Medical Imaging And Texture Analysis In Response Evaluation And Prognosis Prediction For Metastatic Renal Clear Cell Carcinoma Patients Treated With PD-1 Inhibitor Therapy

Part Ⅰ:Radiological Tumor Response Assessment in Metastatic Clear Cell Renal Cell Carcinoma Patients Treated with PD-1 InhibitorsObjective1.To evaluate radiological tumor response patterns in metastatic Clear Cell Renal Cell Carcinoma(mccRCC)patients treated with PD-1 inhibitors.2.To compare the difference in response assessment according to iRECIST versus RECIST 1.1 in mccRCC patients treated with PD-1 inhibitors.Material and methodsThis prospective study enrolled 42 mccRCC patients who received PD-1 inhibitors after standard treatment failure from June 2015 to April 2019.Body CT and head MRI were used to monitor the treatment.Two radiological criteria RECIST 1.1 and iRECIST were used to evaluate the tumor response.Tumor growth rate(TGR)before and after therapy initiation was calculated.Response patterns were analyzed.We assessed the time to progression(TTP)using the Kaplan-Meier estimator.The differences in tumor response assessments were compared between two criteria using the log-rank test.The intra-and inter-observer agreement for iRECIST and RECIST 1.1 were also analyzed by using Kappa analysis.Results1.A total of 42 mccRCC patients treated with PD-1 inhibitors were analyzed.The objective response rate throughout therapy was 40.5%(95%CI:32.1-62.9)by iRECIST and 26.2%(95%CI 13.6-46.4)by RECIST 1.1.Time-to-progression(TTP)by iRECIST was longer than TTP by RECIST1.1(median TTP:not reached vs.12.3 months).2.iRECIST and RECIST 1.1 were discordant in 8 patients,which were evaluated as immune unconfirmed PD by iRECIST and PD by RECIST 1.1.Six patients(14%6/42)were captured as pseudoprogression by iRECIST,of which four demonstrated early pseudoprogression and two are delayed pseudoprogression.3.Significant difference for tumor response assessments between the two criteria was noticed(P<0.001).No patient demonstrated hyperprogression during the study.Conclusion1.Our study confirmed that iRECIST criteria are more capable to capture immune-related atypical responses during immunotherapy,while conventional RECIST 1.1 might underestimate the benefit of PD-1 inhibitors.2.Atypical response is not rare in mccRCC patients during PD-1 inhibitor therapy.Hyperprogression remains an almost undiscovered entity.3.Pseudoprogression demonstrated variability and may last beyond the timeframe recommended by iRECIST with a maximum of 8 weeks,indicating a limitation of the current strategy for tumor response monitoring in immunotherapy.Part Ⅱ:Prognostic value of tumor response dynamics and metastases characters in metastatic renal cell carcinoma patients treated with PD-1 inhibitorsObjective1.To evaluate the tumor response dynamics in advanced renal clear cell carcinoma(CCRCC)patients treated with PD-1 immune checkpoint inhibitor,and to determine the value of a noninvasive imaging biomarker in predicting progressionfree survival(PFS).2.To investigate the site-specific metastases response patterns in advanced CCRCC patients and to clarify the correlation between individual lesion character and prognosis.Material and methodsThis prospective study enrolled 42 mccRCC patients who received PD-1 inhibitors after standard treatment failure from June 2015 to April 2019.Body CT and head MRI were used to monitor the treatment.The tumor burden was calculated as the sum of all measurable lesions on CT or MRI at baseline.Tumor burden and individual lesion dynamics were monitored by imaging every 8-10 weeks.The lesion response rate(LRR)and the overall response rate(ORR)were determined according to immune Response Evaluation Criteria in Solid Tumors(iRECIST).Tumor growth rate(TGR)was calculated to quantify the dynamics of tumor burden and studied for the association with PFS.Multivariate logistic regression was used to examine the correlation between lesion characteristics and tumor responses.ResultsA total of 158 measurable metastases were studied in 42 patients with mccRCC.1.The ORR was 40.5%(17/42),CR in 7 cases(16.67%),PR in 10 cases(23.81%).The median change of total tumor burden was-11%.2.No significant difference in tumor burden between CR Group(N=7)and non-CR group(P>0.05).The tumor burden in the response group(N=17)was significantly lower than that in the non-response group(N=25,95 mm vs 131 mm,P<0.05).The number of metastases in the response group was less than that in the non-response group(3 vs 5,respectively,P<0.0001)2.The diameter change