The CHIP-BCIS3 study, published in the New England Journal of Medicine in March 2026, is to date the only randomized trial directly comparing planned left ventricular unloading with Impella CP against a strategy without planned mechanical circulatory support in high-risk PCI. Given the clinical and health-economic importance of this question, AGIK provides a nuanced statistical and methodological assessment of the results.
The aim of this statement is neither a blanket rejection nor endorsement of Impella therapy, but rather an evaluation of the study’s evidential value, with particular consideration of the primary analytical method, the secondary mortality analyses, and key limitations of the study design.
Study Design and Patient Population
In CHIP-BCIS3, 300 patients with severely impaired left ventricular function and complex multivessel coronary artery disease were randomized in a 1:1 ratio. Eligible participants had an LVEF of no more than 35%, or up to 45% in the presence of severe mitral regurgitation, as well as a BCIS Jeopardy Score of at least 8. Median follow-up was 22 months. The high complexity of the cohort was reflected by a median SYNTAX score of 38 and by high proportions of planned left main interventions, calcified plaque modification, and retrograde CTO procedures.
For clinical interpretation, it is crucial that this was not a purely stable, elective high-risk PCI population. A substantial proportion of participants had an urgent PCI indication; in addition, approximately two thirds of patients were randomized in the context of an acute coronary syndrome, predominantly NSTEMI. This means that the cohort differs substantially from strictly elective CHIP programs, such as those performed in specialized centers with structured preparation, vascular access planning, and selective patient selection.
This heterogeneity is relevant. NSTEMI patients have different pathophysiological conditions than those in a stable elective setting, including activated coagulation, vulnerable plaques, and an increased tendency toward microembolization. As a result, both the potential benefits of hemodynamic unloading and the risks associated with large-bore vascular access, bleeding, and thrombosis may need to be weighted differently. A predefined subgroup analysis comparing stable elective patients with ACS patients is lacking. The available data, however, suggest that the Win Ratio was less favorable in the ACS cohort than in stable participants, although this cannot be assessed confirmatorily due to the sample size.
Methodological Limitations
A major point of criticism concerns the incomplete characterization of relevant comorbidities. The protocol allowed the inclusion of participants with severe mitral regurgitation and an LVEF of up to 45%; however, the published baseline table does not provide systematic information on the prevalence of relevant valvular heart disease. It therefore remains unclear to what extent valvular pathologies contributed to left ventricular dysfunction. A differentiation between ischemic, valvular, and mixed causes of LV dysfunction would have been methodologically important.
The exclusion criteria also appear limited for a cohort of only 300 participants. Among other conditions, dialysis-dependent end-stage renal disease, terminal malignancy, active infections, and severe peripheral arterial disease were not explicitly excluded. Such comorbidities can substantially influence mortality, bleeding risk, and periprocedural complications independently of the revascularization strategy. A sensitivity analysis excluding such high-risk profiles was not reported.
Furthermore, the revascularization strategy must be considered as a potential procedural confounder. Planned staged PCI was performed more frequently in the standard-care arm than in the Impella arm. This suggests that a more comprehensive revascularization in a single session was more often pursued in the Impella arm. In patients with severely impaired LV function and urgent ACS/NSTEMI presentation, such a strategy may be associated with longer ischemic time, higher contrast volume, more extensive lesion manipulation, and increased microembolization. The more frequently observed periprocedural myocardial injury in the Impella arm may therefore be partly explained by differences in the timing and extent of revascularization.
Primary Endpoint and Secondary Analyses
The primary analytical method used was the Win Ratio, a hierarchical pairwise comparison procedure. Each patient in the Impella arm is compared with each patient in the control group. The assessment follows a clinically prioritized hierarchy: death from any cause, disabling stroke, spontaneous myocardial infarction, cardiovascular hospitalization, and periprocedural myocardial injury.
The Win Ratio was 0.85, with a 95% confidence interval of 0.63 to 1.15 and a p value of 0.30. Thus, no statistically significant benefit of the Impella strategy was demonstrated. Numerically, more pairwise comparisons favored standard care than Impella therapy. Formally, neither definite inferiority nor superiority can be inferred from this result. At the same time, the confidence interval does not exclude a clinically relevant disadvantage of the Impella strategy.
The study was powered for the hierarchical composite endpoint, not for mortality as an isolated endpoint. The Kaplan-Meier analyses of all-cause mortality and cardiovascular mortality are therefore secondary, exploratory analyses. All-cause mortality was 32.6% in the Impella arm versus 23.4% with standard care, while cardiovascular mortality was 26.7% versus 14.5%. These signals are clinically relevant, but they do not constitute confirmatory evidence of a mortality harm, as the study lacked sufficient power and no formal adjustment for multiple testing was performed. A direct comparison with studies primarily designed around time-to-event endpoints or mortality is not methodologically appropriate.
The ongoing PROTECT IV study remains of great importance for further evaluation. It is investigating a substantially larger cohort with a classic time-to-event design and clearer exclusion criteria, including those related to relevant valvular disease, dialysis dependency, and limited life expectancy.
AGIK Conclusions
- CHIP-BCIS3 provides no statistically significant evidence of benefit for planned left ventricular unloading with Impella CP in high-risk PCI. The numerical distribution of the primary endpoint tended to favor standard care, although no definitive superiority or inferiority can be derived from this finding.
- Based on the currently available data, routine use of Impella CP in elective high-risk PCI cannot generally be recommended. In selected individual cases, its use may still be reasonable and justifiable in experienced high-risk PCI centers after careful clinical, anatomical, and procedural selection.
- The secondary mortality data should be regarded as a clinically relevant but exploratory warning signal. Since CHIP-BCIS3 was not powered for mortality and no formal adjustment for multiple testing was performed, these data do not allow a confirmatory conclusion regarding mortality harm.
- When communicating the study results, a strict distinction should be made between the primary, non-significant Win Ratio result and the exploratory secondary time-to-event analyses. Direct comparisons with studies primarily designed around time-to-event endpoints, such as DANGER-SHOCK and REVIVED-BCIS2, are not methodologically appropriate.
- The transferability of the CHIP-BCIS3 results to strictly elective, structurally prepared high-risk PCI programs is limited, as the study cohort included a high proportion of urgent ACS/NSTEMI patients.
- The ongoing PROTECT IV study remains decisive for the final evaluation of Impella therapy in elective high-risk PCI, as its larger sample size, classic time-to-event design, and clearer exclusion criteria make it methodologically better suited to answer the unresolved question of prognostic benefit.
More on this Topic
