What is MACE and how do I handle it in meta-analysis?

MACE (Major Adverse Cardiovascular Events) is a composite endpoint typically including cardiovascular death, non-fatal myocardial infarction, and non-fatal stroke (3-point MACE). Some trials use 4-point MACE by adding hospitalization for unstable angina. The critical issue for meta-analysis is that MACE definitions vary across trials. Before pooling, verify that all studies use the same MACE definition. If definitions differ, either analyze individual components separately or clearly state the heterogeneity in your methods.

Should I use HR or OR for cardiovascular meta-analysis?

Use HR (Hazard Ratio) for time-to-event endpoints like MACE, CV death, and heart failure hospitalization, because HR accounts for censoring and the time dimension. Use OR (Odds Ratio) or RR (Risk Ratio) for binary endpoints measured at a fixed time point (e.g., in-hospital mortality, 30-day readmission rate). Most CVOTs report HR as the primary analysis, so HR is the default choice for large cardiovascular outcome trials.

How do I handle ITT vs per-protocol analysis in cardiovascular meta-analysis?

Always use Intention-to-Treat (ITT) analysis as the primary result because it preserves randomization and avoids selection bias from treatment discontinuation. Per-protocol analysis can be used as a sensitivity analysis. In cardiovascular trials, treatment discontinuation and crossover are common (especially in long-duration CVOTs), making ITT particularly important. Document which analysis population each study uses.

How many studies do I need for a cardiovascular meta-analysis?

There is no strict minimum, but practical considerations apply. Heterogeneity tests (I-squared) have low power with fewer than 5 studies. Funnel plots and Egger's test are unreliable with fewer than 10 studies. Meta-regression requires at least 10 studies per covariate. For drug class meta-analyses (e.g., SGLT2 inhibitors), 3-6 large CVOTs may provide sufficient power. Even 2-3 studies can justify pooling if they address the same clinical question with similar designs.

How do I meta-analyze lipid-lowering therapy trials?

For efficacy: pool the percent change in LDL-C as Mean Difference (MD) across trials. For clinical outcomes (MACE, CV death): use HR from each CVOT. Key considerations: (1) ensure baseline LDL-C is comparable, (2) account for background statin therapy, (3) distinguish between primary and secondary prevention populations, (4) report absolute risk reduction (ARR) alongside relative measures. The NNT (Number Needed to Treat) provides clinical context that relative measures lack.

What are the key differences between primary and secondary prevention CV trials in meta-analysis?

Primary prevention trials enroll patients without established CV disease (higher NNT, smaller absolute risk reductions). Secondary prevention trials enroll patients with prior CV events (lower NNT, larger absolute benefits). Mixing primary and secondary prevention populations creates significant heterogeneity because baseline event rates differ dramatically. Always pre-specify this as a subgroup variable. If pooling both, use random-effects models and report the subgroup interaction test.

How do I handle competing risks in cardiovascular meta-analysis?

Competing risks occur when patients can experience events that prevent the outcome of interest (e.g., non-CV death prevents observation of CV death). Standard Cox regression (HR) may overestimate the cumulative incidence of CV events if non-CV mortality is high. Some CV trials report cause-specific HR and subdistribution HR (Fine-Gray). When available, extract the cause-specific HR for consistency. Note competing risk limitations in your discussion, especially for elderly populations where non-CV mortality is substantial.

Can I include observational studies alongside RCTs in cardiovascular meta-analysis?

Yes, but with caveats. Analyze RCTs and observational studies separately first. If results are consistent, you may consider a combined analysis with appropriate sensitivity analyses. Observational studies are subject to confounding (healthy user bias, immortal time bias, indication bias). Use the Newcastle-Ottawa Scale for quality assessment of observational studies. In the GRADE framework, observational evidence starts at 'low' certainty while RCT evidence starts at 'high'.

