This prospective non-randomized study was conducted between November 1995 and May 2001 following approval by the ethical committee of the medical faculty at the University of Leipzig, Germany. A total of 75 patients who underwent cardiac surgery with CPB were analyzed [inclusion criteria: aged 3–18 years, body weight >12 kg; exclusion criteria: missing informed consent of parents, palliative cardiac surgery (e.g. if single ventricle circulation was the aim of surgery, (Glenn, Fontan or total cavopulmonary connection [TCPC]). The surgical procedures included were: closure of atrial septal defect (n = 39) or ventricular septal defect (n = 11); replacement of pulmonary valve by an allogeneic heart valve (n = 18); resection of an aortic subvalvular stenosis resulting from a subaortic membrane or fibrous cap (n = 6); correction of tetralogy of Fallot (n = 1). All children received similar anesthesia, medication and intraoperative and postoperative care and CPB as detailed elsewhere [2]. After delivery to the intensive care unit postoperatively, the incidence of pericardial-, pleural- and/or abdominal-effusion was monitored by echocardiography, chest X-ray or sonogra-phy. If patients developed detectable effusions after removal of the thoracic drainage (which was usually one day after surgery) until discharge they were allocated into the POEE group (n = 29), or into the non-POEE group (no effusion, n = 46). As evaluated visually, all POEE patients had edema of the face and/or hands and/or feet. Incidence of edema was not used for POEE discrimination because quantitative measures of extravascular body fluid volume (such as scintigraphy following labelling of the extravascular fluid by radiolabelled sulphide or bromide) were ethically not feasible in children. Massive generalized edema, CLS or MOD as defined by Seghaye et al. [24] was not observed in any of the patients. However, 65% of the POEE patients fulfilled at least one MOD criterion as defined by Trotter et al. [23]. Postpericar-diotomy syndrome with effusions and fever of non-infectious origin within a week, or later, of surgery [25] was not present in any of the patients.

Blood was obtained one day (median: 20 hours) before surgery in untreated tubes as well as in ethylenediamine-tetracetic acid (EDTA) and heparin tubes, centrifuged at 2800 g for 10 min at 4°C and the supernatant was collected. Urine was sampled in untreated tubes. Within 1 hour after collection, serum, EDTA-plasma and urine samples were stored in aliquots at -80°C. The concentration of the complement components (C3, C4, C5, C1-inhibitor, C3d) and immunoglobulin (Ig)G2 was determined by radial immune diffusion (The Binding Site, Heidelberg, Germany) with serum or EDTA-plasma (C3d) and total hemolytic complement CH100 by lysis of antibody-coated sheep erythrocytes (The Binding Site). All other parameters were quantified using enzyme-linked immunosorbent assay [IgE, interleukin (IL)-1β, TNF-α, interferon-γ, RANTES, histamine: Beckman-Coulter, Krefeld, Germany; IL-4, IL-10, IL-13, soluble intracellular adhesion molecule-1 (sICAM-1), platelet endothelial cell adhesion molecule-1 (PECAM): Bender MedSystems, Vienna, Austria; IL-5, IL-6 high sensitivity, IL-10 high sensitivity, IL-12 p40/p70, soluble leukocytic (sL)-selectin, sE-selectin: R&D Systems GmbH, Wiesbaden, Germany; IL-2, IL-2-receptor, serum and urine neopterin: DPC Biermann GmbH, Bad Nauheim, Germany; IL-4 high sensitivity, IL-11: Natutec, Frankfurt, Germany; IL-12 p70, IL-13: Biozol Diagnostica Ver-trieb GmbH, Eching, Germany; C5a: Behringwerke AG, Marburg, Germany]. The complement fragment ratios C3d/C3, C5a/C5 and immunoglobulin ratio IgE/IgG2, were calculated as measures for complement activation and Th2/Th1 imbalance, respectively. Additionally, routine laboratory and clinical chemistry parameters were determined (cell count, differential blood count, CRP, creatinine, electrolytes, protein, hematocrit, blood coagulation parameters). In total 56 parameters were analyzed per patient including age, gender and body weight.

Data are displayed as mean ± standard deviation (SD). Between-group comparison was undertaken by unpaired Student's t-test or Mann-Whitney U-test as appropriate [Statistical Program for Social Sciences Version 8.0 (SPSS), Knowledge Dynamics, Canyon Lake, TX]. Discrimination of patients into the POEE and control group was tested by data pattern analysis using two different methods as detailed [18]. Classification for individual risk assessment was performed by stepwise multivariate discriminant analysis using SPSS. This classifier was optimized by increasing the F-probability followed by determination of the unstandardized canonical discriminant function. Missing data were substituted by column means, if necessary. No more than one value per patient was extrapolated. In parallel, the triple matrix data pattern analyzer CLASSIF1 [18] was used as an algorithmic data mining approach. With CLASSIF1 no replacement of missing data values and no mathematical assumptions on parameter distributions are required.

Preoperative serological alteration or activation indicates specific pathobiochemical problems. The parameters selected by the two classifiers in this study indicate increased POEE risk for patients with elevated inflammatory response by increased complement and neutrophil activation and coagulation (see Table 4). In different cardiac situations, CRP [15], sE-selectin [16], sICAM-1 and neutrophil adhesion molecule expression [28,29] have been discussed as risk factors. As already suggested by others [19], preoperatively altered blood coagulation values such as partial thrombin time were found to be prognostic for postoperative blood loss. Fibrinogen and fibrin are ligands for Mac-1 [30], inducing neutrophil, monocyte or resting platelet activation. Our study indicates this activation by elevated sL-selectin level as an important discriminant parameter. CPB is associated with major qualitative and quantitative alterations of humoral pathways and changes in leukocyte subsets, generating a systemic inflammatory response [2,4,7,31] with interactions between vascular endothelium, platelets and leukocytes including signal exchanges, adhesion molecule expression and secretion of cytokines or chemokines in a multi-step process. Patients with an altered immune profile before surgery might show a more pronounced or sustained immune response after surgery. In an unstimulated immune system, CPB exposure constitutes the initial stimulus that might prime the system for postoperative complications [32]. In patients with a primed or predisposed immune profile, CPB as the second stimulus may facilitate an enhanced immune response, which, in turn, may lead to POEE, CLS or multiple organ failure.

The main discriminators of at risk patients (elevated levels of complement and activated complement components, TNF-α and IL-10) indicate the significance of complement system and monocyte activation. Activated monocytes liberate TNF-α and IL-10 as important modulators of the inflammatory response. TNF-α stimulates human vascular endothelium, thus mediating leukocyte recruitment to sites of inflammation. IL-10 release is specific to CPB surgery [7] and patients with POEE or MOD release higher quantities of IL-10 [7,23]. Increased IL-10 release as an indicator of MOD or effusions is also supported by the finding that perioperative methyl-prednisolone administration, that enhances IL-10 release during CPB surgery in adults [33], aggravates postoperative effusions and bleeding in children with postcardiotomy syndrome [34]. Elevated preoperative IL-10 concentration was a risk factor in our patients. An observation that contrasts with the finding that children with MOD had reduced IL-10 serum levels [23] prior to CPB. Patients from our study had no gender-related differences in any of the analyzed laboratory parameters. We have no explanation for this discrepancy, but differences in the age distribution and the congenital heart diseases of patients included in our study, as compared to Trotter et al. [23], may play a role.

Severe allergic reactions with cardiac surgery [6,34] and allergic predisposition in adults at risk for cardiovascular death have been reported [35]. The interpretation of allergic/atopic predisposition in POEE risk was indicated by our recent observation of elevated leukocyte function asscoiated molecule-1 (LFA-1) expression on leukocytes of at risk patients [18], as LFA-1 expression is increased on leukocytes of atopic children [36]. We reported earlier that patients at risk for POEE also had increased preoperative histamine and eosinophil counts, among others [7,29]. The results from the present study do not clearly support the hypothesis of risk prevalence for atopic/allergic patients because only few of the selected markers could indicate an atopic/allergic predisposition (e.g. TNF-α and IL-10). We conclude from these differences that both increased inflammatory status and allergic/atopic predisposition are predictors of increased POEE in children.

Taken together, the data indicate at least three risk groups for pediatric POEE. Risk patients might have: (i) latent infection; (ii) atopic/allergic predisposition; or (iii) immune alterations as a result of the congenital heart disease. These hypotheses have to be further scrutinized by future studies.

Because children with postoperative complications usually have a longer stay on the ICU, a longer period of mechanical ventilation and stay longer in hospital, preoperative risk assessment is of clear therapeutic advantage and can be cost-effective by reducing any stay in intensive care. By prospective classification, up to 86% of the patients at risk were correctly identified preoperatively. In view of the fact that such predictions were not possible at all until now, these predictive values are promising. However, the classifier will be optimized by increasing the number of patients enrolled in studies and by combining this serological classifier with additional parameters such as flow-cytometric data [18].

Individual risk assessment before cardiac surgery of this type might open new ways to develop individual treatment strategies with two possible clinical consequences: first, postponement of surgery until the normalization of clinical parameters (e.g. elimination of stress or a latent infection); and, second, application of individual prophylaxis [31] in the case of endogenous reasons for immune system alterations [28,34,35]. The hypothesis that postponement or individual prophylaxis will reduce POEE has to be scrutinized in additional studies.

The proposed serological classifier should permit individual risk assessment in hospitals with lower patient numbers. It is planned to set up and optimize an on-line classifier for POEE risk assessment on the internet. One of the practical consequences of this would be that diseases could be categorized at institutions where no sufficient database can be generated in a reasonable time period. Risk assessments for patients at other institutions can be calculated for test purposes using the indicated SPSS classifier formula (Table 4). Each required parameter value is multiplied with a local data correction factor. The local data correction factor is obtained as a ratio between the parameter mean from non-POEE patients from Table 2 of this study and the mean of the respective parameter from the local non-POEE group of 20 to 40 complication-free patients. The local data correction factor for the establishment of the individual patient's triple matrix for the CLASSIF1 classification is determined in the same way.