Study eligibility criteria

We included randomized controlled trials which directly compared statin monotherapy to fibrate monotherapy in adults with dyslipidemia and that reported mortality or a cardiovascular outcome of interest. Studies that enrolled participants less than 18 years of age or had a follow-up duration of less than 28 days were excluded. No language restrictions were applied.

Outcomes of interest

Primary outcome:

Cardiovascular mortality

Secondary outcomes:

Efficacy

All-cause mortality

Major cardiovascular events, defined as a composite of cardiovascular death, coronary artery disease, myocardial infarction, unstable angina, and stroke

Individual components of major cardiovascular events

Safety outcomes included adverse effects associated with the use of statins or fibrates [13, 14]:

Muscle-related adverse effects included rhabdomyolysis, elevations in creatine kinase (CK), and myalgia

Elevations in serum alanine aminotransferase (ALT)

Renal adverse events, included kidney injury and elevations in serum creatinine

Participant withdrawal due to adverse effects

Number of serious adverse effects

New diagnosis or worsening of diabetes mellitus

Non-cardiovascular mortality

Venous thromboembolism

Exploratory outcomes included the percent reduction from baseline to end of study for the following lipid concentrations:

Total cholesterol (TC)

Low-density lipoprotein cholesterol (LDL-C)

High-density lipoprotein cholesterol (HDL-C)

Non-high-density lipoprotein cholesterol (non-HDL-C)

Triglycerides

Apolipoprotein B (apoB)

Elevations in laboratory outcomes (CK, ALT, and serum creatinine), were defined according to the lowest clinically meaningful threshold as reported by study investigators. Myalgia was defined as the number of participants described as having myalgia or muscle pain. If myalgia or muscle pain was not reported, then we extracted the number of cases with musculoskeletal pain. Because most included studies were designed to assess reductions in cholesterol concentrations, post-hoc we also assessed the surrogate efficacy outcomes of reductions in lipid levels.

Search methods for identification of studies

We systematically searched for published and unpublished studies using Ovid MEDLINE and Epub Ahead of Print, In-Process & Other Non-Indexed Citations and Daily, EMBASE via Ovid, the Cochrane Central Register of Controlled Trials, ClinicalTrials.gov, and the WHO International Controlled Trials Registry Platform, from database inception until October 30, 2019. Search strategies were developed using keywords and medical subject headings for statins, fibrates, and the key efficacy and safety outcomes. We used recommended search terms (filters) which provide the best balance of sensitivity and specificity for studies of treatment [15, 16]. The complete search strategy is described in S1 File.

Two authors (GKYT and JEB) independently screened study abstracts and titles. Relevant full-text articles were then retrieved and assessed independently (GKYT and JEB) for inclusion according to a standard list of exclusion criteria that were applied in priority sequence (Table 1 in S1 File). Discrepancies were resolved by consensus.

Data extraction

Two authors (GKYT and SP or SP and JEB) independently extracted relevant study characteristics and outcomes using a standardized data extraction form. Data were extracted from all identified relevant study reports. If discrepancies between published journal articles and trial registries were identified, we extracted results from ClinicalTrials.gov, since reporting of outcomes and severe adverse events is more complete than in journal publications [17, 18].

Risk of bias assessment

Two authors (SP and JEB) independently assessed the risk of bias within each study using the Cochrane Risk of Bias Tool [19]. For appraisal of performance and detection bias, we broadly grouped outcomes into subjective and objective outcomes.

Data synthesis and statistical analysis

Study level characteristics were pooled and are reported as percentages, or means and standard deviations (SD). Studies with double-zeros, meaning the outcome was not reported or had zero events in both the statin and fibrate arms, were excluded from the selected meta-analysis models, and for single-zero studies, a 0.5 continuity correction was added as the default software setting [20]. After receiving reviewer reports regarding our analysis method for rare events, we changed our primary analysis model to the Mantel-Haenszel (MH) fixed-effect model to estimate odds ratios (OR) for dichotomous outcomes. For continuous lipid outcomes, we used an inverse variance fixed-effect model to estimate mean differences (MD). When we could not extract or calculate a standard deviation for the continuous outcomes, for each treatment group, we imputed the largest standard deviation reported from the included studies for that treatment group [12].

