Radial versus femoral access for coronary angiography and percutaneous coronary interventions in critically ill patients Authors: Alexis Devaux, Alexandre Gamet, Maxime Desgrouas, Piotr Szychowiak, Marc Goralski, Jean-Baptiste Brillant, Thierry Boulain, François Barbier Published: 2026 ABSTRACT AbstractBackgroundEvidence is lacking to appraise whether radial access should be preferred to femoral access for coronary angiography (CAG) in critically ill patients.MethodsPatients admitted to the medical intensive care unit (ICU) of a university hospital over a 5-year period and requiring CAG immediately before or during the ICU stay were retrospectively included. The average treatment effect of the arterial route used for CAG was assessed through a win-ratio methodology with propensity score (PS)-based inverse probability of treatment weighting (IPTW). The primary endpoint was a hierarchical composite outcome of (i) all-cause death, (ii) resuscitated cardiac arrest, (iii) stroke, (iv) myocardial infarction, (v) acute limb ischemia and/or need for vascular surgery on CAG access site and (vi) Bleeding Academic Research Consortium (BARC) type 3 major bleeding, all assessed at day 28. The relationship between access site and the likelihood of all-cause death at day 28 was investigated with a Cox proportional hazard model stratified on the occurrence of cardiac arrest before CAG and adjusted through PS-based IPTW.ResultsAmong the 235 included patients (116 with femoral access and 119 with radial access), 145 (61.7%) were admitted following cardiac arrest and 116 (49.4%) presented with cardiogenic shock at the time of CAG. The primary composite endpoint occurred in 63.4% of patients with femoral access and 47.7% of patients with radial access in IPT-weighted populations (absolute risk difference 15.6%, 95% confidence interval [CI] 1.4% to 29.8%, P = 0.03; win ratio for femoral versus radial access 0.62, 95% CI 0.40 to 0.98, P = 0.04). Access site complications were more frequent in patients with femoral access while major bleedings were similarly observed in both groups. Femoral access was not associated with day-28 mortality (adjusted hazard ratio 1.43, 95% CI 0.90 to 2.26, P = 0.13). Percutaneous coronary revascularisation and implantation of veno-arterial extra-corporeal membrane oxygenation, when required, were not hastened with femoral access.ConclusionIn this study, the use of a femoral rather than radial approach for CAG in critically ill patients was associated with a higher incidence of access site complications but did not correlate with the hazard of major bleeding or all-cause death at day 28.Supplementary InformationThe online version contains supplementary material available at 10.1186/s44158-026-00424-5.Keywords: Coronary angiography, Percutaneous coronary intervention, Cardiac arrest, Cardiogenic shock, Veno-arterial extra-corporeal membrane oxygenation, Intensive care unit Introduction Coronary angiography (CAG) is commonly performed in patients admitted to the medical intensive care unit (ICU). This procedure is especially indicated in those referred for resuscitated cardiac arrest or cardiogenic shock suspected to result from acute coronary syndrome, in whom early revascularisation of the culprit lesion through percutaneous coronary intervention (PCI) improve both short- and long-term outcomes [1–7]. CAG may also be needed for acute coronary syndrome occurring during the ICU stay in patients initially admitted for sepsis or other non-cardiac reasons [8]. The European Society of Cardiology and the American College of Cardiology/American Heart Association guidelines advocate the use of radial rather than femoral access in patients requiring CAG, including for unscheduled procedures [5, 6]. Indeed, several trials conducted in patients with acute coronary syndrome have shown that, compared to the femoral approach, the radial approach is associated with a lower risk of all-cause death, chiefly driven by a reduced incidence of major bleeding and access site complications, with no difference in the hazard of new myocardial infarction or stroke [9–12]. Observational studies and a post-hoc analysis of a large randomised trial also suggested a survival benefit of radial access in patients with cardiogenic shock [13, 14]. However, most studies included patients managed in coronary care units, with relatively low severity of organ failures and overall mortality. Whether a radial approach should be preferred to a femoral approach in critically ill patients has been scarcely investigated so far [15]. Femoral access could enable swifter coronary revascularisation in patients with severe hemodynamic instability—a situation at risk for failure of radial approach [16]—and ease the implantation of veno-arterial extra-corporeal membrane oxygenation (VA-ECMO) by using the CAG cannula in those with peri-procedural refractory cardiogenic shock or cardiac arrest. The primary objective of this single-centre cohort study was to investigate the relationship between