Sepsis represents a significant challenge in clinical wards, especially in the intensive care unit (ICU). Its early recognition is hampered because it is not a definite disease but a syndrome with different pathogen and host factor-associated symptoms (1). However, the time of intervention is fundamental, and a delayed diagnosis is associated with high mortality and morbidity.

Although blood culture has been considered the gold standard for diagnosing sepsis for decades, it has several limitations, including low diagnostic sensitivity with a high rate of false negatives, preanalytical issues, and possible contamination (2).

Literature evidence suggests that circulating biomarkers significantly reflect the systemic host response to infection. Thus, they support the Clinicians in appropriately managing the patient with sepsis (3).

Over the years, spasmodic research for discovering the ideal sepsis biomarker has endured, concluding that no single molecule with optimal sensitivity and specificity for sepsis lasts. Thus, the dream of finding “the troponin” for sepsis is slowly fading. However, promising biomarkers for sepsis screening are emerging. Among these, monocyte distribution width (MDW) has gained much attention in the last few years (4–13). It is a measure of the monocyte’s anisocytosis, and it is calculated by the last generation hemocytometers of Beckman Coulter. It has a great advantage over the other sepsis biomarkers being part of the complete blood count. Another interesting biomarker is presepsin (PSP). PSP, also named soluble CD14 subtype, is the N-terminal fragment of soluble CD14 released from monocyte/macrophage upon activation through proteolysis and exocytosis (14, 15).

Both MDW and circulating PSP levels reflect the monocytes/macrophages activation, which represents a key mechanism in sepsis pathogenesis, especially during the early stages (16).

Beyond early detection of sepsis, monitoring sepsis patients is another critical issue. In such a contest, the kinetic of sepsis biomarkers could provide precious information.

In this study, we explored the accuracy of two new sepsis biomarkers, MDW and PSP, compared to traditional ones, C-reactive protein (CRP) and Procalcitonin (PCT), to predict ICU mortality by analyzing their kinetic at different time points during ICU stay.

In this study, 104 patients (M:F 56%:44%, median [IQR] age 70 [59–78] years) were enrolled (Figure 1). Among these, 30 (29%) had a clinical diagnosis of sepsis while the remaining (71%) were admitted for the following causes: post-operative (n = 20), medical (n = 40), surgical (n = 3), trauma (n = 4), and neurological (n = 7). Table 1 reports levels of CRP, PCT, MDW, and PSP measured at 4 timepoints, respectively, basal (t0), at 24 h (t24), 48 h (t48) and 72 h (t72) after the admission at the ICU in the whole study population and subgrouped according to the presence of sepsis. Patients with sepsis displayed higher levels of PCT and PSP than patients without sepsis at each timepoint; differently, CRP displayed statistically significant differences only at t0, while MDW only at t0 and t24 (Table 1). Noteworthy, the half-life of CRP, PCT and PSP are 20–24 h, 12–24 h, and 3–4 h, respectively.

Data on biomarker trends during the first 72 h, in the whole sample or the sepsis groups, are reported in Table 2. CRP and PCT increased at 24 h, while at 48 h started to decrease (Table 2 and Figure 2); differently, for MDW, a decreasing trend was evident (from t0 to t72), while PSP peaked at 48 h and remained elevated at 72 h (Table 2 and Figure 2).

Overall mortality within the ICU was 44%, with a median survival time of 26 days (IQR 52–12). Although the sepsis group displayed a higher mortality (52 vs. 40%; Fisher’s p = 0.376) and a lower survival time (median 20 vs. 29 days; log rank test p = 0.085) than patients without sepsis, the difference was not statistically significant.

Table 3 reports univariate and multivariate Cox regression for mortality, considering predictors measured at different timepoints. At t0, age was the only independent predictor (p = 0.040), while at t24 only MDW (marginally, p = 0.05) and at t48 only PSP (p = 0.025) (Table 3) were identified. At t72, independent predictors were found to be age (marginally, p = 0.050), CRP (p = 0.003) and PSP (p < 0.001). For PSP and CRP, values have been rescaled to show a meaningful HR, for an increase in 1,000 and 100 units, respectively. Our analysis indicates that age does not exhibit significant correlation with any of the four biomarkers under investigation at admission (Figure 3).

