Recursive Partitioning Analysis Model Research Articles

Development and Internal Validation of an RPA-Based Model Predictive of Pain Flare Incidence After Spine SBRT

Radiotherapy is a standard palliative treatment for spine metastasis, but pain flares (PF) are a common acute toxicity. Prophylactic corticosteroids can reduce the rate of PF in patients receiving traditional palliative regimens, but less is known about prophylaxis for stereotactic body radiotherapy (SBRT). We aim to identify a subset of patients treated with spine SBRT with the highest rate of PF to optimize the use of prophylactic steroids.From December 2007-August 2019, 469 patients received 680 spine SBRT treatments. After exclusion of benign histology, particle therapy, and patients with missing data, the final cohort included 424 patients with 610 treatments. We defined PF as acute worsening of pain at the treatment site requiring initiation of or higher-dose corticosteroids, opiates, and/or hospitalization. First, data was split into 70% training and 30% validation sets with comparable patient characteristics using a random number generator. Feature importance testing was conducted, and a correlation heatmap helped exclude correlated variables. Feature extraction involved selecting the variables used in the highest-fidelity recursive partitioning analysis (RPA) models. Each RPA model was trained, validated, and tested using only the training dataset. Selected variables were verified using the Fisher exact test and univariate t-test.We identified 125 total PF (20%), treated with steroids (14%), opioids (40%), both (39%), or hospitalization (6%). Six variables were identified by RPA. Of these, 5 met significance on Fisher exact test with P < 0.02: renal primary (OR 2.23), soft tissue involvement (OR 1.96), Bilsky > 0 (OR 2.17), SINS > 6 (OR 2.98), and gross tumor volume > 8 cc (OR 2.80). One point was assigned for each variable. The low-risk (LR) group (score = 0, n = 159) had PF rates of 7.0% and 13.6% in the training and validation sets, respectively; the intermediate-risk (IR) group (score = 1, n = 150) had rates of 14.0% and 16.3%; and the high-risk (HR) group (score > 1, n = 301) had rates of 28.8% and 31.3%. A logistic model confirmed the increased pain flare rate in the HR group compared to LR and IR treatments combined (OR = 3.50, 95% CI: 2.06-5.92) or considered separately (OR = 5.41, 2.48-11.78). The concordance index was 0.66 (0.60-0.73) and 0.62 (0.52-0.71) in the training and internal validation sets. Notably, patients with PF after prior spine SBRT had a 61% rate of PF with subsequent spine SBRT, compared to only 5% for patients without PF after prior spine SBRT.Our internally-validated model identifies a high-risk group of patients more likely to develop PF after spine SBRT, for whom prophylactic steroids may confer the greatest benefit. Evaluation in a clinical trial is warranted.

RTOG‐0129 risk groups are reproducible in a prospective multicenter heterogeneously treated cohort

Recursive partitioning analysis (RPA) from the Radiation Therapy Oncology Group (RTOG)-0129 has identified a low-risk group of patients with oropharynx cancer (OPC) who might benefit from therapeutic de-intensification. These risk groups have not yet been reproduced in an independent cohort treated heterogeneously. Therefore, the objective of this analysis was to validate the RPA risk groups and examine the prognostic impact of novel factors. Patients with OPC were enrolled in a prospective study at 3 academic medical centers from 2013 to 2018. Medical record abstraction was used to ascertain clinical variables including staging and survival according to the 7th edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual. Human papillomavirus-positive tumor status was determined by p16 immunohistochemistry and/or HPV RNA in situ hybridization. Kaplan-Meier and log-rank methods were used to compare survival. Cox proportional hazards were used to generate univariate and multivariable hazard ratios (HRs). Median follow-up time was 3.2 years. The low-, intermediate-, and high-risk groups had significant differences in 2-year overall survival (OS, 99.1%; 95% CI, 94.4%-99.9% vs OS, 93.0%; 95% CI, 74.7%-98.2% vs OS, 80.0%; 95% CI, 40.9%-94.6%; Poverall = .0001) and 2-year progression-free survival (PFS, 97.5%; 95% CI, 92.4%-99.2% vs PFS, 89.3%; 95% CI, 70.3%-96.4% vs PFS, 80.0%; 95% CI, 40.9%-94.6%; Poverall < .002). After adjustment for age, sex, and level of educational attainment, OS and PFS were significantly lower for the intermediate- (OS adjusted hazard ratio [aHR], 5.0; 95% CI, 1.0-23.0; PFS aHR, 3.4; 95% CI, 1.0-11.5), and high- (OS aHR, 7.3; 95% CI, 1.4-39; PFS aHR, 5.0; 95% CI, 1.2-21.6) risk groups compared with the low-risk group. Lower education was also independently significantly associated with worse OS (aHR, 8.9; 95% CI, 1.8-44.3) and PFS (aHR, 3.1; 95% CI, 1.0-9.6). In patients with OPC, the RTOG-0129 RPA model is associated with OS and PFS in a heterogeneously treated cohort.

