Mirek Fatyga

Patient Specific Characteristics Are an Important Factor That Determines the Risk of Acute Grade ≥ 2 Rectal Toxicity in Patients Treated for Prostate Cancer with IMRT and Daily Image Guidance Based on Implanted Gold Markers

Aim To model acute rectal toxicity in Intensity Modulated Radiation Therapy (IMRT) for prostate cancer using dosimetry and patient specific characteristics. Methods A database of 79 prostate cancer patients treated with image guided IMRT was used to fit parameters of Lyman-Kutcher-Burman (LKB) and logistic regression Normal Tissue Complications Probability (NTCP) models to acute grade ≥ 2 rectal toxicities. We used a univariate regression model to find the dosimetric index which was most correlated with toxicity and a multivariate logistic regression model with machine learning algorithm to integrate dosimetry with patient specific characteristics. We used Receiver Operating Characteristics (ROC) analysis and the area under the ROC curve (AUC) to quantify the predictive power of models. Results Sixteen patients (20.3%) developed acute grade≥2 rectal toxicity. Our best estimate (95% confidence interval) of LKB model parameters for acute rectal toxicity are exponent n=0.13 (0.1–0.16), slope m=0.09 (0.08–0.11), and threshold dose TD50=56.8 (53.7–59.9) Gy. The best dosimetric indices in the univariate logistic regression NTCP model were D25% and V50Gy. The best AUC of dosimetry only modeling was 0.67 (0.54, 0.8). In the multivariate logistic regression two patient specific variables were particularly strongly correlated with acute rectal toxicity, the use of statin drugs and PSA level prior to IMRT, while two additional variables, age and diabetes were weakly correlated. The AUC of the logistic regression NTCP model improved to 0.88 (0.8, 0.96) when patient specific characteristics were included. In a group of 79 patients, 40 took Statins and 39 did not. Among patients who took statins, (4/40)=10% developed acute grade ≥2 rectal toxicity, compared to (12/39)=30.8% who did not take statins (p=0.03). The average and standard deviation of PSA distribution for patients with acute rectal toxicity was PSAtox = 5.77 ± 2.27 and it was PSAnotox = 9.5 ± 7.8 for the remainder (p=0.01). Conclusions Patient specific characteristics strongly influence the likelihood of acute grade ≥ 2 rectal toxicity in radiation therapy for prostate cancer.

Statins and Metformin Use Is Associated with Lower PSA Levels in Prostate Cancer Patients Presenting for Radiation Therapy

Background A possible association between the level of prostate specific antigen (PSA) and the use of some commonly prescribed medications has been reported in recent studies. Most of these studies were carried out in general populations of men who were screened for prostate cancer using the PSA test. We reported on the association between the initial PSA level and the use of statins, metformin and alpha-blockers in patients who were diagnosed with prostate cancer and presented for radiation therapy. Methods Three hundred and eighty one patients treated between the years of 2000-2005 and 2009-2012 were included in this retrospective study. The information about statin, metformin and alpha-blockers use was recorded immediately prior to treatment. Differences in PSA levels prior to treatment by medication status were estimated using univa-riate and multivariate linear regression on log PSA values. Results Compared with men who were not on these medications, the PSA level at presentation was 20% lower for statin users (p = 0.002) and 33% lower for metformin users (p = 0.004). We did not observe statistically significant associations between the use of statins or metformin and cancer stage, National Comprehensive Cancer Network (NCCN) risk score, or therapy outcome. A statistically significant association between the NCCN risk score and the use of alpha-blockers was observed (p = 0.002). Conclusions We found that statins and metformin were associated with lower PSA levels in prostate cancer patients to an extent that could influence management decisions. We found no statistically significant associations between the use of these medications and treatment outcomes.

