Classifying cancer pathology reports with hierarchical self-attention networks.

We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks - site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data - Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.

Scalable deep text comprehension for Cancer surveillance on high-performance computing.

BACKGROUND: Deep Learning (DL) has advanced the state-of-the-art capabilities in bioinformatics applications which has resulted in trends of increasingly sophisticated and computationally demanding models trained by larger and larger data sets. This vastly increased computational demand challenges the feasibility of conducting cutting-edge research. One solution is to distribute the vast computational workload across multiple computing cluster nodes with data parallelism algorithms. In this study, we used a High-Performance Computing environment and implemented the Downpour Stochastic Gradient Descent algorithm for data parallelism to train a Convolutional Neural Network (CNN) for the natural language processing task of information extraction from a massive dataset of cancer pathology reports. We evaluated the scalability improvements using data parallelism training and the Titan supercomputer at Oak Ridge Leadership Computing Facility. To evaluate scalability, we used different numbers of worker nodes and performed a set of experiments comparing the effects of different training batch sizes and optimizer functions. RESULTS: We found that Adadelta would consistently converge at a lower validation loss, though requiring over twice as many training epochs as the fastest converging optimizer, RMSProp. The Adam optimizer consistently achieved a close 2nd place minimum validation loss significantly faster; using a batch size of 16 and 32 allowed the network to converge in only 4.5 training epochs. CONCLUSIONS: We demonstrated that the networked training process is scalable across multiple compute nodes communicating with message passing interface while achieving higher classification accuracy compared to a traditional machine learning algorithm.

Deep Learning for Automated Extraction of Primary Sites From Cancer Pathology Reports.

Pathology reports are a primary source of information for cancer registries which process high volumes of free-text reports annually. Information extraction and coding is a manual, labor-intensive process. In this study, we investigated deep learning and a convolutional neural network (CNN), for extracting ICD-O-3 topographic codes from a corpus of breast and lung cancer pathology reports. We performed two experiments, using a CNN and a more conventional term frequency vector approach, to assess the effects of class prevalence and inter-class transfer learning. The experiments were based on a set of 942 pathology reports with human expert annotations as the gold standard. CNN performance was compared against a more conventional term frequency vector space approach. We observed that the deep learning models consistently outperformed the conventional approaches in the class prevalence experiment, resulting in micro- and macro-F score increases of up to 0.132 and 0.226, respectively, when class labels were well populated. Specifically, the best performing CNN achieved a micro-F score of 0.722 over 12 ICD-O-3 topography codes. Transfer learning provided a consistent but modest performance boost for the deep learning methods but trends were contingent on the CNN method and cancer site. These encouraging results demonstrate the potential of deep learning for automated abstraction of pathology reports.

Hierarchical attention networks for information extraction from cancer pathology reports.

OBJECTIVE: We explored how a deep learning (DL) approach based on hierarchical attention networks (HANs) can improve model performance for multiple information extraction tasks from unstructured cancer pathology reports compared to conventional methods that do not sufficiently capture syntactic and semantic contexts from free-text documents. MATERIALS AND METHODS: Data for our analyses were obtained from 942 deidentified pathology reports collected by the National Cancer Institute Surveillance, Epidemiology, and End Results program. The HAN was implemented for 2 information extraction tasks: (1) primary site, matched to 12 International Classification of Diseases for Oncology topography codes (7 breast, 5 lung primary sites), and (2) histological grade classification, matched to G1-G4. Model performance metrics were compared to conventional machine learning (ML) approaches including naive Bayes, logistic regression, support vector machine, random forest, and extreme gradient boosting, and other DL models, including a recurrent neural network (RNN), a recurrent neural network with attention (RNN w/A), and a convolutional neural network. RESULTS: Our results demonstrate that for both information tasks, HAN performed significantly better compared to the conventional ML and DL techniques. In particular, across the 2 tasks, the mean micro and macro F-scores for the HAN with pretraining were (0.852,0.708), compared to naive Bayes (0.518, 0.213), logistic regression (0.682, 0.453), support vector machine (0.634, 0.434), random forest (0.698, 0.508), extreme gradient boosting (0.696, 0.522), RNN (0.505, 0.301), RNN w/A (0.637, 0.471), and convolutional neural network (0.714, 0.460). CONCLUSIONS: HAN-based DL models show promise in information abstraction tasks within unstructured clinical pathology reports.

