Text mining of cancer-related information: review of current status and future directions.

PURPOSE: This paper reviews the research literature on text mining (TM) with the aim to find out (1) which cancer domains have been the subject of TM efforts, (2) which knowledge resources can support TM of cancer-related information and (3) to what extent systems that rely on knowledge and computational methods can convert text data into useful clinical information. These questions were used to determine the current state of the art in this particular strand of TM and suggest future directions in TM development to support cancer research. METHODS: A review of the research on TM of cancer-related information was carried out. A literature search was conducted on the Medline database as well as IEEE Xplore and ACM digital libraries to address the interdisciplinary nature of such research. The search results were supplemented with the literature identified through Google Scholar. RESULTS: A range of studies have proven the feasibility of TM for extracting structured information from clinical narratives such as those found in pathology or radiology reports. In this article, we provide a critical overview of the current state of the art for TM related to cancer. The review highlighted a strong bias towards symbolic methods, e.g. named entity recognition (NER) based on dictionary lookup and information extraction (IE) relying on pattern matching. The F-measure of NER ranges between 80% and 90%, while that of IE for simple tasks is in the high 90s. To further improve the performance, TM approaches need to deal effectively with idiosyncrasies of the clinical sublanguage such as non-standard abbreviations as well as a high degree of spelling and grammatical errors. This requires a shift from rule-based methods to machine learning following the success of similar trends in biological applications of TM. Machine learning approaches require large training datasets, but clinical narratives are not readily available for TM research due to privacy and confidentiality concerns. This issue remains the main bottleneck for progress in this area. In addition, there is a need for a comprehensive cancer ontology that would enable semantic representation of textual information found in narrative reports.

Dose-escalated hypofractionated intensity-modulated radiotherapy in high-risk carcinoma of the prostate: outcome and late toxicity.

Background. The benefit of dose-escalated hypofractionated radiotherapy using intensity-modulated radiotherapy (IMRT) in prostate cancer is not established. We report 5-year outcome and long-term toxicity data within a phase II clinical trial. Materials and Methods. 60 men with predominantly high-risk prostate cancer were treated. All patients received neoadjuvant hormone therapy, completing up to 6 months in total. Thirty patients were treated with 57 Gy in 19 fractions and 30 patients with 60 Gy in 20 fractions. Acute and 2-year toxicities were reported and patients followed longitudinally to assess 5 year outcomes and long-term toxicity. Toxicity was measured using RTOG criteria and LENT/SOMA questionnaire. Results. Median followup was 84 months. Five-year overall survival (OS) was 83% and biochemical progression-free survival (bPFS) was 50% for 57 Gy. Five-year OS was 75% and bPFS 58% for 60 Gy. At 7 years, toxicity by RTOG criteria was acceptable with no grade 3 or above toxicity. Compared with baseline, there was no significant change in urinary symptoms at 2 or 7 years. Bowel symptoms were stable between 2 and 7 years. All patients continued to have significant sexual dysfunction. Conclusion. In high-risk prostate cancer, dose-escalated hypofractionated radiotherapy using IMRT results in encouraging outcomes and acceptable late toxicity.

Prospective analysis of patient-reported late toxicity following pelvic radiotherapy for gynaecological cancer.

BACKGROUND AND PURPOSE: As late radiotherapy toxicity impacts negatively on the quality-of-life of cancer survivors and is often under reported, a study was set up to prospectively collect patient-reported data in an unselected series of patients with gynaecological malignancy. Aim 1 - To provide 3 year results for the longitudinal study. Aim 2 - To improve the questionnaire used to collect data by identifying redundant items and modifying for use to collect Common Terminology Criteria for Adverse Events (CTCAE) data. MATERIAL AND METHODS: Aim 1 - Patient reported outcome data were collected prospectively by 226 patients before and up to 3 years following radiotherapy for gynaecological cancer using a questionnaire developed to collect LENT subjective data. Aim 2 - A factor analysis was performed to identify which questions gave the most and least information. RESULTS: Aim 1 - Faecal urgency and incontinence (all grades) peaked at 79% and 24%, respectively at 1 year then settled to 69% and 18% at 3 years, respectively. Urinary urgency (all grades) increased with time and was described in 75% at 3 years. Other symptoms reported at 3 years include diarrhoea in 12%, urinary incontinence in 27% and vaginal dryness in 29%. A third of patients did not feel their sex life had changed following treatment, while a quarter felt that it had. Aim 2 - some questions overlapped and others were non-specific. The questionnaire has subsequently been altered. CONCLUSIONS: The extent of late toxicity is substantial. This detailed information is important for both patients and clinicians in terms of treatment decisions and follow-up care. The LENT questionnaire provides a feasible tool for capture of this information in the clinic.

