Background

Sepsis is a clinical syndrome characterized by organ dysfunction due to dysregulated host response to infection; it is still one of the most common causes of mortality in critically ill patients [1,2,3]. The mortality rate of sepsis is estimated to be between 20% and 40%, and long-term complications are common, including kidney failure, liver failure, depression, and neurocognitive impairment [4,5,6]. It is estimated that there are over 19 million cases of sepsis and 5 million sepsis-related deaths each year, with the majority occurring in low and middle-income countries [7]. Early identification of sepsis is crucial for improving prognosis and preventing relevant complications [8].

Biochemical markers can help practitioners identify sepsis as early as possible. Nevertheless, commonly used biomarkers, including C-reactive protein (CRP) and procalcitonin, have been proven inaccurate in detecting sepsis in previous meta-analyses [9, 10]. CRP showed promising sensitivity in detecting infection, nevertheless, it is only specific if a high cutoff level is used, which it will in turn reduce sensitivity [11]. Procalcitonin (PCT) has been determined to have the highest performance as a biomarker for the diagnosis and prognosis of sepsis. However, it may be elevated in many diseases other than infection, especially after surgery and trauma [12, 13]. Better biomarkers are warranted to increase the value of sepsis detection. Pancreatic stone protein (PSP), a 144-amino-acid glycoprotein, might be a suitable biomarker for sepsis; however, the physiological role of PSP has not been elucidated yet [14,15,16,17]. PSP is mainly secreted by pancreatic acinar cells and also secreted by intestinal and gastric cell subsets [18]. A study illustrated that PSP is an inflammatory mediator that can bind and activate neutrophils, thereby acting as an acute phase protein in response to injury in the early stages of infection [16].

Several investigations have explored the diagnostic capabilities of PSP in detecting sepsis, demonstrating favorable outcomes [19,20,21,22,23]. However, these studies were limited concerning their sample sizes and the population involved. This meta-analysis aimed to draw a comprehensive knowledge of the diagnostic value of PSP in sepsis identification.

Methods

Search strategy and selection criteria

This meta-analysis was conducted in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) [24]. From the inception date until April 20, 2023, a comprehensive search was executed across multiple electronic databases, encompassing PubMed, Web of Science, Embase, the Cochrane Library, and China National Knowledge Infrastructure (CNKI), considering literature in both English and Chinese languages. The search strategy incorporated an array of keywords and synonyms: pancreatic stone protein, PSP, sepsis, Bloodstream Infection, Bloodstream Pyemia, Pyemias, Pyohemia, Septicemia, Blood Poisoning, Blood Severe Sepsis. A flowchart for search, screening, and eligibility identification was constructed based on the PRISMA guidelines. Inclusion criteria were as follows: (1) subjects suspected of having sepsis; (2) clear documentation of a reference standard; (3) a 2 × 2 contingency table can be formed categorizing participants with true positive (TP), false positive (FP), false negative (FN), true negative (TN) results on PSP test. Exclusions pertained to case reports, reviews, editorials, conference abstracts, animal studies, or studies with unextractable data. The entire process of database querying and study selection was independently executed by two reviewers, with any disagreements resolved through iterative discussions until consensus was achieved.

Data extraction and risk of bias assessment

The following study level data were retrieved: name of the first author, publication year, country, sample size, median or mean age, proportion of the female, standard reference, counts of participants with TP, FP, FN, and TN outcomes. The risk of bias assessment of studies was evaluated according to the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) [25]. Two authors independently extracted data from included studies and appraised the risk of bias of these studies; disagreement was solved through consultation with a third investigator.

Statistical analysis

Statistical analyses at the study level were executed using Stata 14.0 software and Meta-DiSc 1.4. The following summary measures were computed, accompanied by their corresponding 95% confidence intervals (CIs): pooled sensitivity, specificity, positive Likelihood Ratio (+ LR), negative Likelihood Ratio (-LR), diagnostic odds ratio (DOR), and the area under the curve (AUC) derived from the summary receiver operating characteristic (SROC) curve. To gauge the heterogeneity among the included studies, Cochran’s Q statistic and the I2 index were employed. According to established thresholds, heterogeneity was categorized as insignificant (I2 = 0–25%), low (I2 = 25–50%), moderate (I2 = 50–75%), or high (I2 = 75–100%) [26]. Publication bias was visually inspected through the construction of funnel plots and further subjected to statistical assessment utilizing Deeks’ method [27]. Additionally, a sensitivity analysis was carried out to gauge the influence of individual studies on the aggregate results. Statistical significance was set at a p-value less than 0.05.

Results

Study selection and characteristics

The literature search yielded 265 articles. Among them, 92 duplicated citations were removed, and another 156 studies were excluded through an initial screening of the title and abstract. The full-text reading of the remaining 17 articles identified nine studies with 1364 participants eligible for inclusion in this study [8, 11, 19,20,21,22,23, 28, 29] and five studies in meta-analysis. Figure 1 displays the flowchart of the database search and study selection. Six studies were performed among adults; the other 3 included studies were conducted among newborns. The coverage of countries included the UK, Switzerland, Netherlands, Spain, Egypt, France, and Italy. Characteristics of enrolled studies are depicted in Table 1. The risk of bias for each included study was assessed as low (Fig. 2 and Figure S1).

Fig. 1
figure 1

Flowchart of the literature search

Full size image
Table 1 Study characteristics
Full size table
Fig. 2
figure 2

Summary of risk of bias in included studies

Full size image

Diagnostic performance of PSP

The overall sensitivity of PSP was 0.88 (95% CI: 0.77–0.94, I2 = 77.4%, p < 0.01), and the pooled specificity was 0.78 (95% CI: 0.65–0.87, I2 = 90.3%, p < 0.01), respectively (Fig. 3). Pooled + LR, -LR, and DOR were 4.1 (2.3, 7.3), 0.16 (0.07, 0.34), and 26 (7, 98). The AUC value for the SROC of PSP was 0.90 (0.87, 0.92) (Fig. 4). The pooled sensitivity, specificity, + LR and - LR, and DOR for PSP among neonates were 0.91 (95% CI: 0.84, 0.96), 0.66 (95% CI: 0.58, 0.74), 3.97 (95% CI: 0.53, 29.58), 0.13 (95% CI: 0.02, 1.00), and 31.27 (95% CI: 0.97, 1004.60) (Table 2). The pooled sensitivity, specificity, + LR and - LR, and DOR for PSP among adults were 0.85 (95% CI: 0.78, 0.90), 0.72 (95% CI: 0.67, 0.776), 4.09 (95% CI: 1.69, 9.92), 0.19 (95% CI: 0.08, 0.49), and 22.74 (95% CI: 4.25, 121.73) (Table 2).

Fig. 3
figure 3

Forest plot of diagnostic performance of PSP in detecting sepsis

Full size image
Fig. 4
figure 4

SROC of PSP in detecting sepsis

Full size image
Table 2 Diagnostic value of PSP in neonates and adults
Full size table

In the study by Garcia de Guadiana-Romualdo et al., results of the ROC curve analysis revealed an AUC of 0.872 for sepsis identification [11]. Klein et al’s analysis of biomarker kinetics (PSP, routine markers) was performed on 90 burned patients, PSP identified between sepsis, infection and sterile inflammation with an AUC 0.89 [28]. An AUC of 0.69 was reported in de Hond et al’s study including 156 participants [8]. Moreover, in Saleh et al’s study, the ROC revealed that the AUC for PSP reached 0.868 for sepsis diagnosis [29].

Publication bias

Deek’s tests for publication bias yielded p-values of 0.42 for the meta-analysis, indicating no statistically significant publication bias (Fig. 5).

Fig. 5
figure 5

Funnel plot of meta-analysis of PSP in detecting sepsis

Full size image

Sensitivity analysis

Results of sensitivity analysis demonstrated that the two included studies investigating the diagnostic value of PSP in neonatal sepsis had a significant impact on the overall effect size, and thus subgroup analysis based on the age of the study population was performed (Figure S2).

