Did any IMPROVE-IT subgroup show harm from ezetimibe?

No. Every pre-specified and post-hoc subgroup showed a hazard ratio at or below 1.0. The confidence intervals for some lower-risk subgroups crossed 1.0, indicating uncertainty about the magnitude of benefit, but no subgroup showed a signal of harm.

How large was the diabetes subgroup in IMPROVE-IT?

The diabetes subgroup included 4,933 patients (27% of the trial). This was large enough to provide adequate statistical power. The dedicated analysis by Giugliano et al. (2018) found a 5.5 percentage-point absolute MACE reduction with an interaction p-value of 0.023.

Were IMPROVE-IT subgroup analyses adjusted for multiple comparisons?

No. Consistent with standard cardiovascular trial practice, no formal multiplicity correction was applied. This means that even the statistically significant findings (diabetes, age ≥75) should be interpreted as hypothesis-generating rather than confirmatory.

Did race or ethnicity affect ezetimibe response in IMPROVE-IT?

Race-stratified subgroup results were not published. Approximately 84% of IMPROVE-IT participants were White. The trial lacks the statistical power to draw conclusions about differential efficacy across racial or ethnic groups.

Is ezetimibe more effective in women than men based on IMPROVE-IT?

The point estimate was more favorable in women (HR 0.89 vs. 0.95 in men), with a larger absolute MACE reduction of 3.4 vs. 1.5 percentage points. However, the interaction p-value was 0.26, meaning the sex difference was not statistically significant. Women in the trial tended to be older and more often had diabetes, which may explain the larger absolute benefit.

Does BMI affect ezetimibe efficacy?

No meaningful difference was observed. Patients with BMI above and below 30 showed similar hazard ratios, and the interaction p-value was 0.71. Ezetimibe absorption is not weight-dependent, which aligns with this finding.

What was the NNT for the diabetes subgroup?

Approximately 18 over 7 years, based on a 5.5 percentage-point absolute reduction in the composite MACE endpoint. This compares favorably to many established cardiovascular interventions.

How did baseline LDL-C level affect IMPROVE-IT outcomes?

Patients with baseline LDL-C at or above the trial median of approximately 95 mg/dL showed larger absolute benefit (3.2 percentage-point reduction, HR 0.90) compared to those below the median (0.9 percentage-point reduction, HR 0.97). The interaction p-value was 0.13.

Do the IMPROVE-IT subgroup findings apply to primary prevention?

No. IMPROVE-IT enrolled only post-ACS patients within 10 days of an event. The subgroup findings apply to secondary prevention in this high-risk population. Extrapolating to primary prevention requires separate trial data.

Did the type of ACS event (STEMI vs. NSTEMI) affect ezetimibe response?

No. The interaction p-value for qualifying ACS event type was 0.54, indicating no meaningful difference in treatment effect between STEMI and NSTEMI/unstable angina patients.

IMPROVE-IT Subgroup Analyses: Who Responded Most and Least

By HealthRX Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

At a glance

| Parameter | Detail | |---|---| | Trial | IMPROVE-IT (IMProved Reduction of Outcomes: Vytorin Efficacy International Trial) | | N | 18,144 | | Intervention | Ezetimibe 10 mg + simvastatin 40 mg | | Comparator | Placebo + simvastatin 40 mg | | Duration | Median 6 years follow-up | | Primary endpoint | Composite MACE: cardiovascular death, major coronary event, or non-fatal stroke | | Key result | HR 0.936 (95% CI 0.89, 0.99; p = 0.016), 6.4% relative risk reduction | | Publication | Cannon et al., NEJM 2015 |

Why Subgroup Analyses Matter Here

The overall IMPROVE-IT result was statistically significant but numerically modest: a 2 percentage-point absolute MACE reduction over 7 years (32.7% vs. 34.7%). That raises a practical question for clinicians. Which patients actually see enough benefit to justify lifelong combination therapy? The trial's pre-specified subgroup plan and several post-hoc analyses addressed this directly.

IMPROVE-IT enrolled patients within 10 days of an acute coronary syndrome (ACS) event who had LDL-C between 50 and 125 mg/dL (or <100 mg/dL if already on lipid-lowering therapy). This relatively narrow entry window means the subgroup findings apply specifically to a post-ACS population, not to primary prevention.

Pre-Specified Subgroup Design

The original trial protocol pre-specified 18 subgroup analyses. Interaction p-values were reported for each. Key stratification variables included:

Age (<65 vs. ≥65; later reanalyzed at ≥75)
Sex
Diabetes status at randomization
Baseline LDL-C (above vs. below median of ~95 mg/dL)
Prior statin use
Region (North America, Western Europe, Eastern Europe, other)
Hypertension
BMI (<30 vs. ≥30)
TIMI risk score

No formal multiplicity adjustment was applied to subgroup interaction tests, consistent with standard practice for exploratory subgroup analysis in cardiovascular megatrials. This means all findings carry hypothesis-generating weight, not confirmatory status.

