Kimberly Thomas

Radiation practice patterns among United States radiation oncologists for postmastectomy breast reconstruction and oncoplastic breast reduction

BACKGROUND For patients requiring radiation therapy following mastectomy or breast reconstruction, there often exist much heterogeneity among practitioners with respect to radiation technique. METHODS AND MATERIALS A 14-question survey was sent nationwide to 1000 active email addresses from the American Society for Radiation Oncology member directory; 271 radiation oncologists completed the survey. RESULTS A total of 75.2% of respondents indicate that they do not routinely deflate the ipsilateral tissue expander (TE) prior to radiation, while 11.5% do routinely deflate (P ≤ .01); 52.2% indicate that they typically use bolus when treating their patients with TEs following mastectomy, 36.7% do not, and 11.1% on a case by case basis (P ≤ .01). Of respondents indicating bolus utilization, 32.8% use a bolus of 0.5 cm every other day; 31.4% indicate a bolus of 0.5 cm every day until tolerated; 20.4% use a bolus of 1 cm every other day; 5.8% indicate a bolus of 1 cm every day until tolerated; and 9.5% indicate a customized bolus approach (P ≤ .01). A total of 22.9% of respondents deliver boost to all patients with TE while 42.9% deliver boost only to select patients, and 33.5% indicate no utilization of boost (P ≤ .01). A total of 33.1% indicate that collaborating surgeons routinely place clips at the lumpectomy cavity at the time of breast reduction or complex tissue rearrangement, while 38.3% indicate that clips are occasionally placed, and 28.6% stated clips are not routinely placed (P = .15); 38.7% of respondents routinely deliver a boost for patients undergoing breast reduction only if clips have been placed in the tumor cavity, while 34.6% indicate that a boost is used regardless of clip placement. CONCLUSIONS Radiation treatments with tissue expanders have become common practice, but details of radiation treatment vary widely. Radiation oncologist and breast surgeons should continue to work to optimize radiation techniques and allow proper localization for radiation boost.

Interruptions of Head and Neck Radiotherapy Across Insured and Indigent Patient Populations

PURPOSE Radiotherapy for head and neck cancer is a cornerstone of care, requiring 30 to 35 days of treatment over 6 to 7 weeks. Diligent patient compliance is crucial, and unplanned treatment interruptions reduce cure rates. We studied interruption rates in private carrier-insured and Medicare-insured populations versus indigent populations served by a single academic health system. MATERIALS AND METHODS A retrospective cohort study of electronic medical and billing records was performed analyzing treatment interruptions between January 2011 and December 2014. The study included 564 patients with head and neck cancer prescribed radiotherapy and referred from clinics run by University of Texas Southwestern Medical Center (UTSW) and the Parkland Health and Hospital System (PHHS), which provides indigent care to Dallas County, Texas. RESULTS Three-hundred sixteen patients (56%) had a treatment break; 114 patients missed a single session, and 202 patients missed multiple treatments. Seventy percent of PHHS patients had treatment delays compared with 47% of UTSW patients ( P < .001). The number of interrupted days in the PHHS population was nearly twice that observed in UTSW patients. PHHS patients most commonly missed treatment for nonmedical or logistical reasons. Delay was predictive for local recurrence ( P < .001) and overall survival ( P < .001). In compliant patients, there was no significant difference in local recurrence ( P = .43) or overall survival ( P = .27) across referral sites. However, among noncompliant patients, there was a higher likelihood for local recurrence in the PHHS cohort ( P = .016). Multivariable modeling suggested treatment interruption to be a key driver of outcome differences across referral sites. CONCLUSION Survival outcomes in our at-risk population were inferior to those in patients insured by commercial carriers or Medicare. Treatment interruption predicted for poor outcome across all patients but was disproportionately experienced by at-risk patients. These results highlight cancer control needs specific to disadvantaged communities at risk for poor treatment compliance.

Deep inspiration breathhold for left-sided breast cancer patients with unfavorable cardiac anatomy requiring internal mammary nodal irradiation

PURPOSE The purpose of this study was to evaluate the utility of moderate deep inspiration breathhold (mDIBH) in reducing heart exposure in left breast cancer patients who have unfavorable cardiac anatomy and need internal mammary lymph node (IMLN) radiation therapy (RT). METHODS AND MATERIALS We used maximum heart distance (MHD), defined as the maximum distance of the heart within the treatment field, >1 cm as a surrogate for unfavorable cardiac anatomy. Twenty-two left breast cancer patients with unfavorable cardiac anatomy requiring IMLN-RT underwent free-breathing (FB) and mDIBH computed tomography simulation and planning. Three-dimensional partially wide tangents (3D-PWTs) and intensity modulated RT plans were generated. Dose-volume histograms were used to compare heart and lung dosimetric parameters. Duration of treatment delivery was recorded for all fractions. RESULTS MHD decreased significantly in mDIBH scans. mDIBH significantly reduced mean heart dose (222.7 vs 578.4 cGy; P < .0001) and percentage of left lung receiving doses ≥20 Gy (V20; 31.93 vs 38.41%; P = .0006) in both 3D-PWT and intensity modulated RT plans. The change in MHD after breathhold reliably predicted mean heart dose reduction after mDIBH. Radiation was effectively delivered in 11.31 ± 3.40 minutes with an average of 10.06 ± 2.74 breathholds per fraction. CONCLUSIONS mDIBH is efficient and can effectively decrease mean heart dose in patients with unfavorable cardiac anatomy who need IMLN-RT, thus simplifying planning and delivery for them. The reduction in mean heart dose is proportional to the reduction in maximum heart distance.