To examine the consistency of our findings, we undertook several subgroup and sensitivity analyses. The first subgroups were primary prevention, secondary prevention, and unreported baseline prevalence of cardiovascular disease. Primary prevention studies were arbitrarily defined as those which enrolled participants with a baseline history of cardiovascular disease of 10% or less, while studies with more than 10% of participants with baseline cardiovascular disease were defined as secondary prevention [4]. We also analyzed studies according to the primary type of dyslipidemia (primary hypercholesterolemia, mixed dyslipidemia, and other), according to fibrate drug, and whether the study included participants with diabetes (pre-specified as > 90% of study participants with a baseline history of diabetes mellitus).

We also undertook sensitivity analyses for dichotomous outcomes by estimating Peto OR, and our original analysis–MH random-effects model for risk ratios (S1 Table). The Peto method may be appropriate and unbiased for the analysis of rare events when the event is rare (<1%), the groups are balanced, and the intervention effects are small [21]. In addition, it may be the least biased method in the presence of a true treatment effect when studies with double-zeroes are excluded [22]. For continuous outcomes, we estimated MD using a random-effects model. To show the effect of imputing standard deviations for the continuous outcomes, we separated studies requiring imputation of standard deviations into a subgroup in the forest plots. Heterogeneity across studies was assessed using the I² statistic. For outcomes reported in 10 or more studies, reporting bias was assessed visually using funnel plots. All statistical analyses were conducted in Review Manager Version 5.3 [20].

Assessment of the certainty of the evidence

The GRADE approach (grading of recommendations assessment, development, and evaluation) was used to assess the certainty of the evidence [11]. Outcomes were categorized into high, moderate, low, and very low certainty of evidence. We imported data from Review Manager into GRADEpro GDT software to create a summary of evidence table for the outcomes judged to be most important (cardiovascular mortality, all-cause mortality, major cardiovascular events, study withdrawal due to adverse effects, serious adverse effects, myalgia, elevated ALT, and elevated serum creatinine) [23].

Description of studies and patient population

Details of the study selection process are shown in Fig 1. After screening titles and abstracts, 244 articles were excluded since they did not report mortality or a cardiovascular event of interest. Five records were identified by reviewing systematic reviews and reference lists of the included studies. The characteristics of the included studies are available in Table 1. Nineteen studies (reported in 24 articles) met the inclusion criteria [24–42], and allocated a total of 7619 participants to either statin or fibrate monotherapy, totaling approximately 4745 person-years of follow-up. Five studies [31, 33, 36–38] had a follow-up duration of 24 weeks or longer, and the longest follow-up was two years [38]. Only one study reported the outcome of cardiovascular events as a primary or secondary outcome of interest [36]; the remainder reported mortality and cardiovascular events as adverse events. Eight studies allowed upward dose titration of statins [24, 25, 31, 33, 35, 39, 40, 42], while all studies assigned participants to a fixed dose of fibrate. Four studies were conducted in primary prevention [30, 32–34], 10 studies assessed secondary prevention [25–27, 31, 35–38, 40, 42], and five studies [24, 28, 29, 39, 41] did not report the number of participants with a baseline diagnosis of cardiovascular disease. Earlier studies tended to enroll participants with primary hypercholesterolemia, while more recent studies enrolled participants with mixed dyslipidemia. Only one study met our pre-specified definition of enrolling participants with diabetes mellitus, thus we did not conduct this subgroup analysis [39].

Fig 1

PRISMA (preferred reporting items for systematic reviews and meta-analyses) flowchart of head-to-head studies evaluating statin and fibrate monotherapy on mortality and cardiovascular outcomes.

Table 1

Summary of the included study characteristics directly comparing statin and fibrate monotherapy.

