Results of meta-analysis of ANXA3 expression from the omnibus datasets
Totally, 51 microarrays datasets from the GEO database and ArrayExpress database fulfilled the entry criteria, including 42 microarrays datasets for diagnosis of sepsis and 19 microarrays datasets for prognosis of sepsis (especially, 11 microarrays datasets contained both data for diagnosis and prognosis of sepsis)[16-56]. The workflow for the study was provided in Figure S1. The basic characteristics of the included datasets were displayed in Table S1 and Table S2.
Of those microarrays, 42 sepsis-related datasets were included to perform meta-analysis on the basis of 2830 sepsis patients and 992 control populations. The results revealed an apparent heterogeneity (p<0.00001, I2=96%), and significant up-regulation of ANXA3 was indicated in the sepsis patients by a random effects model (SMD=2.04 (1.54-2.55); p<0.00001) (Figure 1). As was illustrated in Figure S2A, 30 sepsis-related datasets for adult subgroup compared ANXA3 mRNA levels between 2052 sepsis patients and 743 controls. Significant heterogeneity was observed in the adult subgroup (I2=97%, p<0.00001) by a random effects model, and sepsis patients were with remarkable up-regulation of ANXA3 mRNA compared with controls (SMD=1.98(1.32-2.65); p<0.00001). Meanwhile, a 12-datasets meta-analysis in the children subgroup also suggested ANXA3 mRNA levels in sepsis patients were higher than that of controls (SMD=2.19(1.58-2.79); p<0.00001), including 778 sepsis patients and 249 controls in a random effects model (I2=90%, p<0.00001) (Figure S2B).
Furthermore, 19 related datasets were adopted to investigate the expression levels of ANXA3 mRNA for the diagnosis of sepsis patients. The results indicated slight up-regulation of ANXA3 were demonstrated for the dead patients in sepsis by a fixed effects model (SMD=0.11(0.01-0.22); p=0.03; I2=0%, p=0.47) comprising 1711 sepsis survival patients and 500 sepsis dead patients (Figure 2). In the subgroup analysis, slightly increased ANXA3 mRNA was also observed between 1503 sepsis survival patients and 456 sepsis dead patients (SMD=0.13(0.02-0.23); p=0.02) in the adult subgroup with a no significant heterogeneity by a fixed model (I2=17%, p=0.26) (Figure S3A); but no significant difference was identified between 208 sepsis survival patients and 44 sepsis dead patients (SMD=0.02(-0.13-0.35); p=0.91) in the children subgroup by a fixed model (I2=0%, p=0.93) (Figure S3B).
Above those results, ANXA3 mRNA was apparently up-regulated in the sepsis patients, and ANXA3 might affected the development of sepsis. These results also indicated ANXA3 was up-regulated in the sepsis died patients, but the difference was slight, whether ANXA3 might affected the prognosis of sepsis need further investigation.
General demographic characteristics
To validate the plasma ANXA3 in sepsis, 153 critically ill patients and 57 healthy were recruited in this study. Demographic and clinical characteristics of patients were displayed in the Table 1. The mean age of critically ill patients was 61.69 ± 21.09, 110 (71.89%) were males. Of those critically ill patients, 60 patients (39.22%) were diagnosed with sepsis and 28-day mortality was 21.57% for critically ill patients (33 patients). Pneumonia and primary bloodstream infections were the major infection sites. The median patient APACHE II, SOFA, MOD scores were 17.93 ± 8.20, 6.58 ± 4.37 and 5.36 ± 3.40, respectively.
Plasma ANXA3 concentrations were associated with the sepsis risk in critically ill patients
Firstly, we measured plasma ANXA3 concentrations in 153 critically ill patients on hospital admission. In comparison to a healthy group consisting of 57 blood donors, ANXA3 plasma levels were markedly elevated in critically ill patients (3.42±1.53 ng/ml vs. 1.98±0.75 ng/ml, p<0.001; Figure 3A). Meanwhile, we compared the ANXA3 plasma concentrations in the sepsis (n=60) and non-sepsis groups (n=93), the results indicated ANXA3 plasma concentrations were markedly increased in sepsis patients (4.39±1.39 ng/ml vs. 2.82±1.28 ng/ml, p<0.001; Figure 3B), which were in line with the meta-analysis results. Multiple logistical regression analyses also suggested plasma ANXA3 was significantly correlated with a higher occurrence of sepsis (OR=2.41 (1.75-3.32), p<0.001), adjusted by age, sex.
To further determine the specific role of plasma ANXA3 in sepsis patients, we investigated the correlations among ANXA3 and SOFA, APACHE II and MOD scores. The results suggested plasma ANXA3 was positively correlated with SOFA (r=0.25, p=0.002) (Figure 4A) and MOD scores (r=0.26, p=0.001) (Figure 4B), but not APACHE II (r=0.16, p=0.05, Figure 4C).
Based on the ROC analysis of plasma ANXA3, an AUC of 0.815(0.745-0.886) was gained for the occurrence of sepsis (Figure 5). The optimal cut-off value is 3.39 ng/ml in critically ill patients, and the sensitivity and specificity were 83.33% and 76.34%, respectively (Table 2). The AUCs of IL-6, PCT, SOFA for prediction of sepsis were 0.672(0.585-0.759), 0.673(0.584-0.761), 0.668(0.577-0.759), respectively (Table 2). Our results revealed ANXA3 obtained a better AUC, compared with those biomarkers and scores (IL-6, p=0.011; PCT, p=0.006; SOFA, p=0.01) (Figure 5).
Associations between plasma anxa3 concentrations and mortality
Totally, there were 33 patients died during the 28-day follow-up period in critically ill patients, we intend to investigate the relative associations between plasma ANXA3 and 28-day survival. According to the mean expression of ANXA3 (3.43 ng/ml) in the plasma of critically ill patients, we divided them into high (≥3.43 ng/ml) and low (<3.43 ng/ml) groups. Kaplan-Meier curves (Figure 6A) suggested the mortality of the higher ANXA3 group was significantly higher than that of the lower ANXA3 group (HR=2.16(1.09-4.28); p=0.028). Furthermore, 19 patients died during 60 sepsis patients, then those patients also were divided into high and low groups, Kaplan-Meier curves (Figure 6B) suggested the higher ANXA3 group was with higher mortality rates of sepsis, but the difference was not significantly (HR=1.63(0.65-4.06); p=0.276). More patients need to be recured to investigate whether ANXA3 might affected the prognosis of sepsis patients.