Predicting pathological complete response (pCR) after stereotactic ablative radiation therapy (SABR) of lung cancer using quantitative dynamic [18F]FDG PET and CT perfusion: a prospective exploratory clinical study

BackgroundStereotactic ablative radiation therapy (SABR) is effective in treating inoperable stage I non-small cell lung cancer (NSCLC), but imaging assessment of response after SABR is difficult. This prospective study aimed to develop a predictive model for true pathologic complete response (pCR) to SABR using imaging-based biomarkers from dynamic [18F]FDG-PET and CT Perfusion (CTP).MethodsTwenty-six patients with early-stage NSCLC treated with SABR followed by surgical resection were included, as a pre-specified secondary analysis of a larger study. Dynamic [18F]FDG-PET and CTP were performed pre-SABR and 8-week post. Dynamic [18F]FDG-PET provided maximum and mean standardized uptake value (SUV) and kinetic parameters estimated using a previously developed flow-modified two-tissue compartment model while CTP measured blood flow, blood volume and vessel permeability surface product. Recursive partitioning analysis (RPA) was used to establish a predictive model with the measured PET and CTP imaging biomarkers for predicting pCR. The model was compared to current RECIST (Response Evaluation Criteria in Solid Tumours version 1.1) and PERCIST (PET Response Criteria in Solid Tumours version 1.0) criteria.ResultsRPA identified three response groups based on tumour blood volume before SABR (BVpre-SABR) and change in SUVmax (ΔSUVmax), the thresholds being BVpre-SABR = 9.3 mL/100 g and ΔSUVmax = − 48.9%. The highest true pCR rate of 92% was observed in the group with BVpre-SABR < 9.3 mL/100 g and ΔSUVmax < − 48.9% after SABR while the worst was observed in the group with BVpre-SABR ≥ 9.3 mL/100 g (0%). RPA model achieved excellent pCR prediction (Concordance: 0.92; P = 0.03). RECIST and PERCIST showed poor pCR prediction (Concordance: 0.54 and 0.58, respectively).ConclusionsIn this study, we developed a predictive model based on dynamic [18F]FDG-PET and CT Perfusion imaging that was significantly better than RECIST and PERCIST criteria to predict pCR of NSCLC to SABR. The model used BVpre-SABR and ΔSUVmax which correlates to tumour microvessel density and cell proliferation, respectively and warrants validation with larger sample size studies.Trial registrationMISSILE-NSCLC, NCT02136355 (ClinicalTrials.gov). Registered May 8, 2014, https://clinicaltrials.gov/ct2/show/NCT02136355

Development of a risk classification system combining TN-categories and circulating EBV DNA for non-metastatic NPC in 10,149 endemic cases.

Background:The objective of this study was to construct a risk classification system integrating cell-free Epstein-Barr virus (cfEBV) DNA with T- and N- categories for better prognostication in nasopharyngeal carcinoma (NPC).Methods:Clinical records of 10,149 biopsy-proven, non-metastatic NPC were identified from two cancer centers; this comprised a training (N = 9,259) and two validation cohorts (N = 890; including one randomized controlled phase 3 trial cohort). Adjusted hazard ratio (AHR) method using a two-tiered stratification by cfEBV DNA and TN-categories was applied to generate the risk model. Primary clinical endpoint was overall survival (OS). Performances of the models were compared against American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) 8th edition TNM-stage classification and two published recursive partitioning analysis (RPA) models, and were validated in the validation cohorts.Results:We chose a cfEBV DNA cutoff of ⩾2,000 copies for optimal risk discretization of OS, disease-free survival (DFS) and distant metastasis-free survival (DMFS) in the training cohort. AHR modeling method divided NPC into six risk groups with significantly disparate survival (p < 0.001 for all): AHR1, T1N0; AHR2A, T1N1/T2-3N0 cfEBV DNA < 2,000 (EBVlow); AHR2B, T1N1/T2-3N0 cfEBV DNA ⩾ 2,000 (EBVhigh) and T1-2N2/T2-3N1 EBVlow; AHR3, T1-2N2/T2-3N1 EBVhigh and T3N2/T4N0 EBVlow; AHR4, T3N2/T4 N0-1 EBVhigh and T1-3N3/T4N1-3 EBVlow; AHR5, T1-3N3/T4 N2-3 EBVhigh. Our AHR model outperformed the published RPA models and TNM stage with better hazard consistency (1.35 versus 3.98–12.67), hazard discrimination (5.29 versus 6.69–13.35), explained variation (0.248 versus 0.164–0.225), balance (0.385 versus 0.438–0.749) and C-index (0.707 versus 0.662–0.700). In addition, our AHR model was superior to the TNM stage for risk stratification of OS in two validation cohorts (p < 0.001 for both).Conclusion:Herein, we developed and validated a risk classification system that combines the AJCC/UICC 8th edition TN-stage classification and cfEBV DNA for non-metastatic NPC. Our new clinicomolecular model provides improved OS prediction over the current staging system.

Deep learning-based survival analysis for brain metastasis patients with the national cancer database.

PurposePrognostic indices such as the Brain Metastasis Graded Prognostic Assessment have been used in clinical settings to aid physicians and patients in determining an appropriate treatment regimen. These indices are derivative of traditional survival analysis techniques such as Cox proportional hazards (CPH) and recursive partitioning analysis (RPA). Previous studies have shown that by evaluating CPH risk with a nonlinear deep neural network, DeepSurv, patient survival can be modeled more accurately. In this work, we apply DeepSurv to a test case: breast cancer patients with brain metastases who have received stereotactic radiosurgery.MethodsSurvival times, censorship status, and 27 covariates including age, staging information, and hormone receptor status were provided for 1673 patients by the NCDB. Monte Carlo cross‐validation with 50 samples of 1400 patients was used to train and validate the DeepSurv, CPH, and RPA models independently. DeepSurv was implemented with L2 regularization, batch normalization, dropout, Nesterov momentum, and learning rate decay. RPA was implemented as a random survival forest (RSF). Concordance indices of test sets of 140 patients were used for each sample to assess the generalizable predictive capacity of each model.ResultsFollowing hyperparameter tuning, DeepSurv was trained at 32 min per sample on a 1.33 GHz quad‐core CPU. Test set concordance indices of 0.7488 ± 0.0049, 0.6251 ± 0.0047, and 0.7368 ± 0.0047, were found for DeepSurv, CPH, and RSF, respectively. A Tukey HSD test demonstrates a statistically significant difference between the mean concordance indices of the three models.ConclusionOur results suggest that deep learning‐based survival prediction can outperform traditional models, specifically in a case where an accurate prognosis is highly clinically relevant. We recommend that where appropriate data are available, deep learning‐based prognostic indicators should be used to supplement classical statistics.

Intensity-Modulated Radiotherapy-Based Reirradiation for Head and Neck Cancer: A Multi-institutional Study by Korean Radiation Oncology Group (KROG 1707).