Study	Region(s)	Number of participants	Follow-up duration (weeks)	Dyslipidemia	Lipid test inclusion criteria (mmol/L)	Mean baseline lipid levels (mmol/L)	Primary or secondary prevention (CVD %)	Mean age (SD)	Diabetes (%)	Men (%)	Statin	Fibrate	Funding
Tikkanen 1988	Finland	334	12	Primary hypercholesterolemia	TC > 6.22	TC: 8.9	Secondary (CAD: 43%)	51.6 (NR)	NR	49%	Lovastatin 20–80 mg daily	Gemfibrozil 600 mg twice daily	Merck, Sharp & Dohme
						TG: 2.0
					TG ≤ 3.95	LDL-C: 6.7
						HDL-C: 1.3
Ziegler 1990	France	184	10	Primary hypercholesterolemia	TC ≥ 7.78	TC: 10.0	Secondary (CAD: 43%)	46.0 (NR)	NR	70%	Simvastatin 20–40 mg daily	Fenofibrate 200 mg twice daily	NR
					LDL-C ≥ 5.05	TG: 1.5
	Switzerland				TG < 3.95	LDL-C: 8.1
						HDL-C: 1.1
Arntz 1991	Germany	96	12	Primary hypercholesterolemia	TG < 3.95	TC: 9.4	NR	52.5 (11.7)	0%	50%	Pravastatin 20–40 mg daily	Bezafibrate 400 mg daily	NR
						TG: 1.9
						LDL-C: 7.4
						HDL-C: 1.3
Sanchez 1991	Spain	39	12	Primary hypercholesterolemia	TC ≥ 6.50	TC: 7.8	Secondary (CVD: 72%)	57.5 (9.0)	0%	62%	Lovastatin 20–80 mg daily	Bezafibrate 200 mg three times daily	Merck, Sharp & Dohme
						TG: 1.6
					TG ≤ 3.95	LDL-C: 5.8
						HDL-C: 1.2
D’Agostino 1992	United States	104	18	Primary hypercholesterolemia	TC ≥ 6.21 or LDL-C ≥ 4.92	TC: 7.5	Secondary (CAD: 24%)	55.0 (13.0)	1%	58%	Lovastatin 20–40 mg daily	Gemfibrozil 600 mg twice daily	Merck, Sharp & Dohme
						TG: 1.9
					TC ≥ 6.21 or LDL-C ≥ 4.14 (with definite CAD or 2 CAD risk factors)	LDL-C: 5.5
						HDL-C: 1.2
Wiklund 1993	Sweden	143	12	Primary hypercholesterolemia	TC ≥ 6.00	TC: 7.3	NR	54.1 (NR)	NR	69%	Pravastatin 40 mg daily	Gemfibrozil 600 mg twice daily	Swedish Medical Research Council, Swedish Heart and Lung Foundation, King Gustav V and Queen Victoria Foundation, Bristol-Myers Squibb Company
						TG: 1.8
	Finland				TG < 4.00	LDL-C: 5.2
						HDL-C: 1.2
Greten 1994	Germany	131	12	Primary hypercholesterolemia	LDL-C ≥ 4.10	TC: 7.3	NR	52.4 (NR)	1%	44%	Fluvastatin 40 mg daily	Bezafibrate 400 mg daily	Sandoz AG Nurnberg
						TG: 1.8
					TG ≤ 3.40	LDL-C: 5.2
						HDL-C: 1.2
Lintott 1995	New Zealand	129	24	Moderate hypercholesterolemia	TC 5.20 to 7.80	TC: 6.8	Secondary (CAD: 51%)	56.0 (9.4)	NR	62%	Simvastatin 5–10 mg daily Pravastatin 10 mg daily	Bezafibrate 200 mg three times daily	Merck, Sharp & Dohme
						TG: 1.8
					LDL-C ≥ 4.14	LDL-C: 4.9
					TG ≤ 4.60	HDL-C: 1.1
Sinzinger 1995	Austria	524	24	Moderate hypercholesterolemia	TC 5.20 to 7.80 (no lipid drug therapy) or 6.50 to 7.80 (on lipid drug therapy)	TC: 7.1	Secondary (Angina: 18%)	55.0 (10.5)	8%	45%	Lovastatin 20 mg daily	Bezafibrate 400 mg daily	Boehringer Mannheim Austria
						TG: 2.4
						LDL-C: 4.8
						HDL-C: 1.3
Sweany 1995	United States,	168	18	Primary hypercholesterolemia	LDL-C ≥ 4.90m(no risk factor) or	TC: 6.9	NR	58.5 (NR)	100%	43%	Simvastatin 10–40 mg daily	Gemfibrozil 600 mg twice daily	Merck Research Laboratories
	Austria					TG: 2.3
	Germany					LDL-C: 4.7
	Brazil					HDL-C: 1.2
	New Zealand
Pover 1995	United Kingdom	2467	52	Primary hypercholesterolemia		TC: 8.2	Primary (CHD: 5%)	54.9 (NR)	1%	55%	Pravastatin 20–40 mg daily	Gemfibrozil 600 mg twice daily Bezafibrate 400 mg daily	Bristol-Myers Squibb
					LDL-C ≥ 4.10 (one or more risk factors according to NCEP 1998 guidelines)	TG: 2.5
					TC > 7.80	LDL-C: 5.8
					or > 6.50 and presence of two CV risk factors	HDL-C: 1.2
					TG 1.00 to 5.70 (or < 5.70 mmol/L if already treated with a fibrate)
De Lorgeril 1999	France	64	12	Primary hypercholesterolemia	TC > 6.50	TC: 7.2	Secondary (CAD: 100%)	54.1 (NR)	NR	100%	Simvastatin 20 mg daily	Fenofibrate 200 mg daily	Laboratoire Fournier, Daix, France
						TG: 2.0
					TG ≤ 3.50	LDL-C: 5.1
						HDL-C: 1.2
Skoloudik 2007	Czech Republic	376	104	Hypercholesterolemia	TC > 5.00 and < 8.00	TC: 6.5	Secondary (CAD: 27%)	63.4 (8.5)	23%	63%	Fluvastatin 40 mg daily	Fenofibrate 200 mg daily	NR
						TG: 2.2
					TG ≤ 5.00	LDL-C: 4.2
						HDL-C: 1.3
Goldberg 2009	North America	392	16	Mixed dyslipidemia	LDL-C ≥ 3.37	TC: 6.8	NR	54.8 (10.9)	24%	50%	Atorvastatin 20–80 mg daily	Fenofibrate 135 mg daily	Abbott
					TG ≥ 1.69	TG: 3.2
					HDL-C < 1.02 (men) or	LDL-C: 4.2
					HDL-C < 1.28 (women)	HDL-C: 1.0
Jones 2009	North America	922	16	Mixed dyslipidemia	LDL-C ≥ 3.37	TC: 7.8	Primary (CAD: 8%)	55.1 (10.5)	20%	49%	Rosuvastatin 10–40 mg daily	Fenofibrate 135 mg daily	Abbott
					TG ≥ 1.69	TG: 1.6
					HDL-C < 1.02 (men) or	LDL-C: 5.8
					HDL-C < 1.28 (women)	HDL-C: 1.2
Mohiuddin 2009	North America	419	16	Mixed dyslipidemia	LDL-C ≥ 3.37	TC: 6.7	Primary (CAD: 7%)	54.2 (10.3)	23%	48%	Simvastatin 20–80 mg daily	Fenofibrate 135 mg daily	Abbott
					TG ≥ 1.69	TG: 3.2
					HDL-C < 1.02 (men) or	LDL-C: 4.1
					HDL-C < 1.28 (women)	HDL-C: 1.0
Roth 2010	United States	507	16	Mixed dyslipidemia	LDL-C ≥ 3.37	TC: 6.7	Primary (CAD: 6%)	54.8 (11.0)	28%	42%	Rosuvastatin 5 mg daily	Fenofibrate 135 mg daily	Abbott and Astra Zeneca
					TG ≥ 1.69	TG: 2.7
					HDL-C < 1.02	LDL-C: 3.9
					(men) or
					HDL-C < 1.28 (women)	HDL-C: 1.1
Sano 2010	Japan	274	52	Elevated remnant lipoprotein levels	RLP-C ≥ 5.0 mg/dL	TC: 5.5	Secondary (CAD: 100%)	56.0 (9.4)	34%	88%	Pravastatin 10–20 mg daily	Bezafibrate 200–400 mg daily	Ministry of Education, Culture, Sports, Science and Technology, Health and Labor Sciences Research Grants for Comprehensive Research on Aging and Health, Japan
					TC ≥ 4.66 and <	TG: 2.3
					6.73	LDL-C: 3.3
					TG ≥ 1.69 and <	HDL-C: 1.1
					4.52
Foucher 2015	Czech Republic	346	12	Mixed dyslipidemia	LDL-C 1.81 to 3.37	TC: 4.7	Secondary (CVD: 44%)	60.1 (9.1)	58%	65%	Simvastatin 20–40 mg daily	Fenofibrate 145 mg daily	Abbott
	Mexico				TG ≥ 1.69	TG: 2.7
	Argentina					LDL-C: 2.7
	Poland					HDL-C: 1.2
	Romania
	Russian Federation
	Germany