Designing a public square for research computing.

A set of principles is proposed for sponsors and developers of research computing applications that can increase the likelihood of successful adoption by researchers.

Evaluation of prediagnostic prostate-specific antigen dynamics as predictors of death from prostate cancer in patients treated conservatively.

Prostate-specific antigen (PSA) dynamics have been proposed to predict outcome in men with prostate cancer. We assessed the value of PSA velocity (PSAV) and PSA doubling time (PSADT) for predicting prostate cancer-specific mortality (PCSM) in men with clinically localized prostate cancer undergoing conservative management or early hormonal therapy. From 1990 to 1996, 2,333 patients were identified, of whom 594 had two or more PSA values before diagnosis. We examined 12 definitions for PSADT and 10 for PSAV. Because each definition required PSA measurements at particular intervals, the number of patients eligible for each definition varied from 40 to 594 and number of events from 10 to 119. Four PSAV definitions, but no PSADT, were significantly associated with PCSM after adjustment for PSA in multivariable Cox proportional hazards regression. All four could be calculated only for a proportion of events, and the enhancements in predictive accuracy associated with PSAV had very wide confidence intervals. There was no clear benefit of PSAV in men with low PSA and Gleason grade 6 or less. Although evidence that certain PSAV definitions help to predict PCSM in the cohort exist, the value of incorporating PSAV in predictive models to assist in determining eligibility for conservative management is, at best, uncertain.

Postoperative nomogram predicting the 9-year probability of prostate cancer recurrence after permanent prostate brachytherapy using radiation dose as a prognostic variable.

PURPOSE: To report a multi-institutional outcomes study on permanent prostate brachytherapy (PPB) to 9 years that includes postimplant dosimetry, to develop a postimplant nomogram predicting biochemical freedom from recurrence. METHODS AND MATERIALS: Cox regression analysis was used to model the clinical information for 5,931 patients who underwent PPB for clinically localized prostate cancer from six centers. The model was validated against the dataset using bootstrapping. Disease progression was determined using the Phoenix definition. The biological equivalent dose was calculated from the minimum dose to 90% of the prostate volume (D90) and external-beam radiotherapy dose using an alpha/beta of 2. RESULTS: The 9-year biochemical freedom from recurrence probability for the modeling set was 77% (95% confidence interval, 73-81%). In the model, prostate-specific antigen, Gleason sum, isotope, external beam radiation, year of treatment, and D90 were associated with recurrence (each p < 0.05), whereas clinical stage was not. The concordance index of the model was 0.710. CONCLUSION: A predictive model for a postimplant nomogram for prostate cancer recurrence at 9-years after PPB has been developed and validated from a large multi-institutional database. This study also demonstrates the significance of implant dosimetry for predicting outcome. Unique to predictive models, these nomograms may be used a priori to calculate a D90 that likely achieves a desired outcome with further validation. Thus, a personalized dose prescription can potentially be calculated for each patient.

Pretreatment prostate-specific antigen (PSA) velocity and doubling time are associated with outcome but neither improves prediction of outcome beyond pretreatment PSA alone in patients treated with radical prostatectomy.

PURPOSE: Controversy exists as to whether current pretreatment prostate-specific antigen (PSA) dynamics enhance outcome prediction in patients undergoing treatment for prostate cancer. We assessed whether pretreatment PSA velocity (PSAV) or doubling time (PSADT) predicted outcome in men undergoing radical prostatectomy and whether any definition enhanced accuracy of an outcome prediction model. PATIENTS AND METHODS: The cohort included 2,938 patients with two or more PSA values before radical prostatectomy. Biochemical recurrence (BCR) occurred in 384 patients, and metastases occurred in 63 patients. Median follow-up for patients without BCR was 2.1 years. We used univariate Cox proportional hazards regression to evaluate associations between published definitions of PSADT and PSAV with BCR and metastasis. Predictive accuracy was assessed using the concordance index. RESULTS: On univariate analysis, two of 12 PSADT and four of 10 PSAV definitions were univariately associated with both BCR and metastasis (P < .05). One PSADT and one PSAV definition had a higher predictive accuracy for BCR over PSA alone, and four PSAV definitions improved prediction of metastasis. However, the improvements in predictive accuracy were small, associated with wide CIs, and markedly reduced if additional predictors of stage and grade were included alongside PSA. Modeling with random variables suggests that similar results would be expected by chance. CONCLUSION We found no clear evidence that any definition of PSA dynamics substantially enhances the predictive accuracy of a single pretreatment PSA alone.

Multicenter analysis of effect of high biologic effective dose on biochemical failure and survival outcomes in patients with Gleason score 7-10 prostate cancer treated with permanent prostate brachytherapy.

PURPOSE: To investigate the biochemical control rates and survival for Gleason score 7-10 prostate cancer patients undergoing permanent prostate brachytherapy as a function of the biologic effective dose (BED). METHODS AND MATERIALS: Six centers provided data on 5,889 permanent prostate brachytherapy patients, of whom 1,078 had Gleason score 7 (n = 845) or Gleason score 8-10 (n = 233) prostate cancer and postimplant dosimetry results available. The median prostate-specific antigen level was 7.5 ng/mL (range, 0.4-300). The median follow-up for censored patients was 46 months (range, 5-130). Short-term hormonal therapy (median duration, 3.9 months) was used in 666 patients (61.8%) and supplemental external beam radiotherapy (EBRT) in 620 (57.5%). The patients were stratified into three BED groups: <200 Gy (n = 645), 200-220 Gy (n = 199), and >220 Gy (n = 234). Biochemical freedom from failure (bFFF) was determined using the Phoenix definition. RESULTS: The 5-year bFFF rate was 80%. The bFFF rate stratified by the three BED groups was 76.4%, 83.5%, and 88.3% (p < 0.001), respectively. Cox regression analysis revealed Gleason score, prostate-specific antigen level, use of hormonal therapy, EBRT, and BED were associated with bFFF (p < 0.001). Freedom from metastasis improved from 92% to 99% with the greatest doses. The overall survival rate at 5 years for the three BED groups for Gleason score 8-10 cancer was 86.6%, 89.4%, and 94.6%, respectively (p = 0.048). CONCLUSION: These data suggest that permanent prostate brachytherapy combined with EBRT and hormonal therapy yields excellent bFFF and survival results in Gleason score 7-10 patients when the delivered BEDs are >220 Gy. These doses can be achieved by a combination of 45-Gy EBRT with a minimal dose received by 90% of the target volume of 120 Gy of (103)Pd or 130 Gy of (125)I.

Development of a Virtual Reality Coping Skills Game to Prevent Post-Hospitalization Smoking Relapse in Tobacco Dependent Cancer Patients.

Many hospitalized smokers return to smoking after hospital discharge even though continued smoking can compromise treatment effectiveness, reduce survival, increase risk of disease recurrence, and impair quality of life. After leaving a smoke-free hospital, patients encounter smoking cues at home, such as family members who smoke or emotional triggers such as stress, which can elicit powerful urges to smoke and lead to smoking relapse. Enabling smokers to experience such urges in a controlled setting while providing the ability to practice coping skills may be a useful strategy for building quitting self-efficacy. We are developing a virtual reality coping skills (VRCS) game to help hospitalized smokers practice coping strategies to manage these triggers in preparation for returning home after hospitalization. Our multidisciplinary team developed a prototype VRCS game using Second Life, a platform that allowed rapid construction of a virtual reality environment. The prototype contains virtual home spaces (e.g., living room, kitchen) populated with common triggers to smoke and a "toolkit" with scripted actions that enable the avatar to rehearse various coping strategies. Since eliciting and managing urges to smoke is essential to the game's utility as an intervention, we assessed the ability of the prototype virtual environment to engage former smokers in these scenarios. We recruited eight former smokers with a recent history of hospitalization and guided each through a VRCS scenario during which we asked the patient to evaluate the strength of smoking urges and usefulness of coping strategies. Initial data indicate that patients report high urges to smoke (mean = 8.8 on a 10 point scale) when their avatar confronted virtual triggers such as drinking coffee. Patients rated virtual practice of coping strategies, such as drinking water or watching TV, as very helpful (mean = 8.4 on a 10 point scale) in reducing these urges. With further development, this VRCS game may have potential to provide low-cost, effective behavioral rehearsal to prevent relapse to smoking in hospitalized patients.

12-year outcomes following permanent prostate brachytherapy in patients with clinically localized prostate cancer.

