Major Adverse Cardiovascular Events (MACE)

In plain language

MACE is a shorthand for a serious cardiovascular-event composite. The common version is cardiovascular death, heart attack, or stroke. But different studies use different component lists, so the word MACE is not enough by itself.

MACE is not one universal endpoint. It is a family of cardiovascular composite endpoints whose interpretation depends on the component list, event ascertainment rules, and whether the analysis counts first event only or recurrent components. The most common "3-point MACE" definition combines cardiovascular death, nonfatal myocardial infarction, and nonfatal stroke. Other studies use 4-point, 5-point, or study-specific MACE definitions that add hospitalization for unstable angina, urgent revascularization, heart failure hospitalization, or all-cause death.

In RWE, MACE is an outcome-algorithm problem plus a composite-endpoint problem. Nonfatal MI and stroke are usually identified through inpatient diagnosis algorithms and validated code lists. Cardiovascular death requires cause-of-death linkage; all-cause death is easier but changes the estimand. The composite date is usually the earliest qualifying component date, but the component that fires must be retained because a treatment effect dominated by one component is not a general cardiovascular benefit.

Cross-study comparison is especially risky. "MACE" in a diabetes CV outcomes trial, a device registry, and a claims-based safety study may have different death definitions, claim-position requirements, washout periods, and adjudication standards. RWE reports should spell out the component list in the title, table shell, and methods, not only in the appendix.

Pros, cons, and trade-offs

MACE is attractive because it increases event yield and gives clinicians a familiar serious cardiovascular endpoint. The trade-off is interpretability. Composite precision is not useful if a softer or more frequent component drives the result while cardiovascular death, myocardial infarction, and stroke move differently. Claims-based MACE adds another trade-off: inpatient MI and stroke algorithms may be reproducible, but cause-specific death and out-of-network events require linkage that many datasets do not have.

When to use

Use MACE when the components are clinically coherent, event capture is fit for purpose, and the estimand is time to the first serious cardiovascular event. Keep the first firing component, run component-specific estimates, and state whether the endpoint is 3-point MACE, expanded MACE, MACCE, or a study-specific variant.

When NOT to use - and when it is actively misleading

Do not write "MACE" when the study only observes nonfatal claims events or all-cause death. Do not combine revascularization, unstable angina, heart failure, and death without showing component results. It is actively misleading to compare a nonfatal claims-only composite with a trial-adjudicated 3-point MACE endpoint as if the endpoints were the same.

Index definitions

Source-backed definitions and variants for the index or checklist family.

Worked example

Scenario

A claims study of a diabetes drug defines 3-point MACE as cardiovascular death, nonfatal inpatient MI, or nonfatal inpatient ischemic stroke. One patient has an inpatient MI on day 143, later revascularization on day 151, and death without cause information on day 220.

Dataset

Candidate cardiovascular events for one patient.

Steps

• Apply the MI algorithm and de-duplication window to the inpatient claim.

• Exclude revascularization because the protocol specified 3-point MACE, not expanded MACE.

• Do not classify death as cardiovascular death without cause-of-death evidence.

• Assign first MACE date to day 143 and retain winning component = nonfatal MI.

Result

The time-to-first 3-point MACE endpoint fires on day 143. The later revascularization and cause-unknown death are reported separately or handled in sensitivity analyses.

Runnable example

import pandas as pd

MACE3 = {"cardiovascular_death", "nonfatal_mi", "nonfatal_stroke"}

def first_mace(index, events, followup_end=None):
    candidates = events[events["component"].isin(MACE3)].copy()
    candidates = candidates.merge(index[["person_id", "index_date"]], on="person_id", how="inner")
    candidates = candidates[candidates["event_date"] >= candidates["index_date"]]
    if followup_end is not None:
        candidates = candidates.merge(followup_end[["person_id", "end_date"]], on="person_id", how="left")
        candidates = candidates[candidates["end_date"].isna() | (candidates["event_date"] <= candidates["end_date"])]
    candidates = candidates.sort_values(["person_id", "event_date", "component"])
    out = candidates.groupby("person_id", as_index=False).first()
    out["mace3"] = 1
    out["time_to_mace_days"] = (out["event_date"] - out["index_date"]).dt.days
    return index.merge(
        out[["person_id", "mace3", "event_date", "component", "source", "time_to_mace_days"]],
        on="person_id",
        how="left",
    ).fillna({"mace3": 0})

library(data.table)

first_mace <- function(index, events, followup_end = NULL) {
  setDT(index); setDT(events)
  mace3 <- c("cardiovascular_death", "nonfatal_mi", "nonfatal_stroke")
  cand <- events[component %in% mace3]
  cand <- merge(cand, index[, .(person_id, index_date)], by = "person_id")
  cand <- cand[event_date >= index_date]
  if (!is.null(followup_end)) {
    setDT(followup_end)
    cand <- merge(cand, followup_end[, .(person_id, end_date)], by = "person_id", all.x = TRUE)
    cand <- cand[is.na(end_date) | event_date <= end_date]
  }
  setorder(cand, person_id, event_date, component)
  first <- cand[, .SD[1], by = person_id]
  first[, `:=`(mace3 = 1L, time_to_mace_days = as.integer(event_date - index_date))]
  out <- merge(index, first[, .(person_id, mace3, event_date, component, source, time_to_mace_days)],
               by = "person_id", all.x = TRUE)
  out[is.na(mace3), mace3 := 0L]
  out[]
}