Presepsin values and Prognostic Nutritional Index predict mortality in intensive care unit patients with sepsis: a prospective observational study

Background Sepsis is a major cause of mortality for critically ill patients. This study aimed to determine whether presepsin values can predict mortality in patients with sepsis. Methods A total of 83 adult patients diagnosed with sepsis were prospectively examined. Presepsin values were measured immediately after ICU admission and on Days 2, 3, and 5 after ICU admission. Glasgow Prognostic Score, neutrophil to lymphocyte ratio, platelet to lymphocyte ratio, Prognostic Index, and Prognostic Nutritional Index were also determined at baseline. For category classication, total scores were calculated (hereafter, “inammation-presepsin scores [iPS]”) as follows: a score of 1 was assigned if the presepsin value and inammation-based prognostic scores at baseline were above cut-offs determined by receiver operating characteristic (ROC) curve analysis for 28-day mortality; a score of 0 was assigned if they were below the cut-offs (total score range, 0–2 points). Presepsin values, inammation-based prognostic scores, and iPS were compared between non-survivors and survivors.


Background
The Third International Consensus De nitions Task Force (Sepsis-3) updated their criteria for de ning sepsis in 2016 [1]. Sepsis is de ned as a syndrome that can lead to lethal organ dysfunction caused by a dysregulated host response to infection. Despite efforts to improve early recognition and advances in antibiotics, sepsis remains a major cause of mortality in critically ill patients [2]. Sepsis can evolve to lethal multiorgan failure, for which an increase of ≥ 2 in the Sequential Organ Failure Assessment (SOFA) score is associated with an in-hospital mortality rate as high as 10% [1]. To reduce its high mortality rate, early goal-directed management, i.e., early recognition of severe sepsis, and early initiation of aggressive and supportive therapy are crucial [3,4]. To this end, several biomarkers have been adopted in clinical settings. Procalcitonin (PCT) is a diagnostic marker used for bacterial sepsis with the highest speci city and is widely adopted in clinical settings. However, PCT values can be elevated by non-septic conditions such as severe trauma, invasive surgery, and critical burn injuries, which can in turn lead to false-positive results [5][6][7]. Presepsin, a subtype of soluble CD14 (CD14-ST), is a 13-kDa truncated N-terminal fragment of CD14 [8] which is released into the circulation during monocyte activation upon the recognition of lipopolysaccharide from infectious agents [9]. Presepsin values are correlated with severity of sepsis [10] and are useful for predicting prognosis and monitoring the disease course for sepsis [9,11]. Presepsin can be measured in less than 17 minutes by a chemiluminescent enzyme immunoassay (CLEIA) with a compact fully automated immunoanalyzer (PATHFAST®; Mitsubishi Chemical Medience Corporation, Tokyo, Japan) [12] and has been used as a biomarker to diagnose bacterial sepsis.
In ammation-based prognostic scores, including the Glasgow Prognostic Score (GPS; based on serum Creactive protein (CRP) and albumin levels), neutrophil to lymphocyte ratio (NLR), platelet to lymphocyte ratio (PLR), Prognostic Nutritional Index (PNI; based on albumin and lymphocyte counts), and the Prognostic Index (PI; based on serum CRP and white blood cell counts), have been adopted as signi cant prognostic biomarkers for several types of cancer [13]. However, no study has assessed the association of mortality with presepsin values alone or in combination with in ammation-based prognostic scores (i.e., GPS, NLR, PLR, PNI, and PI) in patients with sepsis as de ned according to the Sepsis-3 de nition.
Accordingly, the present study aimed to test the following hypotheses: 1) presepsin is a useful predictor for early and long-term mortality in sepsis patients; 2) presepsin values are superior to in ammationbased prognostic scores for predicting mortality, and; 3) the ability to predict mortality can be improved by combining presepsin values and in ammation-based prognostic scores, compared to using presepsin values alone.

