The benefit of statins on cardiovascular (CV) end point reduction in both primary and secondary settings has proven to be a consistent finding throughout a large number of randomized controlled trials. Recently, more intensive low-density lipoprotein (LDL)-lowering was shown to be associated with a further reduction of CV events compared with conventional lipid-lowering (1). However, it should be noted that approximately 70% of CV events have not been prevented in statin trials (2). This staggering number has intensified the search for novel therapeutic targets. The choice of a new target has, to some extent, been directed by the worldwide epidemic of obesity, metabolic syndrome, and type II diabetes mellitus (3), all characterized by a non-LDL-cholesterol (C) dyslipidemia consisting of low high-density lipoprotein cholesterol (HDL-C) and increased triglycerides (TG).

Large-scale population studies have demonstrated that HDL-C is a strong and independent inverse predictor of coronary heart disease (CHD), even in subjects with normal LDL-C levels (4). Accordingly, up to 40% of patients with premature CHD are characterized by low levels of HDL-C. Using these data, it has been calculated that coronary risk increases by 1% to 3% for every 1% reduction in HDL-C level. As a consequence, increasing HDL-C has emerged as an attractive tool for CV prevention. In support, several classes of drugs with HDL-increasing effects have been associated with reductions in CV event rates. Thus, fibrates, which mediate their HDL increase predominantly by peroxisome proliferator agonist receptor-alpha activation, have been shown to be associated with an absolute reduction in primary end points of 1.4% and 4.4% in primary (5) and secondary (6) prevention, respectively. In line, nicotinic acid derivatives, mediating HDL increase through a variety of mechanisms (7- 8), have been associated with a reduction of CV events in one large monotherapy trial and two small combination secondary prevention trials (9- 10). Unfortunately, trials evaluating combination therapy of statins with fibrates or niacin have been hampered because of safety concern.

As HDL-increasing interventions are expected to constitute an important therapeutic option for CV prevention in the next decade, we performed a meta-analysis of all randomized controlled trials using monotherapy of two currently available classes of HDL-increasing compounds (fibric acid and nicotinic acid derivatives) in order to provide an accurate and precise estimate of their efficacy as well as their safety.

A literature search of the MEDLINE database (1966 to February 2004) using the medical subject headings fibric acid, fibrates, bezafibrate, ciprofibrate, clofibrate, etofibrate, fenofibrate, gemfibrozil, nicotinic acid, niacin, acipimox, controlled-release, extended-release, immediate-release, sustained-release, lipids, lipoproteins, and CHD was performed. The search was restricted to studies published in English-language journals, conducted in humans, and classified as randomized controlled trials. A manual search was also performed using the authors reference lists from randomized controlled trials and reviews. For inclusion, studies had to meet the following criteria: 1) random allocation of study participants to fibrates or niacin versus a control group or trials with a crossover design; 2) changes in lipids; and 3) intervention duration of ≥3 weeks. The most common reason for exclusion of trials was missing values (mean or standard deviation [SD]) for serum lipids.

All data were abstracted independently by two authors (R.B. and B.H.) in duplicate using a standardized protocol and reporting form. Disagreements were resolved by consensus. Authors were contacted for additional and missing information. When information of multiple doses and different treatment periods was available, we selected the highest dose and the longest treatment period for the calculations. Study characteristics recorded were as follows: 1) first author's name, year of publication, and journal; 2) number of participants (n); 3) mean age, gender, body mass index; 4) type and dosage of study drug; 5) mean total cholesterol (TC), LDL-C, HDL-C, and TG levels at baseline and follow-up; 6) adverse events; 7) coronary end points; 8) treatment duration; and 9) dyslipidemia. Major coronary events recorded during long-term treatment (≥2 years of active treatment) included coronary death (definite), nonfatal myocardial infarction, silent myocardial infarction, coronary revascularization, unstable angina, and sudden cardiac death. In addition, data were abstracted on fatal CHD deaths, stroke, revascularization, non-CV deaths, cancer, and all-cause mortality during treatment. Adverse events were divided in clinical adverse events and laboratory adverse events. Clinical adverse events were: serious adverse events, gastrointestinal symptoms (defined as abdominal pain, diarrhea, dyspepsia, nausea, and vomiting), skin reactions (defined as pruritus and/or rash), flushes, and gout. Laboratory adverse events were hepatotoxicity (defined as >3 times the upper limit of normal serum glutamic oxalocetic transaminase/serum glutamic pyruvic transaminase value), musculoskeletal symptoms (defined as >10 times the upper limit of normal creatine phosphokinase value and/or myalgia), hyperglycemia (defined as fasting glucose of >6.1 mmol/l or HbA1c >7%).

