Design

IMPELLA‑PL was a national, multicenter, retrospective registry launched under the auspices of the Association of Cardiovascular Interventions of the Polish Cardiac Society and coordinated by the First Department of Medical University of Warsaw. The study did not require the consent of the Bioethical Committee of the Medical University of Warsaw due to its retrospective character. The rationale and design of the IMPELLA‑PL registry were published previously.¹¹ The HR‑PCI cohort consisted of hemodynamically stable patients with severe CAD referred for elective or urgent Impella‑assisted HR‑PCI by the local heart team. Data regarding HR‑PCI patient selection, procedure preparation, PCI trajectory, and postprocedural care were collected retrospectively in the form of dedicated, password‑protected, web‑based electronic case reports.

The HR‑PCI patients were retrospectively divided into 2 groups: those treated according to the standardized management algorithm and those who did not fulfil the algorithm criteria (Figure 1). The standardized management was defined as fulfilment of all periprocedural steps according to the ROAD TIP algorithm, including 1) optimal patient selection: at least 3 high‑risk indications and no contraindications; 2) optimal preparation: preprocedural vascular access imaging and heart team discussion; 3) optimal PCI: ultrasound‑guided puncture or contralateral angiography via an additional safety access, along with successful revascularization of all preplanned lesions; and 4) optimal postprocedural care: patient monitoring in the cardiac intensive care unit according to the standardized algorithm (pump and hemodynamic monitoring once per hour, laboratory tests and echocardiography once daily) (Table 1). Nonstandardized management was defined as failure to fulfil at least 1 of the above‑listed criteria.

The study end points included 1) in‑hospital mortality, 2) in‑hospital major adverse cardiovascular events (MACEs; need for cardiosurgical intervention, exacerbation of heart failure [HF], acute myocardial infarction [MI], acute kidney injury [AKI], inflammatory complications, severe bleeding complications as per the operator’s judgement), 3) device‑related complications (access site bleeding, limb ischemia, need for an endovascular or surgical intervention, hemolysis, aortic injury), and 4) 12‑month outcomes, including mortality and MACE after discharge (rehospitalization for HF, acute MI, need for urgent repeated revascularization, stroke, LV assist device [LVAD] implantation, heart transplant). The prespecified end point definitions have been published previously.¹¹

Statistical analysis

Statistical analysis was performed using the IBM SPSS Statistics package, version 24.0 (IBM Corp., Armonk, New York, United States). Categorical variables were reported as numbers and percentages, and were compared using the χ² test or Fisher exact test. Continuous variables were expressed as mean (SD) or median (interquartile range), and were compared using the t test or Mann–Whitney test, depending on the distribution, as verified with the Shapiro–Wilk test. For continuous variables referring to patient characteristics with more than 2 categories, the P value was calculated separately for each category, without adjustment for multiple comparisons. The Kaplan–Meier analysis and the log rank test were used to compare 12‑month mortality. The mortality predictors with a P value below 0.1 in the univariable analysis were included in the final multivariable model in both groups. The results of univariable and multivariable regression analyses are reported as odds ratios and 95% CIs. All analyses were conducted by an independent statistician in a blinded manner. Statistical tests were 2‑sided, with a significance level of 0.05.

Patient and procedural characteristics

Clinical and procedural characteristics of both groups are presented in Table 2. There were no differences regarding baseline characteristics. The mean (SD) patient age was 72.1 (10.1) years and 69.9 (9.6) years, respectively, in the standardized and nonstandardized management groups, and a majority of patients in both groups were men (86.8% vs 87.6%, respectively). The patients managed according to the standardized algorithm group underwent HR‑PCI more often in the setting of acute coronary syndrome (ACS) than chronic coronary syndrome (P <⁠0.001), and had a higher median EuroScore than the nonstandardized management group (P = 0.02). Over 75% of patients in both groups had concomitant hypertension and dyslipidemia, nearly 50% had diabetes mellitus, and approximately 40% had chronic kidney disease (CKD). A history of stroke or transient ischemic attack was more frequent in the standardized than the nonstandardized management group (P = 0.02). About 50% of patients in both groups had a history of MI, about 30% previously underwent PCI, and about 10% previously underwent coronary artery bypass grafting (CABG). Laboratory parameters and echocardiographic characteristics were comparable in both groups, with the median LV ejection fraction (LVEF) of 29% and 26%, respectively, in the standardized and nonstandardized management groups.

