Virtual Library
Start Your Search
K.A. Gold
Author of
-
+
MINI 06 - Quality/Prognosis/Survival (ID 111)
- Event: WCLC 2015
- Type: Mini Oral
- Track: Treatment of Localized Disease - NSCLC
- Presentations: 1
- Moderators:R. Meguid, J. Yoshida
- Coordinates: 9/07/2015, 16:45 - 18:15, 605+607
-
+
MINI06.11 - The Influence of Body Mass Index on Overall Survival following Surgical Resection of Non-Small Cell Lung Cancer (ID 2722)
16:45 - 18:15 | Author(s): K.A. Gold
- Abstract
- Presentation
Background:
Population studies suggest that high body mass index (BMI) correlates with a reduced risk of death from lung cancer. The aim of our study was to evaluate the influence of BMI on long term overall survival (OS) in surgical patients with non-small cell lung cancer (NSCLC).
Methods:
Study population consisted of 1935 patients who underwent surgical resection for lung cancer at MD Anderson Cancer Center between 2000-2014. Patients with perioperative mortality, 90-day mortality, intraoperative transfusion, postoperative ICU days, postoperative pneumonia, and postoperative transfusion were excluded. Study variables included both patient and treatment related characteristics. Univariable and multivariable Cox regression analyses were performed to identify variables associated with overall survival. Propensity matching was performed to compare patients with BMI <25 and BMI≥30 matching on type of surgery, age, gender, histology, and pathological stage.
Results:
On univariable analysis, significant predictors of improved survival were higher BMI, pathologic tumor stage (stage I vs II, III, or IV), type of surgery (lobectomy/pneumonectomy vs wedge resection/segmentectomy), younger age, female gender, and adenocarcinoma histology (vs squamous) (all p<0.05). Patients considered morbidly obese (BMI≥35) had a trend towards better outcomes than those classified as obese (BMI ≥30 and <35), overweight (BMI ≥25 and <30), or healthy weight (BMI<25) (HR 0.727, 0.848, 0.926, and 1, respectively, p=NS). On multivariate analysis, BMI remained an independent predictor of survival (p=0.02, see Table). Propensity matching analysis demonstrated significantly better OS (p=0.008) in patients with BMI≥30 compared to BMI <25 (Figure).Multivariate Cox Regression Model
Figure 1N (%) Overall Survival HR (95% CI) BMI <25 (Reference) ≥25 646 (33.4%) 1289 (66.7%) 1.000 0.833(0.713-0.975) Age Continuous variable Median 66 (13-88) 1.024 (1.015-1.032) Gender Female (Reference) Male 984 (50.9%) 951 (49.1%) 1.000 1.236 (1.061-1.441) Stage I (Reference) II III IV 1149 (59.4%) 431 (22.3%) 299 (15.5%) 56 (2.9%) 1.000 1.839 (1.570-2.271) 2.653 (2.182-3.225) 2.737 (1.934-3.873) Surgery Wedge/Segmentectomy (Reference) Lobectomy/Pneumonectomy 198 (10.2%) 1737 (89.8%) 1.000 0.602 (0.479-0.755) Pre-op therapy No (Reference) Yes 1604 (82.9%) 331 (17.2%) 1.000 1.399 (1.160-1.686) Histology Adenocarcinoma (Reference) Squamous Other 1252 (64.7%) 472 (24.4%) 211 (10.9%) 1.000 1.225 (1.035-1.451) 0.959 (0.747-1.231)
Conclusion:
In a large, single center series, after controlling for disease stage and other variables, higher BMI was associated with improved OS following surgical resection of NSCLC. Further studies are necessary to define the complex relationship between BMI and treatment outcomes.
Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.
-
+
MINI 16 - EGFR Mutant Lung Cancer 2 (ID 130)
- Event: WCLC 2015
- Type: Mini Oral
- Track: Treatment of Advanced Diseases - NSCLC
- Presentations: 1
- Moderators:G.J. Riely, M.C. Garassino
- Coordinates: 9/08/2015, 16:45 - 18:15, Four Seasons Ballroom F3+F4
-
+
MINI16.09 - Design, Execution, and Preliminary Biomarker Results from Paired Tumor Biopsy Cohorts of the AZD9291 AURA Trial (ID 941)
16:45 - 18:15 | Author(s): K.A. Gold
- Abstract
Background:
Epidermal growth factor receptor (EGFR)-mutant non-small cell lung cancer (NSCLC) exhibits sensitivity to EGFR tyrosine kinase inhibitors (TKIs) such as erlotinib and gefitinib; however, acquired resistance eventually develops in most patients. The most common mechanism of TKI resistance is a second-site mutation in the EGFR kinase domain, T790M. AZD9291 is an oral, potent, irreversible EGFR-TKI with potency against both T790M resistance and sensitizing EGFR mutations. In the ongoing Phase I AURA study (NCT01802632), AZD9291 induced durable responses in patients with acquired resistance to EGFR-TKIs. We report results of paired biopsy cohorts of the AURA trial, reviewing modulation of key molecular biomarkers of AZD9291 activity in patient tumor samples.
Methods:
Two cohorts of patients on the AURA trial were consented for collection of paired tumor biopsies. These patients had a pre-study tumor biopsy with T790M positive tumor status confirmed by central laboratory EGFR testing (Cobas™ EGFR Mutation Test). Following 8 to 15 days of once daily AZD9291 treatment (80 or 160 mg), a post-dose tumor biopsy was obtained. Baseline and post-dose tumor tissue was processed for routine histology and pathologic evaluation. More than 100 viable tumor cells per sample were required for subsequent biomarker scoring. Formalin-fixed paraffin-embedded tumor biopsies were profiled by immunohistochemistry with a suite of key pathway and tumor-relevant markers (phospho[p]-EGFR, pERK, pAKT, pS6, PD-L1, CD8). Matching plasma pharmacokinetic samples were also obtained for PK-PD correlations.
Results:
As of February 2015, 58 potential patients with an evaluable baseline biopsy were identified as candidates for post-dose biopsy collection. Sixteen of these patients did not proceed to an on-study biopsy as the identified lesions had regressed too substantially or were no longer considered suitable for re-biopsy, one patient was medically excluded from re-biopsy, and one patient’s sample was not available. In total, 40 patients supplied matched pre- and on-treatment biopsies. As of March 2015, paired tumor samples were available for QC from 26 of these 40 patients. Ten of these 26 biopsy specimens subsequently failed QC due to inadequate tumor content, leaving 16 paired tumor samples available for biomarker analyses, of which five have thus far been evaluated. AZD9291 treatment resulted in the inhibition of EGFR pathway components in the majority of post-treatment tumor biopsies. Tissue biomarker analyses are ongoing and updated data on evaluable biopsy pairs will be reported at the time of the congress.
Conclusion:
The completion of a paired biopsy cohort within the AURA trial was challenging due to the rapid onset of anti-tumor effects of AZD9291. Approximately 29% (17/58) of potentially eligible patients were unsuitable for the post-dose biopsy procedure due to tumor regression and 38% (10/26) of available post-dose biopsies were found to contain too little tumor for analysis. In the evaluable tumor pairs, pharmacodynamic modulation of the EGFR pathway was evident. Further biomarker analyses, including evidence of modulation of immune system markers, may help inform future combination strategies.