Automation of Routine Elements for Spot-Scanning Proton Patient-Specific Quality Assurance

PURPOSE At our institution, all proton patient plans undergo patient-specific quality assurance (PSQA) prior to treatment delivery. For intensity-modulated proton beam therapy, quality assurance is complex and time consuming, and it may involve multiple measurements per field. We reviewed our PSQA workflow and identified the steps that could be automated and developed solutions to improve efficiency. METHODS We used the treatment planning systems (TPS) capability to support C# scripts to develop an Eclipse scripting application programming interface (ESAPI) script and automate the preparation of the verification phantom plan for measurements. A local area network (LAN) connection between our measurement equipment and shared database was established to facilitate equipment control, measurement data transfer, and storage. To improve the analysis of the measurement data, a Python script was developed to automatically perform a 2D-3D γ-index analysis comparing measurements in the plane of a two-dimensional detector array with TPS predictions in a water phantom for each acquired measurement. RESULTS Device connection via LAN granted immediate access to the plan and measurement information for downstream analysis using an online software suite. Automated scripts applied to verification plans reduced time from preparation steps by at least 50%; time reduction from automating γ-index analysis was even more pronounced, dropping by a factor of 10. On average, we observed an overall time savings of 55% in completion of the PSQA per patient plan. CONCLUSIONS The automation of the routine tasks in the PSQA workflow significantly reduced the time required per patient, reduced user fatigue, and frees up system users from routine and repetitive workflow steps allowing increased focus on evaluating key quality metrics.

Data collection of patient outcomes: one institution’s experience

Abstract Patient- and provider-reported outcomes are recognized as important in evaluating quality of care, guiding health care policy, comparative effectiveness research, and decision-making in radiation oncology. Combining patient and provider outcome data with a detailed description of disease and therapy is the basis for these analyses. We report on the combination of technical solutions and clinical process changes at our institution that were used in the collection and dissemination of this data. This initiative has resulted in the collection of treatment data for 23 541 patients, 20 465 patients with provider-based adverse event records, and patient-reported outcome surveys submitted by 5622 patients. All of the data is made accessible using a self-service web-based tool.

Integration of biological and statistical models toward personalized radiation therapy of cancer

Abstract Radiation Therapy (RT) is one of the most common treatments for cancer. To understand the impact of radiation toxicity on normal tissue, a Normal Tissue Complication Probability (NTCP) model is needed to link RT dose with radiation-induced complications. There are two types of NTCP models: biological and statistical models. Biological models have good generalizability but low accuracy, as they cannot factor in patient-specific information. Statistical models can incorporate patient-specific variables, but may not generalize well across different studies. We propose an integrated model that borrows strength from both biological and statistical models. Specifically, we propose a novel model formulation followed by an efficient parameter estimation algorithm, and investigate statistical properties of the estimator. We apply the integrated model to a real dataset of prostate cancer patients treated with Intensity Modulated RT at the Mayo Clinic Arizona, who are at risk of developing the grade 2+ acute rectal complication. The integrated model achieves an Area Under the Curve (AUC) level of 0.82 in prediction, whereas the AUCs for the biological and statistical models are only 0.66 and 0.76, respectively. The superior performance of the integrated model is also consistently observed over different simulation experiments.

Spot scanning proton therapy plan assessment: design and development of a dose verification application for use in routine clinical practice

The use of radiation therapy for the treatment of cancer has been carried out clinically since the late 1800’s. Early on however, it was discovered that a radiation dose sufficient to destroy cancer cells can also cause severe injury to surrounding healthy tissue. Radiation oncologists continually strive to find the perfect balance between a dose high enough to destroy the cancer and one that avoids damage to healthy organs. Spot scanning or “pencil beam” proton radiotherapy offers another option to improve on this. Unlike traditional photon therapy, proton beams stop in the target tissue, thus better sparing all organs beyond the targeted tumor. In addition, the beams are far narrower and thus can be more precisely “painted” onto the tumor, avoiding exposure to surrounding healthy tissue. To safely treat patients with proton beam radiotherapy, dose verification should be carried out for each plan prior to treatment. Proton dose verification systems are not currently commercially available so the Department of Radiation Oncology at the Mayo Clinic developed its own, called DOSeCHECK, which offers two distinct dose simulation methods: GPU-based Monte Carlo and CPU-based analytical. The three major components of the system include the web-based user interface, the Linux-based dose verification simulation engines, and the supporting services and components. The architecture integrates multiple applications, libraries, platforms, programming languages, and communication protocols and was successfully deployed in time for Mayo Clinic’s first proton beam therapy patient. Having a simple, efficient application for dose verification greatly reduces staff workload and provides additional quality assurance, ultimately improving patient safety.