Phase II study of conformal hypofractionated radiotherapy with concurrent gemcitabine in muscle-invasive bladder cancer.

PURPOSE: The aim of this prospective, phase II trial was to determine the response of muscle-invasive bladder cancer (MIBC) to concurrent chemoradiotherapy of weekly gemcitabine with 4 weeks of radiotherapy (RT; GemX). PATIENTS AND METHODS: Fifty patients with transitional cell carcinoma, stage T2-3, N0, M0 after transurethral resection and magnetic resonance imaging, were recruited. Gemcitabine was given intravenously at 100 mg/m(2) on days 1, 8, 15, and 22 of a 28-day RT schedule that delivered 52.5 Gy in 20 fractions. Chemotherapy was stopped for Radiation Therapy Oncology Group (RTOG) grade 3 bladder or bowel toxicity. The primary end points were tumor response, toxicity, and survival. RESULTS: All patients completed RT; 46 tolerated all four cycles of gemcitabine. Two patients stopped after two cycles, and two stopped after three cycles, because of bowel toxicity. Forty-seven patients had a post-treatment cystoscopy; 44 (88%) achieved a complete endoscopic response. At a median follow-up of 36 months (range, 15 to 62 months), 36 patients were alive, and 32 of these had a functional and intact bladder. Fourteen patients died; seven died as a result of metastatic MIBC, five died as a result of intercurrent disease, and two died as a result of treatment-associated deaths. Four patients underwent cystectomy; three because of recurrent disease and one because of toxicity. One patient required a bowel resection for late toxicity. By using Kaplan-Meier analyses, 3-year cancer-specific survival was 82%, and overall survival was 75%. CONCLUSION: Concurrent gemcitabine-based chemoradiotherapy (ie, GemX) produces a high response rate in MIBC and has durable local control and acceptable toxicity, which allows patients to preserve their own bladder. This treatment modality warrants additional investigation in a phase III setting.

Efficacy of data capture for patient-reported toxicity following radiotherapy for prostate or cervical cancer.

We investigated the efficacy of data capture of patient-reported toxicity following radiotherapy by comparing electronic and paper formats. Patient-reported toxicity questionnaires based on items from the NCI Common Terminology Criteria for Adverse Events (CTCAE) were created for patients receiving radiotherapy. Electronic and paper questionnaires had identical questions. Thirty seven gynaecological cancer and 40 prostate cancer patients completed questionnaires. Both questionnaire formats (electronic and paper) were completed by each patient at time points before and after radiotherapy. The average questionnaire and subsection scores for each format were compared directly and by using intra-class correlation (ICC) coefficients. The internal consistency/reliability was assessed by determining Cronbach's alpha coefficient. Patient preference for questionnaire format including clarity and ease-of-use was recorded. 324 questionnaires were collected as part of this study. A similar pattern of average subsection scores was found for the electronic and paper questionnaires. ICC coefficients for the mean overall questionnaire scores and subsection scores were high (>0.8). Cronbach's alpha was generally greater than 0.6, indicating that the reliability was high. Of the patients that responded, 27.3% preferred the electronic format, 25.7% preferred the paper format and 47% had no preference. The average time taken to complete a questionnaire was about 9 minutes for each format. The different questionnaire formats measured toxicity effects consistently and were reliable for both gynaecological cancer and prostate cancer patients. The survey indicated that patients found the questionnaires clear, easy to understand and straightforward to complete. Electronic data capture of patient-reported toxicity for CTCAE is feasible and acceptable.