Regarding angiographic characteristics, the patients managed according to the standardized algorithm had more vessels with significant stenosis, defined as more than 50% stenosis in the LM or more than 70% stenosis in a major coronary vessel (>2.5 mm) on coronary angiography, or 30% to 70% stenosis with fractional flow reserve lower than or equal to 0.8 (P = 0.03). Also, they had more severe calcifications (P <⁠0.001) and presented more often with multivessel disease, including LM disease (P = 0.003). The standardized management group more often underwent LM PCI (P = 0.005), with the procedure guided by intravascular imaging (P <⁠0.001). The proportion of bifurcation PCIs was also higher in the standardized management group (P <⁠0.001), with more frequent use of the 2‑stent technique (P = 0.004). The procedural success rate (revascularization of all preplanned lesions) was lower in the standardized than in the nonstandardized management group (P = 0.003).

All patients were treated with Impella CP, and all received the support either before or during PCI. In the standardized management group, the device was implanted more often before PCI (P = 0.002). Femoral access was less common (P <⁠0.001), whereas subclavian access (P <⁠0.001) and ultrasound‑guided puncture were more common (P <⁠0.001) in these patients, as compared with the nonstandardized management group.

The need for additional cardiopulmonary support, including catecholamines, levosimendan, mechanical ventilation, extracorporeal membrane oxygenation (ECMO), and intra‑aortic balloon pump (IABP), as well as the need for renal replacement therapy were comparable between the groups. The patients managed according to the standardized algorithm group were hospitalized for a longer time than those who did not fulfil the algorithm criteria (P = 0.003). There were no significant differences between the groups with respect to pharmacotherapy at discharge.

Outcomes

In‑hospital and 12‑month outcomes of both groups are presented in Table 3. The in‑hospital mortality rate was comparable between the groups (10.5% vs 7.3%, respectively, in the standardized and nonstandardized management groups). The proportion of patients with in‑hospital MACEs was higher in the standardized than the nonstandardized management group (48.7% vs 27.1%; P <⁠0.001), which was mostly driven by more inflammatory complications (27.6% vs 7.9%; P <⁠0.001). The proportion of device‑related complications, including access site bleeding, limb ischemia, need for endovascular or surgical intervention, hemolysis, and aortic injury, was similar in both groups.

Mortality after discharge was nominally lower in the standardized than the nonstandardized management group (6.6% vs 11.3%) but the difference did not reach statistical significance. The overall 12‑month mortality in both groups was also comparable (hazard ratio [HR], 0.84; 95% CI, 0.42–1.69; P = 0.24; Figure 2A). In a subgroup analysis according to reasons for not fulfilling the criteria for standardized management, there were no differences in 12‑month mortality among the patients who did not receive standardized treatment due to suboptimal patient selection (n = 93) and suboptimal PCI trajectory (n = 80). The subgroups not fulfilling the criteria of optimal patient preparation (n = 2) and postprocedural care (n = 2) were not included in the subanalysis due to a small number of patients.

In the standardized management group, 12 patients (15.8%) experienced 12 MACEs during the 12‑month follow‑up, including 8 rehospitalizations for HF, 2 repeated revascularizations, 1 stroke, and 1 heart transplant. In the nonstandardized management group, 20 patients (11.3%) experienced 32 MACEs, including 17 rehospitalizations for HF, 3 MIs, 6 repeated revascularizations, 3 strokes, 2 heart transplants, and 1 LVAD implantation.

Since the adoption of standardized algorithms is associated with increasing procedure volume and team experience,¹⁵ we also compared the 12‑month mortality outcomes in low- and high‑volume centers participating in the IMPELLA‑PL registry. The high‑volume centers were defined as centers that performed 10 or more Impella‑assisted HR‑PCI procedures over the study period. The ROAD TIP algorithm was used in 23 out of 61 procedures performed in the low‑volume centers (38%) and 53 out of 192 procedures in the high‑volume centers (28%) (P = 0.13). In the low‑volume centers, the 12‑month mortality rate was lower in the standardized than in the nonstandardized management group (HR, 0.39; 95% CI, 0.13–0.96; P = 0.047; Figure 2B), whereas in the high‑volume centers, it was comparable between both groups (HR, 1.21; 95% CI, 0.51–2.86; P = 0.65; Figure 2C). There were no differences in mortality, regardless of the reason for the lack of the nonstandardized management (Supplementary material, Figure S1).