SU-D-204-03: Comparison of Patient Positioning Methods Through Modeling of Acute Rectal Toxicity in Intensity Modulated Radiation Therapy for Prostate Cancer. Does Quality of Data Matter More Than the Quantity?

PURPOSE To determine if differences in patient positioning methods have an impact on the incidence and modeling of grade >=2 acute rectal toxicity in prostate cancer patients who were treated with Intensity Modulated Radiation Therapy (IMRT). METHODS We compared two databases of patients treated with radiation therapy for prostate cancer: a database of 79 patients who were treated with 7 field IMRT and daily image guided positioning based on implanted gold markers (IGRTdb), and a database of 302 patients who were treated with 5 field IMRT and daily positioning using a trans-abdominal ultrasound system (USdb). Complete planning dosimetry was available for IGRTdb patients while limited planning dosimetry, recorded at the time of planning, was available for USdb patients. We fit Lyman-Kutcher-Burman (LKB) model to IGRTdb only, and Univariate Logistic Regression (ULR) NTCP model to both databases. We perform Receiver Operating Characteristics analysis to determine the predictive power of NTCP models. RESULTS The incidence of grade >= 2 acute rectal toxicity in IGRTdb was 20%, while the incidence in USdb was 54%. Fits of both LKB and ULR models yielded predictive NTCP models for IGRTdb patients with Area Under the Curve (AUC) in the 0.63 - 0.67 range. Extrapolation of the ULR model from IGRTdb to planning dosimetry in USdb predicts that the incidence of acute rectal toxicity in USdb should not exceed 40%. Fits of the ULR model to the USdb do not yield predictive NTCP models and their AUC is consistent with AUC = 0.5. CONCLUSION Accuracy of a patient positioning system affects clinically observed toxicity rates and the quality of NTCP models that can be derived from toxicity data. Poor correlation between planned and clinically delivered dosimetry may lead to erroneous or poorly performing NTCP models, even if the number of patients in a database is large.

SU‐E‐T‐803: Verification of QUANTEC Lyman Kutcher Burman (LKB) Model for Grade>=2(2+) Late Rectal Complication Rates Using a Database of 79 Prostate Patients Treated with IMRT

Purpose: QUANTEC review of best parameters for the LKB NTCP model of rectal complications is based exclusively on datasets obtained with 3D conformal techniques. The report suggests that inherent differences in rectal dose distributions obtained with IMRT techniques could require modification of the parameters which provide the best fit to clinical data. Methods: We compiled a database of 79 prostate patients who were treated with an IMRT technique to a dose of 77.4 Gy, 1.8Gy/fx, with an integrated boost to 81–83Gy in a sub-volume of a prostate which was identified on a pre-treatment MRI study. Rectal toxicities were graded according to CTCAE v4 by a physician who retrospectively reviewed patient’s medical records. Late grade 2+ toxicities were defined as toxicities occurring later than 90 days following the end of treatment. We defined the model in terms of parameters, m,n and TD_50, as recommended in the Quantec report. We converted the dose to 2Gy equivalent dose using linear-quadratic model with alpha over beta of 3Gy. We applied QUANTEC model to our data and compared results to our own fit of LKB model using two sample t-test. Results: Grade 2+ late rectal toxicity occurred in 4% (3/79 patients). The best fit of LKB model to data was obtained for n = 0.26, m = 0.25, and TD_50 = 81.53Gy. The two sample t-test yielded p value of 0.67. Average NTCP predicted by our parametrization is 3.8% while average NTCP predicted by QUANTEC parametrization is 3.5%. Conclusion: Predictions of late rectal toxicities in IMRT patients using LKB model with parameters from QUANTEC report match observed toxicity rates. Independent fit of LKB model to IMRT data produces somewhat different parameter set, but two parametrizations are equivalent within statistical uncertainties.

