Individualized Diagnosis And Treatment Strategy Of Pancreatic Cancer Based On Early Recurrence Risk Model And Next Generation Sequencing

IntroductionAs one of the most malignant tumors,pancreatic ductal adenocarcinoma（PDAC）is becoming a major threat to human health.Researchers predict that it will rise to the second leading cancer-related death in 2030.However,the improvement in survival rate over the past 30 years has been limited,with 5-year survival rates of 3%,3.6% and 7.2% in 1975,1995 and 2012,respectively.The 5-year survival rate is still less than 30% even after primary surgery and post-operative chemoradiotherapy under the current standard treatment procedure.Early recurrence after surgery and tumor heterogeneity are the two main reasons for poor prognosis of pancreatic ductal adenocarcinoma after standard treatment.In recent years,neoadjuvant therapy has been increasingly used in gastrointestinal malignant tumors because it is considered to reduce tumor burden,increase R0 resection rate and reduce micrometastasis.Therefore,a better stratification of high-risk subgroups with early recurrence and poor prognosis is essential.It can avoid surgical treatment for high-risk patients with early recurrence,who would benefit from other preoperative neoadjuvant therapy strategies.At present,there is no consensus of the accurate time of early recurrence of pancreatic cancer after radical surgery,therefore the analysis of the risk factors for early recurrence in previous studies only reflects the different research purposes,and the real preoperative high risk factors affected early recurrence have not been found yet.Some previous studies have suggested that tumor size,CA19-9 level,tumor grade,margin of resect,lymph node invasion,nerve invasion,portal vein invasion,and systemic immune status are the main risk factors for predicting early recurrence and prognosis.However,many of the above factors can not be obtained before surgery,so there is no reliable biomarker or recurrence model for clinical treatment decision-making reference;and in recent years,with the rise and promotion of next-generation sequencing technology,it is urgent to find a prognostic biomarker to stratify pancreatic cancer patients more accurately before surgery.ObjectivesIn order to stratify the patients with PDAC more precisely before surgery and help treatment decisions making,we determine the accurate definition of the early recurrence time after PDAC surgery,explore the related risk factors leading to early recurrence,and establish a predictive model of clinical related factors for early recurrence after surgery;also we establish a Chinese PDAC gene mutation landscape based on the results of targeted gene sequencing,and exploring the correlation between mutation status and the outcome of radical PDAC surgery.MethodsPart 1The training set samples are collected from patients who underwent pancreatectomy（including PD/DP with or without PV/SMV/CA）in department of pancreatic surgery of Xx Hospital from 2015 to 2017,and are pathologically confirmed as pancreatic ductal adenocarcinoma,invasive intraductal papillary myxoma（IPMN-related cancer）,adenosquamous carcinoma and undifferentiated carcinoma.The validation set samples are collected from patients who underwent pancreatectomy in other hospital in 2018.The operation methods and pathological confirmation requirements are the same as the training set,and the follow-up time is more than 6 months.Preoperative and post-operative demographic,clinical,pathological and therapeutic data are collected from prospective databases,including gender,age,chief complaints,body mass index（BMI）,family history of cancer,history of alcohol and smoking,history of diabetes,Charlson’s age-adjusted comorbidity index（CACI）,preoperative tumor location in images,preoperative T staging in images,preoperative lymph node enlargement in images,preoperative maximum tumour size in images,preoperative resectability in images,preoperative CEA and CA19-9,platelet count,neutrophil count,lymphocyte count,monocyte count,operation time and mode（if combined with vascular resection）,classification of tumor differentiation（well,moderate and poor）,pathological T and N stages,perineural and Lympho-vascular invasion,resection margin,adjuvant treatment after resection,recurrence site and recurrence time,death time or last follow-up time.Statistical analysis is made on the above data.Minimum P-value method is used to determine the optimum cut-off value of Early,Mid or Late/Non-recurrence time after operation.Chi-square test is used for differences between these two or three groups.After determining the time of early recurrence,the Mid-recurrence and the Late/Non-recurrence groups are included together into the Non-Early recurrence group.Kaplan-Meier curve is used to calculate the best cut-off values of each variable,which are confirmed by Log-Rank test.The preoperative variables whose p value less than 0.1 in univariate analysis are included in COX proportional risk model to explore the independent risk factors affecting early recurrence after surgery.P < 0.05 is statistically significant in the above analysis.Statistically significant independent risk factors associated with early recurrence are included in Nomogram to establish a predictive model for early recurrence after surgery,and validated by the validation set of other hospitals.We use SPSS 23.0 software to analyze Chi-square test,Kaplan-Meier