Dataset Information

A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy.

ABSTRACT: PURPOSE:Development of a supervised machine-learning model capable of predicting clinically relevant molecular subtypes of pancreatic ductal adenocarcinoma (PDAC) from diffusion-weighted-imaging-derived radiomic features. METHODS:The retrospective observational study assessed 55 surgical PDAC patients. Molecular subtypes were defined by immunohistochemical staining of KRT81. Tumors were manually segmented and 1606 radiomic features were extracted with PyRadiomics. A gradient-boosted-tree algorithm was trained on 70% of the patients (N = 28) and tested on 30% (N = 17) to predict KRT81+ vs. KRT81- tumor subtypes. A gradient-boosted survival regression model was fit to the disease-free and overall survival data. Chemotherapy response and survival were assessed stratified by subtype and radiomic signature. Radiomic feature importance was ranked. RESULTS:The mean±STDEV sensitivity, specificity and ROC-AUC were 0.90±0.07, 0.92±0.11, and 0.93±0.07, respectively. The mean±STDEV concordance indices between the disease-free and overall survival predicted by the model based on the radiomic parameters and actual patient survival were 0.76±0.05 and 0.71±0.06, respectively. Patients with a KRT81+ subtype experienced significantly diminished median overall survival compared to KRT81- patients (7.0 vs. 22.6 months, HR 4.03, log-rank-test P = <0.001) and a significantly improved response to gemcitabine-based chemotherapy over FOLFIRINOX (10.14 vs. 3.8 months median overall survival, HR 2.33, P = 0.037) compared to KRT81- patients, who responded significantly better to FOLFIRINOX over gemcitabine-based treatment (30.8 vs. 13.4 months median overall survival, HR 2.41, P = 0.027). Entropy was ranked as the most important radiomic feature. CONCLUSIONS:The machine-learning based analysis of radiomic features enables the prediction of subtypes of PDAC, which are highly relevant for disease-free and overall patient survival and response to chemotherapy.

SUBMITTER: Kaissis G

PROVIDER: S-EPMC6774515 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy.

Kaissis Georgios G Ziegelmayer Sebastian S Lohöfer Fabian F Steiger Katja K Algül Hana H Muckenhuber Alexander A Yen Hsi-Yu HY Rummeny Ernst E Friess Helmut H Schmid Roland R Weichert Wilko W Siveke Jens T JT Braren Rickmer R

PloS one 20191002 10

<h4>Purpose</h4>Development of a supervised machine-learning model capable of predicting clinically relevant molecular subtypes of pancreatic ductal adenocarcinoma (PDAC) from diffusion-weighted-imaging-derived radiomic features.<h4>Methods</h4>The retrospective observational study assessed 55 surgical PDAC patients. Molecular subtypes were defined by immunohistochemical staining of KRT81. Tumors were manually segmented and 1606 radiomic features were extracted with PyRadiomics. A gradient-boost ...[more]

PMID: 31577805

Similar Datasets

Project description:ImportanceFluorouracil, leucovorin, irinotecan, and oxaliplatin (FOLFIRINOX) and gemcitabine plus nanoparticle albumin-bound (nab)-paclitaxel (GA) are first-line chemotherapy regimens for pancreatic cancer. Their relative efficacy in the setting of localized disease is unknown.ObjectiveTo evaluate radiographic and serologic measures of responses associated with first-line chemotherapy with FOLFIRINOX or GA, and to determine the association between these drug regimens, putative measures of response, and survival.Design, setting, and participantsThis case series assessed 485 consecutive patients who were diagnosed as having previously untreated localized pancreatic ductal adenocarcinoma at The University of Texas MD Anderson Cancer Center between January 1, 2010, and December 31, 2017, and who received at least 3 cycles of first-line chemotherapy with FOLFIRINOX or GA. The median (range) follow-up duration was 33 (2-28) months.ExposuresAdministration of FOLFIRINOX (285 patients [59%]) or GA (200 patients [41%]) as first-line chemotherapy.Main outcomes and measuresResection rate, radiographic metrics (Response Evaluation Criteria in Solid Tumors [RECIST], version 1.1, and change in tumor volume or anatomic staging), a serologic metric (serum cancer antigen 19-9 level), and overall survival after administration of first-line chemotherapy.ResultsIn total, 485 patients (266 [55%] male) were included in the analysis. Patients treated with FOLFIRINOX were generally younger (median [range] age at diagnosis: 61 [30-81] vs 71 [36-89] years; P = .001) and had better performance status as indicated by the Eastern Cooperative Oncology Group scale (range 0-4, with lower numbers representing better performance) score of 2 or lower (274 patients [96%] vs 165 patients [82%] P = .001) but more invasive tumors than patients who received GA (91 [32%] vs 90 [45%] resectable tumors; P = .01). After propensity score matching to control for these biases, many objective serologic and radiographic metrics of response associated with administration of FOLFIRINOX or GA-including low rates of local tumor downstaging-did not differ. However, RECIST partial response was more common among patients treated with FOLFIRINOX (27 of 140 patients [19%]) than with GA (8 of 140 patients [6%]; P = .001). Moreover, (chemo)radiation (50% vs 34%; P = .001) was more commonly administered to and pancreatectomy (27% vs 16%; P = .01) was subsequently performed more frequently for patients initially treated with FOLFIRINOX. The overall survival duration of patients treated with either regimen was similar (hazard ratio, 1.48; 95% CI, 0.97-2.26; P = .07).Conclusions and relevanceIn this cohort of patients with localized pancreatic adenocarcinoma who received FOLFIRINOX or GA as their first line of therapy, FOLFIRINOX was associated with higher rates of RECIST partial response and subsequent pancreatectomy than GA, but the overall survival associated with these regimens was similar.

Dataset Information

A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy.

Publications

A machine learning algorithm predicts molecular subtypes in pancreatic ductal adenocarcinoma with differential response to gemcitabine-based versus FOLFIRINOX chemotherapy.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure