Bayesian Network structure learning algorithm for highly missing and non imputable data: Application to breast cancer radiotherapy data

It is not uncommon for real-life data produced in healthcare to have a higher proportion of missing data than in other scopes. To take into account these missing data, imputation is not always desired by healthcare experts, and complete case analysis can lead to a significant loss of data. The algorithm proposed here, allows the learning of Bayesian Network graphs when both imputation and complete case analysis are not possible. The learning process is based on a set of local bootstrap learnings performed on complete sub-datasets which are then aggregated and locally optimized. This learning method presents competitive results compared to other structure learning algorithms, whatever the mechanism of missing data.

Acute skin toxicity of conventional fractionated versus hypofractionated radiotherapy in breast cancer patients receiving regional node irradiation: the real-life prospective multicenter HYPOBREAST cohort

BACKGROUND

Large-scale trials have shown that hypofractionated adjuvant breast radiotherapy was as effective in terms of survival and local control as conventional fractionated radiotherapy, and acute toxicity was reduced with hypofractionated radiotherapy. However, there is a lack of data about the toxicity of breast with regional nodal irradiation (RNI). The aim of this study was to assess the effect of fractionation on radiation-related acute skin toxicity in patients receiving RNI in addition to whole-breast or chest wall irradiation, using real-life data.

METHODS

We conducted a prospective, multicenter cohort study with systematic computerized data collection integrated into Mosaiq®. Three comprehensive cancer centers used a standardized form to prospectively collect patient characteristics, treatment characteristics and toxicity.

RESULTS

Between November 2016 and January 2022, 1727 patients were assessed; 1419 (82.2%) and 308 (17.8%) patients respectively received conventional fractionated and hypofractionated radiation therapy. Overall, the incidence of acute grade 2 or higher dermatitis was 28.4% (490 patients). Incidence was lower with hypofractionated than with conventional fractioned radiation therapy (odds ratio (OR) 0.34 [0.29;0.41]). Two prognostic factors were found to increase the risk of acute dermatitis, namely 3D (vs IMRT) and breast irradiation (vs chest wall).

CONCLUSION

Using real-life data from unselected patients with regional nodal irradiation, our findings confirm the decreased risk of dermatitis previously reported with hypofractionated radiation therapy in clinical trials. Expansion of systematic data collection systems to include additional centers as well as dosimetric data is warranted to further evaluate the short- and long-term effects of fractionation in real life.

Dose-Response Effect and Dose-Toxicity in Stereotactic Radiotherapy for Brain Metastases: A Review

Simple Summary

Brain metastases are one of the most frequent complications for cancer patients. Stereotactic radiosurgery is considered a cornerstone treatment for patients with limited brain metastases and the ideal dose and fractionation schedule still remain unknown. The aim of this literature review is to discuss the dose-effect relation in brain metastases treated by stereotactic radiosurgery, accounting for fractionation and technical considerations.

Abstract

For more than two decades, stereotactic radiosurgery has been considered a cornerstone treatment for patients with limited brain metastases. Historically, radiosurgery in a single fraction has been the standard of care but recent technical advances have also enabled the delivery of hypofractionated stereotactic radiotherapy for dedicated situations. Only few studies have investigated the efficacy and toxicity profile of different hypofractionated schedules but, to date, the ideal dose and fractionation schedule still remains unknown. Moreover, the linear-quadratic model is being debated regarding high dose per fraction. Recent studies shown the radiation schedule is a critical factor in the immunomodulatory responses. The aim of this literature review was to discuss the dose–effect relation in brain metastases treated by stereotactic radiosurgery accounting for fractionation and technical considerations. Efficacy and toxicity data were analyzed in the light of recent published data. Only retrospective and heterogeneous data were available. We attempted to present the relevant data with caution. A BED10 of 40 to 50 Gy seems associated with a 12-month local control rate >70%. A BED10 of 50 to 60 Gy seems to achieve a 12-month local control rate at least of 80% at 12 months. In the brain metastases radiosurgery series, for single-fraction schedule, a V12 Gy < 5 to 10 cc was associated to 7.1–22.5% radionecrosis rate. For three-fractions schedule, V18 Gy < 26–30 cc, V21 Gy < 21 cc and V23 Gy < 5–7 cc were associated with about 0–14% radionecrosis rate. For five-fractions schedule, V30 Gy < 10–30 cc, V 28.8 Gy < 3–7 cc and V25 Gy < 16 cc were associated with about 2–14% symptomatic radionecrosis rate. There are still no prospective trials comparing radiosurgery to fractionated stereotactic irradiation.

