98 R. Murray et al.: J Extra Corpor Technol 2025, 57, 96 – 104
Plasma sample collection and processing
Plasma samples were collected prospectively in EDTA tubes, centrifuged within 30 min at 2500g at 4 ° C for 15 min to make platelet poor plasma, aliquoted into 200 microliter samples and frozen at �80 ° C until analysis. This included a pre-MCS sample, samples from each day of MCS for the first 7 days and weekly samples after the first week of MCS. Samples were typically obtained in the morning( between 8 and 10 AM), in conjunction with clinically scheduled lab draws. The biomarkers sTREM-1, Presepsin, CRP and Procalcitonin were analyzed utilizing commercially available kits. Plasma concentrations of sTREM-1, Presepsin and Procalcitonin were determined in duplicate for each sample using R & D systems human Trem-1 ELISA kit( Catalog number: DTRM10C), LSBio human Presepsin ELISA kit( Catalog number: LS-F55886) and Abcam human Procalcitonin ELISA kit( Catalog number: ab100630), respectively. Plasma samples which yielded a reading over the calibrator range were re-assayed under appropriate dilution. The reliability of duplicate ELISA analysis was evaluated by calculating the coefficients of variation( CV) for each test point. CV values were categorized as follows: < 10 %( excellent), 10 – 15 %( acceptable), and > 15 %( outliers). No outliers( CV > 15 %) were identified for sTREM-1 or Presepsin, while one Procalcitonin sample exceeded this threshold. CVs were calculated using Microsoft Excel. CRP was quantified by chemiluminescence using the Immulite 1000 automated chemiluminometer( Siemens Healthcare Diagnostics, Deerfield, IL) with the high-sensitivity CRP kit( Siemens, LKCR1). All concentrations were calculated using GraphPad 9.0 based on four parameter logistic curve fit.
Statistical analysis
Data is summarized as mean( standard deviation) or median( inter-quartile range and / or range) for continuous variables and frequency( percentage) for categorical variables. Graphical displays of biomarker values over the monitoring timeframe for each biomarker are included. Biomarker values are additionally summarized and presented by MCS modality.
Analysis of biomarker kinetics
The difference between pre and immediately post cannulation values( day 1 of MCS support only) for each biomarker were compared using Wilcoxon rank sum test, while differences between MCS types were compared using Kruskal- Wallis rank sum test. The differences in each biomarker between pre-MCS and uninfected post MCS initiation time points( all days from: day 1 of MCS support to 48 h prior to the diagnosed infection) were tested using linear mixed-effects models for each biomarker.
Analysis of biomarker response to infection
The differences in each biomarker between infected and uninfected time points was tested with linear mixed-effects models. For this analysis we utilized each patient as their own control, with uninfected time points for each patient defined as samples > 48 h in advance of the diagnosed infection while on MCS, and compared biomarker values to the infected time period [ 21 ]. The infected time period for each patient was defined as( the 48 h prior to the time the new positive culture was obtained). This time period, 48 h, was chosen as this is the time in which between 93 and 99 % of cultures will be positive if they will grow a pathogenic organism [ 22 – 25 ]. This time period also represents the potential delay of knowing when a patient has a new infection based on the current gold standard diagnostic mechanism of infection: cultures. Clinically, if a culture is negative at 48 h we would be likely to discontinue antimicrobial therapy, thus this time is the period in which the identification of a biomarker that can predict infection reliably on MCS has the most potential benefit. The linear mixed effects models include random intercepts to account for repeated measures within each patient. Youdens J statistic was used to determine estimated cut points for each biomarker, in an exploratory fashion, with the cut points unadjusted for other covariates [ 26 ]. One patient was found to have much higher presepsin and procalcitonin values compared to other included patients, however met inclusion criteria and on chart review was not found to be different than the other patients in the cohort. A sensitivity analysis was performed, and their removal from the model did not alter results, therefore data points from this patient were included in the overall analysis.
Hypothesis testing was conducted at a 5 % type I error rate( alpha = 0.05). All analyses were conducted in R( Version 4) and all plots were made with ggplot2 [ 21, 27, 28 ]. Some patient time points are missing biomarker values due to the timing of infection, absence of sufficient residual sample to run all four biomarkers, or insufficient sample volumes to repeat analysis at appropriate dilutions. Time points that are missing samples are clearly identified within the Supplemental Tables 1 – 4.
Results
The existing biobank of 81 patients contained 20 patients who developed a new infection while on MCS. Two patients did not have sufficient residual samples for analysis, leaving a total of 18 patients in the final cohort. The demographics of these 18 patients and details about their acquired infections on MCS are presented in Table 1. There were 9 female patients( 50 %), most patients were in the cardiac intensive care unit( 11 / 18, 61 %), with 1 patient( 6 %) in the neonatal intensive care unit and 6 patients( 33 %) in the pediatric intensive care unit, with 3 patients( 17 %) transitioned to MCS following cardiac surgery due to an inability to separate from cardiopulmonary bypass. The majority of patients were able to be separated from MCS for more than 24 h( 11 / 18, 61 %), however 12 / 18( 67 %) of patients did not survive until hospital discharge.
Biomarker values pre and post cannulation are presented in Table 2, with no differences identified before and after the cannulation event. We did not identify a difference in precannulation levels for any of the four biomarkers based on mode of support( Supplemental Table 5).
Pre-cannulation values of each biomarker were compared to average post-cannulation uninfected values, using liner mixed effects modeling. On average post-cannulation CRP was