HLA-DR has been used widely in clinical and basic scientific research, such as the diagnoses of sepsis [11] and the prediction of prognoses [17]. It is used to evaluate the severity and prognosis of burns [18] and to evaluate the immune function of patients after surgical operations; however, thus far, it has not been used in the diagnoses and predictions of disease prognoses.
All 639 patients included in the eight studies met the inclusion and exclusion criteria. In 7 studies, the expression of mHLA-DR was detected using flow cytometry, and in one study, droplet digital PCR was used.
In the first five studies included, the patients in the trial group were admitted within eight hours of blunt trauma, and were hospitalized for 14 to 28 days. All patients were diagnosed with sepsis in accordance with the guidelines of the American College of Chest Physicians/Society of Critical Care Medicine consensus conference [19], with the gold standard for the diagnosis of sepsis being the detection of SIRS plus-positive bacterial cultures. All studies evaluated the mHLA-DR expression in patients in the first 14 days. In three studies conducted by Cheron et al. in 2010, 2012, and 2015, the characteristics of all patients in the experimental group were similar. All patients presented with ISS scores ≥ 25 and had been treated in the ICU. The results of these three studies showed that there were significant differences in the expression of mHLA-DR between patients with sepsis and patients without sepsis, and that the ratios of the expression on days three to four compared with those on days one to two had the highest specificities and sensitivities, with 1.2 being the best cut-off value. Vester et al. studied 24 patients with multiple traumas and compared the occurrence of SIRS, sepsis, and MODS in patients with blunt trauma; eight postoperative patients were considered as the control group. Compared with the patients without MODS, the MFIs of five patients with MODS decreased more on the 3rd and 7th days. In the 17 patients with SIRS, the MFIs decreased on day 3 and increased significantly on day 7 and day 14. Compared with patients without sepsis, the MFIs of patients with sepsis decreased on the 3rd and 7th days after trauma and increased significantly on the 14th day. In our analysis, we used the MFIs of the third day as the cut-off value to diagnose sepsis. Nine patients with SIRS were deemed false-positive, and no false negative cases were found. The sensitivity was high; however, the specificity was low. Ditschkowski et al. not only detected HLA-DR in the monocytes from 1 to 14 days after injury, but they also detected T lymphocytes and soluble HLA-DRs. When the average fluorescence intensity of HLA-DR on T cells measured at channel 161 on the first day was considered as the cut-off value, the sensitivity and specificity for predicting sepsis were 67% and 72%, respectively. Using the cut-off value of the average fluorescence intensity of HLA-DR detected in channel 90 on the second day, the sensitivity and specificity for predicting sepsis were 53% and 76%, respectively. The sensitivity and specificity of detecting the concentration of soluble HLA-DR at 0, 1, 2, 4, and 6 days after injury were 72% and 90%, respectively. Ditschkowski et al. also conducted another study in the same year and found that the sensitivity and specificity for predicting sepsis were 69% and 77%, respectively, when the average fluorescence intensity of HLA-DR in channel 155 on day six was considered as the cut-off value [20]. The patients studied by Vester et al. and Ditschkowsk et al. presented with an ISS ≥ 16. These five studies also showed that the expression of HLA-DR in patients with different degrees of trauma was also significantly different.
In addition to the above-mentioned studies, we also identified other studies that investigated the prediction of post-traumatic sepsis using HLA-DR [21–23]. In these studies, we observed consistent changes in HLA-DR after trauma and sepsis. These studies provided us with a new direction for the early warning of traumatic sepsis, viz., the expression of B cells, T cells, monocytes, and soluble HLA-DR in post-traumatic patients could be considered to explore and establish an optimal cut-off value, and might be combined to predict the occurrence of sepsis.
The other three studies included were studies conducted on sepsis after surgery. In the study reported by Almansa et al., data on PCT and HLA-DR gene expression were combined to detect sepsis in 154 patients after their surgeries and the HLA-DRA expression was quantified using droplet digital PCR. Sepsis-3 criteria was the gold standard for all patients [24]. When PCT alone was used to predict sepsis, the optimal cutoff value was 1.115 ng/ml, the sensitivity was 73.3%, and the specificity was 75.5%. If PCT combined with HLA-DR was used to predict sepsis, then the ratio of PCT concentration and mHLA-Dr concentration was used to predict sepsis. The OOP (optimal operating point) was 0.0003 and the specificity and sensitivity were 75.5% and 80.2%, respectively, indicating that the use of combined predictors could help improve the accuracy of sepsis diagnosis significantly. In the study reported by Asadullah et al. involving 57 neurosurgical patients, blood samples were collected one day before the operation and one to six days after operation. Fourteen patients developed postoperative infection complications. The predictive value of HLA-DR expression of less than 30% for postoperative infection was 0.9. Strohmeyer et al. studied 56 patients subjected to cardiac surgery involving cardiopulmonary bypass, In this study, using a cutoff of 5792 HLA-DR molecules per cell, the sensitivities and negative predictive values of IL-10 and IL-10 combined with HLA-DR expression were found to be 1.0, while the sensitivities and specificities of IL-10 and IL-10 combined with HLA-DR expression were 0.75 and 0.84, 0.75 and 0.94, respectively. Cheadle [25] and Albertsmeier [26] also studied the relationship between HLA-DR and postoperative sepsis. The findings of the studies on postoperative sepsis, and the changes in HLA-DR after surgery and sepsis were consistent.
The analysis of the studies showed that HLA-DR could be considered a useful indicator for the prediction of post-traumatic sepsis, with good sensitivity but low specificity; therefore, our research focus was the combination of other indicators, including clinical indicators and biomarkers, and the use of genomics, proteomics, metabolomics, and bioinformatics research methods to explore early biomarkers, to establish an accurate early warning system for post-traumatic sepsis. Although these findings warrant prospective multi center large sample size verification, we suggest that there is potential applicability in clinics to further improve the success rate of the treatment of traumatic sepsis.
There were some limitations encountered in the present study. First, we searched for relevant literature in various databases, and while many articles met the screening criteria, data of only eight articles were extracted with five articles published on blunt trauma sepsis and three published on surgical trauma sepsis. All of these data are published in English, which will lead to the lack of relevant research data in other languages. Second, in our meta-analysis, no common heterogeneity of the threshold effect was observed; however, heterogeneity of the non-threshold effect could be observed. These studies differed in several ways (e.g., clinical spectra of patients, age, gender, type of admission, and the mHLA-Dr analysis used). There was heterogeneity in the clinical data because of certain unchangeable factors. Third, a few studies used different gold standards to diagnose sepsis. Fourth, while the methodological quality of each study was acceptable, there was only one study with statistical heterogeneity. The statistical heterogeneity of the results was tested by constructing a Galbraith diagram (Fig. 8). However, none of the included studies met the QUADAS-2 criteria completely.