The data needed for the measurement of the weighted mean difference were: 1) the mean change in serum lipids (TC, TG, LDL-C, and HDL-C) from baseline to follow-up in milligrams per deciliter (mg/dl); 2) the SD of the mean difference; and 3) the number in each comparison group (n) at follow-up (11).

The estimate of the principal effect was defined as the mean difference in mg/dl between the change in lipid concentrations during active treatment (mean at follow-up minus mean at baseline) and the change during the control period (mean at follow-up minus mean at baseline). This difference is referred to as the net mean change. When the SD_difference was not available in the source papers, authors were contacted. In case of no response or nonavailability of the requested information, we estimated the variance with the following formula (11): $Standard error(SE)difference=[SDbaseline2/nbaseline+SDfollow-up2/nfollow-up]$

The SD_difference was calculated from the SE_difference with the following formula: $SDdifference=SEdifference×n$ For the computation of pooled effects, each study was assigned a weight consisting of the reciprocal of its variance. Estimates of the average effect of fibrates or niacin on serum lipids and 95% confidence intervals were calculated with models based on both fixed- and random-effects assumptions. Because substantial variability between observations was indicated by preliminary tests for homogeneity, we have presented the results of random-effects models calculated according to the method of DerSimonian and Laird (12).

Relative risk reduction (RRR) and absolute risk reduction (ABR) were used to measure the effect of fibrates on clinical outcomes. The numbers of various outcomes for both the fibrate and control groups were recorded for each study using 2 × 2 tables.

The number of subjects in the treatment group in whom an end point of interest was observed minus the number of subjects in the control group in whom an end point of interest was observed from each trial was summed, and z statistics were used to test whether the total differed from 0. To calculate the pooled absolute risk, each study was weighted by its sample size (N_t × N_c)/(N_t + N_c), (t = treatment; c = control). The number needed to treat (NNT) was calculated by taking the reciprocal of the ABR.

Subgroup analyses were performed to examine the effect of fibrates and niacin on serum lipids in subjects with different dyslipidemias. First, the effects were compared in subjects with all lipid disorders, including secondary dyslipidemia. Then we analyzed the effects in subjects with hypercholesterolemia, combined hyperlipidemia, hypertriglyceridemia, low HDL-C, and type II diabetes mellitus. We used the ATP-III classification to define lipid disorders. Trials classified as Frederickson's type IIa hyperlipidemia or with mean baseline TC above 240 mg/dl, mean baselines LDL-C above 160 mg/dl, and without elevated TGs above 200 mg/dl were considered hypercholesterolemia. A trial was considered combined hyperlipidemia if it was classified as Frederickson's type IIb hyperlipidemia or with mean baseline TC above 160 mg/dl, mean baseline LDL-C above 160 mg/dl, and TG above 200 mg/dl. Trials classified as Frederickson's type IV hyperlipidemia with normal mean baseline TC and LDL-C, but TG levels above 200 mg/dl, were considered as hypertriglyceridemia. Trials with mean baseline HDL-C levels below 40 mg/dl were considered as low HDL-C.

