Purpose: Several biomarkers of response to immune checkpoint inhibitors (ICI) show potential but are not yet scalable to the clinic. We developed a pipeline that integrates deep learning on histology specimens with clinical data to predict ICI response in advanced melanoma. Experimental Design: We used a training cohort from New York University (New York, NY) and a validation cohort from Vanderbilt University (Nashville, TN). We built a multivariable classifier that integrates neural network predictions with clinical data. A ROC curve was generated and the optimal threshold was used to stratify patients as high versus low risk for progression. Kaplan–Meier curves compared progression-free survival (PFS) between the groups. The classifier was validated on two slide scanners (Aperio AT2 and Leica SCN400). Results: The multivariable classifier predicted response with AUC 0.800 on images from the Aperio AT2 and AUC 0.805 on images from the Leica SCN400. The classifier accurately stratified patients into high versus low risk for disease progression. Vanderbilt patients classified as high risk for progression had significantly worse PFS than those classified as low risk (P = 0.02 for the Aperio AT2; P = 0.03 for the Leica SCN400). Conclusions: Histology slides and patients' clinicodemographic characteristics are readily available through standard of care and have the potential to predict ICI treatment outcomes. With prospective validation, we believe our approach has potential for integration into clinical practice.

Despite major improvements in combatting metastatic melanoma since the advent of immunotherapy, the overall survival for patients with advanced disease remains low. Recently, there is a growing number of reports supporting an "obesity paradox," in which patients who are overweight or mildly obese may exhibit a survival benefit in patients who received immune checkpoint inhibitors. We studied the relationship between body mass index and progression-free survival and overall survival in a cohort of 423 metastatic melanoma patients receiving immunotherapy, enrolled and prospectively followed up in the NYU Interdisciplinary Melanoma Cooperative Group database. We analyzed this association stratified by first vs. second or greater-line of treatment and treatment type adjusting for age, gender, stage, lactate dehydrogenase, Eastern Cooperative Oncology Group performance status, number of metastatic sites, and body mass index classification changes. In our cohort, the patients who were overweight or obese did not have different progression-free survival than patients with normal body mass index. Stratifying this cohort by first vs. non-first line immunotherapy revealed a moderate but insignificant association between being overweight or obese and better progression-free survival in patients who received first line. Conversely, an association with worse progression-free survival was observed in patients who received non-first line immune checkpoint inhibitors. Specifically, overweight and obese patients receiving combination immunotherapy had a statistically significant survival benefit, whereas patients receiving the other treatment types showed heterogeneous trends. We caution the scientific community to consider several important points prior to drawing conclusions that could potentially influence patient care, including preclinical data associating obesity with aggressive tumor biology, the lack of congruence amongst several investigations, and the limited reproduced comprehensiveness of these studies.

BackgroundRecent research suggests that baseline body mass index (BMI) is associated with response to immunotherapy. In this study, we test the hypothesis that worsening nutritional status prior to the start of immunotherapy, rather than baseline BMI, negatively impacts immunotherapy response.MethodsWe studied 629 patients with advanced cancer who received immune checkpoint blockade at New York University. Patients had melanoma (n=268), lung cancer (n=128) or other primary malignancies (n=233). We tested the association between BMI changes prior to the start of treatment, baseline prognostic nutritional index (PNI), baseline BMI category and multiple clinical end points including best overall response (BOR), objective response rate (ORR), disease control rate (DCR), progression-free survival (PFS) and overall survival (OS).ResultsDecreasing pretreatment BMI and low PNI were associated with worse BOR (p=0.04 and p=0.0004), ORR (p=0.01 and p=0.0005), DCR (p=0.01 and p<0.0001), PFS (p=0.02 and p=0.01) and OS (p<0.001 and p<0.001). Baseline BMI category was not significantly associated with any treatment outcomes.ConclusionStandard of care measures of worsening nutritional status more accurately associate with immunotherapy outcomes than static measurements of BMI. Future studies should focus on determining whether optimizing pretreatment nutritional status, a modifiable variable, leads to improvement in immunotherapy response.