-
+
MINI 27 - Biology and Other Issues in SCLC (ID 152)
- Event: WCLC 2015
- Type: Mini Oral
- Track: Small Cell Lung Cancer
- Presentations: 1
- Moderators:P.A. Bunn, Jr, J. Sage
- Coordinates: 9/09/2015, 16:45 - 18:15, 605+607
-
+
MINI27.01 - Investigation of Chimeric Antigen Receptor T Cells as a Novel Immunotherapy for SCLC (ID 2901)
16:45 - 18:15 | Author(s): K.A. Gold
- Abstract
- Presentation
Background:
Small cell lung cancer (SCLC) is an aggressive malignancy with an average of 20,000 new cases per year and 16,000 deaths per year. SCLC accounts for about 10-15% of newly diagnosed lung cancers. Even in the face of extensive research, the standard of care- platinum-based combination chemotherapy- has not changed in decades. Yet even with modern chemotherapy formulations, the two year survival rate for advanced disease stages is less than 5%. Complicating treatment is that often at the time of diagnosis, SCLC as already metastasized to the patient’s surrounding lymph nodes. Therefore, a novel therapeutic strategy will have address three disease aspects: (1) reduce primary tumor growth and eliminate metastatic spread; (2) avoid resistance mechanisms used by SCLC to escape radio- and chemotherapies; (3) synergize with or supersede current therapeutic strategies. Chimeric antigen receptor T cells, little explored in SCLC, is well suited to address these aspects.
Methods:
Human SCLC cell lines were analyzed using a 90 gene signature to establish immunological targets. Western blot analysis confirmed the expression of CD56 and other targets on SCLC cell lines. For CAR T cell generation, PBMC were electroporated with the Sleeping Beauty transposase and a transposon containing a CD56R chimeric antigen receptor. CD56R-CAR transduced T cells were cultured for 4 weeks in the presence of K562 cells expressing CD56 and the cytokines IL-2/IL-21 to expand CD56R-CAR T cells. CAR T cells were tested in vitro for killing ability in the presence of three SCLC cell lines using a chromium release assay. CAR T cells were also analysed via FACS to assess CAR expression, T cell phenotype, and memory status.
Results:
An analysis of immune markers in SCLC cell lines revealed that, compared to NSCLC lines, there is a reduction in the expression of suppressive ligands and co-stimulatory ligands, antigen presentation, and natural killer ligands. SCLC cell lines, however, express high levels of CD56. When two CD56-positive and one CD56-negative cell line was tested, CD56-CAR T cells could kill efficiency CD56 expressing cell lines, however there was little killing of the CD56-negative cell line. An analysis of PBMCs cultured after electroporation revealed that a large percentage of CD3+ T cells expressed the CD56 CAR and even after 4 weeks in culture, the CAR T cells displayed a memory phenotype.
Conclusion:
An interrogation of SCLC cell lines versus NSCLC cell lines revealed that SCLC cell lines had reduced expression of checkpoint ligands, NK cell killing ligands, antigen presentation, but consistent with their origin, high expression of CD56. Our conclusion from this analysis is that expansion of SCLC-specific immune responses in vivo or elicitation of de novo responses in vivo will be hindered. Therefore, immunotherapies centered around adoptive transfer of T cell that can kill in an HLA-independent manner maybe better suited for SCLC. In that vein, CD56R-CAR T cells effectively targeted CD56-positive SCLC in vitro, but was unable to kill CD56-negative cells- which indicates a possible escape variant. Our lab is now moving toward testing CD56R-CAR T cell in vivo in both xenograph models and spontaneous ones.
Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.
-
+
ORAL 33 - ALK (ID 145)
- Event: WCLC 2015
- Type: Oral Session
- Track: Treatment of Advanced Diseases - NSCLC
- Presentations: 2
- Moderators:S. Gadgeel
- Coordinates: 9/09/2015, 16:45 - 18:15, Mile High Ballroom 1a-1f
-
+
ORAL33.03 - Updated Efficacy/Safety Data From the Phase 2 NP28761 Study of Alectinib in ALK+ NSCLC (ID 1261)
16:45 - 18:15 | Author(s): K.A. Gold
- Abstract
- Presentation
Background:
ALK gene rearrangements occur in approximately 3–6% of patients with non-small-cell lung cancer (NSCLC). Crizotinib has demonstrated efficacy in ALK+ NSCLC, however many patients experience systemic and/or central nervous system (CNS) disease progression within one year of treatment. Alectinib, a CNS-penetrant and highly selective ALK inhibitor, has shown preclinical activity in the CNS (Ou, et al. JTO 2013) and clinical efficacy in crizotinib-naïve (Ohe, et al. ASCO 2015) and pre-treated (Ou, et al. ASCO 2015; Gandhi, et al. ASCO 2015) ALK+ NSCLC patients. We will present updated efficacy and safety outcomes from the phase II NP28761 study (NCT01871805).