Patients and study design
This single-center prospective study was conducted in a 16-bed ICU. The study protocol was approved by the Ethics Committee of Osaka Medical College (Osaka, Japan). In total, 83 adult patients older than 18 years who were diagnosed with sepsis according to the Sepsis-3 de nition [1] and admitted to the ICU were prospectively examined from December 2017 to August 2019. Written informed consent was obtained from all patients enrolled in this study or their families. Patients who exhibited clinical evidence of anti-in ammatory conditions were excluded, including those undergoing immunosuppressive therapy (e.g., chemotherapy, chronic steroid use, autoimmune disease treatment) within one month of the study.
Arterial blood samples were collected from each patient and presepsin values were measured immediately after ICU admission (baseline) and on Days 2, 3, and 5 after ICU admission. Mortality was categorized as 28-, 60-, 90-, and 180-day mortality.
GPS, NLR, PLR, PI, and PNI were examined at baseline. For category classi cation, total scores were calculated (hereafter, "in ammation-presepsin scores [iPS]") as follows: a score of 1 was assigned if the presepsin value and in ammation-based prognostic scores at baseline were above cut-offs determined by receiver operating characteristic (ROC) curve analysis for 28-day mortality; a score of 0 was assigned if they were below the cut-offs (total score range, 0-2 points). As an example of nomenclature, the combination of presepsin value and NLR is presented as "iPS-NLR." Presepsin values, in ammationbased prognostic scores, iPS, and changes (Δ) in presepsin values relative to baseline values at each sampling point were compared between survivors and non -survivors.

Laboratory assessments
Arterial blood samples collected from sepsis patients at each sampling point were immediately subjected to presepsin measurement. Presepsin concentration was measured by CLEIA (PATHFAST®; Mitsubishi Chemical Medience Corporation, Tokyo, Japan) [12]. Assay results were obtained within 17 minutes. This assay can detect systemic infection (sepsis) within the range of 300 to 500 pg/ml.

Statistical analysis
Categorical data are reported as percentages and compared using Fisher's exact test. Continuous data are reported as medians with inter-quartile ranges and compared using the Mann-Whitney U test. ROC curves were generated for presepsin values, in ammation-based prognostic scores, iPS, and Δpresepsin, and areas under the curve (AUCs), cut-off values, sensitivities, and speci cities were calculated. Presepsin values, in ammation-based prognostic scores, iPS, Δpresepsin, SOFA, and quick SOFA (qSOFA) (with P < 0.05 in univariate analysis) were examined further by multivariate logistic regression analysis for predicting mortality associated with sepsis. P < 0.05 was considered statistically signi cant. JMP software version 11.0.0 (SAS Institute Inc., NC, USA) was used for all statistical analyses.