TH‐AB‐304‐02: Fitting Grade>=2(2+) Acute Rectal Complication Rates in Prostate Cancer Patients to Lyman Kutcher Burman (LKB) and Logistic Regression NTCP Models Using Dosimetry and Patient Specific Characteristics

Purpose: Models of rectal toxicity which include dosimetry only are known to have a relatively low predictive power, at least as measured by the area under the ROC curve (AUC). It has been suggested that the predictive power of models can be improved by including non-dosimetric patient specific characteristics. Methods: We compiled a database of 79 prostate patients who were treated with an IMRT technique to a dose of 77.4 Gy, 1.8Gy/fx, with an integrated boost to 81–83Gy in a sub-volume of a prostate which was identified on a pre-treatment MRI study. Acute grade 2+ rectal toxicities were graded according to CTCAE v4 by a physician who retrospectively reviewed patient’s medical records. We modified the LKB model to include one patient specific variable at a time, and we also used an NTCP model based on logistic regression to perform multi-variate analysis. We used patient specific variables available to us in a retrospective study: age, diabetes, hormonal treatment, Gleason Score, PSA, Statin use, prostate volume, boost volume and rectal volume. Results: Grade 2+ acute rectal toxicity occurred in 20% (16/79 patients). The LKB model with dosimetry alone gives AUC=0.65. Four variables, age, diabetes, PSA, Statin use, increase the AUC in LKB model to a maximum of 0.79. The same four variables in the logistic regression model increase AUC to 0.87. The most significant correlations are with PSA and with Statin use. Conclusion: Including patient specific variables in toxicity models can significantly increase apparent predictive power of a model. Somewhat surprising finding of a strong correlation between rectal toxicity and PSA in our dataset suggests that conclusions from each individual study should be treated with caution, until independently confirmed. Larger databases from prospective studies or meta-analysis of multiple studies may be needed to find patient characteristics that are truly predictive.

SU‐E‐T‐482: Development and Systematic Testing of Dose Analysis Engine for Research and Clinical Applications Using API Interface of Varian Eclipse Treatment Planning System

PURPOSE To validate standalone software that provides dose analysis capabilities of a treatment planning system. The intended use of this software is in research and in clinical applications where commercial solutions are not yet available. METHODS We used Application Programmer Interface (API) of the Eclipse Treatment Planning System (Varian,Inc), to extract dosimetric information from treatment plans. We used dose volume histogram (DVH) object to extract standard dose volume histogram data. We used two API functions to construct 3D representation of anatomy (a bitmap) and compute dose values at all voxels that belong to a structure. We wrote a C++ package which achieves the same reconstruction based on stored contour data and a dose volume. For this work we compared structure volumes and simple dose statistics on dose buffers (minimum, mean, maximum and standard deviation). We used data from 6 Head and Neck patients and examined a total of 102 structures. RESULTS The API based reconstruction of 3D anatomy systematically overestimates structure volumes when compared to volumes reported by the API DVH module. The discrepancy is largest for small structures, and can exceed 20%. Volume estimates obtained by the C++ package show better agreement with the DVH module if partial volume corrections are applied to surface voxels. Among four dosimetric variables, largest discrepancies are observed for the minimum dose. While we observe better than 2% agreement in most cases, we find outliers with discrepancies reaching over 20%, even when we compare the two API based methods. CONCLUSION We find that good validation of a standalone dose analysis package requires testing on a large series of structures obtained from multiple patients. Disagreements between two API based methods show that vendor algorithms have hidden features which may affect dosimetric analysis.