To evaluate a possible role of biomarker kinetic in predicting mortality up to 72 h, 3-time gradients were calculated for each biomarker as [(tX-t0)/t0], where tX indicates the level of the biomarker at 24, 48, and 72 h, while t0 the basal level at admission. Only the gradient [(t48-t0)/t0] for PSP was found to be associated, at the Cox regression, with mortality (Table 4). Based on the results of Cox regression for both PSP point estimates (Table 3) and gradients (Table 4), PSP was further evaluated. PSP gradients at 24, 48, and 72 h were dichotomized into positive (increasing PSP levels) and negative (decreasing PSP levels) values and a Kaplan–Meier analysis was run (Figure 4). Patients with increasing levels of PSP values displayed lower median survival time than patients with decreasing levels; differences reached statistical significance only at 48 h (20 vs. 29 days, log rank test, p = 0.046) (Figure 4B), but not, although marginally, at t24 (p = 0.074) or at t72 (p = 0.058) (Figures 4A, C).

We determined the performance metrics for PSP at 48 h, PSP gradient at 48 h, and MDW at 24 h using decisional cutoffs optimized to maximize the positive predictive value. This approach was chosen to ensure high confidence in the true positivity of events as indicated by the markers, thereby supporting more decisive actions in clinical settings. At the same time, we compared these results with the initial SOFA score. Results are shown in Table 5. From these results, it seems that biomarker metrics are similar to the initial SOFA. To further investigate the potential additional contribution of the biomarkers with respect to the initial SOFA, we performed a separate multivariate analysis including the score. However, none of the covariates (age, SOFA, PCT, and MDW) maintained a significant association with ICU mortality, suggesting a potential confounding effect, likely due to the limited statistical power of the analysis. As a workaround, we conducted a bivariate analysis taking into consideration each biomarker in turn and SOFA. This analysis revealed that MDW at baseline maintains its significance (p = 0.025), as did the SOFA score (p = 0.049), suggesting their independent contributions to ICU mortality. Conversely, PCT did not maintain its significance when analyzed in conjunction with the SOFA score (PCT p = 0.132; SOFA p = 0.009). Similarly, we evaluated MDW at 24 h and PSP at 48 h. Here, MDW at 24 h continued to show significant association, unlike the SOFA score (MDW p = 0.044; SOFA p = 0.052). For PSP at 48 h, both maintained significance (PSP p = 0.020; SOFA p = 0.025). These results suggest that the biomarkers evaluated are indeed independent predictors of mortality, with respect to the initial SOFA score.

Based on the specified decisional threshold and subsequent analysis, our investigation has identified the potential existence of a categorical biomarker integrating both static and dynamic measurements of PSP levels. We have devised a composite threshold termed PSP score, which registers as positive if either PSP at 48 h exceeds 2650 or PSP gradient at 48 h surpasses 0.4; otherwise, it is considered negative.

The contingency table illustrates the distribution of patient outcomes categorized by their PSP score status, depicting a clear distinction between survival and mortality outcomes (Table 6). Employing Fisher’s Exact Test for Count Data, we determined a p-value of 0.005, indicating a statistically significant association between PSP score status and patient outcomes. The odds ratio estimates of 3.79, with a 95% confidence interval ranging from 1.45 to 10.42, further supports this association.

In conclusion, our analysis reveals that patients with a positive PSP score exhibit a 3.79-fold higher likelihood of mortality in the ICU compared to individuals with a negative PSP score. This finding highlights the potential clinical relevance of the PSP score as a prognostic indicator for patient outcomes in critical care settings.

