This is an ancillary analysis of a prospective cohort that aimed to study risk factor for vasoplegia in patients undergoing cardiac surgery with cardiopulmonary bypass. Patients were recruited between October 2017 and July 2021.The inclusion criteria were as follows: all adult patients undergoing cardiac surgery with cardiopulmonary bypass, covered by a social insurance system, who have given oral consent to participate after receiving full information and with at least one blood count at any of the time points of interest. Exclusion criteria were: cardiac graft, pre-operative extracorporeal life support or membrane oxygenation, mechanical cardiac assistance and patient deprived of liberty. Only patients included in Dijon university hospital were considered for the present analysis.

This report was drafted in accordance with the STROBE statement (5). This study was approved by the Institutional Review Board (IRB 00010254 - 2022 - 034). All patients received written information of their inclusion in the study.

Cytometry data were retrospectively extracted from the automated hematology analyzer (Sysmex XN-200, Sysmex, France). No blood sampling was mandatory for the study as blood cell count is usually monitored daily after cardiac surgery in our centre. Those data were merged with the data from the prospective cohort (ClinicalTrials.gov Identifier: NCT03281317). Three time points of interest were defined: pre-operative, post-operative (on the day of surgery) and 5 days after surgery. We choose the pre-operative and post-operative time points to capture the modifications of CPD induce by the surgery. Day 5 was chosen because it allowed to describe the evolution of CPD with the resolution of inflammation without too many lost too follow-up patients.

Blood cell count (CBC) was measured on an automated Sysmex XN2000 analyzer equipped with impedance and fluorescence flow cytometry devices. The optical system of the XN2000 employs a red diode laser producing a light beam. Cell permeabilization is performed by treatment with specific lysis reagents allowing specific dyes to enter the cells, where they bind to nucleic acids in the cytoplasmic organelles and the nucleus. Three mains signals are recorded ( Figure 1 ): forward-scattered light (FSC, cell size), side-scattered light (SSC, internal cell structure and granularity), and side fluorescence light intensity (SFL, reflecting DNA/RNA content). Based on the resulting fluorescence and scattered light characteristics, the cells can be separated. For each type of leukocyte (neutrophil, lymphocyte and monocyte), the cell count and CPD were extracted. Neutrophil to lymphocyte ratio (NLR) was calculated.

The CPD included:

The primary endpoint was inflammation-related complication or death occurring during the hospitalisation (6, 7). This endpoint was defined by a composite criteria (8) that included de novo supra-ventricular arrythmia, stroke, acute renal failure (defined according to KDIGO guidelines (9)) and/or death.

Baseline demographic data, per-operative and ICU data presented were collected prospectively. The EuroSCORE 2, a validated score for post-operative morbimortality of patients undergoing cardiac surgery was calculated (10). The American Society of Anesthesiologists physical status (ASA score) has been evaluated during the pre-operative consultation (11). Complications were monitored until hospital discharge. All those data were recorded in a data base (Excel, Microsoft). Cell population data are recorded prospectively by the automated hematology analyzer and were extracted retrospectively. The two data bases were merged in order to perform this analysis.

Normality was assessed graphically; quantitative data were presented as means (standard deviation/SD) or medians (interquartile range/IQR) and were compared using the Student t-test and the Kruskall-Wallis nonparametric test. Categorical, ordinal, and binary data were presented as frequencies and percentages and were compared using Chi-squared or Fisher’s exact tests if the conditions of validity were not fulfilled. A Bonferroni correction was applied to compensate for alpha risk inflation with repeated measurements. For longitudinal data, changing patterns of CPD with time were assessed using mixed linear modeling. ROC curve analyses were carried out for cell line parameters of interest. Optimal cut-off values were calculated using Youden’s method. An adjusted logistic regression analysis based on the EuroSCORE 2 was also carried out in order to determine whether those associations were independent of classical risks factors of post-operative morbi-mortality included in this validated score. Last, we developed a multivariable model combining the CPD variable of interest with the EuroSCORE 2 to predict post-operative complications. Data were split into a training set (60%) and a validation set (40%). The model was developed using logistic regression on the training data set. Variable selection was stepwise, AIC based. Then the model accuracy and area under the curve were calculated on the validation set. The Hosmer-Lemeshow Goodness-of-fit Test was also carried out. Missing data were considered to be missing at random and were omitted from the analysis. All analyses were performed using R software.

After exclusion of 121 patients ( Figure 2 ) data from 1453 patients had CPD available for at least one of the time points of interest and were analyzed. Out of those 1453 patients, CPD data were available for 639 patients at the pre-operative time point, 907 patients at the post-operative time points and 651 patients 5 days after surgery. Patients’ characteristics are described in Table 1 . Median age was 69 [IQR 61;75], most patients were men (76%) with an ASA score of 3 (71.6%) and the median EuroSCORE 2 was 2.23 [IQR 1.19;4.60]. Sixty eight (4.68%) patients had sepsis, mostly due to endocarditis (76%). Among the 636 (43.8%) patients that developed the composite criteria, 342 (23.6%) developed renal failure, 384 (26.4%) de novo supra ventricular arrhythmia, 58 (4.0%) stroke and 81 (5.6%) died.

The neutrophil count ( Figure 3A ), fluorescence granularity (NE-SCC), intensity (NE-SFL) and size (NE-FSC) increased with surgery ( Figures 4A, C, E respectively). Heterogeneity of neutrophils decreased in terms of fluorescence granularity (NE-WX) and size (NE-WZ) but increased in terms of intensity (NE-WY) ( Figures 4B, D, F respectively).

Patients who developed the composite complication score had higher pre-operative neutrophil counts (5.08 [3.89;6.95] vs 4.76 [3.60;6.13] p = 0.02) and more heterogeneous neutrophils in terms of fluorescence granularity (NE-WX, p< 0.01) and intensity (NE-WY, p< 0.01) ( Figures 4B, D and Table 2 ). Immature granulocytes and the neutrophil to lymphocyte ratio (NLR) increased with cardiac surgery, and pre-operative NLR was associated with occurrence of post-operative complications (p<0.01 for both) ( Supplementary Table 1 ). Using multivariate analysis, NE-WX and NE-WY were associated with post-operative complications independently of the EuroSCORE 2 (adjusted odd ratio of 1.01 [1.00;1.02] and 1.01 [1.00;1.01] respectively) ( Table 2 ).

The lymphocyte count decreased with surgery ( Figure 3B ). While fluorescence granularity (LY-X) and size (LY-Z) increased, lymphocyte fluorescence intensity decreased (LY-Y) ( Figures 5A, C, E ). Lymphocytes were less heterogeneous in terms of their granularity, intensity and size after surgery (i.e., lower LY-WX, LY-WY and LY-WZ respectively).

Patients who developed post-operative complications had lower pre-operative lymphocyte counts (1.49 [1.10;1.14] vs 1.81 [1.39;2.39], p<0.01) and fluorescence granularity (LY-X, p<0.01)) ( Figure 5 ). Using multivariate analysis, LY-X was associated with the occurrence of post-operative complications, independently of the EuroSCORE 2 (adjusted odd ratio of 1.08 [1.02;1.15]) ( Table 2 ).

The monocyte count was unchanged after cardiac surgery ( Figure 3C ). There was no difference between patients developing complications and others in terms of pre-operative monocyte cell parameters ( Supplementary Table 2 ).

After stepwise variable selection (training data set), EuroSCORE 2, NE-WY and LY-X were kept in the model (Supp file 3). The accuracy of this model (testing data set) was of 0.65 [0.59;0.71] and the Area under the curve (AUC) of 0.63 [0.58;0.68]. The p-value of the Hosmer-Lemeshow Goodness-of-fit Test was of 0.33. For the EuroSCORE 2 alone the accuracy was of 0.63 [0.56:0.69] and the AUC of 0.60 [0.55;0.65].