Cardiovascular Meta-Analysis Guide: MACE, Lipid Endpoints & Trial Design (2026)

Why Meta-Analysis Is Essential in Cardiology
Defining Your CV Research Question (PICO)
Understanding Cardiovascular Endpoints
Choosing the Right Effect Size
MACE Composite Endpoints: Definition Challenges
Data Extraction from CV Trials
Handling Heterogeneity in CV Meta-Analysis
Subgroup Analysis by Risk Profile
Step-by-Step: CV Meta-Analysis in MetaReview
Common Pitfalls in Cardiovascular Meta-Analysis

Why Meta-Analysis Is Essential in Cardiology

Cardiovascular medicine generates more clinical trial data than almost any other specialty. Meta-analysis plays a central role because:

CVOTs are large but few — A single CVOT may enroll 10,000+ patients but only 3-5 trials exist for a given drug class. Pooling these trials reveals treatment effects with much greater precision.
Event rates are low — With 3-point MACE rates of 5-10% over 3 years, individual trials may lack power for component endpoints (CV death alone, stroke alone). Meta-analysis provides the statistical power to analyze individual components.
Guideline-shaping — ACC/AHA, ESC, and other bodies base treatment recommendations on systematic reviews and meta-analyses. Your analysis could directly influence clinical practice.
Drug class effects — Meta-analysis can determine whether benefits are drug-specific or represent a class effect (e.g., do all SGLT2 inhibitors reduce heart failure hospitalization, or only empagliflozin?)

Landmark CV meta-analyses that changed practice: CTT Collaboration (statin therapy), EMPA-REG/CANVAS/DECLARE pooled analysis (SGLT2 inhibitors for HFrEF), and antiplatelet therapy duration meta-analyses.

Defining Your CV Research Question (PICO)

Element	Description	CV Examples
P (Population)	CV condition, risk profile, comorbidities	Patients with established ASCVD; T2DM with high CV risk; HFrEF (EF ≤40%); AF on anticoagulation
I (Intervention)	Treatment being evaluated	SGLT2 inhibitors; PCSK9 inhibitors; DOACs; GLP-1 receptor agonists
C (Comparator)	Control treatment	Placebo + standard of care; Warfarin; Statin monotherapy
O (Outcomes)	Primary and secondary endpoints	3-point MACE; CV death; MI; Stroke; HF hospitalization; All-cause death; LDL-C change

Search Strategy for CV Trials

Databases: PubMed, Embase, Cochrane CENTRAL, ClinicalTrials.gov
Conference proceedings: ACC (acc.org), AHA (heart.org), ESC (escardio.org), TCT
MeSH terms: "Cardiovascular Diseases"[Mesh], "Myocardial Infarction"[Mesh], specific drug class terms
Key filter: Publication type = "Randomized Controlled Trial" for highest evidence

Scope management: A PICO of "any anticoagulant for any cardiac condition" is too broad. Narrow to a specific drug class (DOACs), condition (non-valvular AF), and endpoint (stroke prevention). You can expand with subgroup analyses.

Understanding Cardiovascular Endpoints

CV trials use a hierarchy of endpoints ranging from hard clinical outcomes to surrogate markers.

Hard Clinical Endpoints

Endpoint	Definition	Analysis Type	Effect Size
All-cause death	Death from any cause	Time-to-event	HR
CV death	Death attributed to cardiovascular cause	Time-to-event	HR
Non-fatal MI	MI not resulting in death within 28 days	Time-to-event	HR
Non-fatal stroke	Stroke not resulting in death within 28 days	Time-to-event	HR
HF hospitalization	Hospitalization for worsening heart failure	Time-to-event	HR

Composite Endpoints

Composite	Components	Used In
3-point MACE	CV death + non-fatal MI + non-fatal stroke	Most CVOTs (FDA-required)
4-point MACE	3-point MACE + hospitalization for unstable angina	Some older CVOTs
HF composite	CV death + HF hospitalization	HF trials (EMPEROR, DAPA-HF)
Kidney composite	Sustained eGFR decline + ESKD + renal death	Cardiorenal trials

Surrogate Endpoints

LDL-C change (%) — Validated surrogate for CV events (CTT Collaboration). Analyze as Mean Difference (MD).
HbA1c change — For cardiometabolic trials. Not a validated CV surrogate.
Blood pressure change — Validated surrogate for stroke and heart failure.
LVEF change — Surrogate for HF prognosis. Analyze as MD.

Best practice: Always report the composite endpoint result AND individual component results. A composite MACE HR of 0.80 may be driven entirely by MI reduction with no effect on CV death or stroke. Component analysis reveals the true treatment profile.

