Heart Protection Study Subgroup Analyses: Who Responded Most and Least

At a glance

Field	Detail
Trial	Heart Protection Study (HPS)
N	20,536
Intervention	Simvastatin 40 mg daily
Comparator	Matching placebo
Duration	Mean 5.0 years of follow-up
Primary endpoint	First major vascular event (MVE): coronary death, non-fatal MI, stroke, or revascularization
Key result	24% proportional reduction in first MVE (simvastatin 19.8% vs. placebo 25.2%; p <0.0001)
Registration	ISRCTN48489393

Why the Subgroup Story Matters More Than the Headline

The Heart Protection Study was designed from the start to answer a question earlier statin trials could not: does the benefit extend to patients clinicians were still hesitant to treat? By 2002, statins had solid evidence in middle-aged men with elevated cholesterol and established coronary disease. HPS enrolled a deliberately broad population, including women, patients over 70, diabetics without coronary history, cerebrovascular disease patients, and people whose baseline LDL-cholesterol was already below the contemporary treatment threshold of 3.0 mmol/L (roughly 116 mg/dL).

The trial's pre-specified subgroup plan was unusually ambitious. Twelve primary subgroups were defined before unblinding, covering age bands, sex, qualifying diagnosis, baseline lipid levels, and baseline medication use. Post-hoc analyses later extended the picture to BMI categories, renal function, and additional lipid fractions. The collective message reshaped guideline thinking: risk status, not cholesterol level, should drive the prescribing decision.

Pre-Specified Subgroup Results

The HealthRX.com Subgroup Concordance Framework below organizes each pre-specified subgroup by the direction of effect, point estimate, and whether the 99% confidence interval excluded unity. HPS used 99% CIs for subgroups (rather than 95%) to guard against the multiple-comparison problem, a methodological choice worth noting because it makes any surviving signal more credible.

Age

The rate ratios were virtually identical across all three bands. Patients over 70 had the highest absolute risk reduction because their baseline event rate was highest. This was the first large RCT to demonstrate statin benefit in this older group with adequate statistical power. The finding directly influenced the 2004 NCEP ATP III update and, later, the 2013 ACC/AHA cholesterol guidelines that extended moderate-intensity statin recommendations to patients aged 65-75 with elevated cardiovascular risk.

Sex

Women made up about 25% of the cohort. The point estimate was slightly attenuated compared with men, but the interaction test was non-significant (p = 0.56), meaning the apparent difference was consistent with chance. Prior trials (4S, CARE, LIPID) had enrolled far fewer women, so HPS provided the largest female-specific dataset for statin efficacy at the time.

Baseline LDL-Cholesterol

This subgroup generated the most practice-changing signal.

Patients who entered the study with LDL already below the treatment threshold in 2002 guidelines derived the same proportional benefit as those with high LDL. The interaction p-value was 0.90. This result was central to the intellectual shift from "treat the number" to "treat the risk," a principle that now underpins both ACC/AHA and NICE statin guidance.

Qualifying Vascular Condition

HPS enrolled patients across multiple vascular territories, not just coronary artery disease.

The diabetes-only subgroup was the standout. These 3,982 patients had no documented coronary, cerebrovascular, or peripheral artery disease at entry. Their 25% relative risk reduction (rate ratio 0.75) provided the first large-scale evidence for primary prevention with a statin in type 2 diabetes. The 2004 ADA Standards of Care subsequently adopted statin recommendations for diabetic patients aged over 40, citing HPS as a primary evidence source.

Post-Hoc and Extended Analyses

BMI

A post-hoc cut at BMI 30 kg/m² found no meaningful interaction. Obese participants (BMI >30; approximately 22% of the cohort) had a rate ratio of 0.75, non-obese participants 0.76. Body composition did not attenuate the proportional benefit of simvastatin in this trial.

Baseline Triglycerides and HDL

Additional lipid-fraction analyses showed:

• Patients with baseline triglycerides above 2.0 mmol/L: rate ratio 0.77
• Patients with baseline HDL below 0.9 mmol/L: rate ratio 0.77
• Patients with baseline HDL above 1.1 mmol/L: rate ratio 0.76

No lipid subgroup showed differential benefit. The consistency of these findings was later reinforced by the CTT (Cholesterol Treatment Trialists') Collaboration meta-analysis of individual participant data from 14 statin trials, which confirmed that proportional MACE reduction per 1.0 mmol/L LDL decrease did not vary by baseline lipid profile.

Renal Function

A dedicated HPS sub-study published in 2003 examined 1,329 participants with creatinine above 130 µmol/L or estimated GFR below 60. The rate ratio for major vascular events in this renal subgroup was 0.72 (95% CI 0.56-0.93). This was among the earliest RCT evidence supporting statin use in moderate chronic kidney disease. The KDIGO 2013 lipid guidelines cited HPS when recommending statins for CKD patients over 50.

Race and Ethnicity

HPS enrolled participants almost entirely from the United Kingdom, and the published subgroup analyses did not stratify by race or ethnicity. This is a genuine limitation. Extrapolation to non-white populations relies on external data, primarily the CTT meta-analysis and regional trials like MEGA (Japan) and JUPITER (multiethnic cohort). Clinicians should note this gap when applying HPS results to ethnically diverse patient panels.

Subgroups That Appeared to Benefit Least

No pre-specified subgroup showed a statistically significant absence of benefit. Two subgroups had wider confidence intervals that approached the null:

1. Women with diabetes but no vascular disease (n ~ 1,100): The point estimate still favored simvastatin, but the confidence interval was wide enough to include unity. This was a power limitation, not a biological signal. The subsequent CARDS trial (atorvastatin in type 2 diabetes) confirmed similar benefit in both sexes.

2. Patients on baseline beta-blocker therapy: The rate ratio was 0.79, with 99% CI narrowly crossing 1.0 in one early reporting cut. The final analysis resolved this; there was no significant interaction with baseline beta-blocker use.

The HPS investigators explicitly tested for heterogeneity across all 12 primary subgroups. The global interaction test yielded p = 0.7, meaning the data were consistent with a single, uniform treatment effect regardless of subgroup membership.

Methodological Notes for Subgroup Interpretation

Three design choices make HPS subgroup analyses more trustworthy than usual:

Pre-specification with wide 99% CIs. By requiring stronger evidence to declare heterogeneity, the HPS team avoided the garden-of-forking-paths problem that plagues post-hoc subgroup claims in smaller trials.

Run-in phase. All 20,536 randomized patients first completed a 4-6 week active run-in on simvastatin, followed by a 4-6 week placebo run-in. This filtered out non-adherent patients and those with early side effects. Subgroup results therefore apply to a population willing and able to tolerate simvastatin, which inflates efficacy estimates relative to an intent-to-treat-from-first-pill design.

Compliance correction. In the primary analysis, about 18% of placebo patients were taking a non-study statin by year five. The investigators applied a compliance-adjusted analysis suggesting the true risk reduction might be closer to 33% rather than the observed 24%. Subgroup-level compliance adjustment was not published, so differential drop-in across subgroups remains a possible, though unstudied, confounder.

Limitations the Authors Acknowledged

• The run-in design means results do not apply to statin-intolerant patients.
• The 40 mg simvastatin dose produced an average LDL reduction of about 1.0 mmol/L, modest by current standards. Modern high-intensity regimens (simvastatin FDA label at 80 mg, or atorvastatin 40-80 mg) achieve larger reductions. Whether subgroup concordance holds at deeper LDL lowering was later explored by CTT.
• UK-centric enrollment limits generalizability by ethnicity.
• Subgroup analyses, even pre-specified ones, are hypothesis-generating. The diabetes-only result drove CARDS and other confirmatory trials rather than standing as standalone proof.

What This Means for Prescribing

The HPS subgroup data support a simple clinical heuristic: if a patient qualifies for a statin based on cardiovascular risk, age, sex, baseline LDL, BMI, and qualifying vascular territory should not be reasons to withhold therapy. The proportional benefit is remarkably flat across categories. The absolute benefit, however, tracks baseline risk. Patients over 70, diabetics, and those with multi-territory vascular disease gain the most in absolute event reduction per year of treatment.

