| Objective:Hepatic resection(HR)remains the first-line locoregional treatment option for colorectal liver metastasis(CRLM).Due to minimal invasion,repeatability,effectiveness,safety and other advantages,thermal ablation(TA)has gained wide acceptance in the locoregional therapy of CRLM.TA,however,is usually termed as an alternative for ineligible HR therapy.Thus,we conduct this study aiming to compare the efficacy in patients with metachronous CRLM who underwent TA or HR and to explore the potential candidate for TA.Methods:Data of 634 patients with CRLM who intended to receive curative treatment in Cancer Hospital,Chinese Academy of Medical Sciences from November 2007 to January 2021 were retrospectively collected.Patients’ demographics,clinicopathological features of primary tumor,perioperative information of LM,previous treatment records,recurrence and survival status were extracted.Inclusion criteria:(1)LM(confirmed by imaging or pathology)after radical R0 resection of the primary tumor;(2)The number of LM<5 and the largest diameter of LM≤ cm;(3)LM received HR or TA treatment for the first time;(4)Age>18 years old;(5)Karnofsky performance score>80.Exclusion criteria:(1)Synchronous LM;(2)LM was treated with HR and TA at the same time;(3)Combined with other malignant tumors;(4)The largest diameter of LM>5cm;(5)The number of LM>5;(6)The primary tumor has not been surgically removed;(7)LM has already received TA or HR treatment;(8)Patient information is incomplete.To balance the baseline characteristics between the two groups,propensity score matching(PSM)was conducted.Survival analysis was generated by Kaplan-Meier method,and a log-rank test was performed.Univariate and multivariate Cox regression analysis were adopted to identify the risk factor(s)for recurrence and survival.Results:A total of 319 patients were enrolled after the screening process,including 148 in the TA group and 171 in the HR group.By 1:1 ratio to match,there were eventually 92 patients in the TA group and HR group,respectively.Median follow-up period was 60.0(47.0-80.0)months for TA and 44.0(31.0-65.0)months for HR.After matching,the median overall survival(OS)in the thermal ablation group was 49[95%confidence interval 37-76]months,however the median OS of resection group has not reached(P<0.001).The 1-,3-,and 5-year OS rates in the TA group and the HR group were 91.8%,61.0%,42.6%and 96.8%,83.0%,66.3%,respectively.The median disease-free survival(DFS)of TA group and HR group were 10.0[95%CI 8-12],33[95%CI 19-Not reached]months(P<0.001),respectively.The 1-,3-,and 5-year DFS rates in the TA group and the HR group were 38.0%,18.0%,14.5%and 72.7%,49.7%,47.9%.Compared with the HR group,the TA group demonstrated a higher hepatic recurrence rate(59.8%vs.23.9%,P<0.001),but a shorter hospital stay(7 days vs.14 days,P<0.001).Multivariate Cox regression analysis indicated that T3-4 stage of primary tumor,multiple metastases,maximum diameter of metastatic tumor 3-5 cm,preoperative serum CEA level>200 ng/ml and TA were independent risk factors for shortening OS;Lymph node metastasis of primary tumor,multiple metastases and TA were independent risk factors for shortening DFS.Through the minimum P-value approach,the optimal threshold of DFS was obtained which was located at the 10th month,and the population was divided into early recurrence and non-early recurrence accordingly.By comparing the OS receiving TA or HR in the early recurrence patients and non-early recurrence patients,we found that there was no statistical difference between the TA and HR groups in the early recurrence patients(P=0.350),but there was a statistical difference in the non-early recurrence patients(P=0.037).Conclusions:In the treatment of metachronous CRLM,HR still has its efficacy advantage over TA,but the long-term efficacy of TA is non-inferior to HR for early recurrence patients,and it can be considered as a first-line treatment option.Objective:It is quite common to encounter early recurrence after thermal ablation(TA)of metachronous liver metastasis(MLM)patients in the clinical scenario.As ER plays an important role in the survival of MLM patients,it is imperative to ascertain the risk of ER before TA,thus contributing to the clinical decision-making.Radiomics aids disease assessment and clinical decision-making by transforming digital medical imaging data into high-dimensional data,with the advantage of being able to detect occult metastases resulting in microstructural changes.Several previous studies have demonstrated that MRI-based radiomics can predict the risk of developing MLM after colorectal cancer resection.Based on this,we hypothesized that the MRI-based radiomics could predict whether a MLM patient was at risk of ER after TA.Therefore,the purpose of this study is to build a predictive model of ER in MLM patients after TA treatment based on machine learning algorithm.Methods:The 147 MLM patients who received TA treatment from the first part were screened by the exclusion criteria.According to the result from the first part,the optimal threshold for disease-free survival is located at 10th month.Thus,the recurrence within 10 months after TA was defined as early recurrence(ER).No recurrence and recurrence after 10 months were defined as non-early recurrence(NER).Patients in this study were divided into two groups,one group was ER patients after TA,and the other group was NER patients.We select the portal-venous-phase(PVP)in the contrast-enhancement MRI image to extract,transform and segment.1409 quantitative radiomics features were extracted from PVP images.Utilize the variance threshold,SelectKBest,Least Absolute Shrinkage and Selection Operator(LASSO)methods for feature selection.10-Fold cross-validation and K-Nearest Neighbor were performed to build predictive model by ML.Diagnostic evaluation of predictive model was validated by receiver work operating characteristic(ROC)curve,with the validation of Area Under the Curve(AUC),sensitivity and specificity.Statistical analysis of clinical data was the same as the first part,and the rest of statistical analysis was done on the Radcloud platform.Results:After exclusion,a total of 81 patients were eventually enrolled(47 male and 34 female;mean age,59.0 ± 9.5 years.There were 42 patients in the ER group(single metastasis:n=17,40.5%;multiple metastases:n=25,59.5%)and 39 patients in the NER group.Except one variable,the other variables in the baseline data between the ER group and the NER group were no statistical significance.451,21 radiomics features from 1409 features were selected by the variance threshold,SelcetKBest methods,respectively.Finally,8 optimal radiomics features were left after LASSO calculation.The predictive model established by the KNN algorithm has the AUC,sensitivity and specificity in the training and validation sets were 0.83(95%CI:0.75-0.91),0.82,0.77 and 0.85(95%CI:0.51-1.00),0.80,0.75,respectively.Conclusions:The predictive model based on the MRI images at PVP stage in MLM patients before TA treatment in has a good ER after treatment has a certain predictive value and is helpful for clinical decision-making. |
PDF Full Text Request