The contents of 409 (275 fibrates and 134 niacin) abstracts or full-text manuscripts were reviewed to determine eligibility for inclusion. Because trials with etofibrate were scarce, we excluded the only trial with this drug. A total of 83 trials (53 trials for fibrates and 30 trials for niacin) met our criteria and were included in the meta-analysis (Tables 1, 2). The distribution of the 53 trials with fibrates is as follows: 17 trials with gemfibrozil (5- 6,13- 27), 15 trials with fenofibrate (22,28- 41), 13 trials with bezafibrate (22,42- 53), 9 trials with clofibrate (21,54- 61), and 2 trials with ciprofibrate (62- 63). For the 30 trials with niacin, the distribution is as follows: 11 trials with immediate-release niacin (IR-niacin) (54,64- 73), 10 trials with acipimox (72,74- 82), 6 trials with extended-release niacin (ER-niacin) (67,83- 87), and 5 trials with wax-matrix sustained-release niacin (SR-niacin) (64,88- 91). From the 83 trials, 6 trials (21- 22,54,64,67,72) used multiple treatment arms with more than one fibric acid or nicotinic acid derivative, and 15 trials had different study populations (26,28,31,33,38,44,49,51,53,58,62,68,74- 75,81). Overall, in these studies 16,802 subjects were treated with fibrates or placebo, and 4,749 subjects were treated with niacin or placebo. The average length of treatment was 12 and 8 months for fibrates and niacin, respectively. The mean age of patients taking fibrates or niacin was 56 and 58 years, with a male-to-female ratio of 16 to 1 and 7 to 1, respectively. The body mass index of patients in trials with fibrates and niacin was 27 and 28 kg/m², respectively. Twelve trials used random assignment with a crossover design (19,21,26- 27,38,40,42,45,51,57- 58,61). Subjects with the following dyslipidemias and study populations were enrolled into the 83 trials: 29 trials with hypercholesterolemia (5,22- 25,28,33- 34,36,43,49,52- 53,55- 57,59- 60,62,64,67,85- 90,92); 25 trials with type II diabetes mellitus (15,17- 18,20,29,31- 32,37- 39,41,43- 44,46,50,57,65,68,76- 79,81- 83); 16 trials with combined hyperlipidemia (21,27- 28,33,38,44,49,51,58,62- 63,69,72,74- 75,80); 10 trials with hypertriglyceridemia (26,30- 31,38,42,51,58,61,74- 75); 9 trials in subjects with coronary artery disease (6,16,47- 48,54- 55,64,71,73); and 2 trials in subjects with peripheral arterial disease (48,68). Other rare study populations included healthy subjects (45), type 1 diabetics (53), renal failure (35), and cholelithiasis patients (22).

Fibrate treatment induced significant reductions in serum concentrations of TC, TG, and LDL-C, whereas HDL-C levels were increased (Table 3). The net change in TC was a decrease of 25.5 mg/dl (0.66 mmol/l) (95% confidence interval −29.34 to −21.66 mg/dl, p < 0.00001) or 11%. Of the 53 trials, 52 trials (98%) reported a net decrease, and 1 trial (17) (2%) reported a net increase in TC. Although the efficacy of fibrates differed, they all significantly (p < 0.00001) reduced TC: fenofibrate (13%), ciprofibrate (13%), bezafibrate (10%), gemfibrozil (9%), and clofibrate (7%).

The net change of TG levels was a decrease of 70.5 mg/dl (0.80 mmol/l) (95% confidence interval −79.78 to −61.22 mg/dl, p < 0.00001) or 36%. All 53 trials reported significant (p < 0.00001) reductions of TG levels with the following distribution: gemfibrozil (48%), ciprofibrate (45%), fenofibrate (40%), bezafibrate (31%), and clofibrate (18%).

The net change of LDL-C levels consisted of a decrease of 11.7 mg/dl (0.30 mmol/l) (95% confidence interval −17.86 to −5.59 mg/dl, p = 0.0002) or 8%. Bezafibrate (13%; p = 0.04), fenofibrate (11%; p = 0.01), and ciprofibrate (8%; p = 0.005) significantly reduced LDL-C levels, whereas clofibrate (3%; p = 0.53) and gemfibrozil (1%; p = 0.68) had no significant effect on LDL-C. From the 48 trials, 38 trials (79%) showed a net decrease in LDL-C after treatment with fibrates, whereas in 6 (13%) and 4 (8%) trials LDL-C increased and remained unchanged, respectively. All fibrates, with the exception of clofibrate, significantly (p < 0.00001) increased HDL-C concentration by 4.1 mg/dl (0.11 mmol/l) (95% confidence interval 3.34 to 4.91 mg/dl) or 10%. Among the different fibrates, bezafibrate (11%), gemfibrozil (11%), fenofibrate (10%), and ciprofibrate (10%) showed substantial increases in HDL-C levels. From the 53 trials, only 47 trials reported HDL-C outcome. A total of 43 from the 47 eligible trials (91%) showed a net increase of HDL-C levels, whereas 3 trials (6%) with clofibrate and 1 trial with bezafibrate (3%) showed a decrease in HDL-C.