PurposeThe benefits of reirradiation for head and neck cancer (HNC) have not been determined. This study evaluated the efficacy of reirradiation using intensity-modulated radiotherapy (IMRT) for recurrent or second primary HNC (RSPHNC) and identified subgroups for whom reirradiation for RSPHNC is beneficial.Materials and MethodsA total of 118 patients from seven Korean institutions with RSPHNC who underwent IMRT-based reirradiation between 2006 and 2015 were evaluated through retrospective review of medical records. We assessed overall survival (OS) and local control (LC) within the radiotherapy (RT) field following IMRT-based reirradiation. Additionally, the OS curve according to the recursive partitioning analysis (RPA) suggested by the Multi-Institution Reirradiation (MIRI) Collaborative was determined.ResultsAt a median follow-up period of 18.5 months, OS at 2 years was 43.1%. In multivariate analysis, primary subsite, recurrent tumor size, interval between RT courses, and salvage surgery were associated with OS. With regard to the MIRI RPA model, the class I subgroup had a significantly higher OS than class II or III subgroups. LC at 2 years was 53.5%. Multivariate analyses revealed that both intervals between RT courses and salvage surgery were prognostic factors affecting LC. Grade 3 or more toxicity and grade 5 toxicity rates were 8.5% and 0.8%, respectively.ConclusionIMRT-based reirradiation was an effective therapeutic option for patients with RSPHNC, especially those with resectable tumors and a long interval between RT courses. Further, our patients' population validated the MIRI RPA classification by showing the difference of OS according to MIRI RPA class.

Integrating postradiotherapy plasma Epstein–Barr virus DNA and TNM stage for risk stratification of nasopharyngeal carcinoma to adjuvant therapy

After curative radiotherapy (RT) or chemoradiation (CRT), there is no validated tool to accurately identify patients for adjuvant therapy in nasopharyngeal carcinoma (NPC). Post-RT circulating plasma Epstein-Barr virus (EBV) DNA can detect minimal residual disease and is associated with recurrence and survival independent of TNM (tumor-lymph node-metastasis) stage. We aimed to develop and validate a risk model for stratification of NPC patients after completion of RT/CRT to observation or adjuvant therapy. The prospective multicenter 0502 EBV DNA screening cohort (Hong Kong NPC Study Group 0502 trial) enrolled from 2006 to 2015 (n= 745) was used for model development. For internal validation, we pooled independent patient cohorts from prospective clinical studies enrolled from 1997 to 2006 (n= 340). For external validation, we used retrospective cohort of NPC patients treated at Sun Yat-sen University Cancer Center from 2009 to 2012 (n= 837). Eligible patients had histologically confirmed NPC of Union for International Cancer Control (UICC) 7th Edition stage II-IVB who completed curative RT/CRT with or without neoadjuvant chemotherapy, had post-RT EBV DNA tested within 120 days after RT and received no adjuvant therapy. The primary end point was overall survival (OS). We used recursive-partitioning analysis (RPA) to classify patients into groups of low, intermediate, and high risk of death. Combining post-RT EBV DNA level (0, 1-49, 50-499, and ≥500 copies/ml) and TNM stage (II, III, IVAB), RPA model classified patients into low-, intermediate-, and high-risk groups with 5-year OS of 89.4%, 78.5% and 37.2%, respectively. The RPA low-risk group had comparable OS to TNM stage II (5-year OS 88.5%) but identified more patients (64.8% versus stage II 28.1%) that could potentially be spared adjuvant therapy toxicity. The RPA model (c-index 0.712) showed better risk discrimination than either the TNM stage (0.604) or post-RT EBV DNA alone (0.675) with improved calibration and consistence. These results were validated in both internal and external cohorts. Combining post-RT EBV DNA and TNM stage improved risk stratification in NPC.

Optimal adjuvant radiotherapy dose for stage I, II or III Merkel cell carcinoma: an analysis of the National Cancer Database.