Choosing the Right Effect Size

What is your CV endpoint type? ├─ Time-to-first-event (MACE, CV death, MI, stroke, HF hosp) │ └─ Use Hazard Ratio (HR) │ └─ Standard output of Cox regression in CVOTs ├─ Binary at fixed timepoint (30-day mortality, in-hospital events) │ ├─ Common events (>20%) → Risk Ratio (RR) │ └─ Rare events (<20%) → Odds Ratio (OR) └─ Continuous (LDL-C change, BP change, LVEF change) ├─ Same unit across studies → MD └─ Different scales → SMD

Special Considerations for CV Meta-Analysis

Recurrent events: Standard HR only captures time-to-first-event. Some newer CV trials report recurrent event analyses (e.g., total HF hospitalizations). These require different statistical methods (negative binomial, Anderson-Gill, or Wei-Lin-Weissfeld models).
Competing risks: Non-CV death competes with CV endpoints. Standard HR may overestimate cumulative CV event risk in elderly populations. Note this limitation in your discussion.
Absolute vs relative risk: HR = 0.80 is more meaningful in a high-risk population (ARR 4%, NNT 25) than in a low-risk population (ARR 0.5%, NNT 200). Report both.

MACE Composite Endpoints: Definition Challenges

MACE is the most common primary endpoint in CVOTs, but its definition is not standardized across all trials.

MACE Definition Variations

Trial	MACE Definition	Components
EMPA-REG OUTCOME	3-point MACE	CV death + non-fatal MI + non-fatal stroke
CANVAS Program	3-point MACE	CV death + non-fatal MI + non-fatal stroke
LEADER	3-point MACE	CV death + non-fatal MI + non-fatal stroke
SAVOR-TIMI 53	3-point MACE	CV death + MI + ischemic stroke
Older trials	4-point MACE	3-point + unstable angina hospitalization

Handling Definition Heterogeneity

Preferred approach: Pool only studies using the identical MACE definition
If mixing is unavoidable: Conduct sensitivity analysis excluding studies with different definitions
Best alternative: Analyze individual MACE components (CV death, MI, stroke) separately — these are consistently defined

Adjudication matters: Centrally adjudicated events (blinded endpoint committee) are more reliable than investigator-reported events. Prefer studies with independent event adjudication committees.

Data Extraction from CV Trials

Essential Data Points

Study identification: First author, year, trial name/acronym (e.g., "EMPA-REG OUTCOME"), NCT number
Population: Inclusion criteria, established ASCVD vs risk factors only, HF status (HFrEF/HFpEF/no HF)
Baseline characteristics: Mean age, sex distribution, diabetes prevalence, prior MI/stroke, mean LVEF, mean eGFR, baseline LDL-C
Treatment: Drug name, dose, background therapy (statin use %, antiplatelet %, ACEi/ARB %)
Follow-up: Median duration, completion rate
Outcomes: HR + 95% CI for each endpoint, event counts per arm, ITT analysis

Data Extraction Template

Field	Source in Paper	Notes
HR (95% CI) for MACE	Abstract or primary results table	Use ITT population
HR for CV death	Component analysis table or forest plot	May be in supplementary appendix
HR for MI	Component analysis table	Distinguish fatal vs non-fatal
HR for stroke	Component analysis table	Distinguish ischemic vs hemorrhagic
HR for HF hospitalization	Secondary endpoint or supplementary	Often not a primary endpoint
Events / N per arm	Results table	For calculating event rates
Median follow-up	Methods or results	Affects maturity of OS data

Check supplementary appendices. CV trial primary publications often report only composite results. Individual component HRs, subgroup forest plots, and sensitivity analyses are frequently in supplementary materials or companion papers.