Current ACC/AHA guidelines and NICE CG181 both cite HPS subgroup data when defining their risk-based treatment thresholds. The trial remains, over two decades later, the single broadest evidence base for statin prescribing in high-risk patients.

Frequently asked questions

›Did patients over 75 benefit from simvastatin in HPS?

HPS enrolled patients up to age 80 at randomization. The over-70 subgroup (n = 5,538) showed a rate ratio of 0.76, identical to younger groups. Patients aged 75-80 were included within this band, though a separate cut at 75 was not published in the primary paper. The CTT 2019 meta-analysis later confirmed statin benefit extends to at least age 75.

›Was the statin benefit different for women vs. men in HPS?

Women (n = 5,082) had a rate ratio of 0.80 vs. 0.75 for men. The interaction test was non-significant (p = 0.56), meaning the difference was likely due to chance. HPS provided the largest female sample in any statin trial at the time and helped establish that sex alone is not a reason to withhold therapy.

›Did HPS show statin benefit at low LDL levels?

Yes. Patients entering with LDL below 3.0 mmol/L (about 116 mg/dL) had a rate ratio of 0.76, indistinguishable from higher-LDL groups. This was the key finding that shifted guidelines from LDL-target-based to risk-based prescribing.

›How did diabetics without heart disease respond in HPS?

The 3,982 patients with diabetes but no prior vascular disease had a 25% relative reduction in major vascular events (rate ratio 0.75, 99% CI 0.61-0.91). This primary-prevention signal in diabetes was later confirmed by the CARDS trial.

›Were HPS subgroup analyses pre-specified or post-hoc?

Twelve primary subgroups were pre-specified before unblinding, including age, sex, qualifying diagnosis, and baseline LDL. Post-hoc analyses later explored BMI, renal function, and additional lipid fractions. Pre-specified subgroups used 99% confidence intervals to reduce false-positive risk.

›Did obesity reduce the benefit of simvastatin in HPS?

No. Post-hoc analysis showed obese participants (BMI >30) had a rate ratio of 0.75, nearly identical to non-obese participants at 0.76. BMI did not modify the treatment effect.

›What about patients with kidney disease in HPS?

A sub-study of 1,329 participants with impaired renal function found a rate ratio of 0.72. This was among the earliest RCT evidence for statin use in moderate CKD and was cited in the 2013 KDIGO lipid guidelines.

›Did HPS include racially diverse patients?

No. HPS was conducted in the UK with a predominantly white cohort. Published subgroup analyses did not stratify by race or ethnicity. Generalizability to non-white populations relies on external evidence from trials like MEGA and JUPITER.

›How does the run-in phase affect subgroup interpretation?

The active and placebo run-in removed non-adherent patients and those with early intolerance before randomization. This means subgroup results apply only to patients who can tolerate simvastatin. Real-world absolute benefits may be modestly lower due to non-adherence.

›Are HPS subgroup findings still clinically relevant today?

Yes. Both ACC/AHA and NICE guidelines cite HPS subgroup data when defining risk-based statin thresholds. The trial remains the broadest single-trial evidence base for statin prescribing across diverse high-risk populations, even as subsequent meta-analyses have extended the findings.

References

• 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. PubMed
• Cholesterol Treatment Trialists' (CTT) Collaborators. Efficacy and safety of cholesterol-lowering treatment: prospective meta-analysis of data from 90,056 participants in 14 randomised trials of statins. Lancet. 2005;366(9493):1267-1278. PubMed
• Stone NJ, Robinson JG, Lichtenstein AH, et al. 2013 ACC/AHA guideline on the treatment of blood cholesterol. Circulation. 2014;129(25 Suppl 2):S1-S45. PubMed
• Kidney Disease: Improving Global Outcomes (KDIGO) Lipid Work Group. KDIGO clinical practice guideline for lipid management in CKD. Kidney Int Suppl. 2013;3(3):259-305. PubMed
• Grundy SM, Stone NJ, Bailey AL, et al. 2018 AHA/ACC/AACVPR guideline on the management of blood cholesterol. Circulation. 2019;139(25):e1082-e1143. PubMed
• Simvastatin prescribing information. U.S. Food and Drug Administration. FDA Label