Patient Characteristics
In the winter of 2018–2019, 280 patients with suspected influenza and accompanied LRTI from the four hospitals were screened. Of these patients, 173 were excluded: 166 had no laboratory positive test evidence for influenza, and 7 had recently been treated with hormones or immunosuppressants for autoimmune disease or solid organ transplantation. The final enrolled cohort consisted of 107 patients with I-LRTI, 85 from Beijing Chao-Yang Hospital, 9 from Beijing Di-Tan Hospital, 11 from Mi-Yun Teaching Hospital, and 2 from Beijing Huai-Rou Hospital. All the 107 patients were diagnosed with influenza A, 71 (66.4%) of them with the H1N1 subtype. None of them were documented as having received an influenza vaccine. According to the 28-day mortality, the study cohort was divided into survival (80/107) and non-survival (27/107) groups. Flow chart was showed in Fig. 1.
Differential Regulation of Immunoglobulin Proteome and Clinical Indicators between Survival/Non-survival Groups
Each patient in the study cohort was investigated for demographic characteristics (age, sex, and body mass index (BMI)), days from onset to ED, co-morbidities (active cancer, chronic respiratory disease, coronary artery disease, chronic heart failure, chronic hepatopathy, chronic kidney disease, diabetes mellitus), vital signs (heart rate (HR), respiratory rate (RR), mean arterial pressure (MAP), and Glasgow Coma Scale (GCS)), clinical laboratory tests (blood routine examination, blood biochemistry, D-dimer, arterial blood gas analysis, Glycosylated hemoglobin (HbA1C), C-Reactive Protein (CRP), procalcitonin (PCT), and T-cell subset counts), microbiological detections, antiviral administrations, organ supports, and immunoglobulin quantification (IgG, IgG1-4, IgA, IgA1-2, IgM). All of the above indicators were compared between the two groups (Table 1), and the indicators with significant or marginal differences (lymphocyte count (LY), monocytes count (MO), CD3+CD4+T-cell count, CD3+CD8+T-cell count, IgA, IgA1, IgG2, IgG4, CRP, PCT, D-dimer, oxygenation index (OI), HbA1C, lactic acid (LAC), base excess of blood (BEB), lactic dehydrogenase (LDH), and α-hydroxybutyrate dehydrogenase (HBDH)) were also described in Fig. 2. The KS test indicated that LY, BEB, D-dimer, HbA1C, LDH, HBDH, CRP, PCT and IgG2 between survival and non-survival groups are statistically differentially distributed, and meanwhile MO, LAC and IgA are of marginally statistically differentially distributed. To measure effects of age, sex and days from onset on the concentration of Ig subclasses, generalized multivariate linear analysis was performed, in which the concentration of Ig subclass is taken as the dependent variable and age, sex, days from onset and the survival/non-survival group together as the independent variables. The results suggested that IgG2 concentration is statistically impacted with age, and IgA concentration with both age and days from onset.
Table 1
Comparison of clinical features and immunoglobulin proteome quantification between the survivors and non-survivors in patients with I-LRTI
Variables | Total n = 107 | Survivors n = 80 | Non-survivors n = 27 | p value |
Age, years, median(IQRa) | 64(50,79) | 63(49,78) | 67(56,81) | 0.254 |
Age group, n(%) | | | | |
age < 45 | 15(14.0) | 13(16.3) | 2(7.4) | 0.410 |
45 ≤ age < 65 | 42(39.3) | 32(40.0) | 10(37.0) | 0.785 |
age ≥ 65 | 50(46.7) | 35(43.8) | 15(55.6) | 0.288 |
Sex (male), n(%) | 67(62.6) | 51(63.8) | 16(59.3) | 0.677 |
BMI, (kg/m2), median(IQR) | 23.7(21.6,26.6) | 23.7(21.5,26.6) | 24.2(22.0,26.6) | 0.769 |
Co-morbidities, n(%) | | | | |
Active cancer | 3(2.8) | 3(3.8) | 0(0) | 0.570 |
Chronic respiratory disease b | 13(12.1) | 11(13.8) | 2(7.4) | 0.595 |
Coronary artery disease | 25(23.4) | 19(23.8) | 6(22.2) | 0.871 |
Chronic heart failure | 5(4.7) | 4(5.0) | 1(3.7) | 0.999 |
Chronic hepatopathy | 3(2.8) | 2(2.5) | 1(3.7) | 0.999 |
Chronic kidney disease | 10(9.3) | 7(8.8) | 3(11.1) | 0.999 |
Diabetes mellitus | 27(25.2) | 20(25.0) | 7(25.9) | 0.924 |
Days from onset to ED, median(IQR) | 4 (2,7) | 3.5(2,7) | 5(3,7) | 0.124 |
NAIs c administration, n(%) | 90(84.1) | 70(87.5) | 20(74.1) | 0.178 |
NAIs administration within 48 hours, n(%) | 10(9.3) | 9(11.3) | 1(3.7) | 0.434 |