Assessing the effect of a contouring protocol on postprostatectomy radiotherapy clinical target volumes and interphysician variation.

PURPOSE: To compare postprostatectomy clinical target volume (CTV) delineation before and after the introduction of a contouring protocol and to investigate its effect on interphysician variability METHODS AND MATERIALS: Six site-specialized radiation oncologists independently delineated a CTV on the computed tomography (CT) scans of 3 patients who had received postprostatectomy radiotherapy. At least 3 weeks later this was repeated, but with the physicians adhering to the contouring protocol from the Medical Research Council's Radiotherapy and Androgen Deprivation In Combination After Local Surgery (RADICALS) trial. The volumes obtained before and after the protocol were compared and the effect of the protocol on interphysician variability assessed. RESULTS: An increase in mean CTV for all patients of 40.7 to 53.9 cm(3) was noted as a result of observing the protocol, with individual increases in the mean CTV of 65%, 15%, and 24% for Patients 1, 2, and 3 respectively. A reduction in interphysician variability was noted when the protocol was used. CONCLUSIONS: Substantial interphysician variation in target volume delineation for postprostatectomy radiotherapy exists, which can be reduced by the use of a contouring protocol. The RADICALS contouring protocol increases the target volumes when compared with those volumes typically applied at our center. The effect of treating larger volumes on the therapeutic ratio and resultant toxicity should be carefully monitored, particularly if the same dose-response as documented in radical prostate radiotherapy applies to the adjuvant and salvage setting. Prostate cancer, Postprostatectomy, Radiotherapy, Target volume.

Hypofractionated intensity-modulated radiotherapy for carcinoma of the prostate: analysis of toxicity.

PURPOSE: Dose escalation for prostate cancer improves biological control but with a significant increase in late toxicity. Recent estimates of low alpha/beta ratio for prostate cancer suggest that hypofractionation may result in biological advantage. Intensity-modulated radiotherapy (IMRT) should enable dose escalation to the prostate while reducing toxicity to local organs. We report late toxicity data of a hypofractionated IMRT regime. METHODS AND MATERIALS: Eligible men had T2-3N0M0 adenocarcinoma prostate, and either Gleason score >or= 7 or prostate-specific antigen 20-50 ng/L. Patients received 57-60 Gy to prostate in 19-20 fractions using five-field IMRT. All received hormonal therapy for 3 months before radiotherapy to a maximum of 6 months. Toxicity was assessed 2 years postradiotherapy using the RTOG criteria, LENT/SOMA, and UCLA prostate index assessment tools. RESULTS: Acute toxicity was favorable with no RTOG Grade 3 or 4 toxicity. At 2 years, there was 4% Grade 2 bowel and 4.25% Grade 2 bladder toxicity. There was no Grade 3 or 4 bowel toxicity; one patient developed Grade 3 bladder toxicity. UCLA data showed a slight improvement in urinary function at 2 years compared with pretreatment. LENT/SOMA assessments demonstrated general worsening of bowel function at 2 years. Patients receiving 60 Gy were more likely to develop problems with bowel function than those receiving 57 Gy. CONCLUSIONS: These data demonstrate that hypofractionated radiotherapy using IMRT for prostate cancer is well tolerated with minimal late toxicity at 2 years posttreatment. Ongoing studies are looking at the efficacy of hypofractionated regimes with respect to biological control.

External beam boost for cancer of the cervix uteri when intracavitary therapy cannot be performed.

PURPOSE: To assess the outcome of patients treated with radical radiotherapy for cervical cancer who received an external beam boost, in place of intracavitary brachytherapy (ICT), after irradiation to the whole pelvis. METHODS AND MATERIALS: Case notes were reviewed for all patients treated in this way in a single center between 1996 and 2004. Patient and tumor details, the reasons why ICT was not possible, and treatment outcome were documented. RESULTS: Forty-four patients were identified. The mean age was 56.4 years (range, 26-88 years). Clinical International Federation of Gynecology and Obstetrics or radiologic stage for Stages I, II, III, and IV, respectively, was 16%, 48%, 27%, and 7%. A total radiation dose of 54-70 Gy was given (75% received > or =60 Gy). Reasons for ICT not being performed were technical limitations in 73%, comorbidity or isolation limitations in 23%, and patient choice in 4%. The median follow-up was 2.3 years. Recurrent disease was seen in 48%, with a median time to recurrence of 2.3 years. Central recurrence was seen in 16 of the 21 patients with recurrent disease. The 5-year overall survival rate was 49.3%. The 3-year cancer-specific survival rate by stage was 100%, 70%, and 42% for Stages I, II, and III, respectively. Late Grades 1 and 2 bowel, bladder, and vaginal toxicity were seen in 41%. Late Grade 3 toxicity was seen in 2%. CONCLUSION: An external beam boost is a reasonable option after external beam radiotherapy to the pelvis when it is not possible to perform ICT.

X-ray volume imaging in bladder radiotherapy verification.

PURPOSE: To assess the clinical utility of X-ray volume imaging (XVI) for verification of bladder radiotherapy and to quantify geometric error in bladder radiotherapy delivery. METHODS AND MATERIALS: Twenty subjects undergoing conformal bladder radiotherapy were recruited. X-ray volume images and electronic portal images (EPIs) were acquired for the first 5 fractions and then once weekly. X-ray volume images were co-registered with the planning computed tomography scan and clinical target volume coverage assessed in three dimensions (3D). Interfraction bladder volume change was described by quantifying changes in bladder volume with time. Bony setup errors were compared from both XVI and EPI. RESULTS: The bladder boundary was clearly visible on coronal XVI views in nearly all images, allowing accurate 3D treatment verification. In 93.5% of imaged fractions, the clinical target volume was within the planning target volume. Most subjects displayed consistent bladder volumes, but 25% displayed changes that could be predicted from the first three XVIs. Bony setup errors were similar whether calculated from XVI or EPI. CONCLUSIONS: Coronal XVI can be used to verify 3D bladder radiotherapy delivery. Image-guided interventions to reduce geographic miss and normal tissue toxicity are feasible with this technology.

Do differences in target volume definition in prostate cancer lead to clinically relevant differences in normal tissue toxicity?

PURPOSE: Many studies have described the quantitated differences between clinicians in target volume definition in prostate cancer. However, few studies have looked at the clinical effects of this. We aimed to assess the relevance and sequelae of such differences. METHODS AND MATERIALS: Five experienced radiation oncologists were given the clinical details of 5 patients with early-stage prostate cancer and asked to define the clinical target volume, consisting of the prostate and seminal vesicles (CTV1) and the prostate alone (CTV2), on specified planning CT scans of the pelvis. Planning target volumes (PTV1) were generated by automatic expansion of the CTV1 by a 1-cm margin. The PTV2 was defined as the CTV2. The rectum and bladder were defined by a single experienced clinician for each plan without knowledge of the involved clinician marking the CTVs. The Pinnacle planning system was used to generate the plans, using four-field conformal radiotherapy, to deliver 64 Gy in 32 fractions to the PTV1 followed by a boost of 10 Gy to the PTV2 (Medical Research Council RT01 trial protocol). Dose-volume histograms were generated for the whole bladder and rectum for each plan and the volume receiving a specific percentage of the dose (e.g., V(90)) calculated for 74 Gy, followed by estimates of normal tissue complication probabilities (NTCPs) for the bladder and rectum. RESULTS: Statistically significant differences were found in the CTV1 and CTV2 and, consequently, the PTV1 among the 5 clinicians (p < 0.0005). Most of the discrepancies occurred at the delineation of the prostatic apex and seminal vesicles, with the smallest variance noted at the rectum-prostate and bladder-prostate interfaces. No statistically significant differences were found among clinicians for the rectal V(90), V(85), V(80), V(70), or V(50) or for the bladder V(85), V(80), V(70), or V(50). A difference was noted among consultants for the bladder V(90) (p = 0.015), although no correlation was found between the bladder V(90) and the size of the outlined volumes. No statistically significant differences were found between the estimates of bladder (p = 0.1) and rectal (p = 0.09) NTCPs. CONCLUSION: The statistically significant difference in outlined volumes of the CTV1, CTV2, and PTV1 among the 5 clinicians is in keeping with the findings of previous studies. However, the interclinician variability did not result in clinically relevant outcomes with respect to the irradiated volume of rectum and bladder or NTCP. This may have been because the outlined areas in which interclinician differences were smallest (the rectal-prostate and prostate-bladder interfaces) are the areas that have the greatest effect on normal tissue toxicity. For the areas in which the interclinician correlation was lowest (the prostatic apex and distal seminal vesicles), the effects on normal tissue toxicity are smallest. The results of this study suggest that interclinician outlining differences in prostate cancer may have less clinical relevance than was previously thought.