Methods:
North American patients ≥18 years of age with ALK+ NSCLC (by FDA-approved FISH test), disease progression following first-line crizotinib, and ECOG PS ≤2 were enrolled. Patients received oral alectinib (600mg) twice daily until progression, death or withdrawal. The primary endpoint was overall response rate (ORR) by independent review committee (IRC) using RECIST v1.1. Secondary endpoints included investigator-assessed ORR; progression-free survival (PFS); quality of life (QoL); CNS response rate; disease control rate (DCR); and safety.
Results:
At data cut-off (24 October 2014), 87 patients were enrolled in the intent-to-treat population. Median age was 54 years; 74% had received prior chemotherapy; 60% of patients had baseline CNS metastases, of whom 65% (34/52) had prior brain radiation therapy. Median follow-up was 20.7 weeks. ORR by IRC was 48% (95% CI 36–60); median PFS was 6.3 months (Table 1). In patients with measurable CNS lesions at baseline (n=16), IRC CNS ORR was 69% (95% CI 41–89) and CNS DCR was 100% (complete response, 13%; partial response, 56%; stable disease, 31%). In patients with measurable or non-measurable CNS disease (n=52), IRC CNS ORR was 39% (95% CI 25–53) and 11 patients (21%) had complete CNS responses. The most common grade ≥3 AEs were elevated levels of blood creatine phosphokinase (8%), alanine aminotransferase (6%) and aspartate aminotransferase (5%); no GI toxicities leading to treatment withdrawal were reported. Clinically meaningful improvements were seen in EORTC QLQ-C30 items, including Global Health Status. Figure 1
Conclusion:
Alectinib (600mg twice daily) was well tolerated and demonstrated clinical efficacy in patients with ALK+ NSCLC disease who had progressed on prior crizotinib. A clinical benefit with alectinib was also observed in patients with CNS lesions at baseline. These data are preliminary; updated efficacy and safety data from a cut-off date of 27 April 2015 will be presented.
Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.
-
+
ORAL33.06 - Brigatinib (AP26113) Efficacy and Safety in ALK+ NSCLC: Phase 1/2 Trial Results (ID 2125)
16:45 - 18:15 | Author(s): K.A. Gold
- Abstract
- Presentation
Background:
Brigatinib (AP26113), an investigational oral tyrosine kinase inhibitor with FDA breakthrough therapy designation for the treatment of patients with crizotinib-resistant advanced ALK+ NSCLC, has preclinical activity against both rearranged ALK and clinically identified crizotinib-resistant mutant ALK.
Methods:
This is an ongoing phase 1/2, single-arm, open-label, multicenter study in patients with advanced malignancies (N=137; NCT01449461). Patients received escalating total daily doses of brigatinib from 30–300 mg during phase 1. Daily regimens of 90 mg, 180 mg, or 90 mg for 7 days followed by 180 mg were evaluated in phase 2. Safety is reported for all treated patients; antitumor efficacy (ORR and PFS per RECIST v1.1) is reported for ALK+ NSCLC patients.
