Establishment Of A Prediction Model For The Efficacy Of Immune Checkpoint Inhibitors In Chinese Patients With Non-small Cell Lung Cancer And Analysis And Comparison Of Gut Microbiota Structure And Metabolome Characteristics

Research Background:There are many immune checkpoint inhibitors approved in China for the treatment of non-small cell lung cancer,which have greatly improved the prognosis and overall management of NSCLC patients.But less than 20%of patients can benefit from long-term ICIs,and about 50%of patients have immunotherapy-related toxicity.Therefore,it is necessary to guide the choice of immunotherapy population through bio-marker in order to maximize the benefit of NSCLC patients.Research Purposes:1.To summarize the real-world data of immunotherapy for NSCLC population,and analyze the prognostic related factors using common clinical indicators to build a predictive model of curative effect.2.To explore the relationship between the efficacy of immunotherapy and the characteristics of specific tumor mutation genes in NSCLC population.3.To analyze the relationship between the intestinal flora structure and metabolome characteristics of NSCLC immunotherapy and the efficacy of ICIs,in order to discover the characteristics of the flora that can enhance the efficacy of immunotherapy.Experimental Method:Prospective analysis of advanced non-small cell lung cancer patients who visited the Department of Respiratory Medicine of Peking Union Medical College Hospital from March 2018 to June 2019 and were instructed to use PD-1 inhibitors.1.The primary observation endpoints were OS and PFS,and the secondary observation endpoints were ORR,DCR,and safety.Efficacy and toxicity were evaluated in accordance with the Solid Tumor Efficacy Evaluation Standard(RECIST)1.1 and the National Cancer Institute's Toxicity Criteria(NCI-CTC)4.0.LASSO regression was used for univariate analysis,and Cox regression analysis was used for multivariate analysis.The rms26 software package of R version 2.14.1 was used to develop the efficacy prediction line chart.Kaplan-Meier method was used to draw the survival-free curve.2.Dividing patients with sufficient pathological tissues for tumor mutation genes and PD-L1 detection into long-term benefit groups(DCB,CR/PR/SD? 6 months)and non-lasting benefit groups(NDB,CR/PR/SD<6 months).The chi-square test was used to analyze the association between DCB/NDB probability and molecular characteristics.DCB/NDB was used as the outcome variable,and specific gene mutations were used as independent variables for lasso regression,logistic regression,and machine learning random forest analysis to explore independent predictors related to the long-term benefits of immunotherapy.Kaplan-Meier method was used to draw the progression-free survival curve of NSCLC immunotherapy patients with or without specific gene mutations.Draw the ROC curve to evaluate the prediction ability.3.All patients collected stool samples before the start of immunotherapy.All samples were sequenced with Illumina sequencing platform after DNA extraction and library construction,and the sequencing data of all samples entered the biological information data analysis process.All data were divided into 2 groups based on PFS? 6 months and PFS<6 months.The entire analysis includes quality control,assembly,gene prediction and construction of non-redundant gene sets,gene function annotation,gene abundance quantification at the gene,species and functional level,flora composition analysis,diversity analysis,multidimensional analysis,difference analysis,Functional analysis and metabolite analysis.Experimental Results:1.A total of 63 patients were included in the study.The median age was 61 years and 36 patients were over 60 years of age.84.13%of men,46.03%of heavy smoking history,ECOG ?1 point accounted for 92.07%,stage ? accounted for 84.12%.All patients were confirmed pathologically,32 cases received first-line immunotherapy,22 cases were second-line,and 9 cases were third-line and above.Drugs used included pembrolizumab in 42 cases,nivolumab in 4 cases,and sintilizumab in 17 cases.The interval between the diagnosis of the disease and the start of immunotherapy is 1 to 38 months.The best effect was PR in 27 cases(42.8%),SD in 14 cases(22.2%),and PD in 22 cases(34.9%).The objective response rate was 42.8%,and the disease control rate was 65.1%.There were 8 patients with brain metastases and 6 with neurological symptoms.Intracranial lesions were assessed in 3 patients with PR,3 in SD,and 2 in PD after immunotherapy.ORR was 37.5%and DCR was 75%in patients with brain metastases.Thirteen patients had no disease progression by the follow-up deadline.Fourteen patients died at the end of follow-up.The median PFS was 7.0 months(95%CI 5.0-11.0 months)and did not reach the median OS.Significant univariate factors for LASSO regression were smoking index,PD-L1 expression,and NLR.The variables that were significant for LASSO regression were incorporated into the Cox proportional hazards model for multivariate analysis.The results showed that smoking index and NLR were factors that affected the prognosis of immunotherapy PFS.Independent factors.A model line chart containing NLR and smoking index was established.Common adverse reactions are rash,fatigue,digestive system toxicity,pulmonary toxicity,endocrine toxicity,hematological system toxicity,elevated liver and pancreatic enzymes.2.A total of 44 patients had sufficient pathological tissues to detect tumor mutation genes and PD-L1,20 in the DCB group and 24 in the NDB group.Specific gene mutations occurred in TP53 38.64%,KRAS 31.82%,EGFR 20.45%,BRCA 20.45%,ERBB(excluding EGFR)18.18%,PTEN 15.91%,CDK4/6 13.64%,POLE 11.36%,MET 11.36%,PIK3CA 9.10%,FGFR 9.10%,BRAF 9.10%,JAK 9.10%,ALK 6.82%,POLD1 4.55%,BLM 4.55%.Chi-square test results showed significant frequency of KRAS mutation,KRAS+TP53 combination mutation,TP53+PTEN combination mutation,PTEN mutation,JAK1/2/3 mutation,TP53 mutation,EML4 ALK mutation,and Pan-ERRB mutation between DCB and NDB groups.Statistical differences.Taking the long-term benefit as the outcome variable,the gene mutation was the independent variable for lasso regression,and the non-zero gene mutation was KRAS,EML4-ALK,PTEN,TP53+PTEN.The logistic regression showed that the KRAS mutation was statistically significant(P<0.001).Two accuracy indicators,RandomecrameGini and Meandecreaseaccuracy,of random forest classification were used to evaluate the importance of different gene mutations on the outcome.Under two different measures,the ranking of independent variables Not exactly the same,but the most important independent variable is a mutation in the KRAS gene.It is suggested that the mutation of KRAS gene is an independent predictor of the long-term benefit of immunotherapy.3.According to PFS,patients were divided into groups with PFS? 6 months(35 cases)and PFS<6 months(28 cases).Illumina platform was sequenced to obtain a total of 373.5G of original sequencing data.The number of reads in the sample was between 27,060,458-66,836,852,with an average of 43,119,830.Bacteroidetes,Firmicutes,Proteobacteria and Actinobacteria accounted for most of the bacterial communities in the stool samples studied.Compared with the PFS<6 months group,the patients with PFS?6 months had higher ? diversity in the intestinal microbiome at baseline level,which had significant statistical differences.Alpha diversity has a higher trend,but no statistical difference is seen.There are also differences in composition between the two groups.Parabacteroides and methanobrebacter are rich in the PFS?6 months group,while Veillonella,Selenomonadales,Negativicutes are rich in the PFS<6 months group.Using DESeq2 to analyze the relative abundance of each functional group(including KO,COG,and CAZy)protein groups between groups,the results suggest that a total of 390 KO,264 COG,859 CAZy functional group abundances in PFS?6 There was a significant difference between the monthly group and the PFS<6 month group.Based on the relative abundance distribution table,the KO pathway(TOP 10)with significant differences in relative abundance between the two groups was analyzed.The results showed that five groups of metabolic pathways,such as ko02040 Flagellar assembly(flagellate assembly)and ko02030 Bacterial chemotaxis(bacterial chemotaxis),were in PFS.The abundance was higher in the?6 months group.The metabolic pathways such as ko00680 Methane metabolism and ko00362 Benzoate degradation were more abundant in the group with PFS<6 months.Bacterial metabolites suggest that the potential for methanogenesis in the group with PFS<6 months is higher than that in the group with PFS?6 months(P<0.05).Conclusion:1.PD-1 inhibitors are effective and safety in the treatment of advanced non-small cell lung cancer in the real world.Smoking index and NLR are independent factors affecting the prognosis of PFS in immunotherapy.Model alignment charts including NLR and smoking index can predict 1-year progression-free survival of patients.2.The mutation of KRAS gene in tumor tissues is an independent predictor of the long-term benefit of immunotherapy,and the predictive ability is better.3.The intestinal microbiome ? diversity of patients with advanced NSCLC in China is closely related to the anti-PD-1 immunotherapy response.The composition of intestinal flora with PFS?6 months and PFS<6 months,functional group protein family,There are differences in KEGG metabolic pathways and flora metabolites.