Search

Purpose: To evaluate the efficacy of prominent machine learning algorithms in predicting normal tissue complication probability using clinical data obtained from 2 distinct disease sites and to create a software tool that facilitates the automatic determination of the optimal algorithm to model any given labeled data set. Methods and Materials: We obtained 3 sets of radiation toxicity data (478 patients) from our clinic: gastrointestinal toxicity, radiation pneumonitis, and radiation esophagitis. These data comprised clinicopathological and dosimetric information for patients diagnosed with non-small cell lung cancer and anal squamous cell carcinoma. Each data set was modeled using 11 commonly employed machine learning algorithms (elastic net, least absolute shrinkage and selection operator [LASSO], random forest, random forest regression, support vector machine, extreme gradient boosting, light gradient boosting machine, k-nearest neighbors, neural network, Bayesian-LASSO, and Bayesian neural network) by randomly dividing the data set into a training and test set. The training set was used to create and tune the model, and the test set served to assess it by calculating performance metrics. This process was repeated 100 times by each algorithm for each data set. Figures were generated to visually compare the performance of the algorithms. A graphical user interface was developed to automate this whole process. Results: LASSO achieved the highest area under the precision-recall curve (0.807 ± 0.067) for radiation esophagitis, random forest for gastrointestinal toxicity (0.726 ± 0.096), and the neural network for radiation pneumonitis (0.878 ± 0.060). The area under the curve was 0.754 ± 0.069, 0.889 ± 0.043, and 0.905 ± 0.045, respectively. The graphical user interface was used to compare all algorithms for each data set automatically. When averaging the area under the precision-recall curve across all toxicities, Bayesian-LASSO was the best model. Conclusions: Our results show that there is no best algorithm for all data sets. Therefore, it is important to compare multiple algorithms when training an outcome prediction model on a new data set. The graphical user interface created for this study automatically compares the performance of these 11 algorithms for any data set.

Background: Proton therapy (PT) has unique biologic properties with excellent clinical outcomes for the management of localized prostate cancer. Here, we aim to characterize the toxicity of PT for patients with localized prostate cancer and propose mitigation strategies using a large institutional database. Methods: We reviewed medical records of 2772 patients with localized prostate cancer treated with definitive PT between May 2006 through January 2020. Disease risk was stratified according to National Comprehensive Cancer Network guidelines as low [LR, n = 640]; favorable-intermediate [F-IR, n = 849]; unfavorable-intermediate [U-IR, n = 851]; high [HR, n = 315]; or very high [VHR, n = 117]. Descriptive statistics and Kaplan-Meier estimates assessed toxicity and freedom from biochemical relapse (FFBR). Results: Median follow-up was 7.0 years. The median dose was 78 Gy(RBE)(range: 72–79.2 Gy) in 2.0 Gy(RBE) fractions; 63 % of patients received 78 Gy(RBE) in 39 fractions, and 29 % received 76 Gy(RBE) in 38 fractions. Overall rates of late grade ≥3 GU and GI toxicity were 0.87 % and 1.01 %, respectively. Two patients developed grade 4 late GU toxicity and seven patients with grade 4 late GI toxicity. All patients experiencing severe late grade 4 toxicities were treated to 78 Gy(RBE) in 39 fractions with 80 Gy(RBE) dose to the anterior rectal wall and/or bladder neck. The 10-year FFBR rates for patients with LR to U-IR disease were compared between those treated with 76 and 78 Gy(RBE); the rates were 94.5 % (95 % confidence interval [CI] 92.4–96.0 %) and 93.2 % (95 % CI 91.3–95.7 %), respectively (log-rank p = 0.22). Conclusions: Proton therapy is associated with low rates of late grade ≥3 GU and GI toxicity. While rare, late grade 4 toxicities occurred in nine (0.3 %) patients. De-escalation to a total dose of 76 Gy(RBE) yields excellent clinical outcomes for patients with LR to U-IR disease with the potential for significant reductions in grade ≥3 late toxicity.

Purpose: Chemoradiation therapy (CRT) is the standard treatment for squamous cell carcinoma of the anus (SCCA). This study aimed to investigate the relationship between vaginal dosimetry and long-term patient-reported dyspareunia after treatment. We further aimed to use the anterior vaginal wall (AVW) as an organ at risk to define an actionable dosimetric clinical goal to decrease the risk of patient-reported dyspareunia. Methods and Materials: Women with SCCA treated with intensity modulated radiation therapy–based CRT were surveyed at least 2 years after successfully completing therapy. A Female Sexual Function Index (FSFI) pain subscore ≤4 was used to define dyspareunia. Dosimetric parameters were calculated for both the full vaginal canal and AVW. Multivariable linear regression models were created to identify predictors of FSFI pain subscore using backward selection to identify final variables include in the models. An actionable dosimetric predictor for dyspareunia was established using the Youden index method for cutoff optimization. Results: Of 184 women who were contacted, 90 (49%) returned completed surveys. Of those who completed surveys, 51 (56.7%) reported being sexually active, and 47 had dosimetric data available for review. Of sexually active respondents, 32 (68%) had an FSFI pain subscore ≤4. Multiple regression models were generated using the full vaginal canal and AVW as organs at risk, and both models showed similar predictive relationships with volumetric dose parameters emerging as the best dosimetric predictors for dysparenuia. Age over 65 years was also associated with higher FSFI pain subscores (eg, less pain with intercourse) in both models. AVW V35 Gy < 60% was identified as the optimal cutoff to reduce the risk of patient-reported dyspareunia. Conclusions: Increased dose to the vaginal canal is significantly associated with worse patient-reported dyspareunia following CRT for SCCA. Minimizing dose to the AVW to V35 Gy < 60% may reduce the risk of this quality of life-limiting toxicity. Further prospective evaluation is needed to validate these findings.