Single-fraction radiosurgery versus fractionated stereotactic radiotherapy in patients with brain metastases: a comparative study

To compare the local control and brain radionecrosis in patients with brain metastasis primarily treated by single-fraction radiosurgery (SRS) or hypofractionated stereotactic radiotherapy (HFSRT). Between January 2012 and December 2017, 179 patients with only 1-3 brain metastases (total: 287) primarily treated by SRS (14 Gy) or HFSRT (23.1 Gy in 3 fractions of 7.7 Gy, every other day) were retrospectively analyzed in a single center. Follow-up imaging data were available in 152 patients with 246 lesions. The corresponding Biological Effective Dose (BED) were 33.6 Gy and 40.9 Gy respectively for SRS and HFSRT group, assuming an α/β of 10 Gy. Local control (LC) and risk of radionecrosis (RN) were calculated by the Kaplan-Meier method. The actuarial local control rates at 6 and 12 months were 94% and 88.1% in SRS group, and 87.6% and 78.4%, in HFSRT group (p = 0.06), respectively. Only the total volume of edema was associated with worse LC (p = 0.01, HR 1.02, 95% CI [1.004-1.03]) in multivariate analysis. Brain radionecrosis occurred in 1 lesion in SRS group and 9 in HFSRT group. Median time to necrosis was 5.5 months (range 1-9). Only the volume of GTV was associated with RN (p = 0.02, HR 1.09, 95% CI [1.01-1.18]) in multivariate analysis. Multi-fraction SRT dose of 23.31 Gy in 3 fractions has similar efficacy to single-fraction SRT dose of 14 Gy in patients with brain metastases. A slightly higher occurrence of radionecrosis appeared in HFSRT group.

Brain Metastasis and Renal Cell Carcinoma: Prognostic Scores Assessment in the Era of Targeted Therapies

AIM

This study aimed at exploring several brain metastatic prognostic scores in patients with renal cell carcinoma.

PATIENTS AND METHODS

We retrospectively analyzed data of 93 metastatic renal cell carcinoma patients who were diagnosed with brain metastases between October 2005 and July 2016 who received targeted therapy. Potential prognostic factors (RTOG RPA, BS-BM, and a newly developed score CERENAL) were analyzed.

RESULTS

A total of 75 patients received targeted therapy. All scores showed prognostic value in progression-free survival after first-line treatment with CERENAL being the sole independent prognostic factor associated with improved duration of first-line treatment. Both RTOG RPA and CERENAL were potential prognosticators for overall survival, whereas only the CERENAL score was associated with prolonged disease-specific survival.

CONCLUSION

Several prognostic scores can be useful to predict survival of patients with brain metastases from renal cancer, especially the newly developed CERENAL score.

Practical contouring guidelines with an MR-based atlas of brainstem structures involved in radiation-induced nausea and vomiting

BACKGROUND AND PURPOSE

The objective of this project was to define consensus guidelines for delineating brainstem substructures (dorsal vagal complex, including the area postrema) involved in radiation-induced nausea and vomiting (RINV). The three parts of the brainstem are rarely delineated, so this study was also an opportunity to find a consensus on this subject.

MATERIALS AND METHODS

The dorsal vagal complex (DVC) was identified on autopsy sections and endoscopic descriptions. Anatomic landmarks and boundaries were used to establish radio-anatomic correlations on CT and Magnetic Resonance Imaging (MRI). Additionally, delineation of RINV structures was performed on MRI images and reported on CT scans. Next, guidelines were provided to eight radiation oncologists for delineation guidance of these RINV-related structures on DICOM-RT images of two patients being treated for a nasopharyngeal carcinoma. Interobserver variability was computed.

RESULTS

The DVC and the three parts of the brainstem were defined with a concise description of their main anatomic boundaries. The interobserver analysis showed that the DVC, the midbrain, the pons, and the medulla oblongata delineations were reproducible with KI = 0.72, 0.84, 0.94 and 0.89, respectively. The Supplemental Material section provides an atlas of the consensus guidelines projected on 1-mm MR axial slices.

CONCLUSIONS

This RINV-atlas was feasible and reproducible for the delineation of RINV structures on planning CT using fused MRI. It may be used to prospectively assess dose-volume relationship for RINV structures and occurrence of nausea vomiting during intracranial or head and neck irradiation.

