Effects of a Potent and Selective PPAR-α Agonist in Patients With Atherogenic Dyslipidemia or Hypercholesterolemia:  Two Randomized Controlled Trials FREE

Fibric acid derivatives (fibrates) are among the oldest agents used to treat patients with lipid disorders.¹ Clofibrate, the original agent in this class, was introduced in the 1960s but is now rarely prescribed. In current practice, 2 fibrates, gemfibrozil and fenofibrate, are widely used to treat a constellation of lipid abnormalities known as atherogenic dyslipidemia.^2- 4 The features of this syndrome comprise a cluster of risk factors that include elevated serum levels of triglycerides, low levels of high-density lipoprotein cholesterol (HDL-C), and a more atherogenic form of low-density lipoprotein cholesterol (LDL-C), known as small-dense LDL-C.⁵ Fibrates markedly lower serum triglycerides levels and modestly increase levels of HDL-C. Two randomized controlled trials of gemfibrozil demonstrated a reduction in adverse cardiovascular outcomes, particularly in patients with atherogenic dyslipidemia.^6- 7

The mechanism of action of fibrates is mediated through modulation of the peroxisome proliferator-activated receptor α (PPAR-α).^8- 10 PPARs are ligand-activated nuclear transcription factors that modulate gene expression with major effects on lipid and glucose metabolism.⁹ However, currently available fibrates are weak ligands for the PPAR-α receptor and may interact with other PPAR systems. Accordingly, the pharmaceutical industry has sought to develop new, more potent and selective agents within this class. However, none of the novel PPAR-α agonists has achieved regulatory approval. According to a former safety officer in the US Food and Drug Administration (FDA), more than 50 PPAR agonists have been discontinued due to various types of toxicity.¹¹ Nearly all of the new PPAR-modulating agents have been discontinued without scientific publications describing the reasons for termination of the development programs.

We conducted 2 randomized controlled trials to characterize a potent and selective PPAR-α agonist, known as LY518674, which is approximately 10 000 times more potent than fenofibrate.^12- 13 These initial studies assessed the safety and efficacy of this agent at varying doses as monotherapy or in combination with atorvastatin. This report represents the first published description of the potential benefits and risks of a novel, highly potent and selective PPAR-α agonist.

The study design consisted of 2 separate trial protocols intended to evaluate the safety and efficacy of LY518674 in 2 populations, one with atherogenic dyslipidemia and the other with hypercholesterolemia. These 2 trials were multicenter, randomized, double-blind, double-dummy, sequential, parallel, placebo-controlled studies and were conducted concommitantly. The institutional review boards of all participating centers approved the protocols, and all patients provided written informed consent. Both protocols specified enrollment of patients between 18 and 80 years of age.

In the atherogenic dyslipidemia study, patients were required to have an HDL-C level less than 45 mg/dL (1.2 mmol/L) for men or 50 mg/dL (1.3 mmol/L) for women, triglycerides levels between 150 mg/dL and 600 mg/dL (1.7 mmol/L and 6.8 mmol/L), and an LDL-C level of 160 mg/dL (4.1 mmol/L) or less. Statins were permitted, if clinically indicated, but investigators were instructed not to adjust doses during the trial.

Patients were excluded for poorly controlled diabetes defined as a glycosylated hemoglobin level of 8% or greater, or concomitant use of insulin or a thiazolidinedione. Patients were also excluded if they had laboratory measures of thyroid function outside of the normal range; had uncontrolled hypertension, defined as 3 separate measurements exceeding 160/95 mm Hg during screening; or were receiving treatment for congestive heart failure or had a known ejection fraction less than 35%. Patients were also excluded if their levels of aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, or total bilirubin were greater than 1.5 times the upper limit of normal, if their serum creatinine levels were greater than 2 mg/dL (177 μmol/L), or if they had nephrotic syndrome, end-stage renal disease, and used renal replacement therapy such as hemodialysis or peritoneal dialysis.

All patients entered a 4-week placebo lead-in period to evaluate the effect of the National Cholesterol Education Program Therapeutic Lifestyle Changes diet.¹² Patients with triglycerides levels less than 150 mg/dL (1.7 mmol/L) at the end of this period were not permitted to proceed to the active-treatment phase. During a 12-week active-treatment period, patients were randomized to receive treatment with LY518674 (10 μg, 25 μg, 50 μg, or 100 μg) (Eli Lilly and Co, Indianapolis, Ind), fenofibrate (200 mg) (Lofibra; Gates Pharmaceuticals, Sellersville, Pa), or placebo, stratified according to concomitant statin use.

