4S, HPS, JUPITER, IMPROVE-IT: The Big Statin Trials Compared, A Cross-Trial Synthesis

By
Published Updated Last reviewed
Medical lab testing image for 4S, HPS, JUPITER, IMPROVE-IT: The Big Statin Trials Compared, A Cross-Trial Synthesis

Why compare these four trials together?

Statins are among the most prescribed drug classes worldwide, yet the four trials that most shaped prescribing norms were designed in different eras, enrolled vastly different patients, and asked subtly different questions. The Scandinavian Simvastatin Survival Study (4S) asked whether lowering LDL-C in high-risk patients with established disease would reduce death. The Heart Protection Study (HPS) broadened the question to a wider secondary-prevention population, including patients many clinicians had been reluctant to treat. JUPITER moved into primary prevention among people with elevated high-sensitivity CRP but LDL-C levels that, at the time, did not meet treatment thresholds. IMPROVE-IT then asked whether adding a non-statin lipid-lowering agent, ezetimibe, on top of a statin would produce incremental benefit consistent with the LDL hypothesis.

Reading each trial in isolation tells you about one drug in one population. Reading them together, matched column by column, tells you something more useful: what drives outcome differences, what the trials agree on mechanistically, and where the evidence genuinely runs thin.

At a glance

| Trial | N | Drug / Dose | Population | LDL-C Entry Criterion | Follow-up | Primary Endpoint | Primary Result | ARR | NNT | Dropout Rate | Key Adverse Events | |---|---|---|---|---|---|---|---|---|---|---|---| | 4S (1994) | 4,444 | Simvastatin 20-40 mg vs placebo | Secondary prevention; angina or prior MI; TC 5.5-8.0 mmol/L | No explicit LDL floor; TC-based entry | Median 5.4 years | All-cause mortality | 11.5% vs 8.2% (placebo vs statin) | 3.3% absolute mortality reduction | 30 (mortality) | ~10% | Myopathy rare; no excess hepatic events | | HPS (2002) | 20,536 | Simvastatin 40 mg vs placebo | Secondary prevention + high-risk primary; age 40-80; TC ≥3.5 mmol/L | No LDL-C floor (enrolled patients with LDL <3 mmol/L) | Mean 5.0 years | First major vascular event (MI, stroke, revascularization) | 24.7% vs 19.8% | 4.9% ARR in major vascular events | 21 | ~15% | Myopathy 0.01% annual incidence | | JUPITER (2008) | 17,802 | Rosuvastatin 20 mg vs placebo | Primary prevention; LDL-C <3.4 mmol/L; hsCRP ≥2 mg/L | LDL-C <130 mg/dL (upper bound, not lower) | Median 1.9 years (stopped early) | Composite MACE (MI, stroke, UA hospitalization, revascularization, CV death) | 1.36% vs 0.77% per year | ~1.2% ARR over trial period | 95 (composite MACE) | ~15% | Small HbA1c increase; new-onset diabetes signal | | IMPROVE-IT (2015) | 18,144 | Simvastatin 40 mg + ezetimibe 10 mg vs simvastatin 40 mg alone | Secondary prevention post-ACS; LDL-C 50-125 mg/dL on statin or 50-100 mg/dL untreated | Lower bound 50 mg/dL; upper bound 125 mg/dL | Median 6.0 years | Composite CV death, MI, UA hospitalization, coronary revascularization, stroke | 34.7% vs 32.7% | 2.0% ARR | 50 | ~42% (very long trial) | Comparable adverse events between arms |

Population differences: who was actually in each trial?

The most consequential difference across these trials is baseline cardiovascular risk, and it runs almost linearly from 4S through IMPROVE-IT and HPS at the high-risk end, to JUPITER at the low-risk end.

4S enrolled patients who had already survived an acute coronary event or were living with angina, and whose total cholesterol was elevated (5.5-8.0 mmol/L). This is an older, higher-risk European cohort from the early 1990s. Background therapy was modest by modern standards; almost no patients were on aspirin or ACE inhibitors at baseline. That matters because event rates were high, which inflates absolute risk reduction and makes the NNT look more favorable than it would in a contemporary equivalent population.

HPS deliberately widened the net. Its 20,536 participants included patients with prior MI or stroke, but also patients with peripheral arterial disease and diabetic patients without established CVD. HPS enrolled patients regardless of baseline LDL-C, including thousands with LDL below 3.0 mmol/L. The HPS investigators found proportional risk reductions were consistent across all LDL-C subgroups, a finding that directly influenced guidelines that had previously tied treatment eligibility to an LDL-C threshold.

JUPITER is the outlier in this group. Its participants were largely healthy (no prior CVD, low LDL-C), selected specifically for elevated hsCRP, a biomarker of systemic inflammation. Median LDL-C at baseline was 108 mg/dL, well within what most 2008 guidelines called acceptable. The trial was stopped after a median of just 1.9 years by the independent data safety monitoring board when it crossed a pre-specified efficacy boundary. That early termination matters for interpreting the size of absolute benefit. Patients with a 10-year Framingham risk score averaging roughly 8-11% are not the same population as 4S patients; extrapolating the same relative reduction to both groups produces very different absolute gains.

