A flow chart of the study population is presented in Fig. 1. Of the 192 patients clinically diagnosed with sepsis and septic shock, 79 were excluded either due to refusal of study participation (n = 77) or inadequate samples (n = 2). Therefore, 113 blood samples were initially eligible for biomarker measurements (64 for sepsis and 49 for septic shock). After retrospective re-evaluation, 16 subjects in the initial sepsis group were excluded because they did not meet the criteria for sepsis (i.e., no evidence of infection). The final diagnoses of these 16 subjects were congestive heart failure (n = 3), pulmonary thromboembolism (n = 2), acute kidney injury (n = 2), a hyperosmolar hyperglycemic state (n = 2), hepatorenal syndrome (n = 2), chronic obstructive pulmonary disease (n = 2), and others (n = 3). After retrospective re-evaluation, three subjects in the initial septic shock group were recategorized to the sepsis group because they did not meet the criteria for septic shock. The final patient populations were as follows: 51 with sepsis, 46 with septic shock, and 45 as controls. Overall, 142 subjects were enrolled in the present study. From the 46 septic shock patients admitted to our institution, follow-up samples were obtained from 28 patients within 24 hours of discharge to measure IL-6 and PTX3, 15 of whom recovered, and 13 died. The remaining 18 patients were excluded due to refusal to provide a blood sample, transfer to other institutions, sudden death, or undetermined outcomes (still on admission). Baseline characteristics of the study population are shown in Table 1. The most common infection sites were the respiratory (64.9%) and genitourinary systems (33.0%).
Correlations with other biomarkers and severity score
IL-6 levels showed positive correlations with PTX3 (rho = 0.802, P < 0.001), lactate (rho = 0.798, P < 0.001), PCT (rho = 0.752, P < 0.001), CRP (rho = 0.476, P < 0.001), SOFA score (rho = 0.421, P < 0.001), and APACHE II score (rho = 0.407, P < 0.001) by Spearman’s rank analysis.
Diagnostic value of biomarkers
The median IL-6 values (IQR) in the control, sepsis, and septic shock groups were 23.6 (11.2–43.5), 89.9 (45.2–272.6), and 1378.6 (256.4–11062.1) pg/mL, respectively (Fig. 2); those for PTX3 were 4 (2–13), 25 (10–51), and 74 (26–147) ng/mL, respectively; those for PCT were 0.2 (0.1-0.23), 0.3 (0.1-1.1), and 3.4 (1.3-20.1) ng/mL, respectively; and those for lactate were 0.9 (0.5–1.5), 1.9 (0.9–2.9), and 5.5 (3.0–8.1) mmol/L, respectively. The three groups showed significant differences in the levels of these four biomarkers, as determined by Kruskal–Wallis and post hoc tests (P < 0.001). The median CRP values (IQR) in the control, sepsis, and septic shock groups were 3.6 (2.0–5.2), 9.9 (4.9–20.2), and 10.5 (7.3–21.0) mg/dL, respectively. There were significant differences between the control and other groups (P < 0.001); however, no significant difference was found between the sepsis and septic shock groups (P = 0.45).
The AUC of IL-6 to discriminate sepsis from the control group was 0.89 (95% confidence interval [CI], 0.97–1.00; P < 0.001), 0.84 for PTX3 (95% CI, 0.95–0.99; P < 0.001), 0.80 for PCT (95% CI, 0.86–0.96; P < 0.001), and 0.77 for CRP (95% CI, 0.71–0.91; P < 0.001) (Table 2 and Fig. 3). The optimal cut-off value to discriminate sepsis from controls was 52.60 pg/mL for IL-6 (sensitivity, 97.0%; specificity, 97.2%; P < 0.001) (Table 3) and 15.10 ng/mL for PTX3 (sensitivity, 92.6%; specificity, 97.4%; P < 0.001).
The AUC to discriminate septic shock was 0.80 for IL-6 (95% CI, 0.71–0.89; P < 0.001), 0.70 for PTX3 (95% CI, 0.60–0.81; P < 0.001), 0.73 for PCT (95% CI, 0.63–0.83; P < 0.001), and 0.53 for CRP (95% CI, 0.42–0.65; P = 0.603) (Table 2 and Fig. 3). The optimal cut-off value to discriminate septic shock was 348.92 pg/mL for IL-6 (91.8% sensitivity, 63.2% specificity, P < 0.001) (Table 3) and 58.28 ng/mL for PTX3 (93.2% sensitivity, 60.7% specificity, P < 0.001).
Prognostic value of biomarkers
Univariate and multivariate Cox regression analysis results for the risk factors of 28-day mortality are shown in Table 4. The univariate analysis determined that the significant risk factors for 28-day mortality were IL-6, PTX3, lactate, SOFA and APACHE II scores, and septic shock. The significant risk factors determined by multivariate analysis were IL-6 (hazard ratio [HR], 1.0004; 95% CI, 1.0003–1.0005; P = 0.024), SOFA score (HR, 1.128; 95% CI, 1.030–1.211; P = 0.011), and lactate (HR, 1.135; 95% CI, 1.033–1.247; P = 0.009).
Kaplan–Meier curve analyses and log-rank tests were performed to assess cumulative survival rates and compare 28-day survival curves between the high IL-6 (≥ 348.9 pg/mL) and low IL-6 (< 348.9 pg/mL) groups. The optimal cut-off value of IL-6 to predict septic shock in the present study was 348.9 pg/mL. The survival curve of the high IL-6 group significantly differed from that of the low IL-6 group in log-rank tests (P = 0.008) (Fig. 4). Kaplan–Meier curve analyses and log-rank tests were also performed to assess cumulative survival rates and compare the 28-day survival curves between the high PTX3 (≥ 58.28 ng/mL) and low PTX3 (< 58.28 ng/mL) groups. The optimal cut-off value of PTX3 to predict septic shock was 58.28 ng/mL. The survival curve of the high PTX3 group significantly differed from that of the low PTX3 group in log-rank tests (P = 0.043) (Fig. 4).
Among the patients with septic shock at presentation, initial IL-6 levels within 6 hours of clinical diagnosis in the recovered survivors (n = 15) and non-survivors (n = 13) were 444.3 (261.2–5893.5) and 7609.5 (4526.0–12208.4) pg/mL, respectively (P = 0.05), and follow-up IL-6 levels within 24 hours of discharge were 21.5 (10.2–51.7) and 9976.5 (4651.2–71048.3) pg/mL, respectively (P < 0.001). Among the same patients, initial PTX3 levels within 6 hours of clinical diagnosis were 29 (10–75) and 126 (70–147) ng/mL, respectively (P = 0.007) (Fig. 5), and follow-up PTX3 levels within 24 hours of discharge were 4 (2–7) and 188 (101–376) ng/mL, respectively (P < 0.001. Error bars represent the variability of data in Fig. 5. Both IL-6 and PTX3 levels significantly decreased in the recovered survivors (P < 0.001 and P < 0.001, respectively); however, both levels remained high and even significantly increased in the non-survivors (P = 0.03 and P = 0.009, respectively).
Combination of IL-6 and PTX3 to predict severity and mortality
Multivariate logistic regression was performed to model the ability of the biomarker combination (IL-6 and PTX3) to discriminate septic shock patients from sepsis and to predict 28-day mortality among the overall patients. Using the logistic regression equation, the log of probability was converted to the probability of septic shock and 28-day mortality. The AUC values of combined marker score (probability of septic shock) and each biomarker for the discrimination of septic shock are shown in Table 5. The AUC values of combined marker score (probability of 28-day mortality) and each biomarker for the prediction of 28-day mortality are shown in Table 6.