the primary arterial route (radial versus femoral) and the risk of major adverse cardiovascular event (MACE) or major bleeding in critically ill patients undergoing CAG. Time to coronary revascularisation and the duration of ECMO-VA implantation when required were also compared between patients with femoral access and those with radial access. Patients and methods Study setting and patient selection This retrospective cohort study was conducted in the 25-bed medical ICU of a university hospital in France. Over the inclusion period (see below), the CAG unit of the hospital was managed by a total of 8 senior interventional cardiologists trained to both radial and femoral accesses, with a mean number of 3000 procedures per year. This team follows a primary radial approach for patients managed in the coronary care unit; however, the choice of arterial access for those with critically illness is not protocolised and left at the discretion of the interventional cardiologist and attending intensivist. Ultrasound guidance and closure device are routinely used for femoral access. Patients admitted for resuscitated cardiac arrest were managed according to current guidelines, including targeted temperature control [7, 17]. All adult patients admitted over a 5-year period (from June 1, 2018 to June 1, 2023) and who underwent CAG immediately before ICU admission (if initially managed by a pre-hospital medical emergency team) or during the ICU stay were identified through administrative coding databases and eligible for inclusion. Patients with primary humeral approach, brain death (CAG for evaluation before potential heart transplant donation), CAG performed after discharge to hospital wards or missing information regarding the outcomes of interest were not retained in the study population (Fig. S1 in the electronic supplement). Only the first CAG was analysed in patients with more than one procedure during the ICU stay. Only the first ICU stay was analysed in those with multiple admissions over the inclusion period. The study protocol was approved by the ethical committee of the French Society of Anaesthesia and Intensive Care on January 5, 2025 (IRB 00010254‐2024-105); the requirement for informed consent was waived owing to the observational design. Results are reported according to the STROBE guidelines [18]. Data collection, definitions, and study endpoints Variables exposed in the tables were extracted from (i) the patient computerised medical charts including automatedly implemented clinical monitoring data, biological data, and organ support parameters (ICCA, Philips, Amsterdam, the Netherlands) and (ii) the database of the CAG unit (CardioReport 360, MediReport, Paris, France) for detailed information regarding the vascular approach (primary access site and conversion to another site when required), angiography results, modalities of PCI when performed, and procedural time-points (including admission in the CAG room, first arterial puncture, PCI steps, and revascularisation completion). MACE were defined as new myocardial infarction, ischemic or haemorrhagic stroke, acute upper or lower limb ischemia and/or need for vascular surgery on CAG access site, resuscitated cardiac arrest, and death from any cause. Major bleedings were defined as overt or fatal non-coronary artery bypass grafting-related bleedings, that is Bleeding Academic Research Consortium (BARC) types 3 and 5. The primary study endpoint was a hierarchical composite outcome of MACE and major bleedings, namely (i) death from any cause (including BARC type 5 major bleeding), (ii) resuscitated cardiac arrest, (iii) stroke, (iv) myocardial infarction, (v) acute upper or lower limb ischemia and/or need for vascular surgery on CAG access site and (vi) BARC type 3 major bleeding, all assessed at day 28. Components of MACE and BARC type 3 or 5 bleedings at day 28 were individually investigated as secondary study endpoints. Though this event may primarily reflect patient condition, resuscitated cardiac arrest was included in the composite outcome as it could also be related to severe procedural complications (e.g. haemorrhagic shock, PCI failure, or early stent thrombosis), and positioned on the second hierarchical rank owing to its major potential impact on patient outcomes. Exploratory endpoints included (i) door-to-balloon time (from admission to the catheter room to complete revascularisation [including successful dilatation and stent implantation] of the culprit lesion) in patients requiring PCI and (ii) duration of cannulation (from procedure initiation to unclamping) in those requiring VA-ECMO. Statistical analyses Data are expressed as median (interquartile range) for continuous variables and number (percentage) for categorical variables, unless otherwise indicated. The analysis of variance test and the Kruskal–Wallis test were used to compare continuous variables with normal and non-normal distributions, respectively. The Fisher’s exact test or the χ2 test, as appropriate, were used to