Randomized Placebo‐Controlled Trial Assessing the Effect of 24‐Week Fenofibrate Therapy on Circulating Markers of Abdominal Aortic Aneurysm: Outcomes From the FAME‐2 Trial

Study Design

The full trial protocol has been previously published.19 In brief, FAME‐2 was a randomized, double‐blind, multicenter, parallel, and placebo‐controlled clinical trial in which participants were allocated to receive capsules containing fenofibrate (145 mg Lipidil), or inert cellulose powder (placebo) once daily for 24 weeks. The appearance of the capsules and associated packaging were identical for intervention and placebo. The study included men and women aged 60 years and above who had an AAA measuring 35 to 49 mm on computed tomography or ultrasound, and no anticipated requirement for AAA repair within 12 months of recruitment. Participants were excluded if they already took fenofibrate or related fibrates; had a known contraindication to fenofibrate, had previous abdominal aortic surgery, had a mycotic AAA, or if they were currently enrolled in any other drug trial. Patients were randomized to treatment or placebo groups on a 1:1 ratio using a computer‐generated randomization list. Patient randomization was stratified according to recruitment site and initial AAA diameter (35–39, 40–44, and 45–49 mm). All patients and investigators were blinded to treatment allocation. The study was approved by The Prince Charles Hospital Human Research and Ethics Committee (13/QPCH/16), the Therapeutic Goods Administration (Trial number 2013/0415) and was registered with the Australian and New Zealand Clinical Trial Registry before commencing recruitment (ANZCTR12613001039774). All participants provided written informed consent and the trial was conducted with appropriate institutional ethics approvals.

Clinical Risk Factors and Medications

Baseline characteristics collected for each patient included sex, age, history of hypertension, diabetes mellitus, coronary heart disease (CHD), and prescribed medications. Diabetes mellitus and hypertension were defined based on a previous diagnosis and/or treatment for these conditions. Smoking history was classified as never smoker, current smoker, or ex‐smoker (ie not having smoked within 1 month of recruitment). CHD diagnosis was based on a history of angina, myocardial infarction, or coronary artery revascularization. Height and weight were measured in accordance with the guidelines of the International Society for the Advancement of Kinanthropometry. Body mass index was calculated as the body mass in kilograms divided by the square of height in meters.

Quality Assurance

Fenofibrate has an established ability to reduce circulating triglyceride concentrations and is currently prescribed for this indication.16, 24 The effect of fenofibrate on triglyceride concentration was used in addition to capsule counting as measures of drug compliance. For the purposes of the current study, patients were considered compliant if they had taken ≥80% of allocated capsules.

Outcome Measures

The primary outcome of the trial was the difference in circulating concentrations of osteopontin and kallistatin between groups during follow‐up. Secondary outcomes included differences in circulating concentrations of lipids (high density lipoprotein‐cholesterol, low‐density lipoprotein‐cholesterol, very low‐density lipoprotein‐cholesterol, and total cholesterol), MMP‐9, resistin, osteoprotegerin, and D‐dimer. A tertiary outcome was AAA diameter.20, 23, 25

Blood Tests

Blood samples were collected from participants at entry, week 3, and week 24. Circulating concentrations of MMP‐9, resistin, osteoprotegerin, D‐dimer, and kallistatin were measured by commercial ELISA according to manufacturers’ instructions (R&D Systems, Minneapolis, MN and Technozym, Vienna, Austria). Reproducibility for these assays was excellent (intraclass correlation 0.84–1.00) in accordance with accepted guidelines.26 Serum concentrations of triglycerides, high‐density lipoprotein‐cholesterol (HDL‐C), low‐density lipoprotein (LDL)‐C, homocysteine, and creatinine were measured by hospital pathology departments using automated assays as previously described.27, 28 Very low‐density lipoprotein cholesterol (VLDL‐C) concentrations were estimated using the formula total cholesterol—(HDL‐C+LDL‐C) as previously described.27 Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease‐Epidemiology Collaboration group (CKD‐EPI).29

