Prolonged mechanical ventilation (PMV) after heart surgery is correlated with an increased risk of patient morbidity and mortality (range, 4.9–38%), although it occurs in only 3–9.9% of patients [1,2]. Early extubation (within 8–12 h from arrival to postoperative ICU) is associated with better cardiac function, patient comfort and a decrease in respiratory complications and hospital costs as a result of the shortened ICU length of stay, and it should thus be the gold standard [3,4]. To guarantee this critical target, the early and smooth restoration to patient preoperative physiological homeostasis, should be the aim from the operative theatre through return to normothermia, adequate heart function, intravascular volume replacement, electrolyte normalization and optimization of pain relief [3,4]. PMV occurs in high-risk patients who are likely to be identified preoperatively and upon arrival in the postoperative ICU [5–7]. Failure to wean such patients from the ventilator is known to be associated with the activation of the inflammatory cascade associated with cardiopulmonary bypass (CPB) [3,8–12].
However, in the literature, we may find the implication of other variables: preoperative, such as decreased forced expiratory volume rate in the first second, higher BMI, diabetes, increased pulmonary arterial pressure, hypoalbuminemia, history of cerebral vascular disease, older age, New York Heart Association (NYHA) class higher than 2, ejection fraction under 50%, preoperative intraaortic balloon pump and creatinine clearance lower than 50 ml/min. [3,4,13–16]; intraoperative, such as long CPB time, duration (>8 h) and type of surgery (multiple valve replacements, aortic procedures, operative priority and reoperation for bleeding) [3,4,13–16]; postoperative, such as low hematocrit, PaO2/FiO2 ratio and cardiac index [3,4,13–16]. Currently, risk stratification for PMV is an important part of the preoperative evaluation in order to identify those patients who will be likely to undergo it and adopt preemptive strategies.
Unfortunately, there is no consensus in the literature on data regarding the exact definition of PMV rates and independent predictive risk factors [3,4]. In order to provide new insight on this topic, we conducted this study aiming to identify the main preoperative and intraoperative variables associated with PMV in our large cohort of cardiac surgical patients, improve the management quality of such patients and save on total hospital costs.
We considered all patients undergoing cardiac surgery, admitted to our ICU from January 2000 to December 2006, using data prospectively put into an electronic database. The following data was collected from all patients: demographics (age, sex); surgical operation type, gravity scores [clinical risk score (CRS): Higgins score, Canadian Cardiovascular Society (CCS) score, NYHA class] and left ventricular ejection fraction (LVEF) measured with transoesophageal echocardiography (TOE); risk factors likely to have an impact on outcome and underlying illnesses; CPB and aortic cross clamp (ACC) times and intraoperative transfusion of red blood cells (RBCs), fresh frozen plasma (FFP) and vasoactive drugs; ICU supportive techniques (i.e. the presence and duration of mechanical ventilation); and pre-ICU and post-ICU lengths of stay. Patients were ‘a posteriori’ divided into two cohorts: early extubation, undergoing successful extubation within 12 h from ICU admission, and delayed extubation, needing mechanical ventilation over 12 h. Standardized analgesia protocols, mainly based on opioid continuous infusion, were applied. Standardized weaning protocol was adopted, and extubation occurred when clinically indicated, that is, on the basis of metabolic homeostasis and temperature restoration, haemodynamic stability, adequate gas exchange and respiratory mechanics. Early extubation and delayed extubation patients who needed reintubation within 72 h from weaning were censored from the analysis.
SPSS 13.1 software (SPSS Inc., Chicago, Illinois, USA) was used for statistical analysis. For continuous variables, we used median (IQR) as not normally distributed and compared rank values using nonparametric tests (Mann–Whitney U, Wilcoxon W when appropriate). Differences in proportions were compared with X2 test or Fisher's exact test in case of expected frequencies less than five. Under the conditional independence assumption, Cochran's statistic was asymptotically applied for the univariate analysis; the Mantel–Haenszel common OR estimate is shown in the related proportional variable table. A multivariate binary logistic regression model was fitted, using a step-by-step enter approach, to identify independent predictors of PMV duration as outcome (dependent variable assuming the values: yes, if mechanical ventilation ≤12 h and no, if mechanical ventilation >12 h) . As the method used for the multivariate logistic regression was the enter approach, all independent variables were mandatorily forced into the model itself. As a consequence, all variables of the univariate analysis were considered for the multivariate one too. By the enter approach, different models were compared by the likelihood ratio test, using a P value of 0.05 or less as the level of significance. The multivariate binary logistic regression model was set up considering all quantitative variables transformed into dichotomous categorical ones, assumed as ‘yes’ (higher than the cut-off) or ‘no’ (lower than the cut-off) value for all but LVEF, for which the reverse was used. The cut-off was considered as the 50th percentile (median) of the recorded values, if clinically proper and adequate. The median value was included into the lower interval.