Entry criteria required an LDL-C level between 100 mg/dL and 160 mg/dL (2.6 mmol/L and 4.1 mmol/L) with concomitant statin therapy or between 130 mg/dL and 190 mg/dL (3.4 mmol/L and 4.9 mmol/L) in statin-naive patients. Exclusion criteria included the presence of diabetes, any clinical manifestation of coronary heart disease, abnormal thyroid function, heart failure, or uncontrolled hypertension. Patients were also excluded if their levels of aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, or total bilirubin were greater than 1.5 times the upper limit of normal, if their serum creatinine levels were greater than 2 mg/dL (177 μmol/L), or if they had nephrotic syndrome, end-stage renal disease, and used renal replacement therapy such as hemodialysis or peritoneal dialysis.

Following a 4- to 6-week washout period, patients entered a 4-week active-treatment period during which they were randomized to receive placebo or atorvastatin (10 mg or 40 mg). These treatment groups were stratified on the basis of their LDL-C levels into groups with values less than or greater than 160 mg/dL. Patients subsequently entered a 12-week active treatment period during which those within each of the previous treatment groups were randomized to receive placebo or LY518674 (10 μg or 50 μg).

Patients were examined during scheduled clinic visits every 1 to 2 weeks during the treatment phase and at a follow-up visit 2 to 4 weeks following cessation of the study drug. A central laboratory blinded to group assignment performed all biochemical determinations (Covance, Indianapolis, Ind). Lipoprotein levels were measured at 3 intervals during the follow-up period, while blood pressure and safety laboratory measurements were obtained at each study visit.

In the atherogenic dyslipidemia study, the protocol prespecified 2 primary end points (percentage changes in levels of triglycerides and HDL-C from baseline to 12 weeks' follow-up). A sample size of 50 patients per group was selected to provide 90% power to detect a 23% decrease in triglycerides levels (assuming an SD of 35%) and to detect a 13% increase in HDL-C levels (assuming an SD of 20%) for the comparison of each dose of LY518674 with placebo (5% type I error rate for a 2-sided test).

In the hypercholesterolemia study, the prespecified primary end point was the additional effect on LDL-C levels of each dose of LY518674 compared with placebo in patients treated with atorvastatin. Power calculations required a sample size of 23 patients per group to provide 90% power (assuming an SD of 20%) to detect a 20% change in levels of LDL-C (comparing LY518674 with placebo), with a 2-sided test and type I error rate of 5%. Assuming a dropout rate of approximately 30%, enrollment of 300 patients was specified.

Demographic and laboratory characteristics are summarized for all randomized patients taking at least 1 dose of study drug, defined as the intention-to-treat population. Categorical variables are described using frequencies, while continuous variables are reported as mean and SD or as median and interquartile range. The efficacy analyses were performed on all patients in the intention-to-treat population in whom a baseline value and at least 1 postbaseline measurement were available. The analysis of safety was performed for all intention-to-treat patients. Laboratory safety measures are reported as the change from baseline area under the curve for all values to end-of-treatment, divided by the duration of the period (AUC/t).

In the hypercholesterolemia study, the efficacy population excluded patients taking nonstudy lipid-lowering medications during the trial. The methods prespecified a likelihood function–based inference using a mixed model for repeated measures, using all available follow-up data for results with more than 1 postbaseline measurement (a compound symmetry covariance structure was assumed). Imputation of missing data was not considered in the primary analysis. Last observation carried forward was used for those results with only 1 postbaseline measurement. For the analysis of the percentage change in lipids levels, the logarithm of the ratio from end point to baseline was used as a dependent variable. For C-reactive protein level and blood pressure, the change from end point to baseline was analyzed as the dependent variable.

For the atherogenic dyslipidemia study, the model controlled for the baseline value, statin use (yes or no), age (<55 vs ≥55 years), treatment, visit, and treatment × visit interaction. For the hypercholestolemia study, the model controlled for baseline value, atorvastatin dose, LY518674 dose, visit, and all interactions between visit and dose group for atorvastatin and LY518674. For blood pressure and levels of apolipoproteins, age was also included in the model as a covariate. Analyses were performed using SAS version 8.2 (SAS Institute Inc, Cary, NC); P >.05 was used to determine statistical significance.

Between August 10, 2005, and August 22, 2006, 309 patients were enrolled in the atherogenic dyslipidemia study. The disposition of these patients is shown in Figure 1. Between August 29, 2005, and May 10, 2006, 304 patients were randomized in the hypercholesterolemia study. The disposition of these patients is shown in Figure 2.

