Predicting benefit from immune checkpoint inhibitors in patients with non-small-cell lung cancer by CT-based ensemble deep learning: a retrospective study.

BACKGROUND: Only around 20-30% of patients with non-small-cell lung cancer (NCSLC) have durable benefit from immune-checkpoint inhibitors. Although tissue-based biomarkers (eg, PD-L1) are limited by suboptimal performance, tissue availability, and tumour heterogeneity, radiographic images might holistically capture the underlying cancer biology. We aimed to investigate the application of deep learning on chest CT scans to derive an imaging signature of response to immune checkpoint inhibitors and evaluate its added value in the clinical context. METHODS: In this retrospective modelling study, 976 patients with metastatic, EGFR/ALK negative NSCLC treated with immune checkpoint inhibitors at MD Anderson and Stanford were enrolled from Jan 1, 2014, to Feb 29, 2020. We built and tested an ensemble deep learning model on pretreatment CTs (Deep-CT) to predict overall survival and progression-free survival after treatment with immune checkpoint inhibitors. We also evaluated the added predictive value of the Deep-CT model in the context of existing clinicopathological and radiological metrics. FINDINGS: Our Deep-CT model demonstrated robust stratification of patient survival of the MD Anderson testing set, which was validated in the external Stanford set. The performance of the Deep-CT model remained significant on subgroup analyses stratified by PD-L1, histology, age, sex, and race. In univariate analysis, Deep-CT outperformed the conventional risk factors, including histology, smoking status, and PD-L1 expression, and remained an independent predictor after multivariate adjustment. Integrating the Deep-CT model with conventional risk factors demonstrated significantly improved prediction performance, with overall survival C-index increases from 0·70 (clinical model) to 0·75 (composite model) during testing. On the other hand, the deep learning risk scores correlated with some radiomics features, but radiomics alone could not reach the performance level of deep learning, indicating that the deep learning model effectively captured additional imaging patterns beyond known radiomics features. INTERPRETATION: This proof-of-concept study shows that automated profiling of radiographic scans through deep learning can provide orthogonal information independent of existing clinicopathological biomarkers, bringing the goal of precision immunotherapy for patients with NSCLC closer. FUNDING: National Institutes of Health, Mark Foundation Damon Runyon Foundation Physician Scientist Award, MD Anderson Strategic Initiative Development Program, MD Anderson Lung Moon Shot Program, Andrea Mugnaini, and Edward L C Smith.

Effects of glutamine for prevention of radiation-induced esophagitis: a double-blind placebo-controlled trial.

Purpose Acute radiation-induced esophagitis (ARIE) leads to treatment delays, decreased quality of life (QOL), and secondary adverse events such as weight loss. Grade 3 ARIE occurs in 15%-30% of patients undergoing radiotherapy to the esophagus, leading to disruption or discontinuation of treatment. The purpose of this study was to assess the effects of glutamine, a common nutritional supplement, on ARIE in patients with thoracic malignancies. Patients and methods This double-blind, placebo-controlled trial enrolled patients with advanced thoracic malignancies receiving concurrent chemotherapy/radiotherapy or radiotherapy alone, with radiation doses to the esophagus ≥45 Gy. Patients were randomized (1:1) to receive 4 g of glutamine or glycine placebo twice daily. The primary objective was to determine whether glutamine decreases the severity of ARIE in these patients. Secondary objectives included assessment of the effects of glutamine on other measures of ARIE, weight, symptom burden measure assessed by the MD Anderson Symptom Inventory (MDASI-HN) questionnaire and the toxicity profile of glutamine. Results At the time of interim analysis, 53 patients were enrolled: 27 in the glutamine arm and 26 in the placebo arm. There was no difference in the incidence of esophagitis in the first 6 weeks of radiotherapy between the glutamine and placebo arms (74% versus 68%; P = 1.00). There were no significant differences between the two arms for time to onset of esophagitis. The duration of ARIE was shorter (6.3 versus 7.1 weeks; P = 0.54) and median weight loss was lower (0.9 kg versus 2.8 kg; p = 0.83) in the glutamine arm versus the placebo arm. The groups differ significantly in core symptom severity (2.1 vs 1.5, p < .03) but not in head and neck specific symptom severity (1.2 vs 1.1, p < .60) nor in symptom interference (2.1 vs 1.7, p < .22). There was no grade 3 or higher adverse event at least possibly related to glutamine. The study was terminated for futility following interim analysis. Conclusion Oral glutamine was not associated with significant improvement in severity of ARIE, weight loss, head and neck specific symptoms or symptom interference compared with placebo in patients with advanced thoracic malignancies receiving radiotherapy to the esophagus.Clinical trial information. NCT01952847, and date of registration is September 30, 2013.