IMPROVE-IT enrolled patients stabilized after an acute coronary syndrome, with LDL-C between 50 and 125 mg/dL. By enrolling patients already on or eligible for statin therapy, it became the first large trial to isolate the incremental value of a non-statin agent added to background statin treatment. The IMPROVE-IT population was older (mean age 64), with high rates of prior MI, PCI, and diabetes, making it closer to 4S and HPS in absolute risk than to JUPITER.

One generalizability caution applies to all four trials: women were underrepresented. In 4S, only 19% of participants were women. JUPITER had better representation (38%) but was still not balanced. Subgroup analyses in each trial generally showed directionally consistent benefits in women, but none was individually powered to confirm mortality benefit in female participants.

Methodology differences: where the trials diverge by design

Primary endpoint definitions varied enough to complicate direct comparison. 4S used all-cause mortality as the primary endpoint, an unambiguous outcome that requires no adjudication judgment. HPS used a broader composite of major vascular events, which includes revascularization procedures that are more subject to clinical practice variation and local decision-making. JUPITER's composite added unstable angina hospitalization, a softer endpoint than death or MI. IMPROVE-IT used a seven-component composite. As endpoint composites expand, the apparent relative risk reduction can appear larger while the absolute contribution of softer components dilutes clinical meaning.

Blinding and comparator were consistent in approach (double-blind, placebo-controlled in 4S, HPS, and JUPITER), but IMPROVE-IT compared combination therapy against statin monotherapy, introducing a different research question entirely. It is not a placebo-controlled trial in the traditional sense, which is worth keeping in mind when comparing relative risk reductions across the four.

Follow-up duration ranged from 1.9 years (JUPITER, terminated early) to 6.0 years (IMPROVE-IT). Longer follow-up generally allows more events to accumulate, improving power to detect mortality differences. The ACC/AHA cholesterol guidelines have acknowledged that early stopping in JUPITER may have inflated the apparent relative benefit by truncating the curve before event rates in the placebo arm could fully climb.

Statistical approach to subgroups: HPS was uniquely rigorous here, pre-specifying a large matrix of subgroup analyses and powering the trial to detect effects within individual subgroups (e.g., diabetic patients, women, patients with LDL <3 mmol/L). 4S and JUPITER were not powered for most subgroup claims, and several widely cited subgroup findings from JUPITER warrant caution.

Results, matched: head-to-head on shared outcomes

Absolute risk reduction in MACE

This is where the four trials tell their most coherent story. 4S produced an ARR of approximately 8% for major coronary events over 5.4 years in a very high-risk secondary-prevention cohort. HPS produced an ARR of 4.9% for major vascular events. IMPROVE-IT produced an ARR of 2.0% for its composite endpoint, despite a longer follow-up of 6 years, partly because background statin therapy had already lowered baseline risk substantially before randomization. JUPITER produced roughly 1.2% ARR over a median of less than 2 years, in a genuinely lower-risk primary-prevention group.

The pattern is consistent: absolute benefit scales with baseline risk, not with drug potency alone.

Number needed to treat

NNT for MACE over the trial period: 4S approximately 12 for major coronary events, HPS approximately 21 for major vascular events, IMPROVE-IT approximately 50 for its composite, JUPITER approximately 95. These numbers are not interchangeable because denominators, endpoints, and follow-up periods differ. But the rank order is what matters clinically: a secondary-prevention patient with recent ACS gains far more from lipid-lowering than a healthy person with elevated CRP. This is not a reason to withhold statins from primary prevention in appropriate patients; even NNT of 50-100 compares favorably to many widely used preventive interventions. The 2019 ACC/AHA primary prevention guidelines frame this exactly this way.

All-cause mortality benefit

4S demonstrated a clear, statistically significant all-cause mortality benefit: 11.5% placebo versus 8.2% statin, a 30% relative reduction. HPS showed a mortality ARR of 1.8% (14.7% vs 12.9%), which was statistically significant given the large sample size. JUPITER, stopped early and underpowered for mortality, showed a trend toward reduced mortality (hazard ratio 0.80 to 95% CI 0.62-1.03) that did not cross the significance threshold. IMPROVE-IT showed no statistically significant all-cause mortality difference between arms (HR 0.99). So mortality benefit is clearly established in the two large, longer secondary-prevention trials, directionally consistent but not confirmed in JUPITER, and absent in IMPROVE-IT, where all patients were already on statin therapy at baseline.