Improving patient health engagement with mobile texting: A pilot study in the head and neck postoperative setting

Cell phone ownership is nearly universal. Messaging is one of its most widely used features. Texting‐based interventions may improve patient engagement in the postoperative setting, but remain understudied.

Accurate segmenting of cervical tumors in PET imaging based on similarity between adjacent slices

Because in PET imaging cervical tumors are close to the bladder with high capacity for the secreted 18FDG tracer, conventional intensity-based segmentation methods often misclassify the bladder as a tumor. Based on the observation that tumor position and area do not change dramatically from slice to slice, we propose a two-stage scheme that facilitates segmentation. In the first stage, we used a graph-cut based algorithm to obtain initial contouring of the tumor based on local similarity information between voxels; this was achieved through manual contouring of the cervical tumor on one slice. In the second stage, initial tumor contours were fine-tuned to more accurate segmentation by incorporating similarity information on tumor shape and position among adjacent slices, according to an intensity-spatial-distance map. Experimental results illustrate that the proposed two-stage algorithm provides a more effective approach to segmenting cervical tumors in 3D18FDG PET images than the benchmarks used for comparison.

MO-DE-207B-05: Predicting Gene Mutations in Renal Cell Carcinoma Based On CT Imaging Features: Validation Using TCGA-TCIA Datasets

PURPOSE The goal of this work is to investigate the use of contrast enhanced computed tomographic (CT) features for the prediction of mutations of BAP1, PBRM1, and VHL genes in renal cell carcinoma (RCC). METHODS For this study, we used two patient databases with renal cell carcinoma (RCC). The first one consisted of 33 patients from our institution (UT Southwestern Medical Center, UTSW). The second one consisted of 24 patients from the Cancer Imaging Archive (TCIA), where each patient is connected by a unique identi?er to the tissue samples from the Cancer Genome Atlas (TCGA). From the contrast enhanced CT image of each patient, tumor contour was first delineated by a physician. Geometry, intensity, and texture features were extracted from the delineated tumor. Based on UTSW dataset, we completed feature selection and trained a support vector machine (SVM) classifier to predict mutations of BAP1, PBRM1 and VHL genes. We then used TCIA-TCGA dataset to validate the predictive model build upon UTSW dataset. RESULTS The prediction accuracy of gene expression of TCIA-TCGA patients was 0.83 (20 of 24), 0.83 (20 of 24), and 0.75 (18 of 24) for BAP1, PBRM1, and VHL respectively. For BAP1 gene, texture feature was the most prominent feature type. For PBRM1 gene, intensity feature was the most prominent. For VHL gene, geometry, intensity, and texture features were all important. CONCLUSION Using our feature selection strategy and models, we achieved predictive accuracy over 0.75 for all three genes under the condition of using patient data from one institution for training and data from other institutions for testing. These results suggest that radiogenomics can be used to aid in prognosis and used as convenient surrogates for expensive and time consuming gene assay procedures.

Reliable gene mutation prediction in clear cell renal cell carcinoma through multi-classifier multi-objective radiogenomics model

Genetic studies have identified associations between gene mutations and clear cell renal cell carcinoma (ccRCC). Since the complete gene mutational landscape cannot be characterized through biopsy and sequencing assays for each patient, non-invasive tools are needed to determine the mutation status for tumors. Radiogenomics may be an attractive alternative tool to identify disease genomics by analyzing amounts of features extracted from medical images. Most current radiogenomics predictive models are built based on a single classifier and trained through a single objective. However, since many classifiers are available, selecting an optimal model is challenging. On the other hand, a single objective may not be a good measure to guide model training. We proposed a new multi-classifier multi-objective (MCMO) radiogenomics predictive model. To obtain more reliable prediction results, similarity-based sensitivity and specificity were defined and considered as the two objective functions simultaneously during training. To take advantage of different classifiers, the evidential reasoning (ER) approach was used for fusing the output of each classifier. Additionally, a new similarity-based multi-objective optimization algorithm (SMO) was developed for training the MCMO to predict ccRCC related gene mutations (VHL, PBRM1 and BAP1) using quantitative CT features. Using the proposed MCMO model, we achieved a predictive area under the receiver operating characteristic curve (AUC) over 0.85 for VHL, PBRM1 and BAP1 genes with balanced sensitivity and specificity. Furthermore, MCMO outperformed all the individual classifiers, and yielded more reliable results than other optimization algorithms and commonly used fusion strategies.