The results of univariable analysis identifying independent predictors of 12‑month mortality in the standardized and nonstandardized management groups are presented in Supplementary material, Tables S1 and S2, respectively, and the results of multivariable analysis are shown in Table 4. In the standardized management group, age, prior CABG, LVEF, number of vessels with significant stenosis, and Impella removal in the catheterization laboratory were associated with 12‑month mortality (P ≤0.08 for all). However, in multivariable analysis, none of the above parameters turned out to be an independent predictor of 12‑month mortality, with borderline significance for the number of vessels with significant stenosis (P = 0.05). In the nonstandardized management group, dyslipidemia, prior MI, atrial fibrillation, CKD, right ventricular dysfunction, in‑stent restenosis, contrast volume, radiation dose, Impella removal in the catheterization laboratory, and the use of catecholamines were associated with 12‑month mortality (P ≤0.09 for all) in univariable analysis. Of those, prior MI, atrial fibrillation, and CKD were identified as independent predictors of 12‑month mortality (P ≤0.03 for all) in multivariable analysis.

Main findings

We evaluated the association between the implementation of a standardized periprocedural management algorithm (ROAD TIP) and outcomes of patients undergoing HR‑PCI who were enrolled in the IMPELLA‑PL registry. The main finding of our study is that the patients treated with Impella according to the standardized algorithm had a higher‑risk profile at baseline, underwent more complex PCI procedures, and more often experienced in‑hospital complications, as compared with the patients not fulfilling the standardized algorithm criteria, who were generally at a lower risk and underwent less complex procedures. The 12‑month outcomes were comparable between the groups.

Perspective

Over the past decade, the use of pVADs to provide hemodynamic support during HR‑PCI has rapidly expanded, despite a lack of sufficient evidence from large‑scale randomized trials. According to the EAPCI, Impella is the only pVAD that may be considered in highly selected patients undergoing HR‑PCI in the case of acceptable femoral access, whereas IABP and veno‑arterial ECMO (VA‑ECMO) should not be used in this setting.¹³ Data supporting the efficacy and safety of Impella in HR‑PCI come from the PROTECT studies and large‑scale registries, with hitherto no head‑to‑head comparison against other pVADs. The prospective, multicenter PROTECT I trial²⁰ (n = 20) demonstrated that Impella 2.5 can be successfully used during HR‑PCI, providing robust hemodynamic support. In the per protocol analysis of the randomized PROTECT II trial,⁵ the patients treated with Impella 2.5 (n = 226) experienced fewer MACEs than the individuals treated with IABP (n = 226; P = 0.048). In the prospective, single‑arm PROTECT III trial,²¹ HR‑PCI patients supported with Impella 2.5 or CP (n = 504) demonstrated more complete revascularization, lower rates of bleeding complications, and improved 90‑day clinical outcomes, as compared with the historic cohort of PROTECT II patients with severely depressed LVEF (mean, ca. 23%). In other studies, the use of Impella was associated with over 75% lower risk of post‑PCI AKI than expected from current risk models, and lower risk of AKI, as compared with the use of VA‑ECMO, suggesting that Impella might be a protective strategy against AKI during HR‑PCI.^22,23

In line with clinical trial findings, retrospective analyses of data from multicenter national registries, including the IMP‑IT, German registry, IMPELLA‑PL, and cVAD registry,^8,9,12 confirmed acceptable clinical outcomes of Impella in HR‑PCI. However, some authors have questioned these benefits. For example, in a retrospective study of 1680 patients, Khalid et al²⁴ found that HR‑PCI could be successfully performed without a pVAD in 98.2% of patients, with lower mortality and fewer complications at 30‑day follow‑up than reported for Impella‑assisted HR‑PCI (mortality, 1.6%; AKI, 4.6%; stroke, 0.2%; major bleeding, 1.1%). The ongoing randomized PROTECT IV trial, which aims to compare Impella 2.5 or CP vs no support or bailout IABP in patients with complex CAD and reduced LVEF undergoing HR‑PCI (n = 1252) over a 3‑year follow‑up, will help identify patients that may truly benefit from the use of Impella during a percutaneous revascularization procedure. Meanwhile, identification of patients who truly benefit from pVAD support remains problematic.

