TL;DR. DKCRUSH VIII randomized 556 patients with complex coronary bifurcation lesions undergoing DK crush to IVUS-guided vs. angiography-guided PCI at 24 Chinese centers. At 1 year, target vessel failure was 6.1% vs. 14.7% (HR 0.40; 95% CI 0.23–0.71; p=0.0016) — a 58% relative reduction, the largest imaging-guidance effect size reported in any complex PCI trial. Every TVF component moved in the same direction. For two-stent bifurcation strategies, this trial effectively ends the debate: IVUS guidance is not optional.
Background
The DK crush technique, developed by Shao-Liang Chen's group in Nanjing, has accumulated the strongest randomized evidence base of any two-stent bifurcation strategy. DKCRUSH II through VII established its superiority over provisional stenting and culotte for true bifurcation lesions, including unprotected left main distals. But DK crush is also the most technically demanding bifurcation technique: two rewiring steps, two kissing inflations, and a proximal optimization sequence, each with a defined geometric target that angiography assesses poorly.
The rationale for imaging guidance in this setting is mechanistic, not speculative. The failure points of the crush technique — incomplete stent expansion at the ostial side branch, malapposition at the carina, and inadequate proximal overlap zone expansion — are precisely the findings that IVUS detects and angiography systematically misses. Registry data and subgroup analyses (including the ULTIMATE trial's bifurcation subset) had long suggested a benefit, but no dedicated randomized trial had tested imaging guidance specifically in complex bifurcation PCI with a uniform two-stent technique.
DKCRUSH VIII was presented as a Late-Breaking Clinical Trial at ACC.26 and published simultaneously in JACC.
Study design
Investigator-initiated, multicenter, randomized trial at 24 centers in China. 556 patients with complex bifurcation lesions (defined by the DEFINITION criteria) undergoing planned DK crush were randomized 1:1 to IVUS-guided (n=278) or angiography-guided (n=278) PCI. The IVUS arm followed a pre-specified optimization protocol with defined minimum stent area targets for the main vessel, side branch ostium, and polygon of confluence; the angiography arm used standard visual and quantitative angiographic criteria.
The primary endpoint was target vessel failure (TVF) at 1 year: a composite of cardiac death, target vessel myocardial infarction, and clinically driven target vessel revascularization.
Key results
TVF at 1 year occurred in 6.1% of the IVUS-guided arm vs. 14.7% of the angiography-guided arm (HR 0.40; 95% CI 0.23–0.71; p=0.0016). The effect was consistent across every component of the composite: target vessel MI 4.3% vs. 9.4%, spontaneous MI 1.8% vs. 6.1%, and clinically driven TVR 2.9% vs. 7.6%. The direction and magnitude of each component argue against the result being driven by a single soft endpoint.
A number needed to treat of roughly 12 to prevent one TVF event at 1 year, in a procedure already being performed, with a technology available in essentially every contemporary cath lab, is a remarkable return for an imaging catheter.
Critical appraisal
The trial answers a focused question with a clean design: same lesion subset, same stenting technique, single variable changed (imaging guidance), adequately powered for a clinical composite. The pre-specified IVUS optimization protocol — with explicit MSA targets per segment — is a methodological strength that many earlier imaging trials lacked; it converts "IVUS use" from a passive observation into an active, protocolized intervention.
The effect size deserves scrutiny precisely because it is so large. A 58% relative reduction exceeds what ULTIMATE, IVUS-XPL, or the RENOVATE-COMPLEX-PCI trial reported in broader complex PCI populations (typically 30–50% reductions in TVF or TLF). Two factors plausibly explain this: first, DK crush creates more metal, more overlap, and more geometric failure modes than single-stent PCI, so the detectable-and-correctable defect rate is higher; second, the angiography-guided arm's 14.7% TVF confirms how poorly angiography performs in this anatomy. The larger the blind spot, the larger the benefit of removing it.
Generalizability caveats. All 24 centers were in China, operated by teams with deep DK crush experience — the technique's homeland. Whether operators with lower DK crush volumes achieve the same angiography-arm baseline (or the same IVUS-arm optimization quality) is untested. The trial also cannot tell us whether the benefit extends to other two-stent techniques (culotte, T/TAP) or to OCT guidance, although mechanistic reasoning suggests it should.
One-year follow-up is short for a metal-heavy strategy where late events (restenosis within crushed layers, late stent thrombosis) accrue over years. Longer follow-up is planned and will matter.
Take-home for the cathlab
If you perform DK crush — or any two-stent bifurcation strategy — without intravascular imaging, DKCRUSH VIII removes your justification. The imaging step is not an academic refinement; it halves hard clinical events. Practical protocol: IVUS after side branch stenting to confirm ostial expansion before the crush; IVUS again after final kissing and POT, with MSA targets for the polygon of confluence, both ostia, and the proximal main vessel. Correct what you find. The 8 additional minutes of procedure time are the cheapest 58% risk reduction in interventional cardiology.
Combined with ECLIPSE (imaging in calcified lesions) and the established complex-PCI imaging trials, the evidence arc is complete: in complex anatomy, angiography-only PCI is now the intervention that requires justification.
| Domain | Assessment | Notes |
|---|---|---|
| Design | Clean | Single variable; uniform technique; DEFINITION criteria |
| Optimization protocol | Pre-specified | Explicit MSA targets per segment |
| Effect size | Large | HR 0.40; all components concordant |
| Follow-up | Short | 1 year; late events in two-stent PCI accrue beyond |
| Generalizability | Limited | China-only; high-volume DK crush operators |
| Blinding | Open-label | Endpoint adjudication independent; TVR softest component |