Early and dynamic alterations of Th2/Th1 in previously immunocompetent patients with community-acquired severe sepsis: a prospective observational study

We conducted a prospective cohort study in a 60-bed general intensive care unit (ICU) of a tertiary teaching hospital in Nanjing, China. This study was approved by the Institutional Ethics Committee of Zhongda Hospital (Approval Number: 2014ZDSYLL086) before enrolment of the first participant and was in full compliance with Declaration of Helsinki. The study was retrospectively registered with ClinicalTrials.gov, NCT02883218. All the participants provided written informed consent.

Eligible patients between 18 and 90 years old that admitted to the ICU with a diagnosis of community-acquired severe sepsis according to criteria of the American College of Chest Physicians/Society of Critical Care Medicine [15] within 24 h after sepsis-induced organ dysfunction recognition, were recruited as septic group between 18 September, 2014 and 28 September, 2016. Patients were excluded if they had tumors, hematological or immunological disease, or treatment with chemotherapy agents or corticosteroids within 6 months prior to hospitalization.

Heathy volunteers coming for physical examination in and patients admitted to the ICU without a diagnosis of severe sepsis, between 18 and 90 years old, without tumors, hematological or immunological disease, or treatment with chemotherapy agents or corticosteroids within 6 months prior were recruited as healthy control and ICU control, respectively. The sample size of health controls and ICU controls would be calculated with a power of 0.80 and alpha of 0.05 by Power Analysis and Sample Size (PASS) software (PASS 2008. citation: Hintze J (2008). NCSS, LLC. Kaysville, Utah, USA) [16], based on recruitment of severe sepsis and the difference of Th2/Th1 between severe sepsis vs. health controls (2.17 vs. 0.26) and severe sepsis vs. ICU controls (2.17 vs. 0.4), as the Ferguso study showed [14].

We aimed to determine alterations of Th1 and Th2 populations and their associations with 28-day prognosis. Secondly, we stratified the septic patients according to alterations of Th2/Th1 within 1 week: (1) the early recovery group consisted of septic patients with overall decreasing Th2/Th1; (2) the late recovery group with Th2/Th1 starting to decrease after D3; (3) the non-recovery group with an overall increasing Th2/Th1, and then assessed their associations with clinical outcomes.

The onset of severe sepsis was defined as the time of diagnosis of sepsis-induced organ dysfunction. The first 24-h period after enrolment was to be D0; the next 24-h period was D1, etc. We followed the Berlin criteria for acute respiratory distress syndrome (ARDS) [17]. Acute kidney injury (AKI) was defined according to the kidney disease improving global outcomes (KDIGO) criteria [18]. Acute gastrointestinal injury (AGI) was defined according to the European Society of Intensive Care Medicine (ESICM) Working Group [19]. The occurrence of ICU-acquired infections, defined as a first occurrence of bloodstream infection, or pneumonia with a post hoc plausibility of infection rated at least possible according to strict definitions [20] within 28 days.

For septic patients, baseline demographics including age, gender, suspicious infection sites, presence of comorbidities, smoking and alcohol use histories, acute physiology and chronic health evaluation (APACHE) II score and sequential organ failure assessment (SOFA) score [21], and clinical data including systemic inflammatory indicators, such as temperatures, heart rates and count of white blood cell (WBC), as well as absolute lymphocyte cell (ALC), procalcitonin (PCT) and hypersensitive C-reactive protein (hs-CRP) on D0, D3 and D7, occurrence of complicated organ dysfunction and ICU-acquired infection, and prognosis within 28 days were collected from documentations recorded by the treating physician. For ICU controls, baseline demographics were recorded that included age, sex, primary diagnosis, presence of comorbidities, smoking and alcohol use histories, APACHE II score, SOFA score and levels of WBC, ALC, PCT as well as hs-CRP upon enrolment.

Whole blood was collected in heparin sodium anticoagulant tubes on D0, D3 and D7 for septic group and once upon enrolment for controls. T lymphocyte subpopulations were measured by flow cytometry as descripted in Additional file 1. T helper cells gating strategy used for flow cytometric analysis were shown in Fig. 1a. Data were analysed in FlowJo version 10 (Ashland, OR, USA). Levels of IFN-γ, IL-4 and IL-10 in plasma collected upon enrolment in healthy control, ICU control and severe sepsis patients were determined using commercial ELISA kits: EH008-96, EH003-96, EH006-96 (ExCell Bio, Taicang, China) according to manufacturer’s instructions. For severe sepsis group, concentrations of plasma IL-4, IL-10 and INF-γ were consecutively measured on D3 and D7.