Handling Heterogeneity in CV Meta-Analysis

Sources of Heterogeneity

Source	Examples	Approach
Baseline CV risk	Primary vs secondary prevention; HFrEF vs HFpEF	Subgroup analysis
Background therapy	Statin use (30% vs 95%), antiplatelet variation	Meta-regression on % statin use
Follow-up duration	1.5 years (SUSTAIN-6) vs 5.4 years (FOURIER)	Sensitivity analysis, meta-regression
Drug dose	Different doses within the same class	Dose-response meta-analysis
Geographic variation	Regional differences in CV risk profile	Subgroup by region
Endpoint adjudication	Central vs investigator-reported	Sensitivity analysis

Strategies for High Heterogeneity

Pre-specified subgroup analysis — primary vs secondary prevention, with vs without HF, with vs without diabetes
Meta-regression — test moderators like baseline LDL-C, % statin use, mean age, follow-up duration
Sensitivity analysis — restrict to Phase III CVOTs only, exclude open-label studies, ITT only
Component-level pooling — if composite heterogeneity is high, analyze individual components which may be more homogeneous

Expected I² ranges in CV meta-analyses: Drug class CVOTs (same mechanism) typically show I² 0-30%. Cross-class comparisons may show I² 40-70%. Mixed primary/secondary prevention cohorts often exceed I² 50%.

Subgroup Analysis by Risk Profile

Cardiovascular trials are particularly amenable to clinically meaningful subgroup analyses because baseline risk strongly modifies absolute treatment benefit.

Key Pre-Specified Subgroups

Subgroup Variable	Categories	Clinical Rationale
Prevention type	Primary vs secondary	Event rates differ 3-5x, affecting absolute benefit
Heart failure status	HFrEF vs HFpEF vs no HF	SGLT2i show differential HF benefit by EF
Diabetes status	T2DM vs no diabetes	Some drug classes approved only for diabetic patients
Renal function	eGFR ≥60 vs 30-59 vs <30	CV risk increases with CKD stage; some drugs contraindicated
Age	<65 vs ≥65 vs ≥75	Bleeding risk increases with age (anticoagulants)
Baseline LDL-C	Above vs below median	Higher baseline = greater absolute LDL-C reduction
Prior MI	Yes vs no	Post-MI patients have higher event rates

NNT by Risk Profile

Absolute risk reduction (ARR) and Number Needed to Treat (NNT) vary dramatically by baseline risk:

NNT = 1 / ARR = 1 / (Control Event Rate × (1 − HR))

Example for HR = 0.80:

High-risk (20% event rate): ARR = 20% × 0.20 = 4.0%, NNT = 25
Moderate-risk (10% event rate): ARR = 10% × 0.20 = 2.0%, NNT = 50
Low-risk (3% event rate): ARR = 3% × 0.20 = 0.6%, NNT = 167

Statistical caution: Subgroup analyses in individual trials are often underpowered. Meta-analysis of patient-level subgroups requires individual participant data (IPD). Study-level subgroup meta-analysis (using aggregate subgroup HRs) is more feasible but subject to ecological bias.

Step-by-Step: CV Meta-Analysis in MetaReview

Step 1: Select Effect Measure

Open MetaReview and choose:

Hazard Ratio for MACE, CV death, MI, stroke, HF hospitalization
Odds Ratio / Risk Ratio for binary outcomes (30-day mortality, procedural success)
Mean Difference for continuous outcomes (LDL-C change, BP change)

Step 2: Enter Trial Data

For each CVOT, enter the trial name/acronym (e.g., "EMPA-REG 2015"), publication year, and HR + 95% CI from the ITT analysis.

Batch entry: Copy your extraction table from Excel or Google Sheets and paste directly into MetaReview. The tool auto-detects and maps columns.

Step 3: Run Multiple Analyses

Conduct separate meta-analyses for:

Composite MACE
CV death (individual component)
Non-fatal MI (individual component)
Non-fatal stroke (individual component)
HF hospitalization (if reported)
All-cause death

This component-level analysis reveals whether the composite result is driven by one specific endpoint.

Step 4: Assign Subgroups and Re-Analyze

Label studies as "Primary Prevention" or "Secondary Prevention" in the Subgroup column. Re-run to see stratified forest plots with Q-between interaction test.

Step 5: Assess Bias and Quality

Funnel plot + Egger's test for publication bias
Trim-and-Fill to estimate missing study impact
GRADE assessment across 5 domains (risk of bias, inconsistency, indirectness, imprecision, publication bias)
Leave-one-out sensitivity analysis to check if one trial drives the result

Step 6: Export Report

Generate HTML or DOCX reports with all forest plots (composite + components), funnel plots, heterogeneity statistics, and auto-generated Methods paragraph in PRISMA 2020 format.