Bacterial co-infection, n(%) | 17(15.9) | 11(13.8) | 6(22.2) | 0.461 |
S. aureus co-infection, n(%) | 7(6.5) | 3(3.8) | 4(14.8) | 0.066 |
Virus co-infection, n(%) | 23(21.5) | 16(20.0) | 7(25.9) | 0.517 |
Mechanical ventilation, n(%) | 27(25.2) | 9(11.3) | 18(66.7) | < 0.001 |
Vasoactive agents, n(%) | 18(16.8) | 8(10.0) | 10(37.0) | 0.003 |
Heart rate, beats/min, median(IQR) | 97(81,112) | 97(81,112) | 104(81,118) | 0.403 |
Respiratory rate, breaths/min, median(IQR) | 22(20,27) | 21(20,26) | 22(20,29) | 0.690 |
Mean arterial pressure, mmHg, median(IQR) | 95(85,109) | 95(85,107) | 96(82,110) | 0.971 |
Glasgow Coma Scale, median(IQR) | 15(15,15) | 15(15,15) | 15(15,15) | 0.574 |
Clinical parameters, median(IQR) | | | | |
White blood cell, WBC, (× 109/L) | 7.72(4.79,10.79) | 7.94(5.01,10.55) | 6.86(4.43,11.33) | 0.846 |
Neutrophil, NE, (× 109/L) | 6.19(3.77,8.81) | 6.20(3.75,8.75) | 6.07(3.85,10.30) | 0.578 |
Lymphocyte, LY, (× 109/L) | 0.71(0.44,1.24) | 0.86(0.50,1.37) | 0.50(0.38,0.79) | 0.012 |
Monocyte, MO, (× 109/L) | 0.36(0.16,0.62) | 0.40(0.19,0.62) | 0.26(0.09,0.61) | 0.050 |
NK cell, (/µL) | 9.60(5.73,13.15) | 9.55(7.20,16.43) | 9.75(5.43,13.15) | 0.652 |
CD3+CD4+T cell, (/µL) | 199.50(114.25,355.00) | 201.50(130.75,386.50) | 169.00(87.50,238.00) | 0.077 |
CD3+CD8+T cell, (/µL) | 117.50(70.25,208.00) | 123.50(78.50,278.75) | 107.00(45.75,174.50) | 0.080 |
D-dimer, (mg/L FEU) | 1.48(0.89,3.54) | 1.21(0.72,2.18) | 2.43(1.85,13.34) | < 0.001 |
Lactate dehydrogenase, LDH, (U/L) | 355.0(223.0,576.7) | 302.0(216.5,511.1) | 482.0(302.0,642.0) | 0.019 |
α-hydroxybutyrate dehydrogenase, HBDH, (U/L) | 275.5(182.0,423.3) | 236.0(181.0,367.3) | 375.5(259.5,473.0) | 0.008 |
Glycosylated hemoglobin, HbA1C, (%) | 6.1(5.8,6.9) | 6.1(5.7,6.7) | 6.6(6.1,7.7) | 0.015 |
Lactic acid, LAC, (mmol/L) | 1.2(1.0,1.7) | 1.1(0.9,1.7) | 1.5(1.2,1.9) | 0.023 |
Oxygenation index, OI | 263(151,381) | 268(162,393) | 194(116,294) | 0.059 |
HCO3−, (mmol/L) | 21.1(18.6,23.9) | 22.5(20.0,24.9) | 20.0(17.1,22.5) | 0.023 |
Base Excess of Blood, BEB, (mmol/L) | -1.0(-3.8,0.9) | -0.6(-3.3,1.2) | -3.7(-5.0,-0.4) | 0.010 |
C-Reactive Protein, CRP, (mg/dL) | 12.1(4.2,22.1) | 9.6(3.0,22.8) | 19.6(14.2,25.8) | 0.018 |
Procalcitonin, PCT, (ng/mL) | 0.22(0.08,1.01) | 0.16(0.06,0.60) | 0.91(0.23,6.43) | 0.001 |
Plasma immunoglobulin quantification, (µg/mL), mean ± SD | | | | |
IgA | 4277.955517 ± 1751.127383 | 4495.494011 ± 1799.218731 | 3633.397017 ± 1445.182929 | 0.053 |
IgA1 | 3251.871483 ± 1575.089408 | 3411.345318 ± 1646.654368 | 2779.356417 ± 1251.261989 | 0.087 |
IgA2 | 326.738305 ± 148.035244 | 330.15051070 ± 139.051213 | 316.628064 ± 174.505309 | 0.347 |
IgG | 22497.252537 ± 10921.188428 | 22082.874174 ± 10369.593861 | 23725.040279 ± 12547.880435 | 0.596 |
IgG1 | 6682.518079 ± 3470.015276 | 6569.600710 ± 2892.282678 | 7017.088061 ± 4849.447558 | 0.628 |
IgG2 | 167.928721 ± 121.118981 | 189.688541 ± 129.383245 | 103.455179 ± 56.304086 | < 0.001 |
IgG3 | 24.179282 ± 33.797908 | 25.921704 ± 36.743783 | 19.016550 ± 22.755197 | 0.168 |
IgG4 | 21.765171 ± 25.443257 | 21.888248 ± 20.932293 | 21.400501 ± 36.162335 | 0.056 |
IgM | 3515.750491 ± 2696.014576 | 3578.764165 ± 3087.582291 | 3329.043310 ± 786.461713 | 0.683 |
a IQR: inter-quartile range; |
b Chronic respiratory disease: asthma, chronic obstructive pulmonary disease, pulmonary fibrosis, bronchiectasis; |
c NAIs: Neuraminidase inhibitors. |
Co-regulatory Network Analysis of Clinical Indicators and Immunoglobulin Proteome Quantification in the Survival/Non-survival Groups
Co-regulatory network was used to analyze and compare the correlation characteristics across clinical indicators and immunoglobulin quantification for the survivors and the non-survivors respectively, and was showed by hierarchical correlation clustering heatmap (Fig. 3). Red color represents a positive correlation, blue color represents a negative correlation, and the darker the color, the stronger the correlation. In the heatmap of the survival group, different variables were respectively separated into two red-colored positively correlated clusters (Fig. 3A). Cluster 1 included IgA, IgA1-2, IgG, IgG1-4, WBC, NE, MO, NK cell, and age, and these variables were related to