PURPOSE: We reviewed the outcomes in men treated with permanent prostate brachytherapy (PPB). MATERIAL AND METHODS: A total of 1,449 consecutive patients with a mean age of 68 years treated with PPB between 1992 and 2000 and mean pretreatment prostate specific antigen (PSA) 10.1 ng/ml were included in this study. Of the patients 55% presented with Gleason 6 tumors and 28% had Gleason 7 disease. A total of 400 patients (27%) were treated with neoadjuvant hormones and 301 (20%) were treated in combination with external radiation plus PPB. Several biochemical freedom from recurrence (BFR) definitions were determined. Statistical analysis consisted of log rank testing, Kaplan-Meier estimates and Cox regression analysis. RESULTS: Median followup was 82 months with 39 patients at risk at for 144 months. Overall and disease specific survival at 12 years was 81% and 93%, respectively. The 12-year BFR was 81%, 78%, 74% and 77% according to the American Society for Therapeutic Radiology and Oncology (ASTRO), ASTRO-Kattan, ASTRO-Last Call and Houston definitions, respectively. The 12-year ASTRO-Kattan BFR using risk stratification was 89%, 78% and 63% in patients at low, intermediate and high risk, respectively (p = 0.0001). Multivariate analysis identified the dose that 90% of the target volume received (p <0.0001), pretreatment PSA (p = 0.001), Gleason score (p = 0.002), the percent positive core biopsies (p = 0.037), clinical stage (p = 0.689), the addition of hormones (p = 0.655) and the addition of external radiation (p = 0.724) for predicting BFR-ASTRO. Five-year disease specific survival was 44% in patients with a PSA doubling time of less than 12 months vs 88% in those with a PSA doubling time of 12 months or greater (p = 0.0001). CONCLUSIONS: PPB offers acceptable 12-year BFR in patients who present with clinically localized prostate cancer. Implant dosimetry continues as an important predictor for BFR, while the addition of adjuvant therapies such as hormones and external radiation are insignificant. In patients who experience biochemical failure it appears that PSA doubling time is an important predictor of survival.

Local progression among men with conservatively treated localized prostate cancer: results from the Transatlantic Prostate Group.

OBJECTIVES: Men with clinically detected localized prostate cancer treated without curative intent are at risk of complications from local tumor growth. We investigated rates of local progression and need for local therapy among such men. METHODS: Men diagnosed with prostate cancer during 1990-1996 were identified from cancer registries throughout the United Kingdom. Inclusion criteria were age < or =76 yr at diagnosis, PSA level < or =100 ng/ml, and, within 6 mo after diagnosis, no radiation therapy, radical prostatectomy, evidence of metastatic disease, or death. Local progression was defined as increase in clinical stage from T1/2 to T3/T4 disease, T3 to T4 disease, and/or need for transurethral resection of the prostate (TURP) to relieve symptoms >6 mo after cancer diagnosis. RESULTS: The study included 2333 men with median follow-up of 85 mo (range: 6-174). Diagnosis was by TURP in 1255 men (54%), needle biopsy in 1039 (45%), and unspecified in 39 (2%). Only 29% were treated with hormonal therapy within 6 mo of diagnosis. Local progression occurred in 335 men, including 212 undergoing TURP. Factors most predictive of local progression on multivariable analysis were PSA at diagnosis and Gleason score of the diagnostic tissue (detrimental), and early hormonal therapy (protective). We present a nomogram that predicts the likelihood of local progression within 120 mo after diagnosis. CONCLUSIONS: Men with clinically detected localized prostate cancer managed without curative intent have an approximately 15% risk for local progression within 10 yr of diagnosis. Among those with progression, the need for treatment is common, even among men diagnosed by TURP. When counseling men who are candidates for management without curative intent, the likelihood of symptoms from local progression must be considered.

Nomogram incorporating PSA level to predict cancer-specific survival for men with clinically localized prostate cancer managed without curative intent.

