Development And Application Of The Prediction Models For The Prognosis Of H7N9 Infection Based On The Serum Expression Levels Of Cytokines/Chemokines

ObjectiveFirstly,we set out to explore the statistical difference between H7N9-infected patients and healthy controls of cytokines/chemokines in serum;then investigated the pairwise correlation between cytokines/chemokines and clinical characters(viral load,C-reactive protein,procalcitonin,oxygen index,APACHE II scores,CD4+ T cells and CD8+ T cells).Lastly,we established the prediction models of mild and death and provided help for clinical diagnosis and treatment based on the selected cytokines/chemokines.Methods24 H7N9-infected patients and 5 healthy controls were enrolled.The serum of expression levels were measured using antibody array.Depending on the disease development,24 patients were divided into admission(First),discharge or death(Last)and the process of diseases(Other)(due to the hospitalization time of mild patients was short,only the samples of two timepoints were collected(First and Other)).Levels of individual cytokines/chemokines of three disease groups were compared with those of healthy controls to determine statistically changed factors.Using the Spearman correlation coefficient,we analyzed the pairwise correlations between cytokines/chemokines and clinical parameters associated with the severity of disease(viral load,c-reactive protein,procalcitonin,oxygen index,APACHE II scores,CD4+T cells and CD8+T cells).Finally,according to different disease prognosis,the single factor prediction models that could predict fatality were established;furthermore,muiti-factor prediction models were also established with higher predictive accuracy by minimum redundancy maximum correlation(mRMR).Results(1)34 of 80 cytokines/chemokines measured in the first stage were significantly elevated in the plasma samples of H7N9-infected patients using Wilcoxon rank sum test.Similarly,there were 34,35 cytokines/chemokines elevated in the other two stages,respectively.While IGFBP-3 is significantly decreased in three stages.(2)The plasma levels of BLC and IP-10 were found to relate to viral load,C-reactive protein,procalcitonin,oxygen index,APACHE II scores,CD4+ T cells and CD8+ T cells.Among them,the plasma levels of BLC and IP-10 were positively correlated with CRP,PCT and APACHE II scores and negatively correlated with CD4+T cells,CD8+T cells,OI and Ct values.Besides,the plasma levels of IL-8 was positively correlated with CRP,PCT and APACHE II scores and negatively correlated with CD4+T cells,CD8+T cells and OI.The plasma levels of MCP-3 was positively correlated with CRP,PCT and APACHE II scores and negatively correlated with CD4+T cells,CD8+T cells and Ct values.Furthermore,the plasma levels of CCL24,IGFBP-2,OPN,OPG,HGF and IL-10 were correlated to the five of seven clinical parameters.(3)The profile of BLC,IGFBP-1,MIF,OPN,IGFBP-2,IP-10,HGF,MIP-3?,IL-8,SCF and OPG were independent predictors for fatality(AUC?0.80)and the AUC of BLC,IGFBP-1,MIF and OPN were greater than or equal to 0.9.(4)Multi-factor model based on the profile of BLC,IL8,IGFBP1,MIF,GDNF and HGF can predict fatality;Multi-factor model based on the profile of IL8,IGFBP2,Leptin,ENA78,IP10,TIMP2,THPO,IL10 and HGF can predict mildness.Two indexes were established with low error rates.Conclusions(1)The hypercytokinemia occured in H7N9-infected patients.(2)The plasma levels of BLC and IP-10 were correlated to the severity of disease.(3)The profile of BLC,IGFBP-1,MIF,OPN,IGFBP-2,IP-10,HGF,MIP-3?,IL-8,SCF and OPG were independent predictors for fatality.(4)Multi-factor model based on the profile of BLC,IL8,IGFBP1,MIF,GDNF and HGF could predict fatality with zero error rates.