Automatic detection and segmentation of brain metastases on multimodal MR images with a deep convolutional neural network

Stereotactic treatments are today the reference techniques for the irradiation of brain metastases in radiotherapy. The dose per fraction is very high, and delivered in small volumes (diameter <1 cm). As part of these treatments, effective detection and precise segmentation of lesions are imperative. Many methods based on deep-learning approaches have been developed for the automatic segmentation of gliomas, but very little for that of brain metastases. We adapted an existing 3D convolutional neural network (DeepMedic) to detect and segment brain metastases on MRI. At first, we sought to adapt the network parameters to brain metastases. We then explored the single or combined use of different MRI modalities, by evaluating network performance in terms of detection and segmentation. We also studied the interest of increasing the database with virtual patients or of using an additional database in which the active parts of the metastases are separated from the necrotic parts. Our results indicated that a deep network approach is promising for the detection and the segmentation of brain metastases on multimodal MRI.

[Risk of radionecrosis after hypofractionated stereotactic radiotherapy targeting the postoperative resection cavity of brain metastases]

PURPOSE

To investigate the factors that potentially lead to brain radionecrosis after hypofractionated stereotactic radiotherapy targeting the postoperative resection cavity of brain metastases.

METHODS AND MATERIALS

A retrospective analysis conducted in two French centres, was performed in patients treated with trifractionated stereotactic radiotherapy (3×7.7Gy prescribed to the 70% isodose line) for resected brain metastases. Patients with previous whole-brain irradiation were excluded of the analysis. Radionecrosis was diagnosed according to a combination of criteria including clinical, serial imaging or, in some cases, histology. Univariate and multivariate analyses were performed to determine the predictive factors of radionecrosis including clinical and dosimetric variables such as volume of brain receiving a specific dose (V_8Gy-V_22Gy).

RESULTS

One hundred eighty-one patients, with a total of 189 cavities were treated between March 2008 and February 2015. Thirty-five patients (18.5%) developed radionecrosis after a median follow-up of 15 months (range: 3-38 months) after hypofractionated stereotactic radiotherapy. One third of patients with radionecrosis were symptomatic. Multivariate analysis showed that infra-tentorial location was predictive of radionecrosis (hazard ratio [HR]: 2.97; 95% confidence interval [95% CI]: 1.47-6.01; P=0.0025). None V_8Gy-V_22Gy was associated with appearance of radionecrosis, even if V_14Gy trended toward significance (P=0.059).

CONCLUSION

Analysis of patients and treatment variables revealed that infratentorial location of brain metastases was predictive for radionecrosis after hypofractionated stereotactic radiotherapy for postoperative resection cavities.

[Big data and their perspectives in radiation therapy]

The concept of big data indicates a change of scale in the use of data and data aggregation into large databases through improved computer technology. One of the current challenges in the creation of big data in the context of radiation therapy is the transformation of routine care items into dark data, i.e. data not yet collected, and the fusion of databases collecting different types of information (dose-volume histograms and toxicity data for example). Processes and infrastructures devoted to big data collection should not impact negatively on the doctor-patient relationship, the general process of care or the quality of the data collected. The use of big data requires a collective effort of physicians, physicists, software manufacturers and health authorities to create, organize and exploit big data in radiotherapy and, beyond, oncology. Big data involve a new culture to build an appropriate infrastructure legally and ethically. Processes and issues are discussed in this article.

Outcomes in newly diagnosed elderly glioblastoma patients after concomitant temozolomide administration and hypofractionated radiotherapy

This study aimed to analyze the treatment and outcomes of older glioblastoma patients. Forty-four patients older than 70 years of age were referred to the Paul Strauss Center for chemotherapy and radiotherapy. The median age was 75.5 years old (range: 70–84), and the patients included 18 females and 26 males. The median Karnofsky index (KI) was 70%. The Charlson indices varied from 4 to 6. All of the patients underwent surgery. O₆-methylguanine–DNA methyltransferase (MGMT) methylation status was determined in 25 patients. All of the patients received radiation therapy. Thirty-eight patients adhered to a hypofractionated radiation therapy schedule and six patients to a normofractionated schedule. Neoadjuvant, concomitant and adjuvant chemotherapy regimens were administered to 12, 35 and 20 patients, respectively. At the time of this analysis, 41 patients had died. The median time to relapse was 6.7 months. Twenty-nine patients relapsed, and 10 patients received chemotherapy upon relapse. The median overall survival (OS) was 7.2 months and the one- and two-year OS rates were 32% and 12%, respectively. In a multivariate analysis, only the Karnofsky index was a prognostic factor. Hypofractionated radiotherapy and chemotherapy with temozolomide are feasible and acceptably tolerated in older patients. However, relevant prognostic factors are needed to optimize treatment proposals.