Hypofractionated conformal radiotherapy in carcinoma of the prostate: five-year outcome analysis.

PURPOSE: Recent publications have indicated that the alpha/beta ratios for carcinoma of the prostate are much lower than had originally been thought, suggesting that prostate cancer may be highly sensitive to fraction size. We have reviewed our unique experience of the use of 3.13 Gy fractions in a large cohort of men treated homogeneously in a single institute. MATERIALS AND METHODS: The outcome for 705 men with T1-T4, N0, M0 prostate cancer who received conformal radiotherapy between 1995 and 1998 at this center was analyzed. No patient received hormonal manipulation. Mean age was 68 years (range: 49-84 years). Median pretreatment PSA was 13 ng/mL (range: 0.6-270 ng/mL). Disease characteristics were as follows: Stage T1, 125 (18%); T2, 365 (52%); T3/4, 215 (30%); Gleason 2-6, 463 (66%); Gleason 7-10, 242 (34%); pretreatment PSA < or =10 ng/mL, 291 (41%); 10 to < or =20, 228 (32%); >20, 186 (27%). Median follow-up was 48 months (range: 1-82 months). Biochemical-free survival (bNED) was defined by the American Society for Therapeutic Radiology and Oncology consensus definition. Radiotherapy was delivered to a planning target volume (prostate plus all/base of the seminal vesicles dependent on risk criteria with a 1-cm margin) with a 4-field conformal technique to a dose of 50 Gy in 16 daily fractions over 22 days. RESULTS: The 5-year bNED survival was significantly associated (p < 0.001) with pretreatment PSA, stage, and Gleason score. Five-year bNED rates with respect to pretreatment characteristics were as follows: 73% (PSA < or =10), 52% (>10-20), 35% (>20), 64% (Stage T1/2), 38% (T3/4), 61% (Gleason score 2-6), and 46% (Gleason > or =7). When patients were grouped into good (Stage T1/2, PSA < or =10 ng/mL, and Gleason score <7) (n = 181), intermediate (1 raised value) (n = 247), or poor (2 or more raised values) (n = 277) prognostic groups, the bNED was, respectively, 82%, 56%, and 39%. Radiation Therapy Oncology Group Grade > or =2 bowel toxicity was 5% and bladder 9%. CONCLUSIONS: These data indicate that the delivery of a relatively low total dose using a hypofractionated regime results in similar tumor control and normal-tissue toxicity to 65-70 Gy delivered in 1.8-2 Gy fractions. These data suggest that this is an acceptable regime for good-prognosis patients. However, because of the evidence for a dose effect at doses above 70 Gy with "conventional fractionation," we are now treating intermediate- and poor-risk patients within a hypofractionated dose escalation trial to 60 Gy in 20 fractions using intensity- modulated radiotherapy.

Short course para-aortic radiation for stage I seminoma of the testis.