immunity. Cluster 2 included CRP, PCT, D-dimer, LDH, LBDH, HR and RR, and these variables revealed the severity of infection. In the heatmap of the non-survival group, the indicator clusters were re-clustered (Fig. 3B). Cluster 1 included days from onset to ED, IgM, sex, and GCS. Cluster 2 contained indicators of the severity of the infection (LAC, LDH, HBDH, MAP, HR, RR, D-dimer, CRP and PCT). Cluster 3 included partial immunoglobulin family molecules (IgA, IgA1, IgG, IgG1, IgG2, IgG4) and BMI. Cluster 4 included both cell-immune-related indicators (age, WBC, NE, MO, LY, CD3+CD4+T-cell, and CD3+CD8+T-cell), OI, HbA1C and bacterial co-infection. Cluster 5 included NK cell, IgG3, IgA2, BEB and virus co-infection.
Predictive Model of Clinical Outcomes
We investigated the potential of these variables in the prediction of survivors/non-survivors, and calculated the AUCs of these differentially regulated variables. Based on the statistical difference, we performed iterative recursive feature selection and found the optimal combination of features. Logistic regression-based classifier was trained to discriminate survivors and non-survivors. To evaluate the predictive efficacy of the trained model objectively, LOO cross validation was employed to ensure the mutual independence between the training dataset and the testing dataset. In LOO cross validation, the training and validation steps were repeated N times, and for each step one sample was selected for validation and the rest were used for model training iteratively (N denoted with the sample number). Then random sampling (sampling ratio 95%, iteration times 1000) was performed to calculate the 95% confidence intervals (CIs) of the AUC of the prediction model. The AUCs of IgG2, IgA and the combination the two were 0.725 (95% CI 0.702–0.748), 0.575 (95% CI 0.404–0.694) and 0.737 (95% CI 0.714–0.760), respectively (Fig. 4A). The combined prediction model with nine variables (D-dimer, days from onset to ED, IgA, IgG2, LAC, LY, MO, Staphylococcus aureus (S.aureus) co-infection and age) reached the maximum AUC value of 0.810 (95% CI 0.755–0.839) (Fig. 4B). The AUCs of D-dimer, LAC, LY, and MO were 0.722 (95% CI 0.690–0.751), 0.662 (95% CI 0.634–0.684), 0.644 (95% CI 0.607–0.673), and 0.572 (95% CI 0.536–0.609), respectively. All the pairwise comparisons across each of the single-factor models with the combination prediction model were of statistical significance.
To further evaluate the selected optimal indicators for the prediction model, the unsupervised participant-feature hierarchical correlation clustering was also performed and the participants were separated into two clusters. The majority of the actual non-survivors were discriminated into the same cluster, which indicated the consistency between the clustered results and the actual grouping (Fig. 4C). This proved the feature combination’s discriminative efficacy of the survivors/non-survivors.
Validation of the Model by Disease Severity Evaluation
The risk of progression to critical conditions, including mechanical ventilation requirements or vasoactive agent requirements, can also be evaluated using the predictive model and for mechanical ventilation or vasoactive agent apparently there exists statistical difference between survivors and non-survivors in practice. For the requirement of mechanical ventilation, 15 (15/37) predicted non-survivors and 12 (12/70) predicted survivors should receive mechanical ventilation respectively, exhibiting the statistical difference between two predicted groups (Fig. 5A), which is in consistency with the actual facts (In the study cohort, 18 (18/27) non-survivors and 9 (9/80) survivors actually received mechanical ventilation (Fig. 5B)).For the requirement of vasoactive agent, 10 (10/37) predictive non-survivors and 8 (8/70) predicted survivors should receive vasoactive agents respectively, exhibiting the marginally statistical difference between two predicted groups (Fig. 5C), which is in accordance with the actual facts (In the study cohort, 10 (10/27) non-survivors and 8 (8/80) survivors received vasoactive agents, respectively (Fig. 5D)). The results of both mechanical ventilation and vasoactive agents demands proved the potential of the predictive model.