Previous studies consistently showed that a standardized multidisciplinary team approach improves outcomes in patients treated with Impella in the setting of cardiogenic shock.^25-27 However, similar data are not yet available for HR‑PCI. Our results indicate that the standardized management algorithm allowed for 1) selection of patients at a higher baseline risk, including those with ACS, more comorbidities, and more complex CAD (multivessel disease, LM artery involvement, and severe calcifications); 2) performing more complex PCI procedures with a lower procedural success rate, including a higher proportion of LM PCIs, bifurcation PCIs, and use of calcium modification techniques; 3) more frequent Impella implantation before than during PCI; 4) more careful vascular access management, with a higher proportion of ultrasound‑guided punctures and more frequent choice of alternative access (subclavian access, surgical access). Despite a higher baseline risk, greater PCI complexity with subsequent lower procedural success, and more frequent in‑hospital complications, the patients treated according to the standardized algorithm had comparable 12‑month outcomes to those not fulfilling the algorithm criteria, who were at a lower baseline risk and underwent less complex procedures. The higher proportion of inflammatory complications in the standardized algorithm group, defined as confirmed infection treated with antimicrobial agents, might be due to more frequent use of surgical access, and might explain the longer hospital stay.

Importantly, we found an association between the use of a standardized management algorithm and 12‑month mortality depending on the volume of performed HR‑PCI procedures. In the low‑volume centers (<⁠10 Impella‑assisted interventions), 12‑month mortality was lower in the standardized than in the nonstandardized management group, whereas in the high‑volume centers, it was comparable between the groups. This observation confirms the presence of a learning curve in the setting of Impella‑supported HR‑PCI, as previously shown in a prespecified subanalysis of the PROTECT II study.²⁸ Hence, standardized management algorithms seem especially beneficial in low‑volume centers, with a balanced, integrated, team approach being the path toward optimal patient care. However, given the observational and retrospective design, our findings are hypothesis‑generating and should be confirmed in prospective studies.

Limitations

Our study has several limitations. First, the ROAD TIP algorithm has been proposed in the year 2023, whereas the patient inclusion period for the IMPELLA‑PL registry ended in 2022. Hence, we could not prospectively validate the algorithm in the IMPELLA‑PL registry, but only retrospectively evaluate the association between the use of the standardized periprocedural management algorithm and HR‑PCI outcomes. A comparison of the characteristics and outcomes of patients before and after the introduction of the algorithm would be more reliable and remains to be performed. Second, the algorithm was proposed based on the recent expert consensus statements^13,14,17,18 and our experiences. However, a machine learning approach might improve identification of the management steps crucial for patient qualification for Impella‑assisted HR‑PCI and detection of independent predictors of mortality.²⁹ Third, since this was a registry‑based study, it was limited by the availability of complete medical history data. In addition, the outcome events were adjudicated by local committees in every participating hospital, but there was no independent event adjudication committee. Thus, both baseline characteristics and data regarding end points might be prone to under- or overreporting bias, despite prespecified definitions. Fourth, in the current study, Impella was implanted and removed during the same procedure in over 90% of patients, which reflects the overall clinical stability of this cohort. Hence, the results cannot be extrapolated to less stable HR‑PCI patients. Finally, since different end point definitions were used in different national Impella registries, comparisons of our results with data from other registries should be interpreted with caution.

Conclusions

The patients treated according to the standardized algorithm were characterized by a higher‑risk profile and greater procedural complexity than the individuals not fulfilling the algorithm criteria, who were at a lower baseline risk and underwent less complex procedures; however, the 12‑month outcomes were comparable between the groups. A comparison of patient outcomes before and after implementation of the dedicated management algorithm is required to confirm our findings.