BACKGROUND: The prognosis of men with clinically localized prostate cancer is highly variable, and it is difficult to counsel a man who may be considering avoiding, or delaying, aggressive therapy. After collecting data on a large cohort of men who received no initial active prostate cancer therapy, the aim was to develop, and to internally validate, a nomogram for prediction of disease-specific survival. METHODS: Working with 6 cancer registries within England and numerous hospitals in the region, a population-based cohort of men diagnosed with prostate cancer between 1990 and 1996 was constructed. All men had baseline serum prostate-specific antigen (PSA) measurements, centralized pathologic grading, and centralized review of clinical stage assignment. Based on the clinical and pathologic data from 1911 men, a statistical model was developed and validated that served as the basis for the nomogram. The discrimination and calibration of the nomogram were assessed with use of one-third of the men, who were omitted from modeling and used as a test sample. RESULTS: The median age of the included men was 70.4 years. The 25th and 75th percentiles of PSA were 7.3 and 32.6 ng/mL respectively, and the median was 15.4 ng/mL. Forty-two percent of the men had high-grade disease. The nomogram predicted well, with a concordance index of 0.73, and had good calibration. CONCLUSIONS: An accurate tool was developed for predicting the probability that a man with clinically localized prostate cancer will survive his disease for 120 months if the cancer is not treated with curative intent immediately. The tool should be helpful for patient counseling and clinical trial design.

Long-term toxicity monitoring via electronic patient-reported outcomes in patients receiving chemotherapy.

PURPOSE: In cancer treatment trials, clinicians traditionally report patient toxicity symptoms. Alternatively, patients could provide this information directly. PATIENTS AND METHODS: The Common Terminology Criteria for Adverse Events (CTCAE) is the mandated instrument for tracking patient toxicity symptoms in National Cancer Institute (NCI)-sponsored cancer treatment trials. We adapted CTCAE symptom items into patient language and uploaded these to an online platform. Lung cancer outpatients receiving chemotherapy were invited to self-report selected symptoms at visits via waiting area computers or optional home access. Symptom reports were printed for nurses at visits, but no instructions were given with regard to use of this information. RESULTS: From June 2005 through March 2006, 125 patients were invited to participate, and 107 chose to enroll. Mean length of participation was 42 weeks (range, 1 to 71 weeks), by which time 35% died. The average number of clinic visits was 12 (range, 1 to 40 visits). At each consecutive visit, most patients (mean, 78%) logged in without significant attrition. Reasons for failure to log in included having no reminder and having inadequate time. Although 76% of enrollees had home computers, only 15% self-reported from home. Satisfaction with the system was high (90%), but only 51% felt communication was improved. All participating nurses understood the reports and felt this information was useful for clinical decisions, documentation, and discussions. However, only one of seven nurses discussed reports with patients frequently, with insufficient time being the most common barrier to discussions. CONCLUSION: Online patient self-reporting is a feasible long-term strategy for toxicity symptom monitoring during chemotherapy, even among patients with advanced cancer and high symptom burdens. However, without explicit reminders and clinician feedback, patients demonstrated limited voluntary interest in self-reporting between visits.

Major shifts in the treatment and prognosis of prostate cancer due to changes in pathological diagnosis and grading.

OBJECTIVE: To examine data on the changes in the accuracy of the diagnosis of prostate cancer and of Gleason grading in the modern era. PATIENTS AND METHODS: The study comprised a pathological review within a multicentre study of patients with clinically localized prostate cancer diagnosed in the UK from 1991 to 1996 (inclusive) and treated by watchful-waiting or hormonal therapy alone. The clinical follow-up was available, histopathological appearances were reviewed and the Gleason score at diagnosis was compared with the Gleason score as analysed by a panel of genitourinary pathologists using internationally agreed criteria. In all, 1789 patients diagnosed with prostate cancer between 1991 and 1996 were reviewed, with disease-specific survival as the main outcome measure. RESULTS: In all, 133 patients (7%) were reassigned a nonmalignant diagnosis. There was a significant reassignment in the Gleason score for those with cancer, with increases of Gleason score across a wide spectrum. In multivariate analysis the revised Gleason score was a more accurate predictor of prognosis than the original score. CONCLUSION: Misdiagnosis and reassignment of Gleason score at diagnosis would have guided clinicians into large-scale changes in the management of patients. Current rates of misdiagnosis are unknown. If applicable nationally, these changes would have profound effects on the workload of prostate cancer management in the UK.

Pretreatment nomogram predicting ten-year biochemical outcome of three-dimensional conformal radiotherapy and intensity-modulated radiotherapy for prostate cancer.