Overall, 33% of subjects receiving fibrates experienced adverse effects versus 31% of subjects in the control group (relative risk 1.16; 95% confidence interval 1.03 to 1.32, p = 0.02). A larger number of subjects experienced gastrointestinal symptoms in the fibrate drug treatment group (12% vs. 10%) compared with the control group (relative risk 1.37; 95% confidence interval 1.10 to 1.70, p = 0.004). Other adverse effects were skin reactions with a prevalence of 30% and 25% in the treatment and control group, respectively (relative risk 1.03; 95% confidence interval 0.80 to 1.33, p = 0.81), musculoskeletal symptoms 3% and 5% in the treatment and control groups, respectively (relative risk 1.23; 95% confidence interval 0.65 to 2.32, p = 0.52).

Hepatotoxicity occurred in 6.4% of subjects receiving fibrates versus 6.1% in the control group (relative risk 1.23; 95% confidence interval 0.98 to 1.53, p = 0.07). Serious adverse effects occurred in 12% versus 13% of patients, respectively (relative risk 1.02; 95% confidence interval 0.88 to 1.17, p = 0.83), whereas cancer prevalence was 4.1% versus 3.7% of subjects (relative risk 0.93; 95% confidence interval 0.80 to 1.08, p = 0.35). Subject withdrawal was similar (15%) between the treatment and the control group. There was no significant difference in adverse effects between the different fibrates.

Eight long-term clinical end point intervention trials using fibrates were included (5- 6,14,39,47- 48,52,54). Overall, 1,609 major coronary events and 675 coronary deaths occurred in subjects in the control group, whereas 892 events and 398 deaths occurred in those allocated to active treatment (Table 4). Pooled results indicated significant reductions in the risk for major coronary events (p < 0.001), but not for coronary deaths. Treatment with fibrates reduced risk of major coronary events by 25% (95% confidence interval 11% to 37%, p < 0.001). Only the Veterans Affairs High-Density Lipoprotein Intervention trial (VA-HIT) and the Helsinki Heart Study (HHS) showed statistically significant reductions in major coronary events (p < 0.001). Other long-term trials (i.e., Bezafibrate Coronary Atherosclerosis Intervention trial, Bezafibrate Infarction Prevention study, and Bezafibrate treatment in the Lower Extremity Arterial Disease Event Reduction trial) exhibited tendencies toward a statistically significant (p = 0.07) reduction in major coronary events. No statistically significant reductions were found in mortality due to coronary cause or all-cause (Table 4). The NNT for all fibrates to prevent one major coronary event was 33 for four years. Noncardiovascular mortality was similar in active treatment and control groups (relative risk 1.10; 95% confidence interval 0.96 to 1.26, p = 0.18). The total number of enrolled women was 373 compared with 13,945 men.

Niacin treatment was associated with significant reductions in serum concentrations of TC, TG, LDL-C, and HDL-C (Table 5). The net change of TC levels was a decrease of 25.5 mg/dl (0.66 mmol/l) (95% confidence interval −31.80 to −19.13 mg/dl, p < 0.00001), corresponding with 10%. Of the 27 trials with TC outcome, 22 trials (81%) reported a net decrease, and 5 trials (19%) reported a net increase of TC. The SR-niacin and IR-niacin formulations were most effective with a 15% and 13% decrease in TC, respectively.

The net change of TG levels consisted of a decrease of 47.0 mg/dl (0.53 mmol/l) (95% confidence interval −60.72 to −34.67 mg/dl, p < 0.00001), corresponding with 20%. A total of 25 trials (86%) reported a net decrease, and 4 trials (14%) reported an increase of TG levels from a total of 29 trials. Among all the niacin formulations, IR-niacin and ER-niacin showed the most substantial reductions of TG levels of 26% and 20%, respectively.

The net change of LDL-C levels was a decrease of 20.6 mg/dl (0.53 mmol/l) (95% confidence interval −28.24 to −13.02 mg/dl, p < 0.00001), corresponding with 12%. The SR-niacin, IR-niacin, and ER-niacin showed significant reductions of LDL-C concentration of 19%, 15%, and 10%, respectively. Acipimox was the only drug associated with a mean increase of LDL-C of approximately 3% (p = 0.60).

Niacin treatment significantly increased HDL-C with 6.7 mg/dl (0.17 mmol/l) (95% confidence interval 5.10 to 8.44 mg/dl, p < 0.00001), or 16%. Among the different niacin formulations, IR-niacin (23%), ER-niacin (22%), and SR-niacin (13%) had the strongest HDL-increasing capacity (p < 0.00001), whereas acipimox only induced a 7% increase (p = 0.10). From the 25 trials, 21 trials (84%) showed a net increase of HDL-C levels, whereas 4 trials (16%) (3 with acipimox and 1 with SR-niacin) reported a slight decrease in HDL-C concentration.