compare categorical variables. To estimate the average treatment effect of the arterial route used for CAG on the primary and secondary study endpoints while accounting for confounding, we first calculated a propensity score (PS) for primary femoral access using a multivariate logistic regression model handling the first-attempt arterial site as a dependant variable and patient characteristics at admission to the CAG unit (including demographics, comorbid conditions, mean reason for ICU admission, prior cardiac arrest, indication for CAG, and hemodynamic parameters) as independent variables (see Table S1 in the electronic supplement for standardised mean differences of baseline characteristics between the femoral and radial groups). For stability purpose, the PS was capped at the 2.5th and 97.5th percentiles. The inverse probability of treatment weight (IPTW) was defined as the inverse of the PS for patients with femoral access and 1/(1–PS) (i.e. the inverse of 1–PS) for those with radial access. Stabilised weights were calculated from the PS and absolute standardised differences were measured for each covariate to evaluate the success of the model in creating balanced cohorts, with values of more than 0.1 considered as defining imbalance (Fig. S2 in the electronic supplement). A win-ratio methodology was then implemented to investigate the association between the arterial access used for CAG and the occurrence of the composite hierarchical primary endpoint at day 28 in the PS-balanced, IPT-weighted populations [19, 20]. Win ratio and net treatment benefit (win difference, i.e. the proportion of wins minus the proportion of losses) were calculated with their 95% confidence interval (CI). With this approach, a win ratio of less than 1 means that femoral access has fewer wins than losses against radial access in prioritised pairwise comparisons, thereby indicating a benefit (protective effect) of radial access regarding the outcomes of interest. We also compared the proportion of patients with at least one occurrence for each type of event (unhierarchised primary endpoint, and secondary endpoints) between the femoral and radial groups through the Rao-Scott χ2 test, taking into account the individual PS-based weights, with calculation of absolute risk differences and their 95% CI [21]. Lastly, we investigated the relationship between the primary access site and the likelihood of all-cause death at day 28 through a Cox proportional hazard model stratified on the occurrence of cardiac arrest before CAG and adjusted through PS-based IPTW. The proportional hazards assumption was checked using the Grambsch–Therneau test based on Schoenfeld residuals (P = 0.39) [22]. Weighted Kaplan–Meier curves were constructed for each of the 4 subgroups obtained by PS-based IPTW (radial or femoral access with or without cardiac arrest before CAG). A sensitivity analysis was conducted after exclusion of patients with a treatment limitation decision before day 28. All analyses were conducted using the R software version 4.2.3 (http://www.R-project.org). Two-tailed P-values < 0.05 were considered statistically significant. Results Baseline features of the study population A total of 235 critically ill patients were enrolled, including 116 (49.4%) with primary femoral access and 119 (50.6%) with primary radial access (Table 1, and Table S1 and Fig. S1 in the electronic supplement). Non-cardiac comorbid conditions were common, notably lower extremity peripheral artery disease (overall, 14.0%), cerebrovascular disease (9.8%) and chronic renal failure (9.4%). Roughly two thirds of patients (61.7%) were admitted to the ICU following resuscitated cardiac arrest, mostly out-of-hospital cardiac arrest (50.6%). The remaining patients were primarily admitted for cardiogenic shock (10.2%), other acute cardiac diseases (10.2%), and non-cardiac reasons (17.9%). The median SOFA score value at ICU admission was 9 [6–11]. Table 1Study populationAll patients(n = 235)Femoral access(n = 116)Radial access(n = 119)P-valueAge, years65 (53–73)66 (57–74)63 (51–72)0.15Male sex176 (74.9)85 (73.3)91 (76.5)0.68BMI, kg per m−227.4 (24.3–31.0)27.3 (24.2–31.1)27.6 (24.8–30.9)0.74Main cardiovascular risk factors Tobacco use (current or past)138 (58.7)66 (56.9)72 (60.5)0.67 Diabetes mellitus63 (26.8)32 (27.6)31 (26.1)0.91 Dyslipidemia102 (43.4)52 (44.8)50 (42.0)0.76 Hypertension148 (63.0)78 (67.2)70 (58.8)0.23History of cardiac disease Coronary artery disease66 (28.1)41 (35.3)25 (21.0)0.02 Percutaneous coronary intervention44 (18.7)23 (19.8)21 (17.6)0.79 Coronary artery bypass grafting15 (6.4)9 (7.8)6 (5.0)0.5 Non-ischemic cardiac disease57 (24.3)40 (34.5)17 (14.3)0.001Non-cardiac comorbid conditions Lower extremity peripheral artery disease33 (14.0)20 (17.2)13 (10.9)0.23 Cerebrovascular disease23 (9.8)14 (12.1)9 (7.6)0.35 COPD35 (14.9)19 (16.4)16 (13.4)0.65 Chronic renal failure22 (9.4)12 (10.3)10 (8.4)0.77 Chronic haemodialysis6 (2.6)4 (3.4)2 (1.7)0.66 Liver disease6 (2.6)2 (1.7)4 (3.4)0.70 Immune