Purpose: Pelvic chemoradiation therapy (CRT) is the curative therapy for non-metastatic anal cancer, resulting in excellent survival rates. However, anal cancer survivors may experience diminished quality of life (QOL) due to late toxicities from pelvic CRT. Few studies use patient-reported QOL, and few include thematic analyses of the patient experience in their own words. We conducted a survey study with qualitative analysis of free-text responses to explore themes of patient perceptions of their QOL during and after pelvic CRT to inform future interventions, reform patient education, and improve outcomes. Methods and Materials: We surveyed 248 patients with anal cancer treated with definitive intensity modulated radiation and concurrent chemotherapy from 2010 to 2018 who were alive and without recurrence. The survey included the Functional Assessment of Cancer Therapy General 7 item version, questions about satisfaction with preparation and patient education, and an optional free-text response question. Survey free response data were analyzed using the constant comparative method of qualitative analysis. Representative themes were generated. Results: One-hundred and twelve patients (45%) completed surveys. Of these respondents, 84 (75%) answered the free-text question. The median (interquartile range) Functional Assessment of Cancer Therapy General 7 item version score for survey responders (N = 112) was 21 (range, 15-24). Three themes most mentioned by respondents were persistent toxicity effecting QOL (82%), insufficient upfront information about CRT (56%), and gratitude toward care received (35%). Conclusions: Patients described late toxicities that affect QOL after pelvic CRT for anal cancer. This may be partially due to insufficient preparative information and post-treatment support. This study contributes to the literature supporting improved patient education and side effect management to augment long-term QOL for survivors of anal cancer.

Purpose: In this investigation, we aimed to describe trends in time to acceptance (TTA) and time to online publication (TTOP) of research published in leading radiation oncology journals from 2010 to 2019. We further sought to identify journal characteristics that might influence TTA and TTOP. Methods and Materials: We searched the publication history of 5 leading international radiation oncology journals. For all research articles accepted from January 1, 2010, to December 31, 2019, we tabulated the date of article receipt, the date of acceptance, and the date of online publication when available. The TTA was calculated as the number of elapsed days from article receipt to acceptance, and the TTOP was calculated as the number of elapsed days from article acceptance to online publication. Using the Mann-Kendall test, we assessed for monotonic trends over time and used the post hoc Theil-Sen method to estimate rates of change. We created a multiple regression model to identify journal characteristics associated with TTA and TTOP. Results: In total, 10,132 articles were included. Both the TTA and the TTOP decreased significantly from 2010 to 2019 (P = .005 and P < .001, respectively), with an estimated decrease of 1.5 days per year for the TTA and 7.0 days per year for the TTOP. Multiple regression modeling revealed that a higher journal impact factor was independently associated with an increased TTA (P < .001) and a decreased TTOP (P < .001). A higher number of accepted journal articles per year was associated with a decreased TTA (P < .001) and an increased TTOP (P < .001). Conclusions: Radiation oncology research has been accepted and published online at increasingly faster rates during the past decade. The TTA may be longer in higher-impact, more selective journals, possibly suggesting a need for comprehensive peer review and complex editorial decisions. However, these articles are also published online faster after article acceptance. Future work examining patterns of acceptance and publication speed is needed to encourage rapid dissemination of practice-guiding data.

Signet-ring cell carcinoma (SRCC) is a subtype of adenocarcinoma characterized histologically by cells with abundant intracellular mucin production, creating a distinctive “signet-ring” appearance. SRCC typically arises in tissues of the gastrointestinal tract. In the breast, SRCC is a rare subtype of invasive lobular carcinoma defined by the presence of signet-ring cells. Mammary SRCC has unique patterns of metastatic spread compared to more common breast cancers, more often spreading viscerally to involve the serosal surfaces of stomach, urinary tract, and spleen. We report a case of metastatic mammary SRCC presenting in the stomach, thereby mimicking primary gastric SRCC. Recognition of this unusual presentation is critical as treatments for these malignancies are vastly different; misdiagnosis could lead to significant morbidity and mortality for patients.

Comprehensive patient education is necessary for shared decision-making. While patient–provider conversations primarily drive patient education, patients also use published materials to enhance their understanding. In this investigation, we evaluated the readability of 2585 patient education materials published in high-impact medical journals from 1998 to 2018 and compared our findings to readability recommendations from national groups. For all materials, mean readability grade levels ranged from 11.2 to 13.8 by various metrics. Fifty-four (2.1%) materials met the American Medical Association recommendation of sixth grade reading level, and 215 (8.2%) met the National Institutes of Health recommendation of eighth grade level. When stratified by journal and material type, general medical education materials from Annals of Internal Medicine were the most readable ( P < .001), with 79.8% meeting the eighth grade level. Readability did not differ significantly over time. Efforts to standardize publication practice with the incorporation of readability evaluation during the review process may improve patients’ understanding of their disease processes and treatment options.