Results:
Seventy-nine (58%) patients had ALK+ NSCLC. Median age was 54 (29–83) years, 49% were female, 90% had prior crizotinib, and 47% had ≥2 prior chemotherapy regimens. As of February 17, 2015, 45/79 (57%) ALK+ NSCLC patients remained on study, with median time on treatment of 12.6 months (1 day to 35.5 months; n=79); ORR/PFS for evaluable ALK+ NSCLC patients was 74%/13.4 months (additional data shown in Table). In a post hoc independent radiological review of patients with brain metastases at baseline (as of January 19, 2015), 8/15 (53%) patients with measurable brain lesions ≥10 mm had an intracranial response (≥30% decrease in sum of longest diameters of target lesions) and 9/30 (30%) patients with only nonmeasurable lesions had disappearance of all lesions. Treatment-emergent AEs in ≥30% of total patients, generally grade 1/2, included nausea (52%), fatigue (42%), diarrhea (40%), headache (33%), and cough (32%). Early-onset pulmonary events, which occurred ≤7 days after treatment initiation and included dyspnea, hypoxia, and new pulmonary opacities on chest CT consistent with pneumonia or pneumonitis, were reported in 13/137 (9%) patients overall (6/44 [14%] at 180 mg qd; 2/50 [4%] at 90 mg qd [maintained or escalated to 180 mg qd after 7 days]).Response and PFS With Brigatinib All Evaluable ALK+ NSCLC Patients n=78 Prior Crizotinib n=70 No Prior Crizotinib n=8 Response, n(%) OR (CR+PR) 58(74) 50(71) 8(100) [95% CI] [63–84] [59–82] [63–100] CR 7(9) 4(6) 3(38) PR 51(65) 46(66) 5(63) SD 11(14)[a] 11(16)[a] 0 PD 6(8) 6(9) 0 Termination before scan 3(4) 3(4) 0 Median duration of response,[b] mo 11.2[c] 9.9[d] Not reached[e] Median PFS,[b] mo 13.4 13.4 Not reached [a]Includes non-CR/non-PD for 4 patients with no measurable disease at baseline [b]Kaplan-Meier estimate [c]n=55 evaluable [d]n=48 evaluable [e]n=7 evaluable
Conclusion:
Brigatinib has promising antitumor activity in ALK+ NSCLC patients with (71% ORR; PFS 13.4 months) or without (100% ORR) prior crizotinib, including patients with brain metastases (53% ORR in patients with measurable brain lesions). Early-onset pulmonary events were less frequent when starting at 90 vs 180 mg qd. A pivotal global phase 2 trial (ALTA) of brigatinib 90 mg qd vs 90 mg qd for 7 days followed by 180 mg qd in crizotinib-resistant ALK+ NSCLC is ongoing.
Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.
-
+
P2.01 - Poster Session/ Treatment of Advanced Diseases – NSCLC (ID 207)
- Event: WCLC 2015
- Type: Poster
- Track: Treatment of Advanced Diseases - NSCLC
- Presentations: 1
- Moderators:
- Coordinates: 9/08/2015, 09:30 - 17:00, Exhibit Hall (Hall B+C)
-
+
P2.01-041 - MD Anderson Oncology Expert Advisor™ System (OEA™): A Cognitive Computing Recommendations Application (App) for Lung Cancer (ID 3106)
09:30 - 17:00 | Author(s): K.A. Gold
- Abstract
Background:
The OEA[TM] is a clinical support system with a continuous improvement capability. Its objectives are to enable/empower evidence-based decisions/care by disseminating knowledge and expertise to physicians/users tailored to meet the clinical needs of individual patients as if consulting with an expert. Cognitive computing platforms have the potential to disseminate expert knowledge and tertiary level care to patients. This objective is made possible by making available to physicians/providers cognitive computing generated expert recommendations in diagnosis, staging and treatment. The cognitive computing software was trained by MD Anderson experts using currently available consensus guidelines and an iterative feedback process. Here we test the capability of this cognitive computing software program developed at MD Anderson to generate expert recommendations when patients with advanced-stage NSCLC have a targetable molecular aberration.
Methods:
We developed a web based prototype of MD Anderson’s Oncology Expert Advisor (OEA[TM]), a cognitive clinical decision support tool powered by IBM Watson. The Watson technology is IBM’s third generation cognitive computing system based on its unique capabilities in natural language processing and deep QA (question-answer). We trained OEA[TM] by loading historical patient cases and assessed the accuracy of targeted treatment suggestions using MD Anderson’s physicians’ decisions as benchmark. A false positive result was defined as a treatment recommendation rendered with high confidence that was non-correct (less optimal), whereas false negative was defined as a correct or more optimal treatment suggestion listed as a low confidence recommendation.