curve and Log-Rank test,COX proportional risk model between groups,and R3.5 and toolkits（such as grid and rms）to generate and validate Nomogram.Part 2Samples are collected from patients who underwent pancreatectomy（including PD/DP with or without PV/SMV/CA）in department of pancreatic surgery of Xx Hospital from 2015 to 2017,and are pathologically confirmed as pancreatic ductal adenocarcinoma and adenosquamous carcinoma.Targeted gene sequencing is performed after pathological confirmation of the resected samples.Demographic,clinical,pathological and therapeutic informations are extracted from the prospective database,including gender,age,preoperative CA19-9,preoperative tumor location in images,classification of tumor differentiation（well,moderate and poor）,perineural and lympho-vascular invasion,resection margin,recurrence time,death time or last follow-up time.The ratio of four important major driving genes is calculated and compared with the top 20 mutation gene landscape of 591 patients in three studies: ICGC,QCMG and UTSW downloaded from http://www.cbioportal.org.Samples with complete data are selected for subsequent survival analysis.After univariate analysis,P < 0.1 variable was included in COX proportional risk model.The correlation between four major driving gene status and DFS,OS was evaluated by calculating P value,risk ratio（HRs）and 95% CI using multivariate adjusted COX proportional risk model.Subgroups of mutation numbers and G12 status of KRAS were evaluated Univariate analysis.P < 0.05 is statistically significant in the above analysis.SPSS 23.0 software was used for statistical analysis of the above work.ResultsPart 11.General conditions of enrolled patients: Three hundred and twenty-four cases are included in the training set from January 2015 to November 2017,and 68 cases are included in the validation set from January 2018 to June 2018.During the follow-up period,235 patients（72.53%）have recurrence events in the training set.The median time of recurrence is 377 days（95% CI,279.69-474.30）,and the median survival time is 566 days（95% CI,455.08-676.91）.2.The best threshold for early recurrence after radical resection of pancreatic cancer is 162 days.3.The variables associated with early recurrence are: abdominal pain or back pain with jaundice,lymph node enlargement in images,preoperative resectability in images,preoperative CA19-9 more than 210 k U/L,preoperative neutrophil-lymphocyte ratio（NLR）more than 4.2,the p values of the above variables are all <0.05.4.A Nomogram based on these five risk factors can predict the risk of early recurrence in patients with pancreatic cancer after radical surgery,which is validated by cases from other hospitals.Part 21.From January 2015 to May 2018,three hundred and two patients confirmed pancreatic ductal adenocarcinoma,and adenosquamous carcinoma by pathology and 521 tumor-related genes are deeply sequenced.The average sequencing depth is 700×.The mutation rates of four major mutations,KRAS,TP53,SMAD4 and CDKN2 A are 262（86.75%）,171（56.62%）,51（16.89%）,and 53（17.55%）,respectively.There are 116 patients with KRAS_G12D mutation,105 patients with KRAS_G12V mutation,5 patients with KRAS_G12C mutation,and 36 patients with other KRAS mutations in KRAS subgroup,respectively.2.From January 2015 to May 2018,one hundred and fifty patients are included in survival analysis.After adjusting the clinicopathological factors of P < 0.1,including preoperative CA19-9,tumor differentiation,perineural and Lympho-vascular invasion and recection margin,only KRAS mutation is associated with the worse disease-free survival time of the four important gene mutations,P=0.025（HR 2.084,95%CI 1.099-3.952）.There is no significant correlation between the four major gene mutations and the overall survival time.In univariate analysis,the number of mutations in four major genes is not significantly correlated with disease-free survival time and overall survival time,with P values of 0.353（HR 1.137,95%CI 0.866-1.493）and 0.346（HR 1.140,95%CI 0.868-1.498）,respectively.The disease free survival time and overall survival time of patients with KRAS_G12D mutation are worse than those of other mutations in KRAS and those without KRAS mutation,with P values of 0.004（HR 2.589,95%CI 1.361-4.924）and 0.018（HR 2.548,95%CI 1.170-5.549）,respectively.Conclusion1.The best threshold for early recurrence after radical resection of pancreatic cancer is 162 days.2.The variables associated with early recurrence are: abdominal pain or back pain with jaundice,lymph node enlargement in images,preoperative resectability（BR or UR）in images,preoperative CA19-9 more than 210 k U/L,preoperative neutrophil-lymphocyte ratio（NLR）more than 4.2.A nomogram based on these five risk factors can predict the risk of early recurrence in patients with pancreatic cancer after radical surgery,which is validated by cases from other hospitals3.We discover the gene mutation landscape of 302 Chinese patients with pancreatic cancer.The major mutation rates of KRAS,TP53,SMAD4 and CDKN2 A in Chinese pancreatic cancer patients are 262（86.75%）,171（56.62%）,51（16.89%）and 53（17.55%）respectively.The results are similar to those in Western patients.4.Of the four major gene mutations,only KRAS mutation is associated with worse disease-free survival.The patients with KRAS_G12D mutation have worse disease-free survival and overall survival time than those with other KRAS mutations or those without KRAS mutation.