Discussion
Severe sepsis is a syndrome accompanied by lethal organ dysfunction due to dysregulated host responses to infection, and is associated with a mortality rate higher than 50% [14]. Sepsis initiates the activation of both pro-and anti-in ammatory responses [15], both of which are ampli ed by nonimmunologic pathways including cardiovascular, neuronal, autonomic, hormonal, bioenergetic, metabolic, and coagulation pathways [16][17][18]. Early diagnostic and prognostic biomarkers with high sensitivity and speci city are crucial for the early identi cation of high-risk sepsis patients. PCT is a diagnostic marker used for bacterial sepsis with the highest speci city and widely used in clinical settings, but can give false-positive results as its values can increase even during non-septic conditions (e.g., severe trauma, invasive surgical procedure, critical burn injuries) [5][6][7]. These false-positive results may be obstructive toward obtaining information on whether or not to initiate aggressive and supportive therapy in sepsis patients or other critically ill patients within the rst few hours after ICU admission. Presepsin, a subtype of CD14-ST, is secreted from granulocytes by infectious stimuli in an animal sepsis model [19] and can thus be used as a highly speci c biomarker for diagnosing bacterial sepsis. Presepsin values are useful for diagnosing early stage sepsis in emergent patients and are independently associated with the degree of organ dysfunction, coagulation disorders, and ICU mortality [20,21].
The present study demonstrated that presepsin is a better predictor of mortality in septic patients than SOFA or qSOFA (Tables 2, 3, 5, 6 and 8). Furthermore, the AUC of presepsin for predicting mortality showed a range of 0.73-0.91, indicating a higher predictive ability than that observed for iPS and in ammation-based prognostic scores alone ( Table 2). AUC values of presepsin for predicting 28-day mortality were all higher than each of the presepsin AUC values for 60-, 90-, and 180-day mortality.
Moreover, multivariate analysis revealed that only the presepsin value on Day 1 predicted mortality rates. Collectively, these ndings suggest that the AUC of presepsin on Day 1 might be effective in predicting early mortality (Tables 2, 5).
We also investigated the ability of presepsin values on Day 1 to predict early mortality (i.e., death within 7 days of ICU admission). The AUC of presepsin on Day 1 to predict death within 7 days of ICU admission was 0.84, suggesting a greater predictive ability than that observed for presepsin on Day 1 for death after 7 days since ICU admission and in ammation-based prognostic scores alone, iPS, SOFA, and qSOFA for death within 7 days of ICU admission (Tables 6, 7). Multivariate logistic regression analysis for mortality found the presepsin value on Day 1 to be a signi cant independent predictor of death within 7 days of ICU admission (OR, 1.0007; 95%CI, 1.0001-1.0013; p = 0.032; Table 8). Ulla et al. reported that a higher level of presepsin at presentation in the emergency department was correlated to 60-day mortality. [22] Masson et al. found that presepsin levels measured at the time of presentation among patients with severe sepsis or septic shock were higher in non-survivors than in survivors [23], and our ndings were consistent with these results. Masson et al. also found signi cantly higher presepsin levels on Days 2 and 7 in non-survivors when compared with survivors. In the present study, presepsin levels among nonsurvivors were signi cantly higher on all days relative to survivors (Figs. 1, 2).
In the present study, presepsin cut-off values for predicting each designation of mortality were higher than previously reported cut-offs for diagnosing either sepsis (300 to 500 pg/ml) or severe sepsis (500 to 1926 pg/ml [26]. These results suggest the need to adopt a higher presepsin cut-off value for predicting mortality associated with sepsis. The presepsin values for ΔDay 3-Day 1 in 28-or 180-day mortality showed higher speci city for predicting 28-or 180-day mortality compared with presepsin values on Days 1, 2, 3, and 5 alone for all mortality ( Table 2). These ndings suggest that Δpresepsin Day 3-Day 1 may be useful as a "rule in" test for predicting mortality. Log-rank test also revealed Δpresepsin Day 3-Day 1 and Δpresepsin Day 2-Day 1 to be independent predictors of mortality in septic patients (Table 3). Presepsin values on Days 3 and 5 had higher AUCs for predicting 28-day mortality than the presepsin values on Days 1 and 2 ( Table 2). These results suggest the importance not only of measuring presepsin values at the time of ICU admission but also monitoring its changes on Day 2 and thereafter following ICU admission, in order to predict mortality in sepsis patients.
The AUC, sensitivity, and speci city of PNI for predicting 28-day mortality were 0.73, 70%, and 80%, respectively, corresponding to a greater AUC than that observed for in ammation-based prognostic scores alone, iPS, SOFA, and qSOFA ( Table 2). Log-rank test revealed PNI as an independent predictor of all mortality in sepsis patients (Table 3). Previously, we found that NLR is superior to other in ammationbased prognostic scores in predicting mortality among patients with gastrointestinal perforation and pneumonia [27,28]. With regard to 28-day mortality, however, the present study found PNI to be a predictor of mortality in sepsis patients. Pinato et al. [29] reported that PNI is a predictor of poor overall survival in patients with hepatocellular carcinoma. In the Albumin Italian Outcome Sepsis trial, albumin replacement did not improve survival at 28 and 90 days among patients with severe sepsis or septic shock [30]. However, our ndings suggest that, of the many variables that can be investigated for assessing in ammation, hypoalbuminemia and lymphocytopenia (albumin and lymphocyte counts used to calculate PNI) are important variables for predicting mortality in sepsis patients. In the context of predicting 28-day mortality, PNI shows a lower AUC than that of presepsin, but in clinical environments where presepsin values cannot be measured, PNI at presentation may be useful for determining 28-day mortality. PNI can be obtained at low cost and rapidly, and provides crucial information needed to begin aggressive and supportive therapy in sepsis patients within the rst few hours of ICU admission.
No previous study has assessed the association between presepsin values, PNI, and mortality in patients with sepsis according to the Sepsis-3 de nition. Our ndings are novel in this respect. Further studies will be needed to clarify how presepsin values and PNI can predict mortality in sepsis patients.
This study has several limitations. First, the present study was a single-center study with a small cohort.
Second, we used a single biomarker, and no comparisons were made with other biomarkers. Third, presepsin values increase with declining renal function [31]. Thus, the diagnostic accuracy of presepsin values may be in uenced by kidney function.

Conclusions
Presepsin was found to be a predictor of mortality in patients with sepsis. Presepsin is a better predictor of mortality in sepsis patients than SOFA or qSOFA. In particular, the presepsin value on Day 1 is useful for predicting early mortality (i.e., death within 7 days of ICU admission). In addition to measuring presepsin values at the time of ICU admission, any changes in these values on Day 2 and thereafter following ICU admission should be monitored in order to predict mortality in sepsis patients. In the context of 28-day mortality, PNI was found to be a predictor for mortality in sepsis patients. Further studies aimed at understanding the exact role of presepsin values and PNI in predicting mortality in sepsis patients are warranted.   Comparison of presepsin values (Day 1) for predicting early mortality in sepsis survivors versus nonsurvivors.