LDL-C reduction magnitude

4S reduced LDL-C by approximately 35% from a high baseline (mean baseline LDL approximately 4.9 mmol/L). HPS reduced LDL-C by approximately 1.0 mmol/L (about 29%) from a mean baseline of 3.4 mmol/L. Rosuvastatin 20 mg in JUPITER reduced LDL-C by 50% from a baseline of approximately 108 mg/dL, achieving a mean on-treatment LDL of 55 mg/dL. IMPROVE-IT reduced LDL-C from approximately 93 mg/dL to 53 mg/dL in the combination arm versus 70 mg/dL in the statin-only arm, an additional 24% relative reduction from the combination. The IMPROVE-IT result was the first large-trial confirmation that non-statin LDL lowering via a different mechanism (cholesterol absorption inhibition rather than hepatic synthesis reduction) produces proportional clinical benefit, supporting the broader LDL hypothesis rather than a statin-specific mechanism.

What the trials together do and do not establish

What they collectively establish: LDL-C lowering reduces cardiovascular events. The relationship is proportional to baseline risk. The benefit is not statin-specific: IMPROVE-IT showed that ezetimibe-mediated LDL reduction produces clinical benefit consistent with the overall LDL hypothesis. The absolute magnitude of benefit is modest in primary prevention by the numbers in these trials, but consistent. HPS settled the question of whether patients with LDL-C below 3.0 mmol/L benefit, and the answer was yes, provided total cardiovascular risk is elevated.

What they do not establish: Whether there is a lower LDL threshold below which further reduction provides no benefit (IMPROVE-IT achieved LDL of 53 mg/dL, still above the floors now being explored in PCSK9 inhibitor trials). Whether the primary-prevention mortality benefit seen directionally in JUPITER is real and durable over longer follow-up. Whether the new-onset diabetes signal in JUPITER (HR approximately 1.25) modifies the net benefit calculation meaningfully for individual patients, particularly those with metabolic risk factors. The trials also do not establish benefit equivalency across all ethnic groups; HPS included South Asian patients, but 4S was almost entirely Northern European.

The four trials also disagree, in effect if not explicitly, on the question of how much the hsCRP-guided selection strategy in JUPITER adds beyond LDL-C-based risk stratification. Subsequent analyses have suggested that the JUPITER population could have been identified as statin-eligible using traditional risk calculators alone, raising questions about whether hsCRP testing changed prescribing in practice in ways that matter.

Outstanding questions for the next trial

  1. Very low LDL targets and mortality: FOURIER and ODYSSEY OUTCOMES with PCSK9 inhibitors have pushed achieved LDL-C to 20-30 mg/dL. Do the proportional benefits established in these four trials extrapolate linearly to that range, or does the curve flatten? Long-term follow-up data from these newer trials are still maturing.

  2. Primary prevention with formal risk-score entry: Can a trial replicate JUPITER's relative risk reduction in a primary-prevention population selected purely by 10-year ASCVD risk score, without the hsCRP criterion, to determine whether the inflammatory biomarker adds genuine stratification value?

  3. Sex-stratified mortality analysis: None of these four trials was powered for mortality benefit specifically in women. A dedicated trial or prospective individual patient data meta-analysis across these and subsequent trials is needed before the mortality benefit can be stated confidently for female patients.

  4. Duration of benefit beyond 6 years: The longest of these trials ran 6 years (IMPROVE-IT). Whether the NNT in primary prevention improves substantially with 10-15 years of treatment, matching lifetime risk reduction models projected by the ACC/AHA pooled cohort equations, remains untested in randomized trial data.

  5. Statin-associated diabetes: net clinical benefit in at-risk subgroups: The JUPITER diabetes signal requires dedicated evaluation in patients with pre-diabetes or metabolic syndrome, comparing cardiovascular event reduction against incident diabetes incidence within the same trial, with a long enough follow-up to capture diabetes-attributable cardiovascular events.

Frequently asked questions

›
›
›
›

References

  1. Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet. 1994;344(8934):1383-1389. https://pubmed.ncbi.nlm.nih.gov/7968073/

  2. Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of cholesterol lowering with simvastatin in 20,536 high-risk individuals: a randomised placebo-controlled trial. Lancet. 2002;360(9326):7-22. https://pubmed.ncbi.nlm.nih.gov/12114036/

  3. Ridker PM, Danielson E, Fonseca FA, et al. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein (JUPITER). N Engl J Med. 2008;359(21):2195-2207. https://pubmed.ncbi.nlm.nih.gov/18997196/

  4. Cannon CP, Blazing MA, Giugliano RP, et al. Ezetimibe added to statin therapy after acute coronary syndromes (IMPROVE-IT). N Engl J Med. 2015;372(25):2387-2397. https://pubmed.ncbi.nlm.nih.gov/26039521/

  5. 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. https://www.ahajournals.org/doi/10.1161/CIR.0000000000001063

  6. Arnett DK, Blumenthal RS, Albert MA, et al. 2019 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease. J Am Coll Cardiol. 2019;74(10):e177-e232. https://pubmed.ncbi.nlm.nih.gov/30894318/