The baseline characteristics of the patients in the atherogenic dyslipidemia population are shown in Table 1. Characteristics were similar for all treatment groups and are therefore presented in a summary fashion. The mean age was 54 years, and approximately 70% of patients were men. Most patients were obese, with a mean body mass index exceeding 32 (calculated as weight in kilograms divided by height in meters squared). As anticipated by the entry criteria, triglycerides levels were substantially elevated, with a median value approaching 250 mg/dL (2.8 mmol/L), and HDL-C levels were low, averaging less than 40 mg/dL (1.0 mmol/L). Approximately 40% of patients were taking statin drugs at baseline.

The baseline characteristics of the patients in the hypercholesterolemia study are also shown in Table 1. Characteristics were similar for all treatment groups and are presented in a summary fashion. The mean age was 54 years, and approximately half of the study participants were men. Body mass index averaged 29.4. As anticipated by the entry criteria, levels of total cholesterol and LDL-C were elevated (248 mg/dL and 164 mg/dL [6.4 mmol/L and 4.2 mmol/L], respectively). Levels of HDL-C were relatively normal, averaging 49.6 mg/dL (1.3 mmol/L).

Table 2 reports the effects of LY518674 in the atherogenic dyslipidemia study. Both fenofibrate and LY518674 markedly decreased triglycerides levels. For fenofibrate, the reduction was 32.8%, with 95% confidence intervals ranging from 24.8% to 39.9%. LY518674 produced similar reductions, ranging from 34.9% to 41.7%, which were not statistically different from the effect produced by fenofibrate, although there were not statistically significant greater reductions for the 50-μg dose of LY518674 (41.7%; P = .08 compared with fenofibrate). Fenofibrate increased HDL-C levels by 14.4% (95% confidence interval, 7.5%-21.8%). LY518674 showed an unusual dose-response pattern, with the lowest dose (10 μg) producing a modest increase (9.6%) in HDL-C levels, a maximum increase of 15.8% occurring at the 25-μg dose, and lesser increases with higher dosages. The 100-μg LY518674 dose produced a 2.1% increase in levels of HDL-C, which was not significantly different from the effect produced by placebo and smaller when compared with fenofibrate (P = .01).

LY518674 increased levels of LDL-C in this population. The least-square mean percentage increase was 2.3% for fenofibrate but was significantly greater for the 50-μg and 100-μg LY518674 doses, 18.3% and 19.5%, respectively (both P≤.001 compared with placebo and P = .002 compared with fenofibrate). Apolipoprotein (Apo) A-1 levels increased parallel to the increases in HDL-C levels, but Apo A-II levels showed a strong positive dose-response relationship for LY518674, with the increase reaching 30.0% for the 100-μg LY518674 dose.

Table 3 reports the effects of LY518674 as monotherapy; Table 4 reports the effects of LY518674 when added to atorvastatin in patients with hypercholesterolemia. As monotherapy, LY518674 (10 μg and 50 μg) reduced triglycerides levels by 36.9% and 37.5% (P≤.001) and increased HDL-C levels by 15.0% and 12.5%, respectively. Given without atorvastatin, LY518674 (10 μg and 50 μg) decreased LDL-C levels by 13.2% and 15.8%, respectively. When added to atorvastatin, LY518674 (10 μg and 50 μg) produced statistically significant additional decreases in levels of triglycerides, ranging from 20.4% to 42.7%, respectively, depending on the LY518674 and atorvastatin doses. As shown in Table 4, when added to atorvastatin (10 mg), HDL-C levels increased 9.8% for LY518674 (10 μg) and 11.9% for LY518674 (50 μg). When added to atorvastatin (40 mg), HDL-C levels increased 9.2% for LY518674 (10 μg) and 0.6% for LY518674 (50 μg). When added to either atorvastatin dose, LY518674 (10 μg and 50 μg) produced no statistically significant additional reductions in LDL-C levels beyond that provided by atorvastatin alone.

Laboratory safety measures are shown in Table 5 and Table 6. In the dyslipidemia study, administration of LY518674 at the higher dose levels of 50 μg and 100 μg, or administration of fenofibrate, increased the AUC/t in levels of serum creatinine (P≤.001 compared with placebo). Many patients experienced increases in creatinine levels that resulted in values exceeding the upper limits of normal. Fenofibrate, but not LY518674, increased the AUC/t in creatinine phosphokinase measurements (P = .004).