Independent variables entered in the model were divided into the following groups: demographics: sex (male/female), age (yes, age >65 years; no, age ≤65 years); risk factors (presence = yes): diabetes (yes/no), hypertension (yes/no), arteriopathy (yes/no), chronic renal failure (yes/no), chronic obstructive pulmonary disease (COPD) (yes/no); type of heart disease or relative surgery or both (presence = yes): coronary artery disease (CAD) (yes/no), aortic valve disease (yes/no), mitral valve disease (yes/no), tricuspid valve disease (yes/no), CAD and cardiac valve disease (CVD) (yes/no), redo surgery (yes/no), emergency surgery (yes/no); preoperative risk scores (yes: higher than a specified cut-off): NYHA or CCS (yes, >2; no, ≤2), CRS (yes, >8; no, ≤8), LVEF value (yes, ≤30%; no, >30%); intraoperative variables (yes: higher than a specified cut-off): CPB time (yes, >77 min; no, ≤77 min), ACC time (yes, >55 min; no, ≤55 min), RBC and FFP transfusions (yes, >4; no, ≤4), use of vasoactive drugs (i.e. dobutamine, dopamine, epinephrine). We merged the NYHA (for CVD) and CCS (for CAD) class into a single variable, considered as yes, if NYHA or CCS class was higher than 2 and no if it was 2 or less. The considered CRS cut-off was 8 as related to a significant increase in predicted mortality (from 2% if CRS ≤8 to >6% if CRS >8) . The cut-off for ICU length of stay and intraoperative RBC/FFP transfusions was derived from the 75th percentile (4 days and 4 units, respectively) as more clinically appropriate than the median value. For LVEF, the cut-off was 30% as clinically valid to indicate a reasonably depressed left ventricular function.
The Hosmer–Lemeshow goodness-of-fit test was performed to assess the validity of the multivariate binary logistic regression model . Receiver operating characteristic (ROC) curve analysis was used to assess the sensitivity and specificity of statistically significant discrete PMV predictors, whereas Spearman's sensitivity test was applied to define correlation between statistically significant categorical ones and PMV. Kaplan–Meier survival analysis was applied to point out different outcomes between the two cohorts for cumulative hazard of being discharged from ICU to cardiac surgical ward during the first 28 days after surgery, cumulative hazard of being discharged from cardiac surgical ward to rehabilitative one during the first 60 days after ICU stay and hospital survival. Data were compared using the log-rank method (Mantel–Cox test). Deaths were censored from the analysis of discharging probability. Two-tailed P values of less than 0.05 were considered statistically significant.
PMV: duration of mechanical ventilation over 12 h. No patient underwent extubation in the operative theatre. Redo surgery: any surgical patient that had previously undergone a cardiac surgical operation. Emergency surgery: an operation not clinically procrastinable longer than 4 h.
Chronic renal failure: patient with a creatinine clearance 50 ml/min or less for 6 months at least.
Arteriopathy: generalized and unspecified atherosclerosis.
Hypertension: persistently high arterial blood pressure during treatment for over 6 months.
COPD: COPDs diagnosed by lung functional tests.
Aortic, mitral, tricuspid disease: both insufficiency and stenosis, documented by TOE.
CPB time: the duration of CPB.
ACC time: the duration of clamping of the aorta.
CRS: CRS, the Higgins score for cardiac surgical patients .
CCS: CCS (functional classification) score for patients with CAD .
NYHA: NYHA classification (functional classification) for patients with valve diseases .