Dose-response relationship in locoregional control for patients with stage II-III esophageal cancer treated with concurrent chemotherapy and radiotherapy.

PURPOSE: To evaluate the correlation between radiation dose and locoregional control (LRC) for patients with Stage II-III unresectable esophageal cancer treated with concurrent chemotherapy and radiotherapy. METHODS AND MATERIALS: The medical records of 69 consecutive patients with clinical Stage II or III esophageal cancer treated with definitive chemoradiotherapy at the University of Texas M. D. Anderson Cancer Center between 1990 and 1998 were retrospectively reviewed. Of the 69 patients, 43 had received < or =51 Gy (lower dose group) and 26 >51 Gy (higher dose group). The median dose in the lower and higher dose groups was 30 Gy (range, 30-51 Gy) and 59.4 Gy (range, 54-64.8 Gy), respectively. Two fractionation schedules were used: rapid fractionation, delivering 30 Gy at 3 Gy/fraction within 2 weeks, and standard fractionation, delivering > or =45 Gy at 1.8-2 Gy/fraction daily. Total doses of <50 Gy were usually given with rapid fractionation. Cisplatin and 5-fluorouracil were administrated to 93% of the patients. RESULTS: The patient characteristic that differed between the two groups was that patients in the lower dose group were more likely to have had weight loss >5% (46.2% vs. 23.3%). The lower dose group had more N1 tumors, but the tumor classification and stage grouping were similar in the two groups. The median follow-up time for all patients was 22 months (range, 2-56 months). Patients in the higher dose group had a statistically significant better 3-year local control rate (36% vs. 19%, p = 0.011), disease-free survival rate (25% vs. 10%, p = 0.004), and overall survival rate (13% vs. 3%, p = 0.054). A trend toward a better distant-metastasis-free survival rate was noted in the higher dose group (72% vs. 59%, p = 0.12). The complete clinical response rate was significantly greater in the higher dose group (46% vs. 23%, p = 0.048). In both groups, the most common type of first failure was persistence of the primary tumor. Significantly fewer patients in the higher dose group had tumor persistence after treatment (p = 0.02). No statistically significant difference was found between the two groups in the pattern of locoregional or distant failure. The long-term side effects of chemoradiotherapy were similar in the two groups, although it was difficult to assess the side effects accurately in a retrospective fashion. On multivariate analysis, Stage II (vs. III) disease and radiation dose >51 Gy were independent predictors of improved LRC, and locoregional failure was an independent predictor of worse overall survival. CONCLUSION: Our data suggested a positive correlation between radiation dose and LRC in the population studied. A higher radiation dose was associated with increased LRC and survival in the dose range studied. The data also suggested that better LRC was associated with a lower rate of distant metastasis. A threshold of tumor response to radiation dose might be present, as suggested by the flattened slope in the high-dose area on the dose-response curve. A carefully designed dose-escalation study is required to confirm this assumption.

Induction chemotherapy improved outcomes of patients with resectable esophageal cancer who received chemoradiotherapy followed by surgery.