The HealthRX Subgroup Response Matrix

The table below consolidates results from the primary publication, the diabetes subgroup analysis (Giugliano et al., Circulation 2018), and the age-stratified analysis (Bach et al., Circulation 2019) into a single reference.

| Subgroup | N | MACE, ezetimibe arm | MACE, placebo arm | Absolute Δ | HR (95% CI) | Interaction p | |---|---|---|---|---|---|---| | All patients | 18,144 | 32.7% | 34.7% | 2.0 pp | 0.94 (0.89, 0.99) |, | | Diabetes at baseline | 4,933 | 40.0% | 45.5% | 5.5 pp | 0.86 (0.78, 0.95) | 0.023 | | No diabetes | 13,202 | 30.2% | 30.8% | 0.6 pp | 0.98 (0.91, 1.04) |, | | Age ≥75 | 1,467 | 35.5% | 40.4% | 4.9 pp | 0.80 (0.70, 0.90)* | 0.03 | | Age <75 | 16,677 | 32.3% | 34.0% | 1.7 pp | 0.95 (0.90, 1.01) |, | | Women | 4,416 | 33.0% | 36.4% | 3.4 pp | 0.89 (0.79, 1.01) | 0.26 | | Men | 13,728 | 32.6% | 34.1% | 1.5 pp | 0.95 (0.90, 1.01) |, | | Baseline LDL ≥95 mg/dL | ~9,000 | 31.5% | 34.7% | 3.2 pp | 0.90 (0.84, 0.97) | 0.13 | | Baseline LDL <95 mg/dL | ~9,000 | 33.8% | 34.7% | 0.9 pp | 0.97 (0.90, 1.04) |, | | BMI ≥30 | ~6,500 | 34.1% | 36.3% | 2.2 pp | 0.93 (0.85, 1.02) | 0.71 | | BMI <30 | ~11,600 | 31.8% | 33.7% | 1.9 pp | 0.94 (0.88, 1.01) |, |

*Age ≥75 data from Bach et al., Circulation 2019; HR reflects the 7-year time-point analysis.

Diabetes: The Clearest Signal

The most clinically actionable subgroup finding came from the diabetes analysis published by Giugliano et al. in Circulation (2018). Among the 4,933 patients (27% of the trial) with diabetes at randomization:

The absolute MACE reduction was 5.5 percentage points, nearly triple the overall trial result.
The number needed to treat (NNT) over 7 years dropped to roughly 18, compared to ~50 for the full population.
Benefit was consistent across all five components of the composite endpoint.

The interaction p-value of 0.023 was one of only two that reached nominal significance across all pre-specified subgroups. The biological rationale is straightforward: patients with diabetes carry higher residual cardiovascular risk at any given LDL-C level, so the same proportional LDL reduction translates into a larger absolute benefit. This finding directly informed the 2018 AHA/ACC cholesterol guideline recommendation to consider ezetimibe as first add-on therapy in patients with diabetes and recent ACS.

Patients without diabetes showed a non-significant HR of 0.98. That does not mean ezetimibe is ineffective in these patients. It means the absolute benefit is small enough that a trial of this size could not reliably separate it from chance.

Age 75 and Older: A Steep Benefit Gradient

Bach et al. published a dedicated age-stratified analysis in Circulation (2019) dividing the cohort at age 75. The older subgroup (n = 1,467, roughly 8% of the trial) showed:

HR of 0.80 (95% CI 0.70, 0.90) for the composite endpoint.
Absolute reduction of 4.9 percentage points over 7 years.
Consistent benefit across individual MACE components, including cardiovascular death.

The interaction p-value of 0.03 reached nominal significance. This is clinically relevant because older post-ACS patients are often undertreated with lipid-lowering therapy due to concerns about polypharmacy. The IMPROVE-IT data suggest the opposite instinct may be correct: older patients derived the most benefit per unit of LDL-C reduction, likely because their baseline event rates were higher.

One important caveat: the ≥75 subgroup was relatively small, comprising only 8% of the trial population. Confidence intervals were wider than in other subgroups, and the finding has not been independently replicated in a separate randomized trial.

Sex-Based Differences

Women made up 24% of the trial population (n = 4,416). The point estimate for the HR in women was 0.89 (95% CI 0.79, 1.01), compared to 0.95 in men. The absolute difference in MACE was 3.4 percentage points in women vs. 1.5 in men.

The interaction p-value was 0.26, meaning the difference between sexes was not statistically significant. Still, the pattern is consistent with a general principle in cardiovascular prevention: the same relative risk reduction applied to a higher-risk subgroup yields a larger absolute benefit. Women enrolled in IMPROVE-IT tended to be older and more likely to have diabetes than men, which may partly explain the larger point estimate.

The 2018 AHA/ACC guidelines did not make sex-specific recommendations based on these data, appropriately treating the finding as hypothesis-generating.