deficiency 115 (6.4)10 (8.6)5 (4.2)0.26Reason for ICU admission0.83 Cardiac arrest145 (61.7)73 (62.9)72 (60.5) Out-of-hospital119 (50.6)60 (51.7)59 (49.6) In-hospital26 (11.1)13 (11.2)13 (10.9) Cardiogenic shock24 (10.2)10 (8.6)14 (11.8) Cardiac, others24 (10.2)13 (11.2)11 (9.2) Miscellaneous42 (17.9)20 (17.2)22 (18.5)SOFA score value at ICU admission9 (6–11)9 (7–12)8 (5–10)0.004SAPS-2 value at ICU admission64 (46–74)64 (52.5–76.5)59 (40.5–72.5)0.03Direct admission to the CAG unit 2119 (50.6)61 (52.6)58 (48.7)0.65Cardiac arrest before CAG Overall148 (63.0)75 (64.7)73 (61.3)0.69 Within 3 h prior to CAG118 (50.2)58 (50.0)60 (50.4)1 Initial shockable rhythm95/148 (64.2)43/75 (57.3)52/73 (71.2)0.09 Witnessed140/148 (94.6)71/75 (94.7)69/73 (94.5)1 By-stander CPR84/148 (56.8)45/75 (60.0)39/73 (53.4)0.51 No-flow, min 32 (0–6)2 (0–5)1 (0–7)0.69 Low-flow, min25 (12–39)30 (13–41)20 (10–35)0.11Indication for CAG0.58 STEMI140 (59.6)72 (62.1)68 (57.1) NSTEMI82 (34.9)38 (32.8)44 (37.0) Others13 (5.6)6 (5.1)7 (5.9)KILLIP class before CAG I79 (33.6)26 (22.4)53 (44.5)0.006 II11 (4.7)5 (4.3)6 (5.0) III16 (6.8)10 (8.6)6 (5.0) IV116 (49.4)69 (59.5)47 (39.5) NA (no acute myocardial infarction)13 (5.5)6 (5.2)7 (5.9)Antiplatelet therapy before CAG Dual83 (35.3)44 (37.9)39 (32.8)0.25 Single74 (31.5)34 (29.3)40 (33.6)Unfractionated heparin before CAG117 (49.8)55 (47.4)62 (52.1)0.51Intravenous thrombolysis before CAG15 (6.4)8 (6.9)7 (5.9)0.79Admission to the CAG unit Heart rate, per min88 (73–109.50)87 (70–111)88 (75–106)0.69 Systolic blood pressure, mmHg105 (89–125)99 (83–120)110 (93–128)0.03 Diastolic blood pressure, mmHg67 (55–79)65 (54–76)70 (57–83)0.03 Mean arterial pressure, mmHg79 (67.50–93)77 (64.75–91)82 (72–95)0.03 Inotropic agents88 (37.4)52 (44.8)36 (30.3)0.03 Vasopressors124 (52.8)72 (62.1)52 (43.7)0.007 Invasive mechanical ventilation187 (79.6)97 (83.6)90 (75.6)0.17Need for arterial access conversion 45 (2.1)2 (1.7)3 (2.5)1Results of CAG Normal46 (19.6)24 (20.7)22 (18.5)0.79 Coronary artery disease189 (80.4)92 (79.3)97 (81.5)Percutaneous coronary intervention Indicated105 (44.7)44 (37.9)61 (51.3)0.05 Completed, left coronary artery72/105 (68.6)27/44 (61.4)45/61 (73.8)0.20 Completed, right coronary artery26/105 (24.8)13/44 (29.5)13/61 (21.3)0.37 Completed, total number of treated lesions1 (1–1)1 (1–1)1 (1–1)0.36 Door-to-balloon time, min31 (19–43)34 (19–45)31 (19–42)0.71 Failed14/105 (13.3)7/44 (15.9)7/61 (11.5)0.10Unfractionated heparin during CAG91 (38.7)31 (26.7)60 (50.4)0.001Cangrelor or tirofiban during CAG18 (7.7)5 (4.3)13 (10.9)0.08Complications during CAG Need for tracheal intubation12 (5.1)6 (5.2)6 (5.0)1 Cardiac arrest20 (8.5)14 (12.1)6 (5.0)0.09 Death4 (1.7)2 (1.7)2 (1.7)1 Procedure interruption (all causes)13 (5.5)8 (6.9)5 (4.2)0.54IABP implantation during CAG9 (3.8)9 (7.8)00.006Post-CA peak troponin I level, ng per mL8835 (914–48461)8699 (901–58366)8857 (967–42,877)0.95Organ support in the ICU Mechanical ventilation215 (91.5)109 (94.0)106 (89.1)0.27 Inotropic agents145 (61.7)82 (70.7)63 (52.9)0.008 Vasopressors180 (76.6)99 (85.3)81 (68.1)0.003 Renal replacement therapy for AKI30 (12.8)19 (16.4)11 (9.2)0.15Second CAG during the ICU stay19 (8.1)9 (7.8)10 (8.4)1Treatment limitation decision at day 2841 (17.4)19 (16.4)22 (18.5)0.80Length of ICU stay, days7 (3–13)6 (2–10)7 (4–15)0.01Length of hospital stay, days12 (5–24)11 (4–24)13 (7–24)0.09All-cause in-ICU death94 (40.0)53 (45.7)41 (34.5)0.10All-cause in-hospital death103 (43.8)60 (51.7)43 (36.1)0.02Data are expressed as number (%) or median (interquartile range)BMI body mass index, COPD chronic obstructive pulmonary disease, ICU intensive care unit, SOFA sepsis-related organ failure assessment, SAPS-2 simplified acute physiology score, CAG coronary angiography, STEMI ST-segment elevation myocardial infarction, NSTEMI non-ST-segment elevation myocardial infarction, IABP intra-aortic balloon pump, AKI acute kidney injury1Including active cancer, receipt of immunosuppressive drugs, or HIV infection2Via pre-hospital medical emergency team3Missing values, n = 44From radial to femoral access or from femoral to radial access Study population Data are expressed as number (%) or median (interquartile range) BMI body mass index, COPD chronic obstructive pulmonary disease, ICU intensive care unit, SOFA sepsis-related organ failure assessment, SAPS-2 simplified acute physiology score, CAG coronary angiography, STEMI ST-segment elevation myocardial infarction, NSTEMI non-ST-segment elevation myocardial infarction, IABP intra-aortic balloon pump, AKI acute kidney injury 1Including active cancer, receipt of immunosuppressive drugs, or HIV infection 2Via pre-hospital medical emergency team 3Missing values, n = 4 4From radial to femoral access or from femoral to radial access Indications and characteristics of the CAG procedure CAG was performed for ST-segment elevation myocardial infarction, non-ST-segment elevation myocardial infarction and other indications in 140 (59.6%), 82 (34.9%) and 13 (5.6%) patients, respectively (Table 1). Half of patients underwent CAG within 3 h following resuscitated cardiac arrest. Compared to those with radial access, patients with femoral access were more often receiving inotropic agents (44.8% versus 30.3%, P = 0.03) and vasopressors (62.1% versus 43.7%, P = 0.007) at admission to the CAG room, without difference regarding invasive mechanical ventilation (83.6% versus 75.6%, P = 0.17). Myocardial infarction-related cardiogenic shock (KILLIP class IV) was more frequent at the time of CAG in patients with femoral access than in those with radial access (59.5% versus 39.5%, P = 0.003). PCI was indicated in 44 (37.9%) and 61 (51.3%) of patients with femoral and radial access, respectively (P = 0.05), without difference in the number of treated coronary lesions and rate of revascularisation completion (Table 1). Door-to-balloon time did not differ between patients with femoral access and those with radial access (34 [19–45] versus 31 [19–42] minutes, P = 0.71). Details on antiplatelet and anticoagulant therapies around CAG are provided in Table S1. Nineteen patients (8.1%) underwent one or more additional CAG procedure during the ICU stay. Primary and secondary study endpoints PS-based IPTW provided well-balanced populations of patients with femoral access and radial access, with absolute mean differences < 0.10 for all baseline characteristics exploited for PS calculation (Fig. S2). When analyzing the hierarchical primary composite endpoint on 13,804 IPT-weighted patient pairs, radial access produced more wins than femoral access (win ratio 0.62, 95% CI 0.40 to 0.98, P = 0.04; net benefit − 0.16, 95% CI − 0.32 to 0.003, P = 0.05). This benefit was mostly driven by a reduced incidence of resuscitated cardiac arrest, acute limb ischemia and/or the need for vascular surgery and all-cause death at day 28 in patients with radial access (Fig. 1). Win-ratio analysis of the hierarchical composite study endpoint with femoral versus radial access for coronary angiography. IPT, inverse probability of treatment (i.e. femoral access); CI, confidence interval. The number of wins, losses and ties for femoral versus radial access are exposed for each component of the hierarchical composite study endpoint (see the “Methods” section for details). Win ratio and net treatment benefit (win difference, i.e. the proportion of wins minus the proportion of losses) are exposed with their 95% CI. A win ratio < 1 indicates that femoral access has fewer wins than losses against radial access in prioritised pairwise comparisons, corresponding to a protective effect of radial access regarding the occurrence of the primary study endpoint Crude and IPT-weighted incidences of the primary and secondary endpoints are shown in Table 2. The primary composite endpoint occurred in 63.4% of patients with femoral access and 47.7% of patients with radial access in IPT-weighted populations (absolute risk difference 15.6%, 95% CI 1.4% to 29.8%, P = 0.03). Resuscitated cardiac arrest (absolute risk difference 13.9%, 95% CI 4.4% to 23.4%, P = 0.003) and acute limb ischemia and/or the need for vascular surgery on CAG access site (absolute risk difference 2.7%, 95% CI 0.3% to 5.1%, P = 0.03) at day 28 were more common in patients with femoral access while stroke at day 28 was more common in those with radial access (absolute risk difference − 3.6%, 95% CI − 6.8% to − 0.5%, P = 0.03). No significant difference was observed regarding the hazard of all-cause death, BARC type 3 or 5 major bleedings, or new myocardial infarction at day 28. Table 2Crude and weighted incidences of the primary and secondary study endpoints in critically ill patients with femoral or radial access for coronary angiographyWhole study population (unweighted)IPT-weighted population 1All patients(n = 235)Femoral access(n = 116)Radial access(n = 119)P-valueFemoral accessRadial accessAbsolute risk difference, %(95% CI)P-valuePrimary composite endpoint at day 28131 (55.7%)76 (65.5%)55 (46.2%)0.00463.4%47.7%15.6% (1.4% to 29.8%)0.03Components of the primary composite endpoint at day 28 All-cause death97 (41.3%)56 (48.3%)41 (34.5%)0.0447.2%36.8%10.4% (− 3.9% to 24.7%)0.16 Including BARC type 5 major bleeding9 (3.8%)4 (3.4%)5 (4.2%)1.004.1%5.7%− 1.6% (− 8.6% to 5.4%)0.65 Resuscitated cardiac arrest32 (13.6%)24 (20.7%)8 (6.7%)0.00320.0%6.1%13.9% (4.4% to 23.4%)0.003 Stroke5 (2.1%)0 (0.0%)5 (4.2%)0.0803.6%− 3.6% (− 6.8% to − 0.5%)0.03 Myocardial infarction6 (2.6%)3 (2.6%)3 (2.5%)1.003.4%2.8%0.6% (− 5.1% to 6.3%)0.84 Acute limb ischemia and/or need for vascular surgery on CAG access site8 (3.4%)8 (6.9%)00.0032.7%02.7% (0.3% to 5.1%)0.03 BARC type 3 major bleeding21 (8.9%)11 (9.5%)10 (8.4%)0.958.0%8.0%0.0% (− 7.3% to 7.4%)0.99Data are expressed as numbers and percentagesComparisons in the whole study population (unweighted analyses) were performed using the Fisher’s exact test or the χ2 test, as appropriate, while those in the IPT-weighted population were performed using the Rao-Scott χ2 test (i.e. χ2 test weighted by IPT)IPT inverse probability of treatment (i.e. femoral access), CI confidence interval, BARC Bleeding Academic Research Consortium1The numbers of patients in PS-weighted populations were 89.2 in the femoral group and 96.1 in the radial group Crude and weighted incidences of the primary and secondary study endpoints in critically ill patients with femoral or radial access for coronary angiography Data are expressed as numbers and percentages Comparisons in the whole study population (unweighted analyses) were performed using the Fisher’s exact test or the χ2 test, as appropriate, while those in the IPT-weighted population were performed using the Rao-Scott χ2 test (i.e. χ2 test weighted by IPT) IPT inverse probability of treatment (i.e. femoral access), CI confidence interval, BARC Bleeding Academic Research Consortium 1The numbers of patients in PS-weighted populations were 89.2 in the femoral group and 96.1 in the radial group In a Cox proportional hazard model stratified on the occurrence of cardiac arrest before CAG and adjusted on PS-based IPTW, femoral access was not independently associated with the likelihood of all-cause death at day 28 (coefficient for radial versus femoral access − 0.35, robust standard error = 0.235, adjusted hazard ratio [aHR] 1.43, 95% CI 0.90 to 2.26, P = 0.13) (Fig. 2). The interaction between access site (radial versus femoral) and cardiac arrest status, tested by including an interaction term in the Cox model, was not statistically significant (P = 0.15), indicating that the effect of access site did not differ according to this status. Similar results were observed when adjusting the model on the occurrence of cardiac arrest before CAG (aHR for femoral versus radial access 1.43, 95% CI 0.90 to 2.27, P = 0.14; aHR for cardiac arrest versus no cardiac arrest before CAG 3.39, 95% CI 1.93 to 5.96, P < 0.0001) and in a sensitivity analysis excluding patients with a treatment limitation decision before day 28 (coefficient for radial versus femoral access 0.29, robust standard error 0.291, aHR 1.34, 95% CI 0.76 to 2.37, P = 0.32). Kaplan–Meier curves of day-28 survival according to the primary access site and occurrence of cardiac arrest before coronary angiography. Rad, radial access; Fem, femoral access; CA, cardiac arrest The upper panel shows the weighted Kaplan–Meier curves for each group obtained by PS-based IPTW, with 95% confidence intervals. The lower panel shows raw numbers at risk for each group. Adjusted hazard ratio of all-cause death at day 28 with femoral versus radial access, 1.43; 95% confidence interval, 0.90–2.26; P = 0.13 Duration of cannulation in patients requiring VA-ECMO Femoro-femoral VA-ECMO was required during or after CAG in 19 (16.4%) patients with femoral access and 17 (14.3%) patients with radial access (P = 0.58), mostly for refractory cardiac arrest (Table 3 and Table S2). In patients with femoral access, the insertion of the arterial ECMO cannula was performed by using the CAG cannula in 7 out of 19 patients (36.8%). The use of primary femoral access for CAG was not associated with a reduced duration of ECMO-VA implantation (23 [19–44] min from procedure initiation to unclamping versus 28 [19–45] min in patients with radial access, P = 0.97). Table 3VA-ECMO in critically ill patients with femoral or radial access for coronary angiographyAll patients(n = 36)Femoral access(n = 19)Radial access(n = 17)P-valueIndication for veno-arterial ECMO Refractory cardiac arrest26 (72.2)15 (78.9)11 (64.7)0.58 Cardiogenic shock10 (27.8)4 (21.1)6 (35.3)Femoral artery access New puncture29 (80.6)12 (63.2)17 (100)NA Use of CAG cannula7 (19.4)7 (36.8)–Implantation duration, min 127 (19–45)23 (19–44)28 (19–45)0.97ECMO days5 (3–9)6 (2–8)3 (3–11)0.85ECMO-related lower limb ischemia4 (11.1)2 (10.5)2 (11.8)1Data are expressed as number (%) or median (interquartile range). The statistical tests used for these bivariate comparisons are indicated in the methods section. No inverse probability of treatment weighting was appliedVA-ECMO (femoro-femoral) veno-arterial extra-corporeal membrane oxygenation, CAG coronary angiography1From procedure initiation to unclumping VA-ECMO in critically ill patients with femoral or radial access for coronary angiography Data are expressed as number (%) or median (interquartile range). The statistical tests used for these bivariate comparisons are indicated in the methods section. No inverse probability of treatment weighting was applied VA-ECMO (femoro-femoral) veno-arterial extra-corporeal membrane oxygenation, CAG coronary angiography 1From procedure initiation to unclumping Discussion In this study that included 235 critically ill patients undergoing emergent CAG, mostly for cardiac arrest or cardiogenic shock suspected to result from myocardial ischemia, primary femoral access was associated with a higher incidence of a hierarchical composite endpoint of MACE and BARC type 3 or 5 major bleeding at day 28 when compared to primary radial access. This difference was driven by an increased frequency of access site complications and events that may reflect both the patient condition and a procedural effect—that is, new resuscitated cardiac arrest and all-cause death—in patients with femoral access. Major bleedings were similarly observed in the two groups. In a meta-analysis of randomised trials published in 2018 and including the pivotal RIFLE-STEACS [9], RIVAL [10], and MATRIX [11] studies, the use of radial access for CAG and PCI in patients with acute coronary syndrome was associated with a reduced risk of bleeding, access site complication and all-cause death compared to femoral access [12]. More recently, the SAFARI-STEMI trial, which was terminated early following a futility analysis, showed no over-risk of major bleeding or all-cause death with a femoral versus radial approach [23]. Patients with myocardial infarction-related cardiogenic shock were not included or only marginally represented in these studies. Yet, in a post-hoc analysis of the CULPRIT-SHOCK trial, a radial approach correlated with a reduced risk of all-cause death at day 30 that was strikingly no longer significant at one year, without difference regarding major bleedings [14]. Likewise, in an ancillary analysis of the ECLS-SHOCK trial, femoral access was not associated with increased rates of major bleeding, peripheral vascular complication, or all-cause death at day 30 [24]. Despite this substantial amount of research, granular evidence remains scarce regarding the arterial route that should be prioritised for CAG in the most severely ill patients requiring admission to the ICU. Most patients included in our study presented with multi-organ failure complicating post-resuscitation syndrome or cardiogenic shock, were receiving invasive mechanical ventilation and catecholamines at the time of CAG, and exhibited non-cardiac chronic diseases that may further predispose to subsequent MACE and CAG-related adverse events—e.g. lower limb arteriosclerosis. In this population at very high risk for unfavorable outcomes, femoral access was associated with the occurrence of access site complications, namely acute limb ischemia and/or the need for vascular surgery, while no cases were observed with radial access. Besides, the incidence of major bleedings did not differ between the two approaches. This finding may contribute to explain the similar frequency of myocardial re-infarction in both groups since bleeding may impose transient interruption of antithrombotic drugs, thereby promoting secondary ischemic cardiac events. Strokes were only observed in the radial group but remained highly uncommon, precluding any reliable interpretation of this finding to be made. New resuscitated cardiac arrests were more common in patients with femoral access. This finding cannot be attributed to the arterial route chosen for CAG and likely correlates with the greater prevalence of protracted circulatory failure in the femoral group, which may have been incompletely balanced by our PS-based weighting method. Indeed, compared to those with radial access, patients with femoral access received more vasopressors and inotropic drugs during the ICU stay. Moreover, dobutamine, epinephrine and norepinephrine all exert a marked proarrhythmic effect that may facilitate the occurrence of treatment-emergent pulseless ventricular arrhythmia, especially when administered concomitantly [25, 26]. This severity imbalance and the resulting higher incidence of subsequent cardiac arrests suggest residual confounding by indication and could explain the trend toward a higher hazard of all-cause death at day 28 in patients with femoral access despite adjustment for other potential determinants of outcome. These aspects should be considered when appraising the results of our win-ratio analysis. Also, statistical power could have been insufficient to demonstrate a clinically relevant inter-group difference on day-28 mortality. Only one recent single-centre retrospective study compared the radial and femoral approaches in a population of critically ill patients with similar severity and frailty markers. In this work focused on patients with ST-elevation myocardial infarction complicated by out-of-hospital cardiac arrest, femoral access was neither associated with a higher crude incidence of bleedings nor with an increased hazard of in-hospital death—MACE and access site complications were not investigated [15]. A key argument for choosing the femoral route in hemodynamically unstable patients is to obtain an easier and faster access for CAG. Cardiogenic shock is an established risk factor for failure of radial access and need for radial-to-femoral crossover due to poor arterial pulse, along with female sex, low height and/or weight, higher age, and limited physician experience [16, 27, 28]. Consequently, femoral access still appears as the preferred route in this situation [14, 24]. However, in our study, door-to-revascularisation time did not differ between femoral and radial accesses, suggesting that they are equally efficient when performed by interventional cardiologists with prior skill in both approaches. Another theoretical argument for a primary femoral strategy is the potential use of the CAG cannula for faster guidewire insertion and arterial cannulation in patients who necessitate peri-procedural VA-ECMO. This issue was not addressed in the few available studies focused on the feasibility and outcomes of rescue VA-ECMO in the CAG laboratory [29, 30]. In the ECLS-SHOCK trial, the choice of the arterial access for CAG was left at the discretion of the local investigators (femoral, 75%) and the arterial cannula was inserted following a dedicated puncture of the femoral artery in patients randomised in the VA-ECMO arm [31]. In our study, two thirds of patients with femoral access who required VA-ECMO were cannulated through a distinct arterial puncture. Potential explanations include the need for implantation immediately before or during CAG or, conversely, after removal of the CAG cannula. Overall, implantation duration was not reduced in patients with femoral access. Our work has limitations. First, owing to its retrospective design, residual confounding on the study endpoints cannot be ruled out despite the use of granular data and careful adjustment on baseline patient characteristics. Particularly, certain hemodynamic parameters at the time of CAG such as vasopressor dose, number of agents and blood lactate level were not entered in PS calculation. Second, this was a single-centre study, which restrains its external validity; however, CAG and intensive care procedures followed current guidelines, with overall outcomes (notably day-28 mortality) that were consistent with those previously reported in similar populations [14, 15, 31]. Also, our results may not be fully generalisable to centres with different levels of expertise, particularly regarding radial access in hemodynamically unstable patients. Lastly, the long-term incidence of MACE was not compared between the two groups; however, it is unlikely that the arterial access used for CAG exerts an impact on patient outcomes beyond the acute phase at risk for procedure-related complications. Conclusion In this single-centre retrospective study, the use of a femoral rather than radial approach for CAG in critically ill patients was associated with a higher incidence of access site complications, with no independent impact on the hazard of major bleeding or all-cause death at day 28. Time to coronary revascularisation and the duration of ECMO-VA cannulation, when required, were not reduced with a femoral approach. Our results need to be confirmed by prospective multicentre data but suggest that radial access should be preferred to femoral access for CAG in critically ill patients, in line with current guidelines dedicated to patients managed in the coronary care unit. Supplementary Information Supplementary Material 1: Figure S1. Study flow-chart. Figure S2. Unweighted and inverse probability of treatment-weighted absolute mean differences of the covariates included in the propensity score. Table S1. Full characteristics of the study population. Table S2. Main characteristics of patients requiring VA-ECMO according to access site for coronary angiography. Acknowledgements None. Declaration of Generative AI and AI-assisted technologies in the writing process Artificial intelligence was not used for this research or for manuscript preparation. Abbreviations Adjusted hazard ratio Bleeding Academic Research Consortium Coronary angiography Confidence interval Intensive care unit Inverse probability of treatment weighting Percutaneous coronary intervention Propensity score Major adverse cardiovascular event Veno-arterial extra-corporeal membrane oxygenation Authors’ contributions AD and FB designed the study and collected the data. TB performed the statistical analyses. AD and FB interpreted the results and drafted the manuscript. All other authors contributed substantially to data collection and draft revision. FB had full access to all of the data and takes responsibility for the integrity of the data and the accuracy of analyses. All authors approved the final manuscript. Data availability The datasets used and analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The study protocol was approved by the ethical committee of the French Society of Anaesthesia and Intensive Care on January 5, 2025 (IRB 00010254‐2024-105). The requirement for informed consent was waived owing to the observational design. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1: Figure S1. Study flow-chart. Figure S2. Unweighted and inverse probability of treatment-weighted absolute mean differences of the covariates included in the propensity score. Table S1. Full characteristics of the study population. Table S2. Main characteristics of patients requiring VA-ECMO according to access site for coronary angiography. Data Availability Statement The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.