Population general characteristics
Over the study period, we admitted a total of 5123 patients with a median (IQR) age of 67 (59–73) years. Figure 1 shows the description of the patient groups of the studied population. CAD was recorded more frequently among male patients (78.8 vs. 56.1% in women; P = 0.0001) whereas valve disease among female patients (36.6 vs. 18% in men; P = 0.0001). The more frequent CVD were mitral stenosis (68.1%) and aortic insufficiency (61.3%). The most frequent operation was cardiac revascularization (67.1%), followed by valve (19.7%) and combined (9.7%) [coronary artery bypass graft (CABG) and valve] surgery. Other operations occurred in 3.5% of patients, mainly regarding the aorta. Among patients with CAD, triple CABG was the most frequent (39.1%) operation, followed by quadruple CABG (27.6%) and double CABG (19.9%). Among patients with CVD, the main operation was aortic repair/replacement (59.3%), followed by mitral surgery (23.6%) and combined (aortic and mitral) valve replacement (14.5%). Aortic repair/replacement associated with CABG was prevalent (57.1%) among combined interventions.
Tables 1 and 2 show the intergroup comparisons of the perioperative variables.
Multivariate predictive model
Table 3 shows the multivariate binary logistic regression model assessing the effect of the considered independent variable on the duration of mechanical ventilation (yes = not prolonged, a mechanical ventilation of ≤12 h and no = prolonged, a mechanical ventilation of >12 h) as outcome. We identified age more than 65 years, COPD, presence of chronic renal failure, a CCS or NYHA score higher than 2, LVEF 30% or less, undergoing a redo or emergency surgery, CPB time longer than 77 min and need of RBC or FFP transfusions exceeding four units as independent predictors of PMV. To better explain the model, any patient with an LVEF of 25% experiencing a redo surgery has approximately a 4.14 times higher risk of undergoing PMV. Hosmer–Lemeshow test of goodness of fit for the model was performed and showed a χ2 = 129.537, P = 0.896, strengthening the validity of the model itself. ROC curve analysis corroborated the results of multivariate model as it showed the following for quantitative discrete predictors: age more than 65 years (sensitivity, 73.7%; specificity, 70.7%), LVEF 30% or less (sensitivity, 90.4%; specificity, 96.6%), CPB time more than 77 min (sensitivity, 95.2%; specificity, 92.1%), RBC transfusions more than four units (sensitivity, 89.2%; specificity, 70.5%) and FFP transfusions more than four units (sensitivity, 78.3%; specificity, 75.7%). Spearman's sensitivity test showed a direct correlation between each categorical predictor and PMV: chronic renal failure (Spearman ρ, 0.128; P < 0.01), COPD (Spearman ρ, 0.198; P < 0.01), redo surgery (Spearman ρ, 0.150; P < 0.01) and emergency surgery (Spearman ρ, 0.147; P < 0.01). No changes occurred in the multivariate binary logistic regression model results, considering quantitative independent variables as discrete/continuous values without categorizing them into dichotomous ones.
Length of stay and survival analyses
Pre-ICU [early extubation: median (IQR) 2(1–5) vs. delayed extubation: median (IQR) 3(1–5); P = 0.0001], ICU [early extubation: median (IQR) 2(1–2) vs. delayed extubation: median (IQR) 2(2–4); P = 0.0001], post-ICU [early extubation: median (IQR) 7(7–8) vs. delayed extubation: median (IQR) 8(7–11); P = 0.0001] and total hospital length of stay [early extubation: median (IQR) 11(9–13) vs. delayed extubation: median (IQR) 13(10–17); P = 0.0001] were significantly higher in patients in the delayed extubation group. The percentage of failed extubations (within 72 h) (0.5 vs. 1.4%; P = 0.02) as well as the number of readmissions to ICU (0.6 vs. 1.2%; P = 0.03) was significantly higher in the delayed extubation group than in the early extubation one (Fig. 1). No failed extubation was followed by a cycle of noninvasive positive pressure ventilation. The cumulative hazard of being discharged from postoperative ICU to the cardiac surgical ward overlapped for the first ICU day but increased significantly each day in the early extubation group (Fig. 2a) (e.g. on postsurgery day 3, an early extubation group patient showed a cumulative hazard of being discharged from ICU about 2.7 times higher than a delayed extubation one, 37.5 vs. 13.5%). The cumulative hazard of being discharged to rehabilitative ward was similar during the first 5–6 days from postoperative ICU discharge and increased significantly in the early extubation group (Fig. 2b) (e.g. on post-ICU day 5, early extubation group patients presented a cumulative hazard of being discharged from cardiac surgical ward to rehabilitative one 2.25 times higher than a delayed extubation one, 45 vs. 20.7%). Overall recorded hospital mortality was 3.9%. After stratifying mortality in relation to early extubation and delayed extubation cohorts, the former showed a significantly lower mortality (1.9 vs. 17%, P = 0.0001) than the latter. Kaplan–Meier plots show that cumulative survival was significantly higher in the early extubation group 90 days from operation Fig. 3.
