Back to StatUp
Journal Club

CREST-2: A Protocol-Level Deep Dive

Understanding what this trial actually tested, why it matters, and how to interpret the results

Section 1: Why Did We Need This Trial?

Before we can understand CREST-2, we need to understand the landscape that created the question.

1995: ACAS (Asymptomatic Carotid Atherosclerosis Study)

CEA reduced 5-year stroke risk from 11% to 5.1% vs medical therapy. But "medical therapy" meant aspirin and risk factor advice. No statins. No aggressive BP targets.

2004-2010: ACST (Asymptomatic Carotid Surgery Trial)

Confirmed CEA benefit in a larger European cohort. Again, medical therapy was rudimentary by today's standards.

2010: CREST (Original)

Compared stenting to CEA head-to-head in symptomatic and asymptomatic patients. Found them roughly equivalent, but notably did NOT include a medical-therapy-alone arm.

2014-2020s: Medical Therapy Advances

High-intensity statins. PCSK9 inhibitors. Aggressive BP targets (<130 systolic). Better antiplatelet regimens. Smoking cessation programs. The question emerged: Do we still need to fix the artery?

The Central Question

If modern intensive medical management (IMM) has improved so dramatically, does adding CEA or stenting still provide meaningful benefit?

CREST-2 was designed to answer this directly.

Why this matters: Previous trials compared intervention to outdated medical therapy. CREST-2 is the first to pit revascularization against modern intensive medical management.

Section 2: The Protocol - Two Parallel Trials

This is the first critical point many people miss: CREST-2 was not one trial. It was two separate, parallel randomized trials.

Stenting Trial

1,245

patients randomized to
CAS + IMM vs IMM alone

Surgery Trial

1,240

patients randomized to
CEA + IMM vs IMM alone

Critical distinction: These were parallel trials with separate randomization. Patients were NOT randomized between stenting and surgery. You cannot directly compare CAS outcomes to CEA outcomes from this data.

Key Design Elements (from the Protocol Paper)

  • Enrollment period: December 2014 to April 2022
  • Sites: 130 sites in the US and Canada
  • Inclusion: Asymptomatic carotid stenosis 70% or higher by ultrasound (or 60%+ by angiography)
  • Follow-up: Median 4.4 years (stenting) and 4.1 years (surgery)

The Primary Endpoint

A composite of:

  1. Stroke or death from day 0 to day 44 (the "periprocedural" period)
  2. PLUS ipsilateral ischemic stroke from day 45 through 4 years

This composite captures both the upfront risk of intervention AND the long-term stroke prevention benefit.

Why Two Separate Trials?

Some patients are anatomically suitable for stenting but not surgery (or vice versa). Randomizing between procedures would have excluded these patients. The parallel design allowed each procedure to be tested against modern medical therapy in its appropriate population.

Section 3: Who Got In? Operator and Patient Selection

This section is crucial for understanding generalizability. CREST-2 had extensive selection criteria for both operators and patients.

Operator Credentialing

CEA Surgeons

50

consecutive cases required

Combined stroke/death rate <3%

CAS Interventionists

50

total cases required (25 consecutive within 5 years as primary operator)

Combined stroke/death rate <3%

The Companion Registry (C2R)

Many stent operator applicants didn't have sufficient experience to apply for credentialing. The trial created a Companion Registry where operators could receive conditional approval to perform CAS procedures, building their case volume until they qualified for full credentialing.

This ensured procedures were performed by skilled operators at well-resourced sites, while allowing less experienced interventionalists a pathway into the trial.

The implication: Both operator groups had to demonstrate <3% complication rates. But the CAS registry pathway suggests the bar for stenting experience was harder to meet - many applicants needed additional supervised cases before qualifying.

Anatomic Exclusions for CAS

Table 4 from the Protocol: CAS-Specific Anatomic Exclusions

Patients with these features could NOT be randomized to the stenting trial:

  • Severe circumferential calcification of target lesion
  • Presence of visible thrombus at lesion site
  • Target lesion known to be a chronic total occlusion
  • Severe tortuosity of the target vessel (more than two 90-degree angles)
  • Target lesion longer than 15mm
  • "String sign" (near-occlusion with threadlike lumen)
  • Type III aortic arch (highly angulated takeoff)
  • Bovine arch configuration with target on left side
  • Severe tortuosity of the common carotid artery
  • Severe innominate or common carotid stenosis (if stent requires crossing)
What this means: The stenting results apply to anatomically "ideal" patients treated by top-tier operators. These are efficacy results in a best-case scenario, not effectiveness results in typical practice.

All Patients Had to Meet These Criteria

  • Age 35-86 years
  • Asymptomatic (no TIA/stroke in past 180 days)
  • Carotid stenosis 70%+ by ultrasound
  • No prior ipsilateral CEA or CAS
  • No disabling stroke
  • No major surgery or invasive procedure within 30 days

Section 4: Intensive Medical Management (IMM)

Both arms in both trials received identical intensive medical management. This wasn't "aspirin and a handshake" - it was aggressive, protocol-driven therapy.

The IMM Protocol

Component Target
Antiplatelet therapy Aspirin (325mg) + clopidogrel (or alternative)
LDL cholesterol <70 mg/dL (high-intensity statin +/- ezetimibe)
Blood pressure Systolic <140 mmHg (later revised to <130)
Diabetes management HbA1c <7% if diabetic
Smoking cessation Counseling and pharmacotherapy offered
Weight management Diet and exercise counseling

This is the Control Arm

When we say "stenting beat medical therapy," we mean stenting beat this aggressive regimen. The comparison group wasn't left untreated - they received gold-standard preventive care.

But patients weren't perfectly managed. Despite the aggressive protocol:
  • Average LDL was above the <70 mg/dL target in all groups
  • Current smokers: ~21% in most groups (18.8% in stenting arm)

The IMM results reflect what happens with attempted intensive management, not perfect adherence. Your clinic might actually achieve similar results.

The Real Question for Clinical Practice

Are your patients with asymptomatic carotid stenosis on aggressive medical therapy with regular follow-up and target monitoring?

If yes, these IMM results may be achievable. If they're on "a statin and aspirin" with no follow-up, you're not comparing apples to apples.

Section 5: The Results - Headline Numbers

Here's what they found. We'll parse these carefully.

Stenting Trial Results (n=1,245)

CAS + IMM

2.8%

95% CI: 1.5 to 4.3

IMM Alone

6.0%

95% CI: 3.8 to 8.3

Absolute risk difference: 3.2 percentage points (95% CI: 0.6 to 5.9)

Relative risk: 2.13 (95% CI: 1.15 to 4.39)

P = 0.02 - Statistically significant

Surgery Trial Results (n=1,240)

CEA + IMM

3.7%

95% CI: 2.1 to 5.5

IMM Alone

5.3%

95% CI: 3.3 to 7.4

Absolute risk difference: 1.6 percentage points (95% CI: -1.1 to 4.3)

Relative risk: 1.43 (95% CI: 0.78 to 2.72)

P = 0.24 - Not statistically significant

Quick Stats Refresher: Confidence Intervals

A 95% confidence interval gives us a range of plausible values for the true effect. If the CI for a relative risk crosses 1.0, we can't confidently say the treatment is different from no effect.

  • Stenting CI (1.15 to 4.39): Excludes 1.0 → statistically significant
  • CEA CI (0.78 to 2.72): Crosses 1.0 → not statistically significant
  • CEA absolute difference CI (-1.1 to 4.3): Crosses 0 → could be no difference, or up to 4.3% benefit

Want to solidify your understanding of effect sizes and confidence intervals? → Module 3

The Kaplan-Meier curves cross. Within each trial, the intervention and medical therapy curves cross during follow-up. When both trials are plotted together, the curves cross multiple times. This suggests the treatment effects aren't constant over time and adds uncertainty to the conclusions.

Subgroup Analysis: Who Benefits?

Stenting showed significant benefit in:

  • Nondiabetics
  • PSV ≥ 342 cm/s (higher grade stenosis)
  • Nonsmokers
  • CHA₂DS₂-VASc score 0-3

All other subgroups showed no significant difference.

For CEA: No subgroup showed significant benefit when analyzed individually.

Caution: Subgroup analyses are exploratory and prone to false positives. These findings generate hypotheses but don't prove differential effects.

The headline: Stenting + IMM beat IMM alone (p=0.02). Surgery + IMM did not reach statistical significance vs IMM alone (p=0.24).

Expected vs Actual Event Rates

The protocol assumed these rates for power calculations:

Arm Expected Actual
Revascularization 3.6%
(2.0% periprocedural + 1.6% late)
CAS: 2.8%
CEA: 3.7%
IMM Alone 8.4%
(for 85% power)
6.0% (stenting trial)
5.3% (surgery trial)

Medical therapy outperformed expectations. The IMM arm did 2-3% better than anticipated. This likely contributed to the difficulty in reaching significance, particularly for CEA's smaller effect size.

But hold on: Before drawing conclusions, we need to look at where these events came from. The periprocedural period tells a different story.

Section 6: Parsing the Outcome - Periprocedural vs Late Events

The composite endpoint combines periprocedural events (days 0-44) with late ipsilateral strokes. Let's separate these.

Stenting Trial: Events by Period

Period CAS + IMM IMM Alone
Days 0-44
(periprocedural)
7 strokes, 1 death
= 8 events
0 strokes, 0 deaths
= 0 events
Beyond 44 days
(late)
7 ipsilateral strokes 28 ipsilateral strokes

What This Tells Us

Stenting caused 8 early events that wouldn't have happened with medical therapy alone. But it prevented 21 late strokes (28 - 7). The net benefit was 13 fewer events (21 prevented - 8 caused).

Surgery Trial: Events by Period

Period CEA + IMM IMM Alone
Days 0-44
(periprocedural)
9 strokes 3 strokes
Beyond 44 days
(late)
10 ipsilateral strokes 23 ipsilateral strokes

Why Didn't CEA Reach Significance?

CEA had 6 more early events than medical therapy (9 vs 3). It prevented 13 late strokes (23 - 10). The net benefit was 7 fewer events (13 - 6).

With a smaller net benefit, the trial didn't have sufficient power to detect this effect size.

The math: Stenting's larger net benefit (13 vs 7 fewer events) and larger absolute risk reduction (3.2% vs 1.6%) led to significance. CEA showed a trend in the same direction but didn't cross the threshold.

Section 7: How Robust Are These Results?

When a trial has few events, significance can hinge on just a handful of outcomes. This is "statistical fragility."

The Tipping Point Analysis

Researchers calculate how many events would need to change to flip the result from significant to non-significant (or vice versa).

For the stenting trial:

  • Total events in composite endpoint: ~50-60
  • Margin of significance: relatively narrow
  • Tipping point: approximately 3-4 events
What this means: If 3-4 strokes had been classified differently, or occurred in the other arm, the p-value might have crossed 0.05 in the other direction. The statistical significance is real, but fragile.

Why Fragility Matters

This doesn't mean the finding is wrong. It means we should interpret it with appropriate humility. A single site with different outcomes, slightly different follow-up duration, or minor adjudication differences could affect the conclusion.

For clinical practice: The direction of effect is likely real (stenting prevents strokes). But the precision of "p=0.02" overstates our certainty. This is why replication matters.

Section 8: Efficacy vs Effectiveness

This distinction is essential for applying trial results to your patients.

Efficacy

"Can it work?"

Under ideal conditions: best operators, selected patients, rigorous protocols

Effectiveness

"Does it work?"

In real-world conditions: average operators, typical patients, usual care

CREST-2 Is an Efficacy Trial

  • Operators: Only 50% of CAS applicants were credentialed. These were the best.
  • Patients: Extensive anatomic exclusions removed difficult cases.
  • Medical therapy: Protocol-driven with target achievement monitoring.
  • Follow-up: Close surveillance with mandated visits.
The real-world gap: Your hospital's stenting outcomes may differ from CREST-2. When comparing this trial's data to your own practice, be honest with yourself about your institution's overall management quality and procedural outcomes.
How to use this: CREST-2 supports offering stenting to appropriate patients at experienced centers. It does NOT support universal stenting for all asymptomatic carotid disease.

For CEA, the data suggest a trend toward fewer events, but the effect wasn't statistically clear in this trial. That doesn't erase the existing evidence from ACAS and ACST - it adds nuance. CEA likely still benefits some patients, but the magnitude of benefit over modern medical therapy may be smaller than we thought.

Questions to Ask Before Applying These Results

  • Does your interventionalist meet CREST-2 credentialing standards?
  • Does your patient have any of the Table 4 anatomic exclusions?
  • Is your patient actually on (and adherent to) intensive medical management?
  • What are your institution's actual periprocedural complication rates?

Test Your Understanding

Apply what you've learned to interpret CREST-2 correctly.

Key Takeaways

  • 1
    Two parallel trials, not one

    CREST-2 randomized separately to CAS+IMM vs IMM and CEA+IMM vs IMM. These were not head-to-head comparisons of the procedures.

  • 2
    Highly selected operators and patients

    All operators required <3% complication rates. CAS operators needed a companion registry to build experience. 10 specific anatomic exclusions for stenting. This was an efficacy trial in ideal conditions.

  • 3
    Stenting beat modern medical therapy

    2.8% vs 6.0% composite endpoint (p=0.02). The first trial to show intervention benefit over contemporary medical management.

  • 4
    CEA showed a trend but didn't reach significance

    3.7% vs 5.3% (p=0.24). Similar direction, smaller effect size, possibly underpowered.

  • 5
    Periprocedural events matter

    Stenting had 8 early events vs 0 in medical therapy. The late stroke prevention (21 fewer) had to overcome this upfront cost.

  • 6
    Statistical fragility is real

    The significance hinges on a small number of events. The direction is likely correct, but certainty is limited.

The Bottom Line

CREST-2 provides the first evidence that revascularization (specifically stenting) adds benefit to modern intensive medical therapy for asymptomatic carotid stenosis - but only in carefully selected patients treated by expert operators. Medical therapy alone remains reasonable for many patients.

Before You Go: A Quick Poll

You see an asymptomatic 65-year-old female with 70% carotid stenosis. Based on what you've learned, what would you offer her?