Measurement of AAA Diameter

AAAs were imaged by ultrasound at entry and 24 weeks by an experienced sonographer using a published standard protocol.30 Orthogonally oriented cross‐sectional static images of the infrarenal aorta obtained in the anterior–posterior plane were acquired from the site of maximum AAA diameter. All images were measured at the central coordinating center by 2 independent experienced readers blinded to patient allocation using a standard protocol. For each image, 3 measurements were made: (1) inner aortic wall to inner aortic wall; (2) leading edge to leading edge; and (3) outer aortic wall to outer aortic wall. The intraobserver reproducibility of the 2 observers was assessed by repeatedly measuring scans from 50 randomly selected patients at least 24 hours apart, with blinding to the original readings as previously described.31 Intrarater repeatability coefficients for both readers were <3 mm for each measurement type.

Adverse Events

Adverse events (AEs) and serious adverse events (SAEs) were defined based on guidelines published by the Australian National Health and Medical Research Council.32 AEs and SAEs were ascertained at 3‐ and 12‐week phone calls and a 24‐week visit. All AEs were recorded by the trial coordinators and reviewed by site principal investigators. All SAEs were reported to the Human Research Ethics Committee and study sponsor.

Sample Size

Sample size calculations for this trial were determined in line with the anticipated effects of fenofibrate on the primary outcome measures. Previous data suggest that fenofibrate reduces aortic osteopontin expression by ≈95%, and that patients receiving fibrates for 28 days exhibit an ≈30% reduction in circulating osteopontin concentrations.18, 33 Collectively, this led us to anticipate that 24 weeks of fenofibrate therapy would reduce serum osteopontin concentration by at least 30%. Pilot in vitro data led us to hypothesize that serum kallistatin concentrations would increase by 30% in patients allocated to fenofibrate (manuscript under review). These effect sizes were also considered to represent a clinically significant reduction in AAA pathology. It was hypothesized that serum osteopontin and kallistatin concentrations would remain unchanged in patients allocated to placebo. Assuming a dropout rate of 20%, sample size calculations indicated that 60 patients per group were required to demonstrate anticipated biomarker differences with 90% power. Since there were 2 primary outcome measures, a 2‐tailed α of 0.025 was used in these calculations to account for multiple testing. In May 2015 (after ≈50% of the planned participants had completed the protocol), drug compliance was assessed by capsule counting. This assessment suggested that 22% of the patients had taken fewer than 95% of the allocated tablets, or had withdrawn from the study. In view of this, the study sample size was increased by 20 participants (10/group) to minimize the risk of underpowering.

Statistical Analyses

Skewness and kurtosis tests indicated that many continuous variables collected at baseline were not normally distributed. Continuous baseline data are therefore presented as median and interquartile range, and were compared between groups using the Mann–Whitney U test. Nominal variables were compared between groups using the χ² test (SPSS v22; IBM). Any baseline variable with a P<0.100 following these tests was considered to be a potential confounder. Changes in the circulating concentrations of blood tests and AAA diameter were initially compared between groups using random‐intercept unadjusted linear mixed effects (LME) models (R car and nlme packages) as previously described.34, 35, 36, 37, 38 In all models, time (as a factorial variable) and allocation were treated as fixed effects, and patients as random effects. Differences in blood markers and AAA growth between groups were further assessed in multivariable LME models including CHD and prescriptions of statins, β‐blockers, and antiplatelet and anticoagulant medications (because of potential confounding) as additional fixed effects.

For all LME models, fit was assessed through inspection of standardized residual distribution and q‐q norm plots. The response variables were natural‐log transformed to conform to model assumptions where necessary. For all LME analyses, a significant interaction (P<0.025) between time and treatment allocation was considered to denote a difference between groups. For some variables, residual plots suggested the presence of potentially influential outliers, which may have driven positive associations. Where this was the case, subset analyses were performed to exclude the impact of such outliers (defined as observations with a residual >2 standard deviations away from the mean of model residuals) on observed findings. Significant differences were accepted to be true if the interaction between time and allocation was upheld in these subset analyses. Exploratory analyses of the primary outcome measures revealed no difference in biomarker responses between recruitment centers, suggesting that there was minimal study site bias in the current study.

All analyses were conducted by an intention‐to‐treat approach using all available data. According to the study protocol, participants were to provide 3 blood samples (baseline and after 3 and 24 weeks), meaning a total of 420 samples were expected for analysis. Of these, 8 (1.9%) samples were unavailable for measurements of osteopontin, kallistatin, D‐dimer, MMP‐9, and resistin; and 11 (2.6%) samples were unavailable for assessment of serum concentrations of cholesterol, triglycerides, homocysteine, C‐reactive protein, and creatinine. AAA diameter measurements were unable to be performed on 3 (2.1%) patients at 24 weeks because of their death or withdrawal from the study (Figure 1). Missing data were not imputed because they were not randomly distributed throughout the population, and we have highlighted actual numbers included within both groups at each time point. All data analyses were audited by several authors to ensure accuracy before submission.

Patient Flow

A total of 140 patients were randomized to fenofibrate or placebo (n=70/group; Figure 1). Thirteen serious adverse events (SAEs) were reported throughout the trial, and were evenly distributed between groups (6 [8.6%] versus 7 [10.0%]; P=0.771, Table 1). Only 1 SAE was considered to relate to the trial medication, which was a clinically significant decrease in eGFR within 3 weeks of starting fenofibrate. The drug was stopped and eGFR increased. All patients allocated fenofibrate were followed through to the end of the trial, although 12 (17.1%) had stopped taking the allocated intervention. Three (4.3%) patients allocated placebo did not complete the trial (2 withdrawals and 1 death), and an additional 6 (8.6%) had stopped taking trial medication. In total, 137 (97.9%) of the randomized patients were followed to the end of the trial, and 122 (87.1%) remained on trial medication for the duration of the study. There were no protocol violations during the trial.

Patient Characteristics

The demographic characteristics of the randomized patients are shown in Table 1. In accordance with the study design, patients in both groups were matched for initial AAA diameter at recruitment (Table 1). Patients in the placebo group were more likely to have CHD and be prescribed aspirin, other antiplatelet medications, β‐blockers, and statins, and less likely to be prescribed oral anticoagulants. Baseline concentrations of total cholesterol and LDL‐C were slightly lower in the placebo group compared with those allocated fenofibrate (Table 2). No other differences in cardiovascular risk factors or prescribed medications were observed between groups.

Compliance

Capsule‐counting data suggested that 119 (85%) patients had adhered to the medication regimen. No significant difference in compliance rate was observed between groups (81.4% compliant in the fenofibrate group versus 88.6% in the placebo group; P=0.237, Table 1). Serum triglyceride concentrations decreased in participants allocated fenofibrate, and were significantly different between groups at 3 and 24 weeks (interaction between time and allocation P<0.001 for both time‐points following LME). eGFR was significantly lower, and circulating homocysteine concentration higher, in patients randomized to fenofibrate compared with those on placebo 3 and 24 weeks after starting the allocated drug (Figure 2 and Table S1).

Effect of Fenofibrate on Serum Osteopontin and Kallistatin Concentrations

Serum osteopontin concentrations were comparable between groups at the time of recruitment and fluctuated over time for all participants, although no difference in these trends was observed between groups (Figure 3A). No difference in circulating kallistatin concentrations were observed between groups during follow‐up (Figure 3B). Inclusion of other covariates to adjust for potential confounders did not change these findings (Table S1).

Effect of Fenofibrate on Other Circulating Markers and AAA Diameter

No differences in circulating concentrations of MMP‐9, resistin, or D‐dimer were found between groups (Figure 3, Table S1). Serum osteoprotegerin concentrations appeared to decrease over time in patients allocated fenofibrate by comparison to those in the placebo group, although this was not statistically significant (P=0.095 for interaction between time and treatment; Figure 3, Table S1). Fenofibrate reduced serum concentrations of total cholesterol and very low‐density lipoprotein cholesterol, and these reductions remained robust after adjusting for potential confounders and excluding outliers (Figure 4, Table S1). Serum LDL‐C and high‐density lipoprotein‐cholesterol concentrations during follow‐up were similar between groups. No significant difference in AAA growth rate was observed between groups using any of the assessed measurements (Figure 5, Table S2).