The Value Of Radiomics In The Prognosis Prediction Of Targeted Therapy In Patients With Advanced ALK-Positive Non-small Cell Lung Cancer And Related Research

Part 1The value of CT-based radiomics in prognosis prediction of targeted therapy in patients with advanced ALK-positive non-small cell lung cancerObjective:To evaluate the prognostic value of the predictive model established by clinical data combined with CT radiomic signature in advanced anaplastic lymphoma kinase(ALK)-positive non-small cell lung cancer(NSCLC)patients receiving targeted therapy;and to compare the predictive performance of the baseline CT radiomic signature with its early changes(delta radiomic signature)in targeted therapy.Methods:We analyzed the prospectively acquired data from a phase Ⅱ multicenter clinical trial(ClinicalTrials.gov identifier:NCT03215693)about a second-generation ALK inhibitor ensartinib between September 2017 and July 2019.This study included 114 advanced ALK-positive NSCLC patients with a total of 203 tumor lesions,and divided them into progression and non-progression groups according to progression occurrence within median progression-free survival(PFS).Statistical analysis as follows:(1)Analyzed the clinical data and construct a clinical prediction model;(2)Extracted and screened the radiomic features of all eligible target lesions at baseline and the changes after early treatment(delta radiomic features),calculated the average value of each patient,and constructed radiomic prediction models;(3)Constructed clinical-radiomic combined models.The prediction performance and clinical value of the models were evaluated using receiver operating characteristic analysis,decision curve analysis(DCA),and KaplanMeier survival analysis.A nomogram was constructed with the optimal model and its goodness of fit was assessed with the calibration curve.Results:Multivariable analysis demonstrated that only "Additional antitumor therapy between crizotinib and ensartinib"(P=0.04)and "duration of overall response to crizotnib"(P=0.005)were independent factors associated with patients progress within the median PFS.The delta combined model had the highest predictive performance[with area under the curve(AUC)of 0.90 and 0.85 in the training and validation datasets respectively],higher than the clinical model(with AUC of 0.73 and 0.75 in the training and validation datasets respectively)or delta radiomic model(with AUC of 0.86 and 0.79 in the training and validation datasets respectively)alone.The delta radiomic model had a higher predictive efficacy than the baseline radiomic model.In the validation group,the baseline radiomic model(AUC:0.73)had slightly lower predictive power than the clinical model(AUC:0.75).The DCA showed that the clinical value of the delta combined model was superior to other models.The Kaplan-Meier survival curve analysis showed that there were statistically significant differences between the patients grouped based on "liver metastases","duration of overall response to crizotnib≥1 year",baseline radiomic model,delta radiomic model,baseline combined model,and delta combined model.The calibration curve showed a good fit of the nomogram constructed based on the delta combined model.Conclusions:The combined model constructed from clinical data and radiomic features has prognostic value in patients with advanced ALK-positive NSCLC treated with ensartinib.Radiomic-signature early changes in targeted therapy could be more valuable than those at baseline alone.Part 2The value of MRI-based radiomics in prognosis prediction of targeted therapy in ALK-positive non-small cell lung cancer patients with brain metastasesObjective:To evaluate the prognostic value of radiomic signature derived from MRI in anaplastic lymphoma kinase(ALK)-positive non-small cell lung cancer(NSCLC)patients with brain metastasis receiving targeted therapy.Methods:We analyzed the prospectively acquired data from a phase Ⅱ multicenter clinical trial(ClinicalTrials.gov identifier:NCT03215693)about a second-generation ALK inhibitor ensartinib.This study included 87 brain metastatic lesions in 24 ALK-positive lung cancer patients,and divided them into progression group and non-progression group according to whether the patients progressed within the intracranial median progressionfree survival(PFS).Radiomic features were extracted and screened from baseline enhanced MR images.Combined with the selected radiomic features,the Rad-score was calculated with multivariate logistic regression and radiomic prediction model was constructed.The prediction performance and clinical value of the radiomic model were evaluated using receiver operating characteristic analysis,decision curve analysis(DCA),and Kaplan-Meier survival analysis.Results:The prediction model constructed with nine selected radiomic features could predict patients’ progression within median intracranial PFS(with AUC of 0.84 and 0.85 in the training and validation datasets respectively),while clinical and regular MRI characteristics were independent of the progression.The DCA showed that the radiomic prediction model was clinically useful.The Kaplan-Meier analysis showed that the PFS difference between the high-risk and low-risk groups distinguished by the Rad-score(threshold,-0.90)was significant(P=0.017).Conclusions:Enhanced MRI-based radiomics may provide prognostic information and improve pretreatment risk stratification in ALK-positive NSCLC patients with brain metastases undergoing ensartinib treatment.Part 3The correlation between ALK variants and efficacy of targeted therapy in patients with advanced non-small cell lung cancerObjective:We aimed to assess the correlation of anaplastic lymphoma kinase(ALK)variants and ALK resistance mutations with target therapy response duration in non-small cell lung cancer(NSCLC),and explore the potential value of CT radiomic features in predicting progression-free survival(PFS).Methods:We enrolled 88 advanced ALK-positive NSCLC patients with identified ALK variant in a multicenter phase 2 trial(ClinicalTrials.gov identifier:NCT03215693)about a second-generation ALK inhibitor ensartinib,and divided them into progression group and non-progression group according to whether the patients progressed within the median PFS.According to the different fusion partner,divided the ALK variants into three types:V1 variant,V3(V3a/b)variant,other variants(including V2,V5’,V5a variants and nonEML4-ALK fusion variants).This study mainly assessed the impact of ALK variants and ALK resistance mutations on the clinical outcome(response duration)of patients receiving ensartinib.We also established a multifactorial model of clinicopathologic and quantitative CT radiomic features to predict PFS and risk stratification.The prediction performance and clinical value of the models were evaluated using receiver operating characteristic analysis,decision curve analysis(DCA),and Kaplan-Meier survival analysis.A nomogram was constructed with the optimal model and its goodness of fit was assessed with the calibration curve.Results:ALK variant was associated with patient’s progression within median PFS in univariate analysis(P=0.035),but was not retained after multivariate analysis.ALK resistance mutations were adversely associated with PFS both in univariate(P=0.008)and multivariate(P=0.04)analyses and could identify patients at high risk for early progression in the Kaplan-Meier analysis(P=0.002).Among clinical features,only "duration of overall response to crizotnib<1 year"(P=0.002)was the independent risk factor of patients progress within the median PFS.The combined model,composed of clinicopathologic signature and CT radiomic signature,showed good prediction ability with the area under the receiver operating characteristic curve being 0.85,and 0.89 in the training and validation dataset respectively.Conclusions:Our study showed that ALK resistance mutations were adversely associated with ensartinib efficacy,and that ALK variants might not correlate with the PFS.The quantitative radiomic signature provided added prognostic prediction value to the clinicopathologic features.