Abstract C80: Compliance with head and neck radiotherapy in medically insured and uninsured cancer patient populations

Background: Head and neck cancer is a curable malignancy. However, treatment is complex, costly, and toxic. Radiotherapy is a cornerstone of treatment, requiring 30-35 daily treatments over 6-7 weeks. Diligent patient compliance is crucial; unplanned treatment interruptions reduce cure rates. Compliance tracks closely with a complex mix of treatment and human factors, including patient trust in providers, effectiveness of toxicity management, and quality of social support. Previous studies suggest that medically underserved populations experience higher rates of radiation non-compliance. We studied compliance rates and potential confounding factors in medically insured and uninsured populations served by a single academic health system. Methods: We retrospectively analyzed electronic medical and billing records of head and neck radiotherapy patients referred from 1) the University of Texas Southwestern (UTSW) Medical Center and 2) Parkland Hospital (PHHS), which provides indigent care to Dallas County, Texas. Eligible diagnoses included cancers of the nasopharynx, base of tongue, salivary, paranasal sinus, tonsil, oropharynx, oral cavity, larynx, or hypopharynx. We identified 564 analyzable cases from a total of 722 treated from January 2011 through December 2014. For initial analysis, patients who missed any scheduled treatment were considered “noncompliant”. Subsequent analysis included number of missed treatment days, reasons for missed treatment, multivariate modeling of patient/ treatment variables, and downstream survival outcomes. Results: Three-hundred sixteen (56%) of study patients missed treatment; 114 missed a single session, while 202 missed multiple treatments. Median number of missed treatments was 2 days (lowest 25% interquartile = 1 missed day, 50% interquartile = 2 days, 75% interquartile = 5 days). One-hundred fifty two (70%) PHHS patients were non-compliant vs. 163 (47%) UTSW patients. Stepwise logistic regression analyses identified patient age (p=0.016), use of chemotherapy (p=0.021), and referral location (p=0.0002) was predictive for patient cancellation of treatment. UTSW patients enjoyed a 53.7% less frequent cancelation rate than PHHS patients after controlling the effect of age and chemotherapy. Age (p=0.026), chemotherapy (p=0.005), and referral location (p Conclusions: We demonstrate significant disparity in patient compliance to radiation treatment across privately insured and indigent populations managed within a single academic health system. Poor compliance was tracked with expected factors, such as advanced age and treatment intensity (e.g. combined use of chemotherapy with radiation), but was associated most closely with indigent care. Missed treatment predicted for poor local disease control, but only in the PHHS subgroup, suggesting that disease control tracks with high-risk social/treatment factors impacting uninsured patients, rather than secondary time delays themselves. This retrospective study cannot establish causative mechanisms linking uninsured status with non-compliance or disease outcomes, but confirms a pressing need to design interventions to break social/behavioral cycles corrupting cancer care in disadvantaged populations. Citation Format: Kimberly M. Thomas, Travis Martin, Holly Wilhelm, Ang Gao, Chul Ahn, David L. Schwartz. Compliance with head and neck radiotherapy in medically insured and uninsured cancer patient populations. [abstract]. In: Proceedings of the Eighth AACR Conference on The Science of Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; Nov 13-16, 2015; Atlanta, GA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2016;25(3 Suppl):Abstract nr C80.

TU-AB-BRA-02: Predicting Gene Mutations in Renal Cell Carcinoma Using Machine Learning

Purpose: The goal of this project is to investigate the use of software extracted features from contrast-enhanced CT to predict the genetic mutations commonly present in renal tumors using machine learning. Methods: Our initial IRB approved patient database consisted of 33 patients with renal cell carcinoma (RCC). For each patient, the database contained a contrast-enhanced CT image and physician delineated contour outlining the RCC tumor boundary in the CT image. In addition, tumor tissue from each patient was biopsied to determine the presence of gene mutations in BAP1, VHL, and PBRM1. Features based on tumor geometry, intensity, and texture were extracted from the contrast-enhanced CT image of each patient. Features were used to train a support vector machine (SVM) classifier to predict expression of each gene separately. Hyperparameter grid search and feature selection meta-algorithms coupled with cross-validation were employed to protect against overfitting of the SVM model. Results: The average cross-validation accuracy was used to evaluate the predictive model. Average accuracy was 0.87, 0.91, and 0.9 for VHL, BAP1, and PBRM1 respectively. Texture features were the most prominent feature type in the models for all three genes. Conclusion: Using our models we observed predictive accuracy >87% for all three gene mutations evaluated. Accurate predictive models could allow medical images to serve as convenient surrogates for expensive and time consuming gene assay procedures.