Abbreviations: CAD, coronary artery disease; CHD, coronary heart disease; CVD, cardiovascular disease; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; NCEP, National Cholesterol Education Program; NR, not reported; RLP-C, remnant-like lipoprotein particles cholesterol; TC, total cholesterol; TG, triglycerides.

To convert LDL-C, HDL-C, or total cholesterol to mg/dL divide by 0.0259. To convert triglycerides to mg/dL divide by 0.0113.

Risk of bias

Results of the risk of bias assessment across studies and for each study is shown in Figs 1, 2 in S1 File. Most studies did not report allocation concealment and were judged to be at either unclear risk of bias or at high risk of bias. Other reasons for high risk of bias were blinding of participants and personnel and blinding of outcome assessment for subjective outcomes. Four studies were judged to be at high risk of bias because of incomplete outcome data, which was due to higher participant withdrawals in the fibrate group [30, 31], or excessive dropouts in both treatment groups [33, 36]. Reporting bias was assessed graphically using funnel plots for the outcomes of cardiovascular mortality, all-cause mortality, major cardiovascular events, myocardial infarction, withdrawal due to adverse events, myalgia, and elevated CK (Figs 3–9 in S1 File). We strongly suspected reporting bias only for the outcome of participant withdrawal due to adverse events.

Certainty of the evidence

A summary of evidence table for the key outcomes of interest is included in S1 File. The certainty in our estimates for each outcome ranged from very low (unstable angina) to high (elevated serum creatinine). Because many studies had a short duration of follow-up and low numbers of events, all clinical efficacy outcomes and several safety outcomes were judged to have serious imprecision. Surrogate lipid outcomes (percent reduction in TC, LDL-C, HDL-C, non-HDL-C, triglycerides, and apoB) were downgraded for indirectness (surrogate outcomes for cardiovascular events) resulting in moderate certainty of evidence.

Efficacy outcomes

Cardiovascular mortality

A total of 11 studies reported at least one cardiovascular death. We excluded Pover 1995 from the estimation of cardiovascular mortality, since the authors reported that 80% of all deaths in the study were due to acute myocardial infarction, but did not report these deaths according to treatment assignment [33]. The limited number of events resulted in imprecise estimates and there was no evidence of a difference in cardiovascular mortality between statins and fibrates (OR 2.35, 95% CI 0.94–5.86, I² = 0%; ten studies, n = 2657; low certainty; Fig 2).

Fig 2

Forest plot of comparison: Statins versus fibrates, outcome: Cardiovascular mortality.

All-cause mortality

Eleven studies with follow-up ranging from 10 weeks to 2 years, reported a death from any cause (OR 1.67, 95% CI 0.87–3.22, I² = 0%; n = 5124; Fig 3). Evidence was downgraded for imprecision and high or uncertain risk of bias resulting in a low certainty of evidence.