There was no consistent pattern to other adverse events, however, in the hypercholesterolemia study; 16.7% of patients receiving atorvastatin (40 mg) with LY518674 (50 μg) had alanine aminotransferase levels between 1.5 and 3 times the upper limit of normal (P = .10 compared with placebo). In addition to the laboratory values reported in Tables 5 and 6, levels of troponin I were also measured during both trials and showed no significant increases for fenofibrate or LY518674. Investigator-reported adverse events were rare and showed no consistent pattern.

Treatment to reduce levels of LDL-C represents a primary goal for prevention of cardiovascular disease.¹⁴ However, many patients experience adverse cardiovascular outcomes despite treatment to lower LDL-C levels, a phenomenon often attributed to the increasing prevalence of obesity and related factors, including hypertension and a cluster of lipid abnormalities known as atherogenic dyslipidemia.^2,15 Although LDL-C levels are generally not strikingly elevated, this syndrome is characterized by the presence of atherogenic, small-dense LDL-particles, increased triglycerides levels, and low levels of HDL-C.⁵

Intensive efforts have focused on development of pharmacological strategies for treatment of this common lipid disorder. Therapeutic agents that target 3 distinct families of PPAR receptors (α, γ, and ∂) have been considered particularly promising.^9- 10,14 PPAR-α agonists primarily modulate lipid metabolism, lowering serum triglycerides levels and modestly increasing HDL-C levels. The fibric acid derivatives fenofibrate and gemfibrozil are relatively weak ligands for the PPAR-α receptor.³ The PPAR-γ agonists increase insulin sensitivity and are widely used as antidiabetic agents.¹⁰ PPAR-∂ agents are currently the subject of ongoing research.¹⁶

Greater understanding of the chemical structure and function of these nuclear receptors has enabled the pharmaceutical industry to develop increasingly potent and selective PPAR agonists. However, to date, no novel PPAR-α agent has reached an advanced stage of development. According to publicly accessible FDA documents, the pharmaceutical industry has filed at least 50 investigational new drug applications for new PPAR agonists.¹¹ However, in nearly all cases, development was terminated due to toxicity.¹¹ In 1 report, an FDA safety official described the organ systems involved in toxicity as “cardiac, skeletal muscle, renal, and bone marrow.”¹¹ According to FDA officials, 2 dual α-γ agonists were discontinued due to elevations in serum creatinine levels, and rhabdomyolysis occurred in early-stage trials of a single α and another dual α-γ agonist.¹¹ Five dual agonists were discontinued for “multi-species, multi-sex, multi-site increases in tumors with no safety margins for clinical exposures.”¹¹ Another dual α-γ agonist, muraglitazar, was discontinued after publication of a report that this agent increased adverse cardiovascular outcomes.¹⁷ With the exception of muraglitazar, each of these drugs was discontinued without scientific publications describing their safety and efficacy.

In this study, we evaluated a potent and selective new PPAR-α agonist known as LY518674 in patients with atherogenic dyslipidemia or hypercholesterolemia. This report represents, to our knowledge, the first published description of the effects of a potent and selective PPAR-α agonist. Efficacy in an atherogenic dyslipidemia population was generally similar to that of fenofibrate, producing a 35% to 42% decrease in triglycerides levels and a 2% to 16% increase in HDL-C levels.

However, the LY518674 dose-response curves for Apo A-I and HDL-C were unusual, showing the greatest elevation in HDL-C levels (15.8%) at an intermediate dose of LY518674 (25μg) and almost no effect at the maximum tested dose of 100 μg. Since fenofibrate is known to affect both the production and catabolic rates of Apo AI, it is plausible that a potent PPAR-α agonist may have differential effects on these 2 processes at various doses, resulting in complex biphasic changes in HDL-C levels.¹⁸

In atherogenic dyslipidemia, there was a trend toward greater reduction of triglycerides levels with LY518674 at only the 50-μg dose, compared with fenofibrate (41.7% vs 32.8%, respectively; P = .08), but this dose did not result in an optimal increase in levels of HDL-C (11.1%). LY518674 had a unfavorable affect on LDL-C levels, increasing values significantly more than fenofibrate, 18.3% (50 μg) and 19.5% (100 μg) vs 2.3% for fenofibrate (P = .002 for both comparisons). In the hypercholesterolemia study, LY518674 (10 μg or 50 μg), when added to atorvastatin, produced no significant additional reductions in LDL-C levels, increased HDL-C levels by 0.6% to 11.9%, and decreased triglycerides levels an additional 20.4% to 42.7% (Table 4).