We performed an analysis of the National Cancer Database to determine optimal doses of conventionally-fractionated adjuvant radiotherapy for patients with stage I/II or III Merkel cell carcinoma. The cohort included 2735 patients with resected Merkel cell carcinoma of the head and neck, trunk or extremities receiving radiotherapy. Exclusion criteria included doses of radiotherapy <30 or>80Gy, or dose per fraction >200 or<180cGy. Recursive partitioning analysis and spline models were used to select dose thresholds. Multivariable Cox regression was performed to validate thresholds with respect to overall survival. Recursive partitioning analysis models defined a threshold of 57Gy for stage I/II Merkel cell carcinoma, above which 3-year overall survival rate was decreased (P<0.0001). The 3-year overall survival rate for patients receiving 50.0-57.0Gy (81.2%) was greater compared to doses of 30.0-49.9Gy (75.3%) or>57.0Gy (70%, P <0.0001). Doses > 57.0Gy were associated with an increased hazard of death (1.31, confidence interval 1.07-1.60) with respect to doses of 50.0-57.0Gy. Doses < 50.0Gy for stage III Merkel cell carcinoma were associated with worsened 3-year overall survival (P<0.0001) and increased hazard of death (2.01, confidence interval 1.43-2.82) with respect to doses between 50.0 and 57.0Gy. Our results support doses of 50-57Gy for most patients with stage I/II Merkel cell carcinoma receiving conventionally-fractionated adjuvant radiotherapy. In contrast to a prior National Cancer Database analysis, our results suggest doses ≥ 50Gy should be strongly considered for patients with stage III Merkel cell carcinoma regardless of anatomic subsite.

The impact of timing of adjuvant therapy on survival for patients with glioblastoma: An analysis of the National Cancer Database

The optimal timing of adjuvant chemoradiation after surgical resection of glioblastoma remains unknown. The National Cancer Database was queried for all patients diagnosed with glioblastoma from 2004 to 2012 who underwent surgical resection followed by adjuvant chemoradiation. Cox proportional hazards modeling was used to determine factors influencing survival. Optimal thresholds for time to initiation (TTI) of adjuvant therapy were determined using recursive partitioning analysis (RPA). A total of 16,335 patients were included. The RPA model identified an optimal threshold of 61 days to initiation of adjuvant therapy which was confirmed with randomly selected training and validation sets using conditional inference trees repeated 1000 times for robustness. Compared to patients who initiated adjuvant therapy within 61 days (median overall survival: 14.1 months, 95% CI 13.8–14.3 months), patients who initiated adjuvant therapy after day 61 (MOS: 12.0, CI: 10.7–12.9) had higher risk of death (p < 0.0001). For patients with glioblastoma, large increases in time to the initiation of adjuvant therapy leads to inferior survival with a threshold of 61 days or less following surgical resection.

P2.12-10 Novel Prognostic Model for Limited-Stage Small-Cell Lung Cancer

There is currently no prognostic model designed specifically for patients with limited-stage small cell lung cancer (SCLC). The objective of this study was to construct a novel prognostic model for this patient population using baseline characteristics readily available in clinic. An institutional retrospective consecutive patient database was constructed by reviewing the charts of patients diagnosed with limited-stage SCLC between January 2000 and December 2013. Baseline patient characteristics, treatment details and outcomes were extracted. The primary endpoint for the prognostic model was overall survival (OS) and the secondary endpoint was progression-free survival (PFS). Univariable and multivariable Cox regression analyses were performed to identify variables associated with OS and PFS. Prognostic models were generated using recursive partitioning analysis (RPA). A total of 398 patients were eligible for analysis. Median follow-up was 8.2 years. The median age was 67 (range: 40-87). Overall, 94% of patients received chemotherapy, 82% received radiotherapy and 65% received both concurrently. Both hypofractionated (40 Gy in 15 fractions) and conventionally-fractionated (50-66 Gy in 25-33 fractions) radiotherapy were used. Prophylactic cranial irradiation was used in 45% of patients. The median OS for the entire cohort was 15.4 months, with a 5-year OS of 19.7%. Overall median PFS was 9.3 months, with a 5-year PFS of 14.1%. On multivariable Cox regression, age, Eastern Cooperative Oncology Group (ECOG) performance status, tumor location and baseline presence of any pleural effusion (non-malignant or unknown status) were prognostic of OS, while performance status and T-stage were prognostic of PFS. RPA model of OS divided patients into favorable (ECOG 0-2, no pleural effusion, middle/upper lobe location: 5-year OS 31%), intermediate (ECOG 0-2, no pleural effusion, non-middle/upper lobe location: 5-year OS 21%) and unfavorable (ECOG 3-4 or any pleural effusion: 5-year OS 5-8%) groups (log-rank p < 0.001). RPA for PFS divided patients into favorable (ECOG 0-2, <=T3: 5-year PFS 21%) and unfavorable (ECOG 3-4 or T4: 5-year PFS 5-9%) groups (log-rank p < 0.001). Novel prognostic factors for OS and PFS based on baseline patient characteristics were successfully identified for patients with limited-stage SCLC. RPA prognostic models were able to separate patients into distinct risk groups for OS and PFS. Our results will benefit from further external validation and can be useful in stratifying patients in future prospective studies.