Institutional, retrospective analysis of 777 patients with brain metastases: treatment outcomes and diagnosis-specific prognostic factors

PURPOSE

To retrospectively evaluate the prognostic factors and survival of a series of 777 patients with brain metastases (BM) from a single institution.

METHODS AND MATERIALS

Patients were treated with surgery followed by whole-brain radiation therapy (WBRT) or with WBRT alone in 16.3% and 83.7% of the cases, respectively. The patients were RPA (recursive partitioning analysis) class I, II, and III in 11.2%, 69.6%, and 18.4% of the cases, respectively; RPA class II-a, II-b, and II-c in 8.3%, 24.8%, and 66.9% of the cases, respectively; and with GPA (graded prognostic assessment) scores of 0-1.0, 1.5-2.0, 2.5-3.0, and 3.5-4.0 in 35%, 27.5%, 18.2%, and 8.6% of the cases, respectively.

RESULTS

The median overall survival (OS) times according to RPA class I, II, and III were 20.1, 5.1, and 1.3 months, respectively (P<.0001); according to RPA class II-a, II-b, II-c: 9.1, 8.9, and 4.0 months, respectively (P<.0001); and according to GPA score 0-1.0, 1.5-2.0, 2.5-3.0, and 3.5-4.0: 2.5, 4.4, 9.0, and 19.1 months, respectively (P<.0001). By multivariate analysis, the favorable independent prognostic factors for survival were as follows: for gastrointestinal tumor, a high Karnofsky performance status (KPS) (P=.0003) and an absence of extracranial metastases (ECM) (P=.003); for kidney cancer, few BM (P=.002); for melanoma, few BM (P=.01), an absence of ECM (P=.002), and few ECM (P=.0002); for lung cancer, age (P=.007), a high KPS (P<.0001), an absence of ECM (P<.0001), few ECM and BM (P<.0001 and P=.0006, respectively), and control of the primary tumor (P=.004); and for breast cancer, age (P=.001), a high KPS (P=.007), control of the primary tumor (P=.05), and few ECM and BM (P=.01 and P=.0002, respectively). The triple-negative subtype was a significant unfavorable factor (P=.007).

CONCLUSION

Prognostic factors varied by pathology. Our analysis confirms the strength of prognostic factors used to determine the GPA score, including the genetic subtype for breast cancer.

Early evaluation predicts pain relief of irradiated bone metastases: a single-center prospective study

Background

Radiation therapy is a well-recognized, effective modality used for palliative care. Most studies completed to date have endpoints of one month or greater after treatment completion. This study analyzed the response rates at different time points during the first month after treatment.

Methods

From May 2010 to November 2011, 61 patients treated for 74 metastases were included in the study. The end points were defined as the completion of treatment (CT) and d8, d15 and d30 after the completion of treatment. The response rate was measured by the worst pain in the last 24 hours and the administered opioid dose. Patient assessment was performed during consultations and phone appointments.

Results

The overall response rate significantly improved from the CT (38%) to d8 (53.8%), d15 (53.8%) and d30 (57.1%) (respectively p < 0.001; p < 0.001 and p = 0.001). The improvement peaked at d8. Patients responding to the treatment at d8 had a significative longer pain relapse free survival (PRFS) compared to patients not responding (3.38 weeks vs 0.3 weeks; p < 0.001). From the beginning of treatment to the CT and at d8 , d15 and d30, oral morphine equivalent dose (OMED) did not significantly differ. However, the pain decrease did not result in a performance status improvement, which declined over time (p < 0.001).

Conclusion

Radiation therapy is an efficient treatment method for providing pain relief. This relief peaked at d8 after treatment, and the response at d8 is predictive of the response at 4 weeks. Pain management alone is not enough to improve performance status; further studies are needed to evaluate a more global supportive care approach.

Outcome Improvement in RPA I or II Patients With 1 or 2 Brain Metastases by Combined Surgery and Radiotherapy

Background

To evaluate the role of surgery and postoperative radiotherapy in the management of brain metastases (BM): a retrospective analysis for overall survival (OS), local and brain control (LC and BC) of a series of 329 patients with recursive partitioning analysis (RPA) I or II with 1 or 2 BM in a single institution.

Methods

Patients were treated either with combined surgical resection and whole brain radiation therapy (WBRT) in 104 cases (31.6%) or with WBRT alone in 225 cases (68.4%). Ninety-five patients (91.4%) who underwent surgery and WBRT and 147 (65.3%) who underwent WBRT alone benefited from a radiation boost to the metastatic site.