PURPOSE: To determine the outcome in men with Stage I seminoma treated with low-dose para-aortic radiation. MATERIALS AND METHODS: Between January 1988 and December 2000, 431 men with Stage I seminoma were treated with para-aortic radiation to a midplane dose of 20 Gy in 8 fractions over 10 days. RESULTS: At a median follow-up of 62 months, 15 patients (3.5%) had relapsed, with a median time to relapse of 13 months (range: 9 to 39 months). Nine patients had pelvic nodal relapse; in addition, 1 patient had para-aortic involvement, and 2 had distant disease. Four had metastatic disease only (mediastinum 2, lung 2). One patient had scrotal recurrence, and 1 was treated for progressive rise in human chorionic gonadotrophin without identifiable disease. Initial treatment at relapse was chemotherapy (12), radiation (2), and surgery (1). One patient died from progressive disease. Thirteen men (3%) have developed second malignancies, including 7 contralateral testicular tumors, 5 solid malignancies, and 1 leukemia. The overall 5-year survival was 98%, and the estimated recurrence-free survival at 5 years was 96.3%. On log-rank univariate analysis, lymphovascular invasion, involvement of the tunica, and a preoperative human chorionic gonadotrophin level of greater than 5 were found to be of prognostic significance for recurrence. CONCLUSIONS: These data support short-duration, limited-field radiation as an optimal safe and effective protocol in the management of Stage I seminoma patients.

Prediction of the benefits from dose-escalated hypofractionated intensity-modulated radiotherapy for prostate cancer.

PURPOSE: To estimate the benefits of dose escalation in hypofractionated intensity-modulated radiotherapy (IMRT) for prostate cancer, using radiobiologic modeling and incorporating positional uncertainties of organs. MATERIALS AND METHODS: Biologically based mathematical models for describing the relationships between tumor control probability (TCP) and normal-tissue complication probability (NTCP) vs. dose were used to describe some of the results available in the literature. The values of the model parameters were then used together with the value of 1.5 Gy for the prostate cancer alpha/beta ratio to predict the responses in a hypofractionated 3 Gy/fraction IMRT trial at the Christie Hospital, taking into account patient movement characteristics between dose fractions. RESULTS: Compared with the current three-dimensional conformal radiotherapy technique (total dose of 50 Gy to the planning target volume in 16 fractions), the use of IMRT to escalate the dose to the prostate was predicted to increase the TCP by 5%, 16%, and 22% for the three dose levels, respectively, of 54, 57, and 60 Gy delivered using 3 Gy per fraction while keeping the late rectal complications (>/=Grade 2 RTOG scale) at about the same level of 5%. Further increases in TCP could be achieved by reducing the uncertainty in daily target position, especially for the last stage of the trial, where up to 6% further increase in TCP should be gained. CONCLUSIONS: Dose escalation to the prostate using IMRT to deliver daily doses of 3 Gy was predicted to significantly increase tumor control without increasing late rectal complications, and currently this prediction is being tested in a clinical trial.

Scoring of treatment-related late effects in prostate cancer.

BACKGROUND AND PURPOSE: To assess the correlation between different general and organ specific quality of life and morbidity scoring methods in a cohort of men treated with radical radiotherapy for prostate cancer. MATERIALS AND METHODS: Men who had been treated with radical radiotherapy (50 Gy in 16 fractions over 21 days) for localized prostate cancer more than 3 years previously and who had no evidence of recurrent disease were invited to take part in the study. A total of 101 of 135 invited patients agreed and completed LENT/SOMA, UCLA Prostate Cancer Index, and 36 item RAND Health survey questionnaires. RESULTS: The patients had comparable results with other published series with respect to the UCLA and SF-36 indices. There was significant correlation between the corresponding parts of the UCLA and LENT/SOMA scales (P<0.0005). However, for the same symptoms, a patient tended to score lower (worse) on the UCLA scale in comparison to LENT/SOMA. The relationship between the average LENT/SOMA score and maximum score was also not straightforward with each set of data revealing different information. CONCLUSIONS: The LENT/SOMA questions were, in the main, more wide-ranging and informative than the UCLA index. It is helpful to give both the overall and maximum LENT/SOMA scores to most efficiently use all of the data. There may need to be a further LENT/SOMA question to allow both symptoms of tenesmus and faecal urgency to be fully addressed.