Data of patients that could not get consecutive T helper cell measurements on D0, D3 and D7 were excluded. T helper cells populations were expressed as numbers in CD3+ CD8− T lymphocytes (%). Data were analysed using SPSS Version 23 (IBM, Chicago, Ill, USA) and GraphPad PRISM Version 5.3 (San Diego, CA, USA). Descriptive statistics, including the mean ± standard deviation (SD), median (interquartile range [IQR] defined as the 25th and 75th percentile), median ± mean squared error (SEM) and mean ± SEM were used as appropriate. Variables that failed tests of normality were presented and analysed in a nonparametric fashion. Comparisons of characteristics among cohorts, between 28-day survivors and non-survivors, among subgroups according to Th2/Th1 alterations were performed using unpaired t tests, Mann–Whitney U tests, or Chi squared tests, as appropriate. All tests were two-tailed, and a value of p < 0.05 was considered statistically significant. Correlations between Th1 populations and plasma INF-γ levels, and Th2 populations and plasma IL-4 levels were analysed with Pearson correlation test and presented as R square, when R square is more than 0.64 with a p value < 0.05, it would be regarded as a significant correlation.

Univariate analysis of cox regression was performed initially, using SPSS, on the septic cohort data (n = 71) to identify variables that were independently associated with 28-day mortality. The variables entered included demographics, severity score, inflammatory and immune indicators including WBC, ALC, Th1 and Th2 populations, PCT as well as hs-CRP and alterations of these indicators. Variables identified with a threshold of p < 0.05 were investigated for associations with 28-day mortality in a cox-proportional hazards model. Specific cox models were conducted for each variable that were mathematically coupled or collinear with each other, such as WBC and ALC, and T helper population and its alterations on D0, D3 and D7 respectively. Hazard ratios were provided for each variable included in the final model with 95% confidence intervals (CIs).

Receiver-operating characteristic (ROC) curve analysis was performed to assess the ability of the severity scores, inflammatory and immune indicators as well as its alterations to predict mortality. CIs on areas under the curves (AUCs) were calculated using nonparametric assumptions. The best threshold was determined using the Youden index with sensitivity and specificity for variables with p < 0.05 and AUC above 0.5. Kaplan–Meier analysis was performed to determine the survival lifetimes of subgroups for 28-day survival, and a log-rank test was used to compare curves.

APACHE II and SOFA scores at admission in septic patients were significantly higher than that in ICU controls, indicating a higher level of severity. By 28 days, 18 septic patients (25.4%) had died (Table 1).

Compared to septic patients who survived, patients who died presented a higher APACHE II at admission and SOFA score on D3 and D7. APACHE II score at admission and SOFA scores on D3 and D7 was associated with mortality in univariate analysis of septic cohort (Additional file 3: Table S2).

Compared to healthy controls, community-acquired severe sepsis patients had increased Th2 populations (p < 0.001) along with higher Th2/Th1 (p < 0.001) at admission (Fig. 1b, c). Compared to ICU controls, community-acquired severe sepsis patients had higher levels of PCT (p = 0.008), hs-CRP (p = 0.020), lower Th1 population (p = 0.004) and higher T helper 2 cells (p < 0.001) along with higher Th2/Th1 (p < 0.001) at admission (Table 1 and Fig. 1). Absolute counts of lymphocyte were lower (p = 0.004) whereas white blood cell counts were similar (Table 1). Levels of plasma IFN-γ, IL-4 and IL-10 in blood serum collected upon enrolment in healthy control, ICU control and severe sepsis patients were measured. Compared with the healthy control, the IFN-γ levels of septic group were significantly higher (p < 0.05). There was no significant difference of IFN-γ between ICU controls and septic groups. In addition, IL-4 and IL-10 levels were higher in septic group than that of ICU controls and healthy controls (p < 0.05, see Additional file 3: Table S1).

As Table 1 showed, inflammatory indicators such as PCT and hs-CRP were markedly elevated and gradually decreased thereafter. Patients who died within 28 days had lower levels of ALC on D3 (p = 0.02) and D7 (p = 0.02) than those of 28-day survivors with a gradual recovery within a week. By univariate analysis, ALC on D7 was a protective factor of 28-day prognosis (p = 0.013) while levels of WBC on D3 and D7 were risk factors (p = 0.005 and 0.001, respectively).

Compared to survivors, non-survivors hold significantly higher Th2 populations and levels of serum IL-4 and IL-10 on D3 and D7 as well as higher Th2/Th1 than that of 28-day non-survivors on D0 (p = 0.044), D3 (p < 0.001) and D7 (p < 0.001), showing a sign of a more pronounced anti-inflammatory host response with Th2 dominance (Fig. 2 and Additional file 3: Table S1). However, no significant correlation was found between Th1 population and INF-γ levels. In terms of Th2 populations and plasma IL-4 levels, R square of correlation is less than 0.5 with a significant p value on D3 and D7 (Additional file 2: Figure S2).

By univariate analysis, populations of Th1 on D7, Th2 on D0, D3 and D7 and values of Th2/Th1 on D3 and D7 as well as alterations of Th2 within 7 days and Th2/Th1 within 3 days and 7 days were identified to be associated with 28-day prognosis (see Additional file 3: Table S2).

The variables of severity scores, inflammatory and immune indicators of peripheral blood cell count and T helper populations identified above associated with 28-day mortality entered different cox-proportional hazards models. Finally, fourteen specific cox models were conducted for including the variables that had no interaction with the others on D0, D3 and D7 respectively. The p value and HRs with 95% CI were illustrated in Fig. 3. It demonstrated that APACHE II score at admission, SOFA score on D7, Th2 populations on D0, D3 and D7, and Th2/Th1 on D3 and D7 as well as its increase to D3 and D7 from enrollment were independently associated with risk of 28-day mortality while ALC on D7 could be protective for 28-day prognosis.

More importantly, using the ROC curves to predict 28-day prognosis, we observed that Th2/Th1 on D3 and D7 demonstrated relatively high AUCs of 0.831 and 0.869 (Fig. 4). The best cut-off values of Th2/Th1 to predict 28-day mortality were 2.95 on D3 and 2.74 on D7, which hold a sensitivity of 66.1% and 75.0% as well as a specificity of 92.5% and 95.1%, respectively (Additional file 3: Table S3).

Patients were divided into subgroups based on alterations of Th2/Th1 (Fig. 5a). The early recovery group showed a decreasing trend of Th2/Th1 values with 2.51 ± 0.46, 1.11 ± 0.22 and 0.78 ± 0.19 while those in late recovery group were 0.95 ± 0.28, 2.39 ± 0.46 and 0.96 ± 0.31 and those in the non-recovery group were 1.26 ± 0.27, 2.33 ± 0.41 and 3.52 ± 0.76 on D0, D3 and D7, respectively (Fig. 5b). As the baseline characteristics of patients in subgroups showed (Table 2), patients in early recovery group had lowest SOFA scores on D3.

All patients have obvious systemic inflammatory responses on admission (Additional file 2: Figure S3). Patients in late recovery group had higher WBC than those in early recovery at the beginning. During the following week, lymphocyte counts recovered in early recovery group and were significantly higher than those of late recovery group on D3 (p = 0.009). Patients in late recovery and non-recovery groups were presented with higher HR on D7 than that in early recovery group. It was demonstrated a higher risk of AGI in late recovery and non-recovery groups compared to early recovery (OR 3.07, 95% CI 1.32–7.14, p = 0.006; OR 2.745, 95% CI 1.132–6.657, p = 0.023, respectively, see Additional file 3: Table S4).

Patients with non-recovery of Th2/Th1 had a higher incidence of ICU-acquired infections compared with those with early recovery (64.7% vs. 22.9%, p = 0.012, Fig. 5c). In addition, there was a significant difference of survival curves among subgroups (Fig. 5d). Patients with Th2/Th1 non-recovery had the highest mortality (47.1%), followed by 31.6% of late recovery of Th2/Th1, and the lowest mortality of 11.8% in early recovery group.