Fig 3

Forest plot of comparison: Statins versus fibrates, outcome: All-cause mortality.

Major cardiovascular events

In the eligible studies, no evidence of a difference for statins and fibrates was observed for major cardiovascular events (OR 1.15, 95% CI 0.80–1.65, I² = 13%; 19 studies, n = 7619; low certainty; Fig 4). Individual cardiovascular outcomes were rare resulting in imprecise estimates, and there was no evidence of a difference for myocardial infarction (OR 0.78, 95% CI 0.49–1.24, I² = 0%; 15 studies, n = 6362; Fig 5), coronary artery disease (OR 0.98, 95% CI 0.34–2.78, I² = 0%; six studies, n = 2505; Fig 6), unstable angina (OR 2.38, 95% CI 0.90–6.24, I² = 52%; four studies, n = 1200; Fig 7), and stroke (OR 2.04, 95% CI 0.86–4.82, I² = 36%; three studies, n = 1157; Fig 8).

Fig 4

Forest plot of comparison: Statins versus fibrates, outcome: Major cardiovascular events.

Fig 5

Forest plot of comparison: Statins versus fibrates, outcome: Myocardial infarction.

Fig 6

Forest plot of comparison: Statins versus fibrates, outcome: Coronary artery disease.

Fig 7

Forest plot of comparison: Statins versus fibrates, outcome: Unstable angina.

Fig 8

Forest plot of comparison: Statins versus fibrates, outcome: Stroke.

Surrogate efficacy outcomes

There were greater reductions in percent change from baseline for TC (MD -11.49%, 95% CI -12.20 to -10.77, I² = 96%; 15 studies, n = 6002; S1 Fig), LDL-C (MD -19.63%, 95% CI -20.70 to -18.55, I² = 96%; 15 studies, n = 5795; S2 Fig), non-HDL-C (MD -20.94%, -22.46 to -19.41, I² = 93%; four studies, n = 2008; S3 Fig), and apoB (MD -16.83%, 95% CI -18.10 to -15.56, I² = 86%; nine studies, n = 3003; S4 Fig) among statin therapy than fibrate therapy. Fibrates reduced triglyceride levels by 15.34% (95% CI 13.52 to 17.15, I² = 71%; 15 studies, n = 5922; S5 Fig) and increased HDL-C concentrations (MD 8.15%, 95% CI 9.23 to 7.07, I² = 69%; 15 studies, n = 5850; S6 Fig) more than statins. Pooled results for studies which we imputed SD were of similar magnitude, but tended to be closer to the null (except for triglyceride levels) as compared to studies with no SD imputation.

Safety outcomes

Tolerability

Statins were associated with a lower risk of study withdrawal due to adverse effects (OR 0.71, 95% CI 0.55–0.93, I² = 4%; 16 studies, n = 4680; low certainty; Fig 9) and serious adverse effects (OR 0.57, 95% CI 0.36–0.91, I² = 0%; nine studies, n = 3749; moderate certainty; Fig 10). Six studies included in the outcome of withdrawal due to adverse effects permitted dose increases in statin treatment at specified intervals [25, 31, 35, 39, 40, 42]. Dose titration may influence the reported adverse effects as participants initiated or maintained on lower doses of statins may be less likely to experience adverse effects, as compared to those initiated on a standard dose of fibrates and was considered in the risk of bias assessment.

Fig 9

Forest plot of comparison: Statins versus fibrates, outcome: Study withdrawal due to adverse effects.

Fig 10

Forest plot of comparison: Statins versus fibrates, outcome: Serious adverse effects.

Muscle-related adverse effects

There was no clear evidence of a difference for myalgia (OR 1.32, 95% CI 0.95–1.83, I² = 0%; ten studies, n = 6090; low certainty; Fig 11) or for elevations in CK (OR 1.43, 95% CI 0.99–2.06, I² = 0%; 14 studies, n = 6762; S7 Fig). No study reported rhabdomyolysis in participants receiving statin or fibrate monotherapy.

Fig 11

Forest plot of comparison: Statins versus fibrates, outcome: Myalgia.

Hepatic and renal adverse effects

In the primary analysis, statins increased the risk of elevated ALT (OR 1.43, 95% CI 1.03–1.99, I² = 44%; seven studies, n = 5225; low certainty; Fig 12). Statins greatly reduced the risk of elevated serum creatinine (OR 0.17, 95% CI 0.08–0.36, I² = 0%; six studies, n = 2553; high certainty; Fig 13). Because of the small number of events and different outcome descriptions, we could not pool studies for the outcome of kidney injury. Three studies reported kidney injury related outcomes as renal failure [30, 43], renal impairment [27], or renal dysfunction [36]. A total of four cases of kidney injury were reported in the fibrate group and zero in the statin group.

Fig 12

Forest plot of comparison: Statins versus fibrates, outcome: Elevated alanine aminotransferase.

Fig 13

Forest plot of comparison: Statins versus fibrates, outcome: Elevated serum creatinine.

Other outcomes

A limited number of events did not permit pooling of studies for the outcomes of new onset diabetes mellitus, venous thromboembolism, and non-cardiovascular mortality. Two studies [39, 42] reported the outcome of increased fasting blood glucose, with one case occurring in a participant assigned to fibrate [39] and the other in a participant assigned to statin therapy [42]. In a third study, one case each of diabetes mellitus, decompensated type 2 diabetes mellitus, and inadequate control of diabetes mellitus were reported (two participants in the statin group and one participant in the fibrate group) [27]. A single case of deep vein thrombosis was reported in the fibrate group [28, 44]. Non-cardiovascular deaths were reported in two studies [33, 39].

Subgroup and sensitivity analyses

The results of subgroup analyses and are mostly consistent with the main analyses (S1 File). Evidence of differences between subgroups was apparent for major cardiovascular events and elevated ALT. Statins appeared to reduce the risk of major cardiovascular events in the primary prevention and mixed dyslipidemia subgroups, while fibrates reduced the risk in the secondary prevention and other dyslipidemia subgroup (Fig 12 in S1 File). A high degree of heterogeneity between studies for elevated ALT resulted in different estimates by fibrate drug (Fig 17 in S1 File).

The results of sensitivity analyses are shown in S1 Table. Cardiovascular mortality, all-cause mortality, stroke, unstable angina, and elevated ALT were all sensitive to model choice and should be considered as hypothesis generating results. Estimates using the Peto method were generally consistent with the main analysis, despite not always meeting all three suggested criteria for analysis of rare events [21].

Strengths and limitations

This study has several strengths. First, we included trials of head-to-head comparisons of statins and fibrates that have not usually been included in previous systematic reviews. Second, a careful review of full-text articles by two independent authors, allowed us to identify all cardiovascular event outcomes for study inclusion. However, this study is limited by the eligible randomized controlled trials. The short duration of follow-up and rare events resulted in reduced power to detect differences between groups, and some estimates were sensitive to the choice of meta-analysis model and should be considered as hypothesis generating. Lastly, in the context of secondary prevention patients, the choice of statin drug and statin dose intensity in many of the eligible trials does not align with current guideline recommendations for treatment with high-intensity statins.

Future research on statins versus fibrates

Given the more robust body of evidence for statins in reducing the risk of cardiovascular events and an improved safety profile, further research is needed to understand the role of fibrates for the prevention of cardiovascular events. We identified one ongoing study, the Pitavastatin or Bezafibrate Intervention, Assessment of Antiarteriosclerotic Effect study (PIONEER), that is directly comparing pitavastatin to bezafibrate monotherapy with the primary outcome of change in mean carotid intima-media thickness [53]. The results of PIONEER may add additional information to the potential role of fibrate monotherapy particularly in patients with both elevated LDL-C and triglyceride concentrations. In addition, observational studies with adequate sample size and long-term follow-up can complement the available data from randomized controlled trials.