Discussion
Sepsis is a common problem in critically ill patients. Early diagnosis and early treatment are essential for the clinical course and the outcome of patients with sepsis. Given that a large proportion of critically ill patients exhibit SIRS, the ability to accurately distinguish between SIRS and sepsis, which is defined as SIRS as a result of bacterial infection, has become one of the holy grails of critical care medicine.
Currently, no recommendation can be given for the use of biomarkers to differentiate sepsis from non-infectious SIRS. Many biomarkers, particularly procalcitonin (PCT) and C-reactive protein (CRP), are widely used to identify sepsis in current clinical practice. In comparison with CRP, PCT seems to be a better marker to differentiate sepsis from non-infectious SIRS. However, two published meta-analyses concluded that the SEN and SPE of PCT varies in the diagnosis of sepsis, leading to questions of the ability of PCT to distinguish sepsis from SIRS. Presepsin was recently identified as a molecule involved in the inflammatory response and represents a promising diagnostic biomarker with high SEN and SPE.
The primary finding of our meta-analysis is that presepsin exhibits very good diagnostic accuracy for distinguishing patients with sepsis from those with systemic inflammatory disease. Specifically, in our primary analysis involving all studies that evaluated presepsin in adults with and without sepsis, the area under the SROC curve was 0.89, which was greater than the results of published meta-analyses of the use of PCT and soluble triggering receptor expressed on myeloid cells-1 (sTREM-1) for the diagnosis of sepsis. The pooled SEN and SPE were 86 % and 78 %, respectively. To our knowledge, presepsin exhibited the highest sensitivity among the proposed biomarkers in differentiating sepsis form other non-infectious SIRS. To date, no biomarkers have exhibited sufficient (greater than 90 %) sensitivity to distinguish sepsis from SIRS in these critically ill adult patients.
The rescue principles indicate that the infection foci of patients with sepsis should be detected within 6 hours, followed by antibiotic treatment within 1 hour after the diagnosis of sepsis. Generally, PCT increases 4 hours after infection, slowly reaching a plateau at 8–24 hours and peaking one day after infection. Compared with PCT, presepsin increases at 2 hours post-infection in the cecal ligation and puncture (CLP) sepsis model and peaks at 3 hours. Presepsin can be detected in the early stage of infection using rapid dosage methods based on chemiluminescence enzyme immunoassay, which are available and permit automated measurements in 1.5 hours.
The exclusion of reviews, letters, commentaries, correspondence, case reports, conference abstracts, expert opinions, editorials and reports of animal experiments may have contributed to publication bias. However, we tested for this, and no significant publication bias was observed in our study (Fig. 3).
Marked statistical heterogeneity was present in all analyses, a fact that must not be overlooked in the interpretation of the above findings. We observed significant heterogeneity in SEN and SPE among the studies analysed. Consecutive patient recruitment, sample size, setting and excluded patients substantially affected the SEN of the diagnosis of sepsis in the meta-regression analysis, and none of the variables affected the SPE of diagnosis of bacterial infection in the meta-regression analysis. According to Sackett and Haynes' classification, the index test in group 1 is developed in an ideal situation against a validation set (group 2) in which the performance is tested in a more realistic clinical context. Group 2 studies are the most informative for clinical practice, as they are designed to resemble the real-world setting of routine clinical practice by restriction to patients who are the most likely to be encountered by clinicians. In this meta-analysis, six of the included studies were classified as group 2 studies. Additionally, we performed subgroup analysis restricted to group 2. The pooled sensitivity, DOR and AUC of the six studies (11, 14–18) were similar to corresponding values in the other studies. In particular, the pooled AUC indicated high diagnostic accuracy (AUC ≥0.9). All of the studies included involved a prospective cohort, and two were multicentre studies, underlining the high quality of these studies. The test results obtained by the prospective recruitment (PR) study design method achieved more realistic results than that obtained by the consecutive recruitment (CR) method. Therefore, satisfactory results could be achieved in the future by implementing more prospective studies.
Likelihood ratios and PTPs are also relevant for clinicians, as they provide information on the likelihood of a patient with a positive or negative test actually exhibiting sepsis. In our study, with a hypothetical pretest probability of 20 % and a PLR of 3.8, detecting presepsin for sepsis diagnosis would raise the PTP to 49 %, with an NLR of 0.18. Detecting presepsin reduced the PTP to 4 % (Fig. 7), demonstrating that the application of the presepsin test was advantageous in the diagnosis of sepsis. Additionally, all included studies recruited a mix of medical and surgical patients. Our findings can therefore be generalized to patients from different countries as well as to different admission categories.
Delayed resuscitation is reportedly significantly associated with a high risk of death, and rapidly initiating the appropriate therapeutic interventions upon the patient's arrival to the ED is critical. Thus, we performed subgroup analysis restricted to ED patients to evaluate the diagnostic accuracy of presepsin. The pooled SEN and SPE were 0.85 (95 % CI: 0.77, 0.92) and 0.78 (95 % CI: 0.69, 0.88), respectively, nearly equal to the overall results.
Several limitations should be considered when interpreting the findings of this meta-analysis. First, despite the extensive literature search, the number of included studies was small; however, the number of patients enrolled was satisfactory (n = 1,815), thereby decreasing type II error. Second, we could not determine the optimized cutoff value because we failed to obtain the raw data to map the ROC curve. In all studies, the optimal cutoff point was retrospectively determined based on the ROC curve. The cutoff points varied greatly among the studies, despite using the same presepsin assay. A reason for this discrepancy may be differences in study design, especially the patient inclusion criteria. Falsely elevated values of presepsin or PCT are observed in conditions of chronic renal failure or a history of resuscitation and trauma. One study excluded patients with these comorbidities, but the others did not. Thus, future research should be designed in consideration of how comorbidities may influence presepsin levels to confirm an optimal cutoff point for clinical use. Third, due to the small number of eligible studies and the lack of necessary data reported in the original publications, we could not specifically analyse patients with different conditions (e.g., different severities of sepsis or different sites of infection) to distinguish the sepsis, nor could we determine the therapeutic decisions in the individual patient. Last, it is possible that presepsin may perform differently in sepsis caused by gram-positive, gram-negative, or fungal pathogens. Hence, both the clinical characteristics of the enrolled patients as well as the local microbiological profile in included studies are likely to affect the value of presepsin in predicting sepsis. However, we were unable to explore this further because the necessary information was largely unavailable in the studies.