A Comparative Analysis of the Usefulness of Survival Prediction Models for Patients with Glioblastoma in the Temozolomide Era: The Importance of Methylguanine Methyltransferase Promoter Methylation, Extent of Resection, and Subventricular Zone Location

Several survival prediction models for patients with glioblastoma have been proposed, but none is widely used. This study aims to identify the predictors of overall survival (OS) and to conduct an independent comparative analysis of 5 prediction models. Multi-institutional data from 159 patients with newly diagnosed glioblastoma who received adjuvant temozolomide concomitant chemoradiotherapy (CCRT) were collected. OS was assessed by Cox proportional hazards regression and adjusted for known prognostic factors. An independent CCRT patient cohort was used to externally validate the 1) RTOG (Radiation Therapy Oncology Group) recursive partitioning analysis (RPA) model, 2) Yang RPA model, and 3) Wee RPA model, Chaichana model, and the RTOG nomogram model. The predictive accuracy for each model at 12-month survival was determined by concordance indices. Calibration plots were performed to ascertain model prediction precision. The median OS for patients who received CCRT was 19.0 months compared with 12.7 months for those who did not (P < 0.001). Independent predictors were:1)subventricular zone II tumors (hazard ratio [HR], 1.6; 95% confidence interval [CI], 1.0-2.5); 2) methylguanine methyltransferase promoter methylation (HR, 0.36; 95% CI, 0.2-0.6); and 3) extent of resection of >85% (HR, 0.59; 95% CI, 0.4-0.9). For 12-month OS prediction, the RTOG nomogram model was superior to the RPA models with a c-index of 0.70. Calibration plots for 12-month survival showed that none of the models was precise, but the RTOG nomogram performed relatively better. The RTOG nomogram best predicted 12-month OS. Methylguanine methyltransferase promoter methylation status, subventricular zone tumor location, and volumetric extent of resection should be considered when constructing prediction models.

Heterogeneity and irregularity of pretreatment 18F-fluorodeoxyglucose positron emission tomography improved prognostic stratification of p16-negative high-risk squamous cell carcinoma of the oropharynx

ObjectivesHuman papillomavirus-negative oropharyngeal squamous cell carcinoma (OPSCC) has unfavorable survival outcomes. Two outcomes have been identified based on smoking history and tumor stage. We investigate the prognostic role of pre-treatment positron emission tomography (PET) in high-risk OPSCC. Materials and MethodsWe enrolled 147 M0 OPSCC patients with p16-negative staining and a history of heavy smoking (>10 pack-years) or T4 disease. All patients completed primary chemoradiotherapy, and 42% maximum standard uptake values (SUVmax) were used as the threshold for primary tumor. Patients were classified into training and validation cohorts with a ratio of 1:1.5 according to the PET date. Heterogeneity and irregularity indices were obtained. PET parameters with significant impact on progression-free survival (PFS) in receiver operating characteristic curves and univariate Cox models were identified and included in recursive partitioning analysis (RPA) for constructing a prognostic model. The RPA-based prognostic model was further tested in the validation cohort using multivariate Cox models. ResultsFifty-eight and 89 patients were in the training and validation groups, respectively. Heterogeneity parameter, SUV-entropy (derived from histogram analysis), and irregularity index, and asphericity were significantly associated with PFS. The RPA model revealed that patients with both high SUV-entropy and high asphericity experienced the worst PFS. Results were confirmed in the validation group. The overall concordance index for PFS of the model was 0.75, which was higher than the clinical stages, performance status, SUVmax, and metabolic tumor volume of PET. ConclusionsPET prognostic model provided useful prediction of PFS for patients with high-risk OPSCC.