Discussion

Results of this study demonstrated an overall sensitivity of 0.88 (95% CI: 0.77–0.94) and a pooled specificity of 0.78 (95% CI: 0.65–0.87) with an overall AUC value of 0.90 (0.87, 0.92). Compared to other biomarkers, the pooled sensitivity in this meta-analysis was higher than the pooled sensitivity (0.79) in Chen et al’s meta-analysis investigating neutrophil to lymphocyte ratio (NLR) in the diagnosis of sepsis, while the specificity was lower than that (0.91) in Chen et al’s meta-analysis [30]; the comparison between PSP and calprotectin (sensitivity: 0.88 vs. 0.77, specificity: 0.78 vs.0.85) was similar to NLR [31]; Poggi’s meta-analysis assessing the accuracy of presepsin for the sepsis diagnosis showed that the pooled sensitivity and specificity were 0.93 (95% CI, 0.86–0.95) and 0.91 (95% CI, 0.85–0.95), respectively, which were higher than PSP [32]; the overall sensitivity and specificity were higher than those of CRP and PCT [33].

Several research findings suggest that PSP is involved in the early defense mechanism of sepsis [16, 19, 28]. In these studies, PSP has been shown to be associated with the severity of inflammation and can activate neutrophils by upregulating activation markers CD11b and CD62L [16]. In addition to the activation of neutrophils, PSP possesses antibacterial functions; it can induce bacterial aggregation, which may help prevent bacteria from penetrating the intestinal barrier [34]. Moreover, PSP levels rise 72 h before the onset of clinical symptoms of sepsis [8]. The results of this meta-analysis verified that PSP displayed favorable diagnostic performance in detecting sepsis. The results were consistent with previous studies regarding the diagnostic value of PSP in sepsis [8, 11, 29]. Notably, neonatal sepsis is difficult to diagnose due to the nonspecific clinical signs in response to sepsis [35]. Schlapbach et al.‘s study are the first to investigate PSP in neonatal sepsis; results demonstrated that the level of PSP in infected infants was significantly higher than in uninfected ones with an AUC of 0.69 [20]. Subgroup analysis of this meta-analysis showed favorable pooled sensitivity but low specificity for PSP alone in the diagnosis of neonatal sepsis, suggesting the necessity for combining different biomarkers to detect sepsis in this specific population. More well-designed prospective studies are required to clarify this opinion.

Interestingly, the pooled sensitivity for adults was lower than that in neonates, while the difference in specificity is the opposite. The underlying reason needed to be clarified based on the current meta-analysis. However, it was reported that maximum PSP levels in neonates were lower than those in adults with sepsis, which may be one of the reasons for the different diagnostic values for PSP in newborns and adults. Relevant studies are warranted to investigate this difference.

Strengths and limitations

To our knowledge, this is the first meta-analysis to evaluate the diagnostic value of PSP in the context of sepsis in general. Data from previously published citations were synthesized to enhance the statistical power of the diagnostic value of PSP. Results of this study were favorable and promising, which may serve as both advanced level of evidence and reference for practitioners to make decisions on the diagnosis of sepsis in their clinical practice.

Like other meta-analyses, there are several limitations of this meta-analysis. The study protocol was not registered on PROSPERO. Significant heterogeneity existed between included studies; it may be attributed to differences in the study population, standard reference, and cutoffs of PSP in component studies. Although Deek’s funnel plots asymmetry test revealed no statistically significant publication bias in the meta-analysis, bias caused by published and unpublished studies inherently existed because this study is only focused on published articles. The number of included studies was limited owing to the inclusion criteria of this meta-analysis; aside from subgroup analysis based on the study population, subgroup analysis on other covariates was not performed. The interpretation of findings from this study ought to be with caution; more similar studies are needed to specify the diagnostic value of PSP in detecting sepsis.

Conclusion

In this meta-analysis, evidence suggests that PSP was a promising biomarker for diagnosing patients suspected of sepsis. According to the findings presented in this meta-analysis, specifically designed studies on different populations are needed to ascertain the validity of PSP measurement to guide early empirical antibiotic treatment, particularly in neonates.