Results:
In our preliminary analyses, OEA[TM] demonstrated four core capabilities: 1) Patient Evaluation through interpretation of structured and unstructured clinical data to create a dynamic case summary with longitudinal view of the pertinent events 2) Treatment and management suggestions based on patient profile weighed against consensus guidelines, relevant literature, and MD Anderson expertise, which included approved therapies, genomic based therapies as well as automated matching to appropriate clinical trials at MD Anderson, 3) Care pathway advisory that alerts the user for anticipated toxicities and its early identification and proactive management, and 4) Patient-oriented research functionalities for identification of patient cohorts and hypothesis generation for future potential clinical investigations. Detailed testing continues and the accuracy of standard-of-care (SOC) treatment recommendations of OEA[TM], as well as false positivity and negativity rates will be presented in detail at the meeting.
Conclusion:
OEA[TM] is able to generate dynamic patient case summary by interpreting structured and unstructured clinical data and suggest personalized treatment options. Live system evaluation of OEA[TM] is ongoing and the application of OEA[TM] in clinical practice is expected to be piloted at our institution.
-
+
P3.03 - Poster Session/ Treatment of Locoregional Disease – NSCLC (ID 214)
- Event: WCLC 2015
- Type: Poster
- Track: Treatment of Locoregional Disease – NSCLC
- Presentations: 1
- Moderators:
- Coordinates: 9/09/2015, 09:30 - 17:00, Exhibit Hall (Hall B+C)
-
+
P3.03-032 - MD Anderson Oncology Expert Advisor™: A Cognitive Clinical Decision Support Tool for Evidence-Based Multi-Disciplinary Lung Cancer Care (ID 3039)
09:30 - 17:00 | Author(s): K.A. Gold
- Abstract
Background:
The majority of patients diagnosed with non-small cell lung cancer (NSCLC) receive care in the community setting with limited access to multidisciplinary management common in tertiary care centers. The availability of genomics allows tailored treatments for patients; and with novel, rapidly emerging therapeutic options, it is challenging for busy clinicians to maintain familiarity with current therapy recommendations. Therefore, to empower practicing oncologists in community settings to offer the optimal management at the first intervention, we have developed the MD Anderson Oncology Expert Advisor™ (OEA) application for multi-disciplinary management of lung cancer patients. As the first multi-disciplinary solution for providing comprehensive management of lung cancer, the objective of OEA™ Lung is to leverage cognitive analytics on vast and ever evolving clinical care information and patient big data to disseminate knowledge and expertise, thus enabling physicians to provide evidence-based care and management tailored for the individual patient, similar to consulting an expert. Further, we aimed to create a system for sharing knowledge from more experienced experts to provide care pathways and management recommendations for physicians globally.
Methods:
Using cognitive computing, our cancer center partnered with IBM Watson to develop an expert system designed to provide physicians with the tools needed to process high-volume patient and medical information and to stay up-to-date with the latest treatment and management options, so that they can make the best evidence-based treatment decisions for their lung cancer patients. The OEA™ application for lung was built upon core capabilities of the OEA™ applications for leukemia and molecular/targeted therapies. Experts in multiple disciplines including thoracic surgery, medical oncology, and radiation oncology met regularly to design and provide specialized input to the IBM technical team in an agile development cycle. This system was powered to utilize both structured and unstructured data from validated sources; to thoroughly evaluate and stage patients; and to offer eligible clinical trials and personalized therapeutic options. In addition to delivering evidence-based, weighted therapy recommendations, OEA™ Lung provides care pathways for management of toxicities for each treatment modality (surgery, radiation, and medical oncology).
Results:
The OEA™ Lung application supports three core functions: 1) dynamic patient summary assimilating complete (structured and unstructured) data to show demographics, labs, genotype, treatment history, and previous treatment responses; 2) weighted evidence-based, multimodality treatment options, with recommendations based on literature support which is provided, along with screening for relevant trials; 3) care pathway advisories, to manage treatment related toxicities for each modality. Further, this product improves quality of care by optimizing outcomes with access to trials and care pathways.