Because the high drop-out rate (75%) of the largest IR-niacin trial would dominate the outcome of the safety analysis (54) compared with newer niacin formulations, we did not pool safety data of this trial with other niacin formulations. The most common adverse effect in subjects who received niacin was skin flushing. Overall, 70% of subjects receiving niacin experienced flushes, in contrast with 4% of subjects who received placebo (relative risk 7; 95% confidence interval 3.98 to 12.26, p < 0.00001). The distribution of flushes in the various niacin formulations showed that subjects with IR-niacin had the highest prevalence of flushing up to 85% (relative risk 14; 95% CI 4.94 to 38.90). Other formulations with a high rate of flushes included ER-niacin 66% (relative risk 5; 95% confidence interval 3.36 to 7.51), whereas flushes seemed to occur less in SR-niacin (26%) and acipimox (22%). Flushing was the main reason for subject withdrawal.

Approximately 13% of the subjects in the niacin group withdrew from the study, compared with 6% in the placebo group (relative risk 1.87; 95% confidence interval 1.19 to 2.93, p = 0.006). The subject withdrawal rate was distributed as follows: IR-niacin (20.2% in the niacin group vs. 8.6% in the control group), ER-niacin (19.7% vs. 10.2%), acipimox (4.3% vs. 4.6%), and SR-niacin (2.9% vs. 0%), respectively; IR-niacin had the highest prevalence for subject withdrawal (relative risk 4.06; 95% confidence interval 1.44 to 11.44, p = 0.008). The relative risk for subject withdrawal during ER-niacin was lower than during IR-niacin (relative risk 1.51; 95% confidence interval 0.63 to 3.61, p = 0.35).

Gastrointestinal symptoms had a prevalence of 23% and 15% in the treatment and control group, respectively (relative risk 1.57; 95% confidence interval 1.05 to 2.34, p = 0.03) ER-niacin 35% vs. 24% (relative risk 1.95; 95% confidence interval 0.91 to 4.14, p = 0.08); IR-niacin 20% vs. 20% (relative risk 1.24: 95% confidence interval 0.47 to 3.29, p = 0.66); SR-niacin 15% vs. 5% (relative risk 5.21; 95% confidence interval 0.36 to 75.18, p = 0.23); and acipimox 11% vs. 8% (relative risk 1.48; 95% confidence interval 0.79 to 2.78, p = 0.22). Other adverse effects were skin reactions, 11% and 5% (relative risk 2.71; 95% confidence interval 1.48 to 4.97, p = 0.001), and musculoskeletal symptoms, 1% and 0.1% (relative risk 2.87: 95% confidence interval 0.49 to 16.91, p = 0.24), respectively. Hepatotoxicity occurred in 2.1% of subjects receiving niacin versus 0.3% in the control group (relative risk 3.15; 95% confidence interval 1.85 to 7.85, p = 0.01).

Hyperglycemia had an occurrence rate of 2.3% versus 0.4% in the niacin and the control group, respectively (relative risk 3.04; 95% confidence interval 1.28 to 7.21, p = 0.01). From all the niacin derivatives, ER-niacin and IR-niacin had the highest prevalence of, respectively, 4% (relative risk 5.67; 95% confidence interval 1.06 to 30.30, p = 0.04) and 2.3% (relative risk 2.79; 95% confidence interval 0.89 to 8.73, p = 0.08).

Comparison of the fixed-effects with the random-effects model resulted in comparable changes for lipid efficacy. For fibrates, the numbers were −15 versus −25 mg/dl for TC, −51 versus −70 mg/dl for TG, −13 versus −12 mg/dl for LDL-C, and +4 versus +4 mg/dl for HDL-C, respectively. The numbers for niacin were −24 versus −25 mg/dl for TC, −36 versus −47 mg/dl for TG, −20 versus −21 mg/dl for LDL-C, +7 versus +7 mg/dl for HDL-C, respectively. Subgroup analyses for serum lipids by fibrates and niacin in subjects with all lipid disorders, hypercholesterolemia, combined hyperlipidemia, type II diabetes mellitus, and low HDL-C showed moderate differences (Table 6).

Statin therapy has been accepted as standard therapy to lower CV risk. However, because the majority of CV events cannot be prevented with statins (2), despite their extensive LDL-lowering and weak HDL-increasing (5%) properties, this has prompted us to focus the present analysis on efficacy and safety of HDL-increasing compounds offering the potential for future combination therapy. The choice for HDL-increasing strategies as second target after LDL-lowering reflects both solid epidemiologic data as well as the advent of numerous options for selective HDL-increase within the next few years. To date, two compounds, currently available for clinical use, fit this profile: fibric acid and nicotinic acid derivatives.

A consistent HDL-increasing effect of approximately 10% was found for fibrates. This effect was more pronounced in patients with combined hyperlipidemia and/or hypercholesterolemia ranging from 11% up to 16%. The HDL increase by niacin exceeded that of fibrate therapy approximately 1.6-fold, ranging from 7% up to 23%. The effects of IR-niacin (23%) and ER-niacin (22%) were more pronounced than those observed during SR-niacin (13%) or acipimox (7%) therapy. Of note, the consistency of this observation seriously questions the use of the latter compounds as a means of achieving HDL increases.

With regard to TG, significant reductions were achieved by all fibrates, ranging from 18% up to 48%. This effect of fibrates was consistent throughout the different forms of dyslipidemia included in the present analysis. The efficacy of niacin or its derivatives toward TG lowering was less pronounced when compared with fibrate therapy, ranging from 17% to 26%. The reductions in TC and LDL-C during fibrate therapy ranged between 7% and 13% and 8% and 13%, respectively. Niacin and its derivatives were shown to be slightly more effective in reducing LDL-C levels than fibrates.

The incidence of adverse effects in subjects receiving fibrates was increased compared with placebo (relative risk 1.16, p = 0.02). The most common side effects during fibrate therapy were skin reactions (30%) and gastrointestinal symptoms (12%). Other adverse events like hepatotoxicity (6%) and musculoskeletal symptoms (3%) occurred much less frequently. The withdrawal rate from the trials was 15% in both the fibrate and the placebo group. Previous reports have addressed concerns with regard to using fibrates in view of a potential increase in the risk of death from noncoronary causes and/or the induction of cancer (93). In the present meta-analysis, we could not find any increase in noncoronary mortality and/or the prevalence of cancer. Notably, in view of the predefined inclusion criteria for studies with regard to available lipid profile data and treatment arms with the combination of both drugs, we were unable to include the data from the World Health Organization cooperative trial on primary prevention of CHD by clofibrate. Because adverse effects on overall mortality in previous analyses were largely explained by the results of the latter study, it is in line with expectations that we did not find a negative effect. In view of the large number of subjects included in the present meta-analysis, it appears safe to conclude that fibrate therapy is well-tolerated without conveying additional risks.

As expected, flushes were the most common adverse effect with niacin. Most studies, however, report a decline in the incidence of flushing during prolonged use. Although the SR-niacin and acipimox formulations had a lower incidence of flushes, the HDL-raising efficacy of these compounds is significantly less than the other compounds. There was also a modest increase in gastrointestinal symptoms in the niacin group (23%) when compared with the control group (15%). We found relatively little evidence to support the concerns for hepatotoxicity with SR-niacin, because only two of six trials with SR-niacin reported modest elevations of liver enzymes (90- 91). Although niacin treatment has been associated with worsening of insulin sensitivity (assessed using euglycemic, hyperinsulinemic clamp technique), only few subjects experienced this side-effect (2%). More importantly, potential changes in glycemic control did not result in a need to change the regimen of diabetic control. These findings are consistent with the Arterial Disease Multiple Intervention Trial (ADMIT) study, in which glycemic control was affected only temporarily by IR-niacin therapy in 125 diabetics (68).

In contrast with the overwhelming number of patients included in statin studies, the number of studies evaluating the effect of fibrates and/or niacin derivatives on CV outcome is limited. Two crucial studies with fibrates comprise the VA-HIT study and the HHS. In the VA-HIT study, the primary end point (nonfatal myocardial infarction or coronary death) frequency was reduced from 21.7% in the control group to 17.3% in the gemfibrozil group, whereas the HDL increase upon gemfibrozil treatment largely predicted CV benefit in multivariate analysis (94). In the HHS, the primary end point (CHD events) was significantly reduced in the gemfibrozil group (2.7%), compared with the control group (4.1%). Other studies using fibrate therapy have yielded less convincing results (Table 4), most likely related to power and patient selection. It remains to be established whether and to what extent the combination of statins and fibrates confers additive protection against CHD. Currently, two studies addressing this issue are the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) trial and the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial.