OBJECTIVES: To update our previously published nomogram predicting for biochemical outcome with 10-year data from a larger cohort of patients treated with three-dimensional conformal radiotherapy (RT) or intensity-modulated RT for localized prostate cancer. METHODS: From 1988 to 2004, 2253 patients were treated with three-dimensional conformal RT or intensity-modulated RT for clinical Stage T1-T3 prostate cancer. Prescription doses ranged from 64.8 to 86.4 Gy. The median follow-up time was 7 years. The nomogram was developed using a proportional hazards regression model predicting for the probability of biochemical relapse after RT according to the nadir plus 2 ng/mL definition of prostate-specific antigen (PSA) relapse. RESULTS: The 10-year PSA relapse-free survival rate was 62%. The nomogram incorporated the following variables to predict likelihood of PSA failure after RT: pretreatment PSA level, Gleason score, radiation dose, use of neoadjuvant androgen deprivation, and clinical stage. The concordance index of this long-term nomogram was 0.72. CONCLUSIONS: A nomogram predicting the 10-year probability of biochemical control after three-dimensional conformal RT or intensity-modulated RT for prostate cancer was reasonably accurate and discriminating. The nomogram also provided evidence that long-term biochemical control can be achieved after conformal RT for the treatment of localized prostate cancer.

Customized dose prescription for permanent prostate brachytherapy: insights from a multicenter analysis of dosimetry outcomes.

PURPOSE: To investigate the biochemical control rate in patients undergoing permanent prostate brachytherapy as a function of the biologically effective dose (BED) and risk group. METHODS AND MATERIALS: Six centers provided data on 3,928 permanent brachytherapy patients with postimplant dosimetry results. The mean prostate-specific antigen level was 8.9 ng/mL. (125)I was used in 2,293 (58%), (103)Pd in 1,635, and supplemental external beam radiotherapy in 882 (22.5%) patients. The patients were stratified into low- (n = 2,188), intermediate- (n = 1,188), and high- (n = 552) risk groups and into three BED groups of < 140 Gy (n = 524), 140-200 Gy (n = 2284), and >200 Gy (n = 1,115). Freedom from biochemical disease progression (biochemical freedom from failure [bFFF]) was determined using the American Society for Therapeutic Radiology Oncology and Phoenix definitions and calculated using the Kaplan-Meier method, with factors compared using the log-rank test. RESULTS: The 10-year prostate-specific antigen bFFF rate for the American Society for Therapeutic Radiology Oncology and Phoenix definitions was 79.2% and 70%, respectively. The corresponding bFFF rates for the low-, intermediate-, and high-risk groups was 84.1% and 78.1%, 76.8% and 63.6%, and 64.4% and 58.2%, respectively (p < 0.0001). The corresponding bFFF rate for the three BED groups was 56.1% and 41.4%, 80% and 77.9%, and 91.1% and 82.9% (p < 0.0001). The corresponding bFFF rate for the low-risk patients by dose group was 69.8% and 49.8%, 86% and 85.2%, and 88.1% and 88.3% for the low-, intermediate, and high-dose group, respectively (p <0.0001). The corresponding bFFF rate for the intermediate-risk patients by dose group was 52.9% and 23.1%, 74.1% and 77.7%, and 94.3% and 88.8% for the low-, intermediate-, and high-dose group, respectively (p < 0.0001). The corresponding bFFF rate for high-risk patients by dose group was 19.2% and 41.7%, 61.8% and 53.2%, and 90% and 69.6% for the low-, intermediate-, and high-dose group, respectively (p < 0.0001). CONCLUSIONS: These data suggest that permanent brachytherapy dose prescriptions can be customized to risk status. In low-risk patients, achieving a BED of >or=140 Gy might be adequate for prostate-specific antigen control. However, high-risk disease might require a BED dose of >or=200 Gy.

Evaluation of an online platform for cancer patient self-reporting of chemotherapy toxicities.

The current mechanism for monitoring toxicity symptoms in cancer trials depends on a complex paper-based process. Electronic collection of patient-reported outcomes (PROs) may be more efficient and accurate. An online PRO platform was created including a simple data entry interface, real-time report generation, and an alert system to e-mail clinicians when patients self-report serious toxicities. Feasibility assessment involving 180 chemotherapy patients demonstrated high levels of use at up to 40 follow-up clinic visits per patient over 16 months (85% of patients at any given visit), with high levels of patient and clinician acceptance and satisfaction (>95%). Alerts were used as the basis for delayed chemotherapy treatments, dose modifications, and scheduling changes. These results demonstrate that online patient-reporting is a feasible strategy for chemotherapy toxicity symptom monitoring, and may improve safety and satisfaction with care. Ongoing multi-center research will evaluate the impact of this approach on clinical and administrative outcomes.