Results

The median OS was higher for patients RPA I compared to RPA II: 21.3 and 5.9 months (P < 0.0001), as well as for the surgical group compared to the radiation group: 20.2 vs 5.3 months (P < 0.0001), respectively. After the multivariate analysis, the improved OS was significantly associated with control of primary tumor (P = 0.0002) after surgical resection and with type of primary tumor (P = 0.002), absence of extracranial metastases (ECM) (P = 0.006), and high Karnofsky performance status (90 - 100 vs 70 - 80) (P = 0.003) after radiotherapy alone. The 12-, 24- and 36-months LC rates were 91.1%, 91.1% and 83.9%, respectively, after surgical resection and 81.2%, 63.1% and 57.3%, respectively, after radiotherapy alone (P = 0.005). In a univariate analysis, improved LC for the surgical group was also associated with the absence of ECM (P = 0.01) and for the radiation group, with a radiation boost (P = 0.01). The BC rates at 12, 24 and 36 months were 73.2%, 66.9% and 56%, respectively, in the surgical group and 75.7%, 49.6% and 42.4%, respectively, in the radiation group (P = 0.2). In our univariate analysis, improved BC after surgical resection was associated with control of primary tumor (P = 0.02). For patients in the radiation group, gender (P = 0.03) and a radiation boost (P = 0.0003) were significant prognostic factors in a univariate analysis. In our multivariate analysis, only the radiation boost was significant (P = 0.001).

Conclusions

Surgical resection followed by WBRT leads to a better outcome compared to WBRT alone for RPA I or II patients with 1 or 2 BM.

An institutional retrospective analysis of 93 patients with brain metastases from breast cancer: treatment outcomes, diagnosis-specific prognostic factors

To evaluate the prognostic factors and indexes of a series of 93 patients with breast cancer and brain metastases (BM) in a single institution. Treatment outcomes were evaluated according to the major prognostic indexes (RPA, BSBM, GPA scores) and breast cancer subtypes. Independent prognostic factors for overall survival (OS) were identified. The median OS values according to GPA 0-1, 1.5-2, 2.5-3 and 3.5-4, were 4.5, 9.5, 14.2 and 19.1 months, respectively (p < 0.0001) and according to genetic subtypes, they were 5, 14.2, 16.5 and 17.1 months for basal-like, luminal A and B and HER, respectively (p = 0.04). Using multivariate analysis, we established a new grading system using the six factors that were identified as indicators of longer survival: age under 60 (p = 0.001), high KPS (p = 0.007), primary tumor control (p = 0.05), low number of extracranial metastases and BM (p = 0.01 and 0.0002, respectively) and triple negative subtype (p = 0.002). Three groups with significantly different median survival times were identified: 4.1, 9.5 and 26.3 months, respectively (p < 0.0001). Our new grading system shows that prognostic indexes could be improved by using more levels of classification and confirms the strength of biological prognostic factors.

Waiting times before initiation of radiotherapy might not affect outcomes for patients with glioblastoma: a French retrospective analysis of patients treated in the era of concomitant temozolomide and radiotherapy

Purpose of this study was to determine the effect of waiting time for radiotherapy on overall survival of patients with glioblastoma treated in the EORTC-NCIC trial at 18 centers in France. A total of 400 adult patients with glioblastoma who were treated between January 1, 2006 and December 31, 2006 were included. There were 282 patients with "minimum criteria" according to the EORTC-NCIC trial: (i) concurrent chemotherapy with temozolomide; and (ii) age between 18 and 70 years old. Among these patients, 229 were treated with adjuvant temozolomide and were classified as "maximal criteria". One-hundred and eighteen patients were in the "without minimal criteria" group. Waiting time from the first symptom (FS-RT), pathology diagnosis (P-RT), multidisciplinary meeting (MM-RT), surgery (S-RT), and CT scan for delineation (CT-RT) until the start of radiotherapy were recorded. Median follow-up for all patients was 327 days. Overall, median FS-RT, P-RT, MM-RT, CT-RT, and S-RT times were 77, 36, 32, 12, and 41 days, respectively. Median, and 12 and 24-month overall survival were 409 days, and 56.3 ± 2.1 % and 27.6 ± 2.6 %, respectively. Univariate analysis failed to reveal a difference in survival, irrespective of the delay. In multivariate analysis, independent favorable prognostic factors for overall survival were age (p ≤ 0.0001) and type of surgery (p = 0.0006). In this large series treated during the EORTC-NCIC protocol period, waiting time until radiotherapy did not seem to affect patient outcome.