Timely weaning of cardiac surgical patients may improve cardiopulmonary function, anticipate ambulation, lead to reduced ICU length of stay and save healthcare costs [21–24]. However, patients undergoing heart surgery often have multiple concurrent diseases that need to be adequately managed to reduce the risk of PMV. No consensus can be found in literature for the definition of PMV for cardiac surgical patients as it ranges range from 8 h to 7–14 days. For the purpose of this observational prospective study, we, reasonably, considered PMV as mechanical ventilation lasting over 12 h, as it is clinically reasonable and such value coincides with the median of total population [1–6]. Moreover, predicting the occurrence of PMV might help the cardiac anaesthetists to both optimize the preoperative patient management and rationalize ICU workload and costs [12,25,27] and the surgeons to better inform their patients about the risk–benefit balance associated with the intended surgical procedure. Second, some of the predictors of PMV identified may be preemptively managed to minimize their impact on clinical outcome. In this controversial field, our cohort study may provide new helpful insight. The multivariate model identified age older than 65 years, CCS or NYHA scores higher than 2, LVEF 30% or less, COPD, chronic renal failure, undergoing a redo or emergency surgery, CPB time more than 77 min and RBC or FFP transfusions more than four units as independent PMV predictors.
Clinical implications of the predictive model
Patients with a CCS and NYHA score higher than 2 are likely to suffer from coronary artery or valve diseases for a longer period of time, their clinical conditions are likely to be worse and they are likely to undergo heart surgery after one or more episodes of acute cardiac failure or during its onset. Renal function is expected to deteriorate in patients with level of creatinine clearance below 50 ml/min because of volume depletion and administration of nephrotoxic drugs (i.e. radiographic contrast for coronary artery angiography, antibiotics, vasoactive drugs, etc.) . CPB represents a nonphysiological circulation affecting peripheral tissue perfusion, and a long CPB time increases the likelihood of cardiac dysfunction and impairs oxygenation . The pump–oxygenator apparatus and its intrinsic processes of haemodilution, hypothermia and anticoagulation affect temporary physiologic homeostasis in organ system functions, exacerbating blood trauma, produces abnormal capillary membrane permeability and predisposes the patient to tissue anoxia and pulmonary complications [26–28]. CPB may initiate a systemic inflammatory response potentially harmful to the cardiac patient, determined by blood contact with foreign surfaces and responsible for the activation of multiple inflammatory mediators (e.g. complement, thrombin, cytokines, endothelin and endotoxins) that are likely to cause respiratory failure [28–31]. In about one-half of patients, mild pulmonary oedema is relatively common due to low plasma colloid osmotic pressure and elevated extravascular lung water (>10 ml/kg).
Moreover, pulmonary radiographic and ventilatory abnormalities may result, at least in part, from atelectasis rather than increased permeability oedema [28–31]. The variables, RBC or FFP transfusion units or both and redo and emergency surgery, are to be considered the strongest predictors, synergistically interacting, as ‘the formers’ are likely to be considered as ‘epiphenomenon’ of moderate-to-severe bleeding and the latter as a clinical indicator of a higher probability of bleeding correlated with greater surgical technical difficulty. In this sense, the higher the number of RBC and FFP transfusion units, the more likely they may be considered as clinical predictors of severe bleeding, surgical complexity, prolonged CPB time and reduced chances of reversing haemodilution. As a final synthesis, the older the age, the higher the probability for a patient to have concurrent risk factors affecting mechanical ventilation weaning.
Future consequences on our daily clinical practice
It is paramount to properly assess PMV predictors for both cardiac anaesthetist team and surgeons because of its strong impact on patient ICU and post-ICU length of stay and mortality. A patient on the ventilator for over 12 h was still correlated with a cumulative hazard of a total length of stay in ICU of 6 times higher than the early extubation group a week after surgery. To identify patients with any PMV predictor may help us to better preemptively manage them in order to rationalize the ICU workload and costs (each day of extra length of stay is associated with a median additional cost of about 1250€). To this day, we implemented a shared team policy of individuating patients with statistically significant preoperative and intraoperative predictors of PMV; a shared preoperative policy of better volume management in patients with renal insufficiency, increasing protection strategies towards preoperative radiographic contrast agents ; our strategies (e.g. use of intraoperative steroids such as 1 mg/kg of methylprednisolone or use of CPB circuit treated with surface-modifying additives or both) to control the inflammatory response correlated with cardiac surgery and CPB as much as possible [13,14,28–30].
Limitations of our cohort study
The most critical methodological weakness of our study was defining PMV. We opted for ‘12 h’ mainly because it seems clinically reasonable for such type of surgical patients. Unfortunately, we did not find any help from literature, in which no consensus can be found in terms of PMV definition. On the contrary, its reported values range from 8 h to 7–14 days [4–11], and such variation is too wide to fit the hypothesis that all approaches are appropriate. Methodological corroboration to the validity of our choice may come by univariate analyses showing that the two patient cohorts were significantly different in terms of almost all the considered perioperative variables.
Our study has two other weaknesses; the first is its observational nature that obviously does not allow us to get strong reproducible general conclusions, although the high number of considered patients permitted us to start modifying our practice, thus improving quality of assistance and reducing costs of hospitalization. However, the multivariate model finds its intrinsic validity both in the results of Hosmer–Lemeshow test and in the narrow CIs that give the predictors a good level of prediction accuracy. ROC curve analysis and Spearman's sensitivity test corroborated that as well. The second is that the high number of investigators (eight cardiac anaesthetists) could affect the validation of data entry as it might be most likely responsible of a higher percentage of data entry errors. We did not separate valve insufficiency from stenosis; however, patients with the former disorder may be different from the latter one in terms of clinical picture. A bias may be created in the model because we considered only the intraoperative RBC and FFP unit transfusions. We did not collect any perioperative data on patient haemodynamic status and management but LVEF and use of vasoactive drugs. Although LVEF may be considered as a good indicator of left ventricular performance, it is associated with numerous limitations as it is a volume computed on the basis of bidimensional measures, and its reproducibility is linked to both operator and echocardiography machine. Finally, LVEF may be absolutely normal even in some patients with heart failure, thus reflecting the limited sensitivity of the index itself.
Despite its limitations, our study suggests that in heart surgery, PMV is associated with higher hospital length of stay, mortality and, consequently, costs. Given the limited resources and the ever-increasing demand of medical care, the independent predictors of PMV identified by our cohort study may help optimize the perioperative management of cardiac surgical patients in order to achieve a higher quality of assistance.
We do thank Dr Miryam Shuman for her precious help in language consultancy.
1 Prakash O, Jonson B, Meij S, et al
. Criteria for early extubation after intracardiac surgery in adults. Anesth Analg 1977; 56:703–708.
2 Nilsson J, Ohlsson M, Thulin L, et al
. Risk factor and mortality prediction in cardiac surgery using artificial neural network. J Thorac Cardiovasc Surg 2006; 132:9–12.
3 Reddy SL, Grayson AD, Griffiths EM, et al
. Logistic risk model for prolonged ventilation after adult cardiac surgery. Ann Thorac Surg 2007; 84:528–536.
4 Ranucci M, Bellucci C, Conti D, et al
. Determinants of early discharge from the intensive care unit after cardiac operations. Ann Thorac Surg 2007; 83:1089–1095.
5 Rajakaruna C, Rogers CA, Angelini GD, Ascione R. Risk factors for and economic implications of prolonged ventilation after cardiac surgery. J Thorac Cardiovasc Surg 2005; 130:1270–1277.
6 Serrano N, Garcia C, Villegas J, et al
. Epidemiological Project for ICU Research and Evaluation (EPICURE). Prolonged intubation rates after coronary artery bypass surgery and ICU risk stratification score. Chest 2005; 128:595–601.
7 Oliver WC Jr, Nuttall GA, Orszulak TA, et al
. Hemofiltration but not steroids results in earlier tracheal extubation following cardiopulmonary bypass: a prospective, randomized double-blind trial. Anesthesiology 2004; 101:327–339.
8 Pappalardo F, Franco A, Landoni G, et al
. Long-term outcome and quality of life of patients requiring prolonged mechanical ventilation
after cardiac surgery. Eur J Cardiothorac Surg 2004; 25:548–552.
9 Dunning J, Au J, Kalkat M, Levine A. A validated rule for predicting patients who require prolonged ventilation post cardiac surgery. Eur J Cardiothorac Surg 2003; 24:270–276.
10 Gopalan PD, Burrows RC. Critical care of the vascular surgery patient. Crit Care Clin 2003; 19:109–125.
11 Branca P, McGaw P, Light R. Factors associated with prolonged mechanical ventilation
following coronary artery bypass surgery. Chest 2001; 119:537–546.
12 Engoren M, Buderer NF, Zacharias A. Long-term survival and health status after prolonged mechanical ventilation
after cardiac surgery. Crit Care Med 2000; 28:2742–2749.
13 HaR RH, Zacharias A, Engoren M. Determinants of prolonged mechanical ventilation
after coronary artery bypass grafting. Ann Thorac Surg 1996; 62:1164–1171.
14 Lo Cicero J 3rd, McCann B, Massad M, Joob AW. Prolonged ventilatory support after open-heart surgery. Crit Care Med 1992; 20:990–992.
15 Peters RM, Brimm JE, Utley JR. Predicting the need for prolonged ventilatory support in adult cardiac patients. J Thorac Cardiovasc Surg 1979; 77:175–182.
16 Sthor IM, Albes JM, Franke U, et al
. Outcome of patients after cardiac surgery transferred to other hospitals following prolonged intensive care stay. Thorac Cardiovasc Surg 2002; 50:329–333.
17 Geissler HJ, Holzl P, Kuhn-Régnier F, et al
. Risk stratification in heart surgery: comparison of six score systems. Eur J Cardiothorac Surg 2000; 17:400–406.
18 Kleinbaum DG, Mitchel K. Logistic regression a self learning text. 2nd ed. New York: Spinger Verlag; 2002.
19 Campeau L. Letter: grading of angina pectoris. Circulation 1976; 54:522–523.
20 The Criteria Committee of the New York Heart Association. Nomenclature and criteria for diagnosis of diseases of the heart and great vessels
. 9th ed. Boston, Massachusetts: Little, Brown & Co; 1994. pp. 253–256.
21 Miyamoto T, Kimura T, Hadama T. The benefits and new predictors of early extubation following coronary artery bypass grafting. Ann Thorac Cardiovasc Surg 2000; 6:39–45.
22 Cheng DC. Pro: early extubation after cardiac surgery decreases intensive care unit stay and cost. J Cardiothorac Vasc Anesth 1995; 9:460–464.
23 Bezanson JL, Deaton C, Craver J, et al
. Predictors and outcomes associated with early extubation in older adults undergoing coronary artery bypass surgery. Am J Crit Care 2001; 10:383–390.
24 Legare JF, Hirsch GM, Buth KJ, et al
. Preoperative prediction of prolonged mechanical ventilation
following coronary artery bypass grafting. Eur J Cardiothorac Surg 2001; 20:930–936.
25 Nakasuji M, Nishi S, Nakasuji K, et al
. Duration of dialysis is a significant predictor of prolonged postoperative mechanical ventilation in dialysis-dependent patients undergoing cardiac surgery. Anesth Analg 2006; 102:2–7.
26 Kouchoukos NT, Masetti P, Murphy SF. Hypothermic cardiopulmonary bypass and circulatory arrest in the management of extensive thoracic and thoraco-abdominal aortic aneurysms. Semin Thorac Cardiovasc Surg 2003; 15:333–339.
27 Rothenburger M, Tjan TD, Schneider M, et al
. The impact of the pro- and anti-inflammatory immune response on ventilation time after cardiac surgery. Cytometry B Clin Cytom 2003; 53:70–74.
28 Levy JH, Tanaka KA. Inflammatory response to cardiopulmonary bypass. Ann Thorac Surg 2003; 75:S715–S720.
29 Peterson ED, Coombs LP, Ferguson TB, et al
. Hospital variability in length of stay after coronary artery bypass surgery: results from the Society of Thoracic Surgeon's National Cardiac Database. Ann Thorac Surg 2002; 74:464–473.
30 Rothenburger M, Soeparwata R, Deng MC, et al
. Prediction of clinical outcome after cardiac surgery: the role of cytokines, endotoxin, and antiendotoxin core antibodies. Shock 2001; 16(Suppl 1):44–50.
31 Bond BR, Dorman BH, Clair MJ, et al
. Endothelin-1 during and after cardiopulmonary bypass: association to graft sensitivity and postoperative recovery. J Thorac Cardiovasc Surg 2001; 122:358–364.
32 Tepel M, van der Giet M, Schwarzfeld C, et al
. Prevention of radiographic-contrast-agent-induced reductions in renal function by acetylcysteine. N Engl J Med 2000; 343:180–184.