We also performed a post hoc analysis to explore the increase in LDL-C levels observed with LY518674 in the atherogenic dyslipidemia study but not in the hypercholesterolemia study. The increase in LDL-C levels with LY518674 showed a positive correlation with baseline levels of triglycerides (Figure 3). Accordingly, in the atherogenic dyslipidemia study, in which triglycerides levels were elevated at baseline, LDL-C levels increased in response to LY518674 therapy. In contrast, in the hypercholesterolemia study, in which triglycerides levels were less elevated, LY518674 moderately reduced levels of LDL-C.

In general, both fenofibrate and LY518674 were well tolerated, but some safety concerns emerged during treatment. In the dyslipidemia study, administration either of LY518674 at higher doses or of fenofibrate increased the AUC/t in levels of serum creatinine (P≤.001 compared with placebo). In some patients, serum creatinine levels increased above the upper limit of normal (38% in the fenofibrate treatment group and a maximum of 37.3% in the LY518674 treatment group [100-μg group in the dyslipidemia study]). Fenofibrate, but not LY518674, increased the AUC/t in levels of creatine phosphokinase (P = .004). However, no patients experienced rhabdomyolysis or acute renal failure. Elevation of serum creatinine has been reported with fenofibrate previously, although the etiology is controversial.^19- 22 One report in 3 renal transplant patients demonstrated reversible allograft dysfunction with biopsy evidence of tubular toxicity.²² However, another study reported that increased metabolic production of creatinine, rather than renal dysfunction, contributed to the increase in serum creatinine levels.²⁰ Although fenofibrate increased levels of creatine phosphokinase in our study, this agent has only rarely caused rhabdomyolysis in the absence of concomitant statin administration.²³

The effects of LY518674 on lipoproteins raise some additional concerns. The dose-dependent increases in LDL-C levels when LY518674 was administered as monotherapy are clearly undesirable. LY518674 also produced substantial, dose-dependent increases in levels of Apo A-II, ranging up to 30% for the highest LY518674 dose (100 μg). While increases in Apo A-I levels are usually linked to clinical benefit, the impact of increases in Apo A-II levels is more controversial.^24- 27 Some studies show protection from atherosclerosis, whereas others have linked increases in Apo A-II levels to adverse cardiovascular outcomes.^24- 25 Some reports have suggested that the higher the ratio of Apo A-II to Apo A-I, the greater the risk of adverse vascular events.²⁶ The precise impact of altering the relative proportion of Apo A-II to Apo A-I on the functionality of HDL-C remains to be determined.

These findings have implications for the interpretation of clinical trials that have investigated the efficacy of fibrates.^{6- 7,28- 30} Although gemfibrozil has been associated with improved cardiovascular outcomes in 2 randomized controlled trials, results of outcomes studies with other fibrates have been inconsistent. Bezafibrate, although not available in the United States, produced no benefit on clinical outcomes.²⁹ A post hoc analysis suggested clinical benefit in patients with hypertriglyceridemia at baseline. A recent large study of fenofibrate in patients with diabetes showed no significant reduction in morbidity and a trend toward increased all-cause mortality.³⁰ Whether this potential increase in mortality is derived from compound-specific toxicity of fenofibrate or is an adverse effect of PPAR-α activation remains uncertain. The short-term nature of our study (ie, 12 weeks of therapy in both trials) and the lack of data on clinical events during that period limit our ability to help clarify this issue. Accordingly, the net clinical benefit of existing and novel PPAR-α agonists remains to be determined.^31- 32

The results of this pair of trials demonstrate the challenges in developing new PPAR agonists as therapeutic agents. These drugs modulate activity of a large number of genes, some of which produce unknown effects. Accordingly, the beneficial effects of PPAR activation appear to be associated with a variety of untoward effects, which may include oncogenesis, renal dysfunction, rhabdomyolysis, and cardiovascular toxicity. Recently, the FDA began requiring 2-year preclinical oncogenicity studies for all PPAR-modulating agents prior to exposure of patients for durations longer than 6 months.¹¹ Progress toward understanding these issues has been hampered by the absence of published reports of the benefits and hazards of novel agents. The promise of PPAR agonists stems primarily from the hope that more active agonists would produce more robust benefits. However, the current pair of studies do not support this hypothesis. Despite a potency approximately 10 000 times greater than fenofibrate and greater selectivity, LY518674 did not produce more favorable effects on lipoproteins compared with fenofibrate.

These findings underscore the importance of prompt and thorough reporting of clinical trials of novel pharmaceutical agents. Such studies advance the understanding of risks and benefits of new approaches and enable subsequent development programs to refine future study designs. Prompt publication also allows better protection of patient safety by providing study sponsors, investigators, and data and safety monitoring committees with a better appreciation of any potential risks that may require close monitoring.