym University Research Fund 2020.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.-H.J. conceptualized and designed the study and wrote and reviewed the manuscript. H.N., Jeongwoo Lee, S.P., Jungkuk Lee, and D.-S.K. contributed to the statistical analysis, supported software, and reviewed the manuscript draft. S.-H.J. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.