With the awareness that the ideal sepsis biomarker does not exist, the research is focused on identifying reliable biomarkers to assist Clinicians in the appropriate patient management, from early detection to prognostication. Although hundreds of sepsis biomarkers are currently under investigation, PSP and MDW are the most promising in terms of both accuracy and advantages over the traditional (9, 19). A recent meta-analysis by Paraskevas et al. (20), including twenty-nine studies, showed that PSP has a good accuracy for diagnosing sepsis with an area under the curve of 0.875 and a sensitivity of 81%. Previous meta-analyses also displayed good diagnostic value of PSP for sepsis (21–24). Similarly, Motawea et al. (25), showed that MDW has good accuracy for diagnosing sepsis, superior to PCT (25). Also, Huang et al. (10) and Agnello et al. (7) performed a meta-analysis to evaluate the diagnostic accuracy of MDW in adult patients with sepsis concluding that MDW is a reliable biomarker of sepsis (7, 10). Although the accuracy for detecting sepsis is comparable to CRP and PCT in most studies, PSP and MDW hold several advantages over CRP and PCT. Indeed, altered PSP and MDW levels indicate monocyte activation, and their increase is independent of the causative pathogen (26, 27). They can be easily measured. In addition, MDW is part of cell blood count (CBC) and, as such, is always available to Clinicians. On the other hand, PSP may be measured at the bedside by a point-of-care analyser. Thus, MDW and PSP have a rapid turn-around time (< 5 and < 20 min, respectively), aligned with the imperative for swift diagnosis, which is critical in acute settings. Finally, both MDW and PSP have low costs.

In this study, we evaluated the absolute value and the kinetic of MDW, PSP, CRP, and PCT at several timepoints during ICU stay. At admission, MDW, PSP, and PCT were significantly increased in patients with sepsis than without, in line with literature evidence. Interestingly, PSP was an independent predictor of ICU mortality at 48 and 72 h after hospital admission. Also, the kinetic of PSP had prognostic value, with increased values at 48 h after admission being associated with reduced survival.

In survival analysis using Cox regression, the hazard ratio (HR) quantifies the effect of a one-unit increase in a predictor variable on the hazard. Evaluating the results, it is crucial to consider the scale of the predictor, especially when it spans a wide range, such as thousands, like PSP does. When a predictor variable covers such extensive values, interpreting the HR for a single unit increase may not be practically informative or may lead to misleading conclusions. For example, the HR of PSP at 48 h is 1.0002, and it may seem irrelevant, but it means that a PSP rise of 1000 corresponds to an HR of 1.20, which means an increase in risk of 20%, which is not negligible. To enhance the interpretability and relevance of the HR in these cases, we also showed the rescaled predictor variable for a value compatible with its dynamic range. For instance, by scaling PSP that varies in the thousands to reflect changes per thousand units, the resultant HR of 1.24 directly conveys the impact of substantial and realistic increments in the PSP, providing a more meaningful understanding of its influence on the hazard. In our opinion this approach not only aids in clearer interpretation but also helps in comparing and communicating the effects of such predictors in a clinically relevant manner.

Thus, our findings support the role of PSP and its kinetic as a predictor of ICU mortality. Previously, some Authors assessed the prognostic value of PSP in patients with sepsis showing its accuracy in predicting mortality (28, 29). However, only a few evaluated its kinetic during a hospital stay. Similarly to our study, Shimoyama et al. (30) measured PSP longitudinally in ICU patients. The authors showed that PSP values on days 3 and 5 after ICU admission were independent predictors of 28-day mortality. The ΔPSP Day 3-Day 1 of ICU stay showed the best performance for predicting mortality compared to absolute values. Similarly, Hassan et al. (31) explored the clinical value of PSP kinetic in critically ill patients, measuring it at admission and the third day of hospitalization. They showed an increased trend from admission to day 3 in non-survivor patients. Also, Masson et al. (32) displayed an increasing trend in PSP levels in non-survivors within the first week of hospitalization.

Overall, these findings indicate the importance not only of evaluating the absolute PSP value but also of monitoring its changes over the first days of hospital stay to predict mortality. Noteworthy, all studies highlight the better prognostic power of PSP for predicting mortality than the other biomarkers, especially CRP and PCT. In a recent meta-analysis including 60 studies and a total of 15,681 critically ill patients with sepsis, Molano-Franco et al. (33) showed that PSP, but not CRP and PCT, was an independent predictor of mortality.

In our effort to develop a categorical variable that captures both the absolute value and the dynamic changes over time, we introduced a PSP score defined as positive by either a high absolute value or a significant gradient (Table 6). This approach resulted in a straightforward and effective cutoff, exploring a new avenue in this area of research. We hope that our findings could serve as a useful reference for other research groups engaged in similar studies. Biomarkers must not be used as a stand-alone test. They should be considered as one piece of a comprehensive approach to the patient with sepsis, based on the integration among laboratory medicine, i.e., CRP, PCT, PSP, and MDW detection, clinical features, and microbiological findings, keeping in mind that each single dowel has limitations and strengths.

Recent literature debunked the myth that PCT is a reliable biomarker of sepsis (34–36). Although thousands of articles have been published on PCT and sepsis, an agreed decisional cut-off has yet to be established. Additionally, PCT levels may significantly increase in several non-infectious inflammatory conditions (37). It has been shown that the evaluation of PCT kinetics has poor diagnostic and prognostic accuracy (38). The strength of PCT is the high association with bacterial infection, with an elevated negative predictive value to rule it out (39). Also, the accuracy of CRP in detecting sepsis is not optimal, aggravated using different decisional cut-offs ranging from 2 to 10 mg/L (40, 41).

The main limitation of this study is that it was a single-center small sample size. Our analysis revealed a statistically significant difference with the current number of participants. This indicates that the observed effect is strong enough to be detected even with a relatively small sample size. We acknowledge that any study with a limited sample size has limitations regarding the generalizability of the results. However, the significant findings obtained serve as a preliminary basis for future research, justifying larger studies that can further explore these observations in more extensive samples. Thus, larger multicentre cohort studies are required to confirm our findings. Another limitation of our study is the absence of detailed patient histories prior to ICU admission, which precludes us from determining whether the sepsis originated in the community or within a hospital setting. It is plausible that the biomarker dynamics differ between these two types of sepsis. Stratifying patients based on the origin of sepsis could have potentially revealed more nuanced patterns and improved the performance of our biomarker analysis. Another limitation of the study is that some patients in the control cohort developed sepsis post-admission potentially diluting the contrast between the groups in the timepoints after baseline. This could lead to an underestimation of the true differences in biomarker levels between groups. Despite this, the observed biomarker differences were statistically significant, suggesting a real, and potentially underestimated difference even in the face of potential cohort contamination. In contrast, the undetected differences might actually be there but not detected due to a decreased power of the study caused by this dilution effect.

In this study, we first compared the accuracy of two new promising sepsis biomarkers, i.e., MDW and PSP, in relation to the traditional ones, i.e., CRP and PCT. Beyond the traditional and widely used biomarkers of sepsis, MDW and PSP could be implemented in clinical practice for detecting and monitoring sepsis, respectively. MDW, being part of the CBC, could help detect sepsis early, also when it is not suspected, while PSP kinetic could help monitor sepsis and identify patients who would benefit from intensive treatment.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Comitato Etico Palermo 1, University of Palermo. The study was conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

LA: Conceptualization, Data curation, Writing – original draft. AMC: Conceptualization, Data curation, Writing – original draft. FDB: Statistical analysis, Writing – original draft. CMG: Data curation, Writing – review & editing. CS: Data curation, Writing – review & editing. AG: Formal analysis, Writing – review & editing. GB: Formal analysis, Writing – review & editing. AC: Data curation, Writing – review & editing. BLS: Data curation, Writing – review & editing. MC: Conceptualization, Supervision, Validation, Writing – review & editing.