Conclusion:
The OEA™ application for lung is a cognitive expert system designed to assimilate multidisciplinary recommendations for care and management of lung cancer patients based on current consensus guidelines and expert recommendations from a quaternary referral cancer center to the community practice setting. By democratizing knowledge from our specialty cancer center, we have taken steps toward achieving an important goal of ending cancer for all, by providing global access to optimal cancer care for patients with this disease. Further evaluation of outcomes following implementation are warranted.
-
+
P3.04 - Poster Session/ Biology, Pathology, and Molecular Testing (ID 235)
- Event: WCLC 2015
- Type: Poster
- Track: Biology, Pathology, and Molecular Testing
- Presentations: 1
- Moderators:
- Coordinates: 9/09/2015, 09:30 - 17:00, Exhibit Hall (Hall B+C)
-
+
P3.04-031 - Combining CT Texture Analysis with Semantic Imaging Descriptions for the Radiogenomic Detection of EGFR and KRAS Mutations in NSCLC (ID 2965)
09:30 - 17:00 | Author(s): K.A. Gold
- Abstract
Background:
Existing literature suggests quantitative texture features derived from CT imaging can differentiate tumor genotypes and phenotypes. We combined CT texture analysis with semantic imaging descriptions provided by radiologists, and evaluated their ability to identify EGFR and KRAS mutation status in NSCLC.
Methods:
We retrospectively reviewed CT images from 628 patients from the GEMINI (Genomic Marker-Guided Therapy Initiative) cohort. Included were NSCLC patients whose biopsies included genetic testing for EGFR or KRAS mutations, and who underwent contrast-enhanced CT imaging within 90 days of biopsy. Excluded were patients who had undergone therapy or biopsy of their primary tumor before imaging, or whose tumors weren’t segmentable. All CT images were contrast-enhanced, with body kernel reconstruction, and slice thicknesses of 1.25-5mm. Tumor segmentation was done in 3DSlicer (Harvard University, Cambridge MA) using a semi-automatic segmentation algorithm. Image pre-processing and textural feature extraction was performed using IBEX (MDACC, Houston TX). Semantic descriptions of the tumors were recorded by a thoracic radiology fellow and a board-certified thoracic radiologist in consensus. For each patient a set of textural features was calculated, based on the GreyLevel Co-Occurrence Matrix, Run-Length Matrix, voxel intensity histogram, and geometric properties of the tumor. Feature selection was based on existing literature, prior research experience, and excluded those features previously found to be poorly reproducible in lung tissue. These were combined with semantic descriptions (e.g. presence or absence of features such as spiculations, air bronchograms, and pleural effusions), for a total of 51 textural and geometric features, and 11 semantic features. When available, the SUVmax for the tumor was also included. To detect correlations with genetic mutations, these features were combined to train a Random Forest machine learning algorithm. This algorithm output a prediction for the mutation status of each tumor, and the predictive accuracy was assessed based on 10-fold cross-validation.
Results:
Included were 121 patients, 113 tested for KRAS mutations (26 positive) and 118 tested for EGFR mutations (31 positive). Maximum tumor dimensions ranged from 1.2–15.5cm (mean 5.6cm). Individual semantic features found to correlate with mutation status included tumor cavitation, pleural effusion, presence of ground glass opacity, and the nature of tumor margins (all p-values <0.05). Used collectively in a Random Forest classifier, textural features alone showed a sensitivity and specificity for KRAS detection of 50% and 81% respectively, with 74% overall accuracy. This increased modestly to a sensitivity and specificity of 50% and 84% respectively when semantic features were added, with accuracy increasing to 77%. For EGFR detection, textural features had sensitivity and specificity of 48% and 77% respectively, giving 69% accuracy. Detection of EGFR did not improve with inclusion of semantic features.
Conclusion:
Texture analysis correctly identified EGFR and KRAS mutation status in most patients. Although some semantic features correlated with mutation status, when combined with textural features they provided little or no improvement in predictive accuracy. One possible explanation is that textural features may already be capturing the information contained in the semantic features. Our results suggest oncogenic drivers of NSCLC are associated with distinct imaging features that can be detected radiographically.