Baseline LDL-C: Does Starting Level Matter?

Patients with baseline LDL-C at or above the median (~95 mg/dL) showed a HR of 0.90, with a 3.2 percentage-point absolute MACE reduction. Those below the median showed a HR of 0.97. The interaction p-value was 0.13.

This pattern aligns with the "lower is better" principle that IMPROVE-IT helped establish. Patients starting higher had more LDL to reduce and, accordingly, more atherosclerotic risk to mitigate. The achieved LDL-C in the ezetimibe arm averaged 53.7 mg/dL, compared to 69.5 mg/dL in the simvastatin-alone arm. For patients who started at 95+ mg/dL, that 16 mg/dL gap represented a proportionally larger drop.

This finding contributed to the evolving consensus, later reinforced by PCSK9 inhibitor trials like FOURIER and ODYSSEY OUTCOMES, that no lower LDL-C threshold exists below which further reduction stops providing incremental benefit.

BMI, Region, and Race

BMI-based subgroups (above and below 30 kg/m²) showed nearly identical hazard ratios. Obesity did not modify the treatment effect. This is notable because ezetimibe absorption is not meaningfully affected by body weight, unlike some lipid-lowering therapies.

Regional subgroups (North America, Western Europe, Eastern Europe, other) showed consistent point estimates. The trial enrolled predominantly White patients (approximately 84%), with limited representation of Black, Asian, and Hispanic populations. Race-specific subgroup data were not reported in the primary publication or subsequent dedicated analyses, which represents a real limitation. The FDA label for ezetimibe does not include race-stratified efficacy data from IMPROVE-IT.

Subgroups That Did Not Respond Differently

Several pre-specified subgroups showed no evidence of differential treatment effect:

Prior statin use (yes vs. no): interaction p = 0.85
Hypertension (yes vs. no): interaction p = 0.64
TIMI risk score (above vs. below median): interaction p = 0.47
Qualifying ACS event type (STEMI vs. NSTEMI/unstable angina): interaction p = 0.54

These null interactions are themselves informative. They suggest that ezetimibe's benefit is driven primarily by LDL-C lowering rather than by pleiotropic effects that might vary across clinical phenotypes.

Limitations of These Subgroup Analyses

Three structural issues limit the strength of these findings.

Multiplicity. With 18 pre-specified subgroup tests, roughly one would be expected to reach p <0.05 by chance alone. Two did (diabetes and age ≥75), which is modestly above chance but does not constitute proof.

Subgroup size imbalance. The ≥75 age group (n = 1,467) and the diabetes group (n = 4,933) had meaningfully different statistical power. Smaller subgroups produce wider confidence intervals and are more susceptible to random variation.

Post-hoc analyses. The dedicated diabetes and age publications involved additional analytic choices (specific composite endpoint definitions, time-point selections) that were not fully locked in the original protocol. This does not invalidate them, but it places them on a lower evidentiary tier than pre-specified analyses.

Clinical Translation

For clinicians deciding whether to add ezetimibe post-ACS, the subgroup data suggest a straightforward prioritization:

Highest priority: Patients with diabetes and recent ACS. The NNT of ~18 over 7 years is competitive with many standard cardiovascular interventions.
Strong consideration: Patients aged 75 and older, where the absolute benefit was nearly as large as in the diabetes subgroup.
Reasonable consideration: Patients with baseline LDL-C above 95 mg/dL who have not reached goal on statin monotherapy.
Lower absolute benefit (but no harm): Non-diabetic patients under 75 with baseline LDL already below 95 mg/dL. The benefit exists but is small in absolute terms.

This hierarchy is consistent with the 2018 AHA/ACC cholesterol guidelines, which recommend ezetimibe as first-line add-on for patients who have not reached their LDL-C threshold on maximally tolerated statin therapy, with diabetes and high baseline risk as key factors favoring early addition.

Frequently asked questions

›

References

Cannon CP, Blazing MA, Giugliano RP, et al. Ezetimibe added to statin therapy after acute coronary syndromes. N Engl J Med. 2015;372(25):2387-2397. PubMed
Giugliano RP, Cannon CP, Blazing MA, et al. Benefit of adding ezetimibe to statin therapy on cardiovascular outcomes and safety in patients with vs without diabetes mellitus. Circulation. 2018;137(15):1571-1582. PubMed
Bach RG, Cannon CP, Giugliano RP, et al. Effect of simvastatin-ezetimibe compared with simvastatin monotherapy after acute coronary syndrome among patients 75 years or older. JAMA Cardiol. 2019;4(9):846-854. PubMed
Grundy SM, Stone NJ, Bailey AL, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol. Circulation. 2019;139(25):e1082-e1143. PubMed
Sabatine MS, Giugliano RP, Keech AC, et al. Evolocumab and clinical outcomes in patients with cardiovascular disease. N Engl J Med. 2017;376(18):1713-1722. PubMed
Ezetimibe prescribing information. FDA Label