PURPOSE: To investigate the effect of induction chemotherapy (CHT) before trimodality therapy on the outcome of patients with resectable cancer of the esophagus. METHODS AND MATERIALS: This retrospective study included 81 consecutive patients with resectable cancer of the esophagus who received neoadjuvant chemoradiotherapy followed by esophagectomy between January 1990 and December 1998 (inclusive). Thirty-nine patients underwent chemoradiotherapy followed by esophagectomy (CHT/RT+S), 42 received additional induction CHT followed by CHT/RT+S (CHT+CHT/RT+S). Of the 81 patients, 47 were entered in institutional or national prospective trials (6 in the CHT/RT+S and 41 in the CHT+CHT/RT+S group). Induction CHT consisted of three courses of 5-fluorouracil (5-FU), cisplatin, and paclitaxel given in 28-day cycles in 37 patients (88.1%). Concurrent CHT was 5-FU and platinum based. The median radiation dose for patients treated with CHT/RT+S was 30 Gy (range, 30-50.4 Gy) delivered in a median of 10 fractions (range, 10-28 fractions) and 45 Gy (range, 30-45 Gy) in a median of 25 fractions (range, 10-25 fractions) for patients treated with CHT+CHT/RT+S. Esophagectomy was performed 6-8 weeks after completion of concurrent chemoradiotherapy. Most patients underwent transthoracic esophagectomy (n = 66, 82.5%). RESULTS: The pretreatment characteristics were well balanced between the two groups except for age. The median follow-up time was 29 months (22 months for the CHT/RT+S group and 38.5 months for the CHT+CHT/RT+S group) for all patients and 49 months for living patients. The actuarial overall survival (OS), disease-free survival (DFS), locoregional control (LRC), and distant metastasis-free survival (DMFS) rate at 5 years for the entire group was 46%, 36.6%, 70.7%, and 53.2%, respectively. Statistically significant differences in the OS, DFS, and LRC rates between the two groups were detected. Specifically, the 5-year OS rate was 22.8% and 71.1% in the CHT/RT+S and CHT+CHT/RT+S group (p = 0.0001), respectively. The 5-year DFS rate was 27.6% and 56.6% in the CHT/RT+S and CHT+CHT/RT+S group (p = 0.003), respectively. The 5-year LRC rate was 64.2% and 85.6% in the CHT/RT+S and CHT+CHT/RT+S group (p = 0.007), respectively. The difference in the DMFS rate between the two groups was statistically significant, with a 2- and 5-year actuarial rate of 63.9% and 51.9%, respectively, in the CHT/RT+S group and 76.9% and 74.1%, respectively, in the CHT+CHT/RT+S group (p = 0.04). The statistically significant differences persisted when patients who received >/=45 Gy in each group were compared. Among those patients, the 5-year OS, DFS, LRC, and DMFS rates were 23.1%, 15.4%, 58.6%, and 39.2%, respectively, for those receiving CHT/RT+S, and 71.4% (p = 0.001), 55.8% (p = 0.0008), 84.6% (p = 0.005), and 77.3% (p = 0.009), respectively, for those receiving CHT+CHT/RT+S. The pathologic complete response (pCR) rate was greater in the CHT+CHT/RT+S group compared with in the CHT/RT+S group (p = 0.008). In univariate analysis, young age, good Karnofsky performance status, Stage II disease, total radiation dose, multiple drug regimen for concurrent CHT, pCR, R0 resection, distant disease progression, and CHT+CHT/RT+S treatment proved to be prognostic factors for OS. Lower esophageal/gastroesophageal junction tumor location, pCR, R0 resection, and CHT+CHT/RT+S treatment were favorable prognostic factors for LRC. Neither the total radiation dose nor multiple drugs for concurrent CHT were negative prognostic factors for LRC. In multivariate analysis, pCR, R0 resection, and treatment with CHT+CHT/RT+S were independent positive predictive factors for OS, and distant recurrences were negative predictive factors for OS. R0 resection, CHT+CHT/RT+S treatment, and lower esophageal/gastroesophageal junction tumor location were positive predictive factors for LRC. The radiation dose was not identified as an independent prognostic factor for either OS or LRC in the multivariate analysis. Meaningful multivariate analysis could not be performed when the multiple drug vperformed when the multiple drug variable was included in the model because of the small number of patients. CONCLUSION: Significantly greater LRC, DFS, OS, and DMFS were found in patients treated with CHT+CHT/RT+S compared with those treated with CHT/RT+S. The pCR rate was significantly higher in the CHT+CHT/RT+S group. Induction CHT was an independent favorable prognostic factor for both LRC and OS for the population included in this study. Our data suggest that a randomized trial comparing CHT+CHT/RT+S and CHT/RT+S is warranted to assess further the merits of this treatment in patients with this currently very lethal cancer.