The previously reported interaction between fibrates, particularly gemfibrozil, and statins asks for some degree of caution while implementing combination therapy with these compounds (95).

The number of studies addressing the effect of monotherapy with nicotinic acid derivatives on CV event outcomes is limited to one study, the coronary drug project. This study reported a modest 11% reduction of CV mortality and a 27% reduction of major coronary events after a follow-up period as long as 10 years (54). However, the use of immediate-release preparations in that study, characterized by multiple daily dosing and a large number of side-effects, resulted in a dropout rate larger than 70% in addition to not-quantified effects due to noncompliance. In this respect, two recent smaller trials evaluating niacin combined with either a bile acid sequestrant or a statin reported impressive HDL increases ranging from 25% to 41% with an unprecedented reduction of CV event rates ranging from 60% to 72% (9- 10). Notably, these impressive reduction rates fit nicely to the estimated reduction rates based on the obtained HDL increase in these studies (i.e., 1% HDL increase being associated with a 1% to 3% reduction in CV events 94). These data warrant longer and adequately powered studies to verify these promising observations. In line with these results, the combination of clofibrate and niacin showed impressive reductions of 36% to 60% (p < 0.01) in CV mortality, which was directly related to serum TG-lowering in the Stockholm Heart study (96). Unfortunately, in view of the predefined inclusion criteria for studies with regard to the treatment arm, we were unable to include the data from this study.

Most studies did not report the SD_difference. This was the most common reason to contact the authors. Because the response rate and data availability was disappointing, we had to calculate the SD_difference with the formula mentioned in the Methods section. This resulted in a slightly overestimated SD_difference, which in turn, marginally decreased the weight of the studies. Although this is not the ideal situation, this formula provides a generally accepted method to calculate the SD_difference. Another issue is the weighing of continuous data, which is in favor of small studies with homogeneous populations upon pooling of all studies with different numbers of subjects, dyslipidemias, and agents. In the present analyses, we limited this phenomenon by applying a random-effects model.

It must also be noted that, in almost all fibrate intervention trials, mostly men were included, with a male-to-female ratio of 16:1. Awaiting ongoing combination trials using statins and fibrates, one has to be careful before extrapolating all findings toward female patients.

The global epidemic of CHD in combination with the inability of statin monotherapy to prevent more than 25% to 30% of CV events emphasizes the need for combination therapies; HDL increase offers a promising target for future CV prevention trials. First, strong observational studies validate the antiatherogenic potential of HDL-C. Second, the prognostic value of HDL-C on CV outcome remains present, even in patients using statin therapy (97). Third, the currently available intervention studies with niacin support an additive beneficial effect of these compounds on top of statin therapy (9- 10), and we believe that fibrates may also add to event reduction in combination with statins. Finally, new HDL-increasing strategies with weekly apoAI-Milano infusions are able to induce a 4.2% reduction in plaque volume in a period of five weeks (98). These findings outperform the reduction in plaque volume established after 18 months of intensive LDL-lowering therapy with 80 mg of atorvastatin in the Reversal of Atherosclerosis with Aggressive Lipid Lowering (REVERSAL) study, reaching only 0.9% (99). Because the latter changes might be extrapolated to result in a significant improvement in CV outcome during intensive lipid-lowering therapy, the expectations for potent HDL increase combined with intensive LDL-lowering have been raised significantly. The recent reports on significant HDL increases up to 60% during cholesterol ester transfer protein-inhibition with 120 mg of torcetrapib (100) might bring us a step closer to executing this combined intervention strategy.

In summary, both fibrates and niacin provide a safe and effective way of increasing HDL-C, the latter being the most potent one. Whereas for both, particularly for fibrates, intervention data have emerged to show beneficial effects on CV outcome; their ability to enhance CV event reduction when added to statin monotherapy has been shown in small trials with niacin and may be shown with larger ongoing trials with fenofibrate.

The authors wish to thank all authors who responded and provided the requested information to make this work possible. The authors also thank Dr. Mark McGovern from Kos Pharmaceuticals, Miami, Florida, for his efforts and contribution to this study.