Stratification according to recursive partitioning analysis predicts outcome in newly diagnosed glioblastomas.

Glioblastoma accounts for more than half of diffuse gliomas. The prognosis of patients with glioblastoma remains poor despite comprehensive and intensive treatments. Furthermore, the clinical significance of molecular parameters and routinely available clinical variables for the prognosis prediction of glioblastomas remains limited. The authors describe a novel model may help in prognosis prediction and clinical management of glioblastoma patients. We performed a recursive partitioning analysis to generate three independent prognostic classes of 103 glioblastomas patients from TCGA dataset. Class I (MGMT promoter methylated, age <58), class II (MGMT promoter methylation, age ≥58; MGMT promoter unmethylation, age <54, KPS ≥70; MGMT promoter unmethylation, age >59, KPS ≥70), class III (MGMT promoter unmethylation, age 54-58, KPS ≥70; MGMT promoter unmethylation, KPS <70). Age, KPS and MGMT promoter methylation were the most significant prognostic factors for overall survival. The results were validated in CGGA dataset.This was the first study to combine various molecular parameters and clinical factors into recursive partitioning analysis to predict the prognosis of patients with glioblastomas. We included MGMT promoter methylation in our study, which could give better suggestion to patients for their chemotherapy. This clinical study will serve as the backbone for the future incorporation of molecular prognostic markers currently in development. Thus, our recursive partitioning analysis model for glioblastomas may aid in clinical prognosis evaluation.

A prognostic nomogram for overall survival after neoadjuvant radiotherapy or chemoradiotherapy in thoracic esophageal squamous cell carcinoma: a retrospective analysis

BackgroundCurrently, the AJCC staging system or pathological complete response (pCR) are considered not sufficiently accurate to evaluate the survival of patients with esophageal squamous cell carcinoma after neoadjuvant radiotherapy or chemoradiotherapy. This study aimed to establish a nomogram and a recursive partitioning analysis (RPA) model to estimate prognosis and to provide advice for subsequent treatments.MethodsWe analyzed retrospectively 407 patients that were diagnosed with thoracic esophageal squamous cell carcinoma (TESCC) and received neoadjuvant radiotherapy or chemoradiotherapy. Hazard ratios and 95% confidence intervals of categorical clinicopathological characteristics with overall survival (OS) were calculated using the Cox proportional hazard model. The nomogram and RPA model were then established and total scores according to each variable were calculated and stratified to predict OS.ResultsPatients were followed-up over a median 49.9 months. AJCC did not perform well in distinguishing OS among each stage except for IIB and IIIA. Patients were divided into 4 groups according to the total scores based on nomogram (low risk: ≤180; intermediate risk: 180-270; high risk: 270-340; very high risk: >340). The 5-year OS was 57.3%, 40.7%, 18.3%, 6.1% respectively (p<0.05). RPA model also divide the patients into 4 groups, though group2 and group3 were not statistically significant (p=0.574).ConclusionThe nomogram is a good evaluation model for estimating the prognosis of patients with TESCC after neoadjuvant radiotherapy or chemoradiotherapy compared with the AJCC and RPA. The results of this study also suggested that the high-risk subgroups need further treatments.

Molecular-Based Recursive Partitioning Analysis Model for Glioblastoma in the Temozolomide Era

There is a need for a more refined, molecularly based classification model for glioblastoma (GBM) in the temozolomide era. To refine the existing clinically based recursive partitioning analysis (RPA) model by incorporating molecular variables. NRG Oncology RTOG 0525 specimens (n = 452) were analyzed for protein biomarkers representing key pathways in GBM by a quantitative molecular microscopy-based approach with semiquantitative immunohistochemical validation. Prognostic significance of each protein was examined by single-marker and multimarker Cox regression analyses. To reclassify the prognostic risk groups, significant protein biomarkers on single-marker analysis were incorporated into an RPA model consisting of the same clinical variables (age, Karnofsky Performance Status, extent of resection, and neurologic function) as the existing RTOG RPA. The new RPA model (NRG-GBM-RPA) was confirmed using traditional immunohistochemistry in an independent data set (n = 176). Overall survival (OS). In 452 specimens, MGMT (hazard ratio [HR], 1.81; 95% CI, 1.37-2.39; P < .001), survivin (HR, 1.36; 95% CI, 1.04-1.76; P = .02), c-Met (HR, 1.53; 95% CI, 1.06-2.23; P = .02), pmTOR (HR, 0.76; 95% CI, 0.60-0.97; P = .03), and Ki-67 (HR, 1.40; 95% CI, 1.10-1.78; P = .007) protein levels were found to be significant on single-marker multivariate analysis of OS. To refine the existing RPA, significant protein biomarkers together with clinical variables (age, Karnofsky Performance Status, extent of resection, and neurological function) were incorporated into a new model. Of 166 patients used for the new NRG-GBM-RPA model, 97 (58.4%) were male (mean [SD] age, 55.7 [12.0] years). Higher MGMT protein level was significantly associated with decreased MGMT promoter methylation and vice versa (1425.1 for methylated vs 1828.0 for unmethylated; P < .001). Furthermore, MGMT protein expression (HR, 1.84; 95% CI, 1.38-2.43; P < .001) had greater prognostic value for OS compared with MGMT promoter methylation (HR, 1.77; 95% CI, 1.28-2.44; P < .001). The refined NRG-GBM-RPA consisting of MGMT protein, c-Met protein, and age revealed greater separation of OS prognostic classes compared with the existing clinically based RPA model and MGMT promoter methylation in NRG Oncology RTOG 0525. The prognostic significance of the NRG-GBM-RPA was subsequently confirmed in an independent data set (n = 176). This new NRG-GBM-RPA model improves outcome stratification over both the current RTOG RPA model and MGMT promoter methylation, respectively, for patients with GBM treated with radiation and temozolomide and was biologically validated in an independent data set. The revised RPA has the potential to contribute to improving the accurate assessment of prognostic groups in patients with GBM treated with radiation and temozolomide and to influence clinical decision making. clinicaltrials.gov Identifier: NCT00304031.

231P Nomogram to predict prognosis in thoracic esophageal squamous cell carcinoma after neoadjuvant radiotherapy or chemoradiotherapy

Background It is considered not accurate enough to adopt AJCC staging system to evaluate the survival of patients with esophageal squamous cell carcinoma after neoadjuvant radiotherapy or chemoradiotherapy. Our study aims to establish a nomogram for prognosis estimating and instruction about successive treatment. Methods We retrospectively reviewed 407 patients who diagnosed with thoracic esophageal squamous cell carcinoma(TESCC) and received neoadjuvantradiotherapy or chemoradiotherapyfrom 1980 to 2014. Hazard ratios (HRs) and 95% confidence intervals (95%CIs) of categorical age, sex, tumor length, treatment response, lymph node status, resection margin, proximal margin length and anastomotic leakage with overall survival (OS) were calculated using Cox proportional hazard model. Then, the nomogram and recursive partitioning analysis (RPA) model were established respectively, total scores according to each variables were calculated and stratified to predict OS respectively. Results Four hundred and seven patients were followed-up over a median 26.0 months (49.9monthsforcensorcases).ThefiveyearOSanddiseasefreesurvival(DFS)were 36.7%,36.1%withmediansurvivaltime31.0and23.0monthsrespectively.AJCC2009 staging system did not performance well in distinguishing OS except IIB and IIIA(p=0.005).Patientsweredividedinto4groupsaccordingtothetotalscoresbased on nomogram (group A: <180; group B: 180-270; group C: 270-340; group D: >340). The5yearOSwas57.3%,40.7%,18.3%,6.1%,respectivelyandDFSwas57.4%,40.8%, 18.3%, 6.0%, each group shows statistically different prognosis. RPA model indicated that lymph node status, proximal margin length and treatment response were the best prognostic factors, but group 2 and 3 in the 4 groups were not statistically significant (p=0.574). Conclusions The nomogram is a good predictor for prognosis in patients with TESCC after neoadjuvant radiotherapy or chemoradiotherapy. AJCC cancer staging system does not accurately identify each groups. RPA model is not good as nomogram in this patient group. