Design: systematic review and random-effects meta-analysis of randomised trials, pre-specified geographic interaction
Population: 14,033 patients in 17 trials of drug-eluting stent PCI, 10,155 in East Asian and 3,878 in non-Asian trials
Primary outcome: cardiac death, with the geographic interaction test as the primary analysis
Effect: pooled RR 0.71 (95% CI 0.53–0.94); East Asian 0.56 (0.46–0.69) vs non-Asian 1.23 (0.85–1.76), interaction P < 0.001

Intravascular ultrasound guidance for percutaneous coronary intervention has an evidence base that looks convincing in aggregate and inconsistent in detail: East Asian trials have been positive, several recent European trials neutral. This meta-analysis made that comparison the pre-specified primary hypothesis rather than a post hoc observation, pooling seventeen randomised trials of IVUS-guided against angiography-guided drug-eluting stent implantation in 14,033 patients — ten East Asian trials with 10,155 patients and seven non-Asian trials with 3,878.

Pooled across everything, IVUS guidance reduced cardiac death, with a risk ratio of 0.71 (95% CI 0.53–0.94). But the geographic interaction was strong: 0.56 (0.46–0.69) in East Asian trials against 1.23 (0.85–1.76) in non-Asian trials, interaction P < 0.001. The same divergence appeared for target-vessel myocardial infarction, target-vessel revascularisation and major adverse cardiovascular events. One outcome behaved consistently everywhere — definite stent thrombosis, risk ratio 0.39 (0.19–0.79), interaction P = 0.433. All-cause death did not differ, 0.89 (0.76–1.04).

The authors' explanation is that what operators do with the IVUS images differs, not that the imaging itself works differently in different populations. That is the useful reading for an Indian catheter laboratory, where the question is not whether to believe the pooled estimate but whether your own optimisation criteria match those used in the trials that showed benefit. Buying the probe is not the intervention; acting on post-deployment minimal stent area and edge dissection is.

Agree a written stent optimisation protocol — minimal stent area, expansion, edge dissection — rather than leaving it to the operator
Audit how often an IVUS run actually changes what is done, not how often it is performed
Expect the stent thrombosis benefit to hold in your practice; expect the mortality benefit only if you optimise
Record the pre-specified criteria that triggered post-dilatation, so the audit is possible
Do not use the neutral non-Asian trials as an argument to abandon intravascular imaging

Why it matters

It shifts the question from whether to use intravascular imaging to whether your laboratory does anything different when it does.

The statistics, in plain English

An interaction P value below 0.001 means the difference between the two sets of trials is very unlikely to be chance variation — this is a genuine effect modification, not the usual weak subgroup claim, and it was pre-specified rather than found by inspection. Note also that the non-Asian estimate of 1.23 has an interval from 0.85 to 1.76 that includes 1.0, so those trials showed no benefit rather than proven harm. The contrast with stent thrombosis, where the interaction P of 0.433 indicates no meaningful difference by region, is what points to technique rather than biology.