In this cohort study, we analyzed retrospectively, from a prospective database, all patients who were admitted consecutively with CS and received treatment with NIV in the 18-bed intensive care unit (ICU) of the Morales Meseguer University Hospital in Murcia, Spain, between January 1997 and December 2021. The study was approved by the ethics committee of the institution.

Cardiogenic shock was defined by systolic blood pressure <90 mmHg for at least 30 minutes despite appropriate fluid resuscitation or need for vasopressor support to maintain systolic pressure above this level, plus signs or symptoms of hypoperfusion (altered mental status, cool extremities, oliguria with urine output less than 0.5 milliliters/kilogram/hour for at least 6 hours, serum lactate >2 mmol/L). The cardiac etiology of the shock required confirmation by echocardiography or invasive monitoring of pulmonary pressures and cardiac output.

Inclusion criteria were respiratory failure of cardiac cause defined by the presence of a respiratory rate >30 breaths per minute or use of respiratory accessory muscles, in addition to PaO2/FiO2 <250 mmHg, or respiratory acidosis (arterial pH <7.35 with PaCO2 >45 mmHg). The exclusion criteria were uncontrollable ventricular arrhythmias, shock due to right ventricular dysfunction, severe multiorgan failure at admission, cardiorespiratory deterioration or cardiac arrest requiring immediate ETI, decreased cardiac contractibility due to septic shock and transfer to another hospital for coronary surgery in the first 12 hours of ICU admission.

Cardiogenic shock was classified by its etiology as secondary to ACS-AMI or other cause (chronic coronary artery disease, valvular disease, arrhythmia). The diagnosis of ACS-AMI was made according to the criteria in force at the time of admission.21–24 Before the first universal definition of AMI, when the patient presented a clinical condition compatible with cardiac ischemia, accompanied by elevation of cardiac markers or suggestive electrocardiographic changes, AMI was considered. In relation to cardiac markers, pathological an elevation of CPK above twice the considered normal value and a CPK-Mb above 10% of total CPK, were the criteria. Once the use of troponin I became standard in 2001, the cut-off point used was 0.04 ng/ml. During the early years of the study, patients were classified as myocardial infarction with and without Q wave. These patients were subsequently reclassified as ST segment elevation myocardial infarction (STEMI) and non-ST segment elevation myocardial infarction (NSTEMI) or indeterminate (presence of pacemaker or complete left bundle branch block). All electrocardiograms were evaluated by at least two intensivist physicians with extensive experience in coronary patients, and doubtful cases had the opinion of a third intensivist and subsequently were corroborated by the cardiologist on the ward. The assignment of the severity of CS was also carried out subsequently, using the patients’ data from the first 24 hours of ICU admission.25 The reclassification was carried out blinded to the outcome of the patients.

The patients under analysis were classified as de novo HF if they presented typical symptoms and signs of HF for the first time and there was no previous history of HF, and as acute decompensated HF if they showed worsening of a preexisting HF.26

The clinical management of patients is shown in the Supplementary Material. All patients underwent a bedside echocardiography in the first 12 hours of ICU admission, assessing the impact of the global and segmental contractility of the left ventricle, dimensions of the cardiac cavities, possible mechanical complications in patients with ACS, involvement of the contractibility and size of the right ventricle, and, if possible, the filling pressures of the cardiac cavities.

An indication for non-invasive ventilation indication was established by the attending physician. All patients were continuously monitored by electrocardiogram, pulse oximetry, blood pressure and respiratory rate. NIV protocol is shown in Supplementary Material.

In all patients, an arterial blood gas control was performed in the first hour after NIV initiation. Subsequent arterial blood gas controls were performed according to the attending physician, considering the clinical status and evolution of the patient.

Non-invasive ventilation failure was defined as the need for ETI in a patient with deterioration of respiratory parameters despite NIV optimization, and/or death in the ICU. In case of NIV failure due to deterioration, the patient was intubated and connected to mechanical ventilation, according to previously published criteria.19,27 Patients with a do not intubate (DNI) order received non-invasive support until improvement or death.

Demographic, clinical and laboratory variables were recorded at NIV initiation. At ICU admission, body mass index was calculated for all patients. The degree of disease severity during the first 24 hours in the ICU was obtained by calculating the Simplified Acute Physiology Score II (SAPS II), and the degree of multiorgan dysfunction through the Sequential Organ Failure Assessment (SOFA) score, measured daily. To quantify patient comorbidities, the Charlson Comorbidity Index was calculated. Anemia was defined as Hb <12 g/dl in women and <13 g/dl in men. Classification CS severity was performed at the beginning of NIV and in the first 24 hours after admission while receiving NIV. The Society for Cardiovascular Angiography and Interventions (SCAI) classification of cardiogenic shock was used, subsequently assigning the patients to each of the stages.28,29 The SCAI Shock is divided into five stages: A: patients at risk of cardiogenic shock; B: pre-shock state, patients with arterial hypotension but with normal tissue perfusion; C: tissue hypoperfusion without treatment, or requiring pharmacological hemodynamic support or mechanical intervention; D: tissue hypoperfusion that does not respond to an adequate trial of an initial supportive intervention; and E: refractory shock that does not respond despite escalation of treatment and the patient is in severe cardiovascular collapse including arrest in progress.28,29 The HACOR score was measured at 1 hour after starting NIV.30 The HACOR score integrates the results of the hemodynamic (heart rate), respiratory (respiratory rate, arterial pH and ratio PaO2/FiO2), and neurological parameters (Glasgow Coma Scale) of the patient. The usefulness of HACOR in the prediction of NIV failure has been clearly established in hypoxemic ARF of different etiologies.30 Definitions of the complications analyzed are shown in Supplementary Material. Mortality was assessed in ICU and during hospital stay, and at one year.

Qualitative or categorical variables were expressed as absolute and relative frequencies, and comparisons between them were made using Pearson's chi-square test or Fisher's exact test. Quantitative variables were expressed as mean±standard deviation or median (interquartile range – IQR), and the comparison between independent groups using the Student's t-test or Mann–Whitney U test if the variable did not follow normal distribution. The measures of association analyzed were odds ratio (OR) with 95% confidence intervals (95% CI). Independent risk factors for developing NIV failure were analyzed using logistic regression. Variables identified as predictors of NIV failure in the univariate analysis, together with those considered relevant, were included in a multivariate logistic regression model using the stepwise forward method (PIN\0.10, POUT\0.05) to correct collinearity.

Propensity score matching was produced using nearest-neighbor model without replace, with a 1:1 ratio.31 Each patient with CS-NIV was matched to one control patient with the nearest-neighbor matching within calipers determined by the propensity score. Variables used for matching included: year to admission, age, gender, acute coronary syndrome, SAPS II, initial SOFA, Charlson Index, left ventricular ejection fraction (LVEF), location of the patient before admission to the ICU, type of AHF–de novo or acute on chronic HF, NYHA class and DNI order. A caliper width of 0.1 of the standard deviation of the logit of the propensity score was used for the matching. To determine the effectiveness of the propensity score matching for controlling the differences between patients with and without NIV, standardized mean differences (SMDs) were calculated for each variable before and after matching. SMDs <10% indicated successful propensity scores matching and balancing between the two groups. The analysis was performed using the SPSS 25.0® program (IBM™, Armonk, NY) and R version 3.4.0® (Copyright 2017 The R Foundation for Statistical Computing Platform™).

The main sociodemographic and clinical background and characteristics are shown in Table 1. There was a slight predominance of male gender (57%), advanced age (mean of 73.8 years) and low comorbidities, measured by the Charlson Index. Of the variables analyzed, SAPS II index, ward origin, comorbidity index and the presence of DNI order were higher in patients with NIV failure.

The main characteristics related to heart disease are shown in Table 2. The most frequent etiology was ACS-AMI (54.7%). One hundred and fifteen patients (38.3%) had de novo AHF, while the rest had previous chronic HF. The presence of ACS and de novo AHF was related to higher NIV failure rate. Of the 164 patients with ACS-AMI, coronary angiography was performed in 155 (94.5%), immediate in 48.4%, early in 43.2% and delayed in 8.4%. Nine patients died without having undergone coronary angiography. Primary revascularization was performed in 48.8% of the patients and coronary revascularization in 83.4% during the hospital stay. Severity of CS was related to NIV failure: among patients with NIV success, 68% were C stage, 28.6% stage D and 3.4% stage E, while in the NIV failure group, 18.3%, 32.7% and 49%, respectively (p<0.001). The cardiological treatments used during the ICU stay are shown in Table S1 (Supplementary Material).

Respiratory, hemodynamic, and neurological parameters, before and after 1 hour of NIV, are shown in Table 3. Initial values reflected severe hemodynamic deterioration, with mixed metabolic and respiratory acidosis, elevated lactate levels, high heart, and respiratory rates, altered mental status and severe hypoxemia. Most parameters at the start of NIV did not differ between NIV success and failure. The presence of a positive fluid balance in the first 24 hours of ICU stay as well as the total volume of fluids were higher in NIV failure (p=0.006 and p=0.001, respectively).

Two hundred seventy-nine (93%) patients were treated with NIV in bilevel mode, and the rest in CPAP mode (Table 4). The mean level of IPAP and EPAP at baseline was 15±2 cmH2O and 7±1 cmH2O, respectively, and maximum levels reached were 18±3 cmH2O and 9±1 cmH2O, respectively. IPAP and EPAP levels were higher in patients with NIV failure (p<0.001). The median duration of NIV was 22 hours (IQR=12–38). Patients with NIV success used NIV for shorter periods (median 20 hours), than patients in the NIV failure group (median 22 hours) (p=0.004). Eighty-seven (25%) patients presented complications related to ventilatory support, skin lesions being the most frequent. Patients with complications related to NIV had higher NIV failure rate (34.6% vs. 23.1%; p=0.028). The HACOR score measured at 1 hour after NIV initiation had an average value of 9.4±4.4 in the NIV failure group and 4.6±2.2 in the success group (p<0.001). The discriminating ability of the HACOR score, at 1 hour after starting NIV, to predict NIV failure showed an area under the curve of 0.863 (95% CI 0.819–0.900) (Figure S1 of Supplementary Material).

Multiple variables were related to NIV failure in multivariate analysis (Figure 2). The independent predictive factors related to greater NIV failure were being initially admitted to the ward (OR=2.39), ACS-AMI as a cause of SC (OR=2.71), SAPS II index (OR=1.04 per point), higher HACOR score at 1 hour after starting NIV (OR=1.44 per point in the score), development of NIV related complication (OR=2.14) and, the most important, the severity of CS (OR=7.26 to D/E stage vs. C stage).

Figure 2.

Independent risk factors related NIV failure (Hosmer–Lemeshov test p=0.798. AUC=0.909; 95% CI=0.877–0.941).

(0,15MB).

The relation between NIV failure and patient outcome is shown in Table 5 and Table S2. ICU length of stay was longer in patients with NIV success comparing to NIV failure, median 6 days (IQR 5–7) and 5 days (IQR 4–8) respectively (p<0.001), as well as hospital stay, 16 days (IQR 10–22) and 6 days (IQR 4–11) respectively (p<0.001). ICU and hospital mortality was 44% and 50%, respectively. Fifteen patients (10.2%) with NIV success died in the hospital, compared to 135 (88.2%) in NIV failure group (p<0.001). Hospital mortality was related to the severity of CS: it was 10.2% in C stage, 48.9% in D stage and 92.5% in E stage (p<0.001). In patients with NIV failure and need for intubation, there was a longer time between the start of NIV to ETI in those who died in the hospital (Figure 4). Mean duration of NIV use was 22 hours (IQR 6–44) in hospital deceased patients and 6 hours (IQR 3–12) in patients discharged alive (p=0.001).

In the paired sample using propensity score matching (Table 6), patients in the ETI-IMV group showed a worse prognosis, increased nosocomial infection (OR=2.51, 95% CI=1.33–4.74), ICU mortality (OR=1.75, 95% CI=1.12–2.73), hospital mortality (OR=1.69, 95% CI=1.09–2.63) and 1 year mortality (OR=1.61, 95% CI=1.04–2.51). Only C stage benefited from NIV (hospital mortality 10.1% vs. 32.9%; p<0.001). In patients with E stage, mortality is higher, although without significant differences, in those receiving NIV than ETI-IMV as first respiratory support (Figure 3).

Figure 3.

Comparison hospital mortality between NIV and EIT-IMV according to severity CS. EIT-IMV: endotracheal intubation and invasive mechanical ventilation; NIV: non-invasive ventilation.

(0,15MB).

Figure 4.

Time to intubation and mortality. Horizontal bars represent median (IQR) values of survivors and non-survivors for intubated patients.

(0,13MB).