36DNA-based molecular assays for determining mutational status in melanomas are time-37 consuming and costly. As an alternative, we applied a deep convolutional neural network 38 (CNN) to histopathology images of tumors from 257 melanoma patients and developed a 39 fully automated model that first selects for tumor-rich areas (Area under the curve 40 AUC=0.98), and second, predicts for the presence of mutated BRAF or NRAS. Network 41 performance was enhanced on BRAF-mutated melanomas 1.0 mm (AUC=0.83) and on 42 non-ulcerated NRAS-mutated melanomas (AUC=0.92). Applying our models to 43 histological images of primary melanomas from The Cancer Genome Atlas database also 44 demonstrated improved performances on thinner BRAF-mutated melanomas and non-45 ulcerated NRAS-mutated melanomas. We propose that deep learning-based analysis of 46 histological images has the potential to become integrated into clinical decision making 47 for the rapid detection of mutations of interest in melanoma. 48 49 50 Mutations in the BRAF oncogene are found in 50-60% of all melanomas 1 , while NRAS 51 mutations comprise an additional 15-20%. With the development of targeted therapies 2, 52 3 , determining the mutational status of BRAF and NRAS has become an integral 53 component for the management of Stage III/IV melanomas. DNA molecular assays such 54 as Sanger sequencing, pyrosequencing, and next generation sequencing (NGS) are the 55 current gold standard to determine mutational status 4 . However, these methods are costly 56 and time-consuming. Immunohistochemistry, real-time polymerase chain reaction (PCR), 57 and automated platforms 5, 6, 7 are rapid and less expensive alternatives, but are limited to 58 screening for specific mutations, such as BRAF-V600E/K or NRAS-Q61R/L, and may 59 potentially fail to identify rare mutational variants in patients that might have otherwise 60 benefited from adjuvant targeted therapy.61 62 Deep Convolutional Neural Network (CNN) methods to predict mutational status have 63 been demonstrated in other solid tumors. CNNs utilize multiple layers of convolution 64 operations, pooling layers, and fully connected layers to perform classification of images 65 to classes of interest through identification of various image features often not directly 66 detectable by the human eye. Deep CNNs, which utilize non-linear learning algorithms, 67have been successful in manipulating and processing large data sets, particularly for 68 image analysis 8 . Using images from The Cancer Genome Atlas (TCGA), a collaborative 69 cancer genomics database 9 , our group has previously developed a machine learning 70 algorithm that can predict for 6 different genes, including EGFR and STK11, in lung 71 carcinoma 10 . In breast cancer, deep learning applied to tumor microarray images has 72 been shown to predict for ER status with an 84% accuracy 11 . 73 74 In this study, we adapt our previous deep learning algorithm to a different dataset 75 comprised of histopathology images of primary melanomas resected from patients 76 pros...

Background The American Joint Committee on Cancer (AJCC) maintains that the eighth edition of its Staging Manual (AJCC8) has improved accuracy compared with the seventh (AJCC7). However, there are concerns that implementation may disrupt analysis of active clinical trials for stage III patients. We used an independent cohort of melanoma patients to test the extent to which AJCC8 has improved prognostic accuracy compared with AJCC7. Methods We analyzed a cohort of 1315 prospectively enrolled patients. We assessed primary tumor and nodal classification of stage I–III patients using AJCC7 and AJCC8 to assign disease stages at diagnosis. We compared recurrence-free (RFS) and overall survival (OS) using Kaplan-Meier curves and log-rank tests. We then compared concordance indices of discriminatory prognostic ability and area under the curve of 5-year survival to predict RFS and OS. All statistical tests were two-sided. Results Stage IIC patients continued to have worse outcomes than stage IIIA patients, with a 5-year RFS of 26.5% (95% confidence interval [CI] = 12.8% to 55.1%) vs 56.0% (95% CI = 37.0% to 84.7%) by AJCC8 (P = .002). For stage I, removing mitotic index as a T classification factor decreased its prognostic value, although not statistically significantly (RFS concordance index [C-index] = 0.63, 95% CI = 0.56 to 0.69; to 0.56, 95% CI = 0.49 to 0.63, P = .07; OS C-index = 0.48, 95% CI = 0.38 to 0.58; to 0.48, 95% CI = 0.41 to 0.56, P = .90). For stage II, prognostication remained constant (RFS C-index = 0.65, 95% CI = 0.57 to 0.72; OS C-index = 0.61, 95% CI = 0.50 to 0.72), and for stage III, AJCC8 yielded statistically significantly enhanced prognostication for RFS (C-index = 0.65, 95% CI = 0.60 to 0.70; to 0.70, 95% CI = 0.66 to 0.75, P = .01). Conclusions Compared with AJCC7, we demonstrate that AJCC8 enables more accurate prognosis for patients with stage III melanoma. Restaging a large cohort of patients can enhance the analysis of active clinical trials.