Common Pitfalls in Cardiovascular Meta-Analysis

Pitfall 1: Inconsistent MACE Definitions

Pooling 3-point MACE with 4-point MACE inflates heterogeneity. Studies adding unstable angina hospitalization will have higher event rates.

Solution: Pool only studies with identical MACE definitions. Alternatively, analyze individual MACE components where definitions are standardized.

Pitfall 2: Mixing ITT and Per-Protocol Analyses

ITT preserves randomization but may dilute treatment effects due to discontinuation. Per-protocol inflates effects by excluding non-compliant patients.

Solution: Use ITT results as the primary analysis. Report per-protocol as sensitivity analysis. Document which population each study uses.

Pitfall 3: Ignoring Background Therapy Evolution

Older CV trials had lower statin use (30-50%), while modern CVOTs have near-universal statin use (90%+). This affects the baseline event rate and the potential for additional benefit.

Solution: Record background therapy rates and use meta-regression to explore whether statin use modifies the treatment effect.

Pitfall 4: Double-Counting from Multiple Publications

Large CVOTs often produce multiple publications: primary results, extended follow-up, subgroup analyses, and mechanistic sub-studies. Using data from multiple timepoints of the same trial creates bias.

Solution: Use the most recent (most mature) data cutoff for each trial. Track trial registration numbers (NCT IDs) to identify duplicate publications.

Pitfall 5: Neglecting Absolute Risk Measures

A pooled HR of 0.80 sounds impressive, but if the baseline event rate is only 2%, the ARR is just 0.4% (NNT = 250). Relative measures alone can overstate clinical importance.

Solution: Report ARR and NNT alongside HR. Calculate NNT at different baseline risk levels to help clinicians apply results to their patient populations.

Pitfall 6: Publication and Reporting Bias in Industry CVOTs

FDA-mandated CVOTs are designed to demonstrate non-inferiority for safety, with superiority as a secondary objective. Positive results receive prominent publication; neutral results may be reported with less emphasis.

Solution: Search ClinicalTrials.gov for all registered CVOTs with the drug class. Compare registered primary endpoints against published endpoints to detect outcome switching. Include industry-funded and independently-funded studies.

Start Your Cardiovascular Meta-Analysis Now

Enter HR and 95% CI from CVOTs. MetaReview handles pooling, forest plots, component analysis, and report generation. Free, no coding required.

Open MetaReview

See live demo: Aspirin vs Placebo meta-analysis (7 RCTs) →

Stay Updated

Get notified about new features and meta-analysis tips.

No spam. Unsubscribe anytime.

Cardiovascular Meta-Analysis Guide

Table of Contents

Why Meta-Analysis Is Essential in Cardiology

Defining Your CV Research Question (PICO)

Search Strategy for CV Trials

Understanding Cardiovascular Endpoints

Hard Clinical Endpoints

Composite Endpoints

Surrogate Endpoints

Choosing the Right Effect Size

Special Considerations for CV Meta-Analysis

MACE Composite Endpoints: Definition Challenges

MACE Definition Variations

Handling Definition Heterogeneity

Data Extraction from CV Trials

Essential Data Points

Data Extraction Template

Handling Heterogeneity in CV Meta-Analysis

Sources of Heterogeneity

Strategies for High Heterogeneity

Subgroup Analysis by Risk Profile

Key Pre-Specified Subgroups

NNT by Risk Profile

Step-by-Step: CV Meta-Analysis in MetaReview

Step 1: Select Effect Measure

Step 2: Enter Trial Data

Step 3: Run Multiple Analyses

Step 4: Assign Subgroups and Re-Analyze

Step 5: Assess Bias and Quality

Step 6: Export Report

Common Pitfalls in Cardiovascular Meta-Analysis

Pitfall 1: Inconsistent MACE Definitions

Pitfall 2: Mixing ITT and Per-Protocol Analyses

Pitfall 3: Ignoring Background Therapy Evolution

Pitfall 4: Double-Counting from Multiple Publications

Pitfall 5: Neglecting Absolute Risk Measures

Pitfall 6: Publication and Reporting Bias in Industry CVOTs

Start Your Cardiovascular Meta-Analysis Now

Stay Updated

Related Guides