Evaluating Metagenomic Next Generation Sequencing For Diagnosing Pathogens Of Lower Respiratory Tract Infection And Investigation Of Etiology Spectrum Among ICU Immunocompromised Pneumonia

Objective Aim to study the clinical feature,etiology and outcome of patients with rheumatic diseases admitted to ICU due to respiratory failure and also to evaluate the performance of metagenomic next generation sequencing(mNGS)using adequate criteria for detection of pathogen in lower respiratory tract(LRT)samples with paired comparison to conventional microbiology tests(CMT).MethodsPART Ⅰ A retrospective study included patients with rheumatic diseases admitted to ICU with confirmed pneumonia between 2014 and 2018.We collected demographics,type of rheumatic diseases and daily management,lab investigation for respiratory pathogen,ICU intervention and outcome.Etiology of respiratory failure was classified.Independent risk factors of hospital mortality were identified by multivariate logistic regression model.PART Ⅱ A retrospective study reviewed patients with suspected pneumonia from 4 different ICUs in mainland China during 2018,whose LRT samples sent for both mNGS and CMT to investigate causative agents.We also collected demographics,comorbidities and daily management,lab results indicating immune status,ICU intervention and outcome.RPMsample/NTC and SDSMRN were chosen to construct mNGS positive criteria.McNemar test was used for paired comparison analysis between mNGS and CMT.Outliers seeking and machine learning logistic regression model were applied to set up cut-off value of RPM in order to identify potential pathogen from non-sterile specimens by mNGS technique.ResultsPART Ⅰ 205 patients were counted into final analysis.Patients with systemic lupus erythematosus accounted for the largest proportion of this study group(n=65,31.7%).The top 3 pathogens were revealed as Pneumocystis jirovecii(n=103,50.2%),followed by Aspergillus spp.(n=86,42%),and Cytomegalovirus(n=60,29.3%).127 patients(61.9%)died during hospital stay.Polymyositis/dermatomyositis(OR 3.750,95%Cl 1.190 to 11.815,P=0.024),sever ARDS(PaO2/FiO2<100 mmHg)(OR 3.952,95%CI 1.900 to 8.218,p<0.001),CD4+T cell count<200*106/L(OR 3.59595%CI 1.488 to 8.685,p=0.004)and ICU acquired infection(OR 2.496 95%CI 1.188 to 5.244,p=0.016)were 4 independent predictors for hospital mortality.PART Ⅱ 149 cases were counted into final analysis.RPMsample/NTc criterion performed better with a higher accuracy for bacteria,fungi and virus than SDSMRN criterion(bacteria[RPMsample/NTC vs.SDSMRN],65.1%vs.55.7%;fungi,75.8%vs.71.1%;DNA virus,86.3%vs.74.5%;RNA virus,90.9%vs.81.8%).mNGS was superior in bacteria detection only if using SDSMRN>3 as positive criterion with a paired comparison to culture(SDSMRN positive,92/149[61.7%];culture positive,54/149[36.2%];p<0.001),but outperformed with significantly more fungi and DNA virus identification by choosing both criteria for positive outliers(fungi[RPMsampie/NTC vs.SDSMRN vs.culture],23.5%vs.29.5%vs.8.7%,p<0.001;DNA virus[RPMsample/NTc vs.SDSMRN vs.PCR],14.1%vs.20.8%vs.11.8%,p<0.05).Based on outlier analysis,we put parameters like logRPM,rank,Whether a respiratory pathogen or not,whether fungus or not,whether respiratory virus or not into the machine learning logistic regression model(LRM).Our model predicted microbe mNGS detected as pathogen with AUROC of 0.85.After computing logRPM with LR score diagram,we reached agreement that RPM<10 could better distinguish colonized microbe when applying mNGS for LRTI aetiology investigation.Conclusion Opportunistic infection was the most common reason for respiratory failure among immunocompromised host,and often resulted in high hospital mortality.A better diagnostic strategy would help initiate anti-infective treatment on time and improve clinical outcome of these high-risk patients.When treating suspicious pneumonia in critically ill immunocompromised patients,and where there is a limitation of conventional microbiology tests,mNGS is valuable detecting potential fungal by wisely chosen RPMsample/NTC criterion to optimize its accuracy.RPM parameter would allow clinicians to interpret mNGS reports regardless of different sequencing platforms and bioinformatic pipeline.Implementing LRM model could also make RPM possible for colonization distinguishment in LRT samples.