in individual lesion was-100%to 1015%from baseline,median-11%.LRR was 44%(42/96,95%CI:34-54,iCRL 5 cases,iPRL 37 cases).3.The mean diameter of iCRL was significantly smaller than that of non-iCRL(22.81 mm vs 31.61 mm,P<0.05)4.There was a significant difference in the mean diameter between the response lesion group and the non-response lesion group(26.6 mm vs 36.27 mm,P<0.05)5.PFS was longer in patients with lung metastases than in patients without(median PFS 30.1 vs 7.3,P<0.05),and shorter in patients with liver metastases than in patients without(median PFS 21.6 vs 6.95,P<0.05);6.Patients with TGR<70 had longer PFS than those with TGR≥7(median PFS was 23.4 vs 7.3,P<0.05).Conclusion1.A TGR<70 at first follow-up during anti-PD-1 therapy was associated with longer PFS,indicating a promising noninvasive imaging biomarker of treatment benefit.2.Metastases in different anatomical locations display distinct response patterns and are also associated with PFS in mccRCC patients under PD-1 inhibitor therapy.Specific sites of metastases may hold unique mechanisms of tumor response and call for more personalized treatment.Part Ⅲ:CT texture analysis in predicting the efficacy of PD-1 inhibitor therapy in metastatic renal clear cell carcinomaObjective1.To investigate the predictive efficacy of texture analysis techniques based on contrast-enhanced CT at baseline for tumor response and pseudoprogression in patients with metastatic renal clear cell carcinoma.2.To establish and validate a CT texture-based risk prediction model for progression in renal clear cell carcinoma patients treated with PD-1 inhibitors.Material and methodsThis prospective study enrolled 42 mccRCC patients who received PD-1 inhibitors after standard treatment failure from June 2015 to April 2019.Contrast-enhanced CT was performed at baseline before treatment and follow up every 8 to 10 weeks.Tumor response and TTP were assessed according to iRECIST.The patients and target lesions were classified as response/non-response,CR/non-CR,pseudoprogression/non-pseudoprogression groups according to tumor response evaluation.CT texture features were compared between each group.Univariate and multivariate logistic regression analysis was used to screen independent predictors of tumor response,and a predictive model was developed.The receiver operating characteristic curve(ROC)and the area under the curve(AUC)was used to evaluate the predictive efficacy of independent predictors and the model.To predict patient outcomes,a risk stratification model based on CT texture features was also developed by Cox regression analysis.ROC was used to evaluate the efficacy of this model.Results1.There were significant differences in texture features between iCRL group and non-iCRL group,the OR values were 0.071(95%CI:0.021-0.235),3.532(95%CI:1.591-7.841),0.342(95%CI:0.152-0.768),respectively,the AUC was 0.928(95%CI:0.8645-0.9915),0.7984(95%CI:0.6291-0.9676)and 0.757(95%CI:0.59670.9177)by ROC analysis.2.There were significant differences in texture features between the response lesion group and the non-response lesion group.The results of the ROC analysis showed that the AUC were 0.645(95%CI:0.5091-0.7817)and 0.703(95%CI:0.5731-0.8341),respectively.3.There were significant differences in texture features between the pseudoprogression group and the non-pseudoprogression group.The model was established by binary logistic regression,OR value was 0.045(95%CI:0.009-0.227),the AUC of the model was 0.9068(95%CI:0.8171-0.9966)by ROC analysis.4.There were significant differences in texture features between the response and non-response groups,the CR and non-CR groups,and between the pseudoprogression group and the non-pseudoprogression groups at the patients’ level.6.Multiple texture features can be used as independent predictors to predict TTP.Based on this,the risk stratification model C-index was 0.701(95%CI:0.548-0.853).Conclusion1.This study firstly developed and validated a CT texture-based model for the tumor response prediction in patients treated with PD-1 inhibitors.2.There were significant differences in texture features between the response/non-response,CR/non-CR,pseudoprogression/non-pseudoprogression lesions.3.At the patient level,there were statistical differences in texture features between the response/non-response,CR/non-CR,pseudoprogression/non-pseudoprogression groups.4.The risk stratification prediction model based on CT texture features showed a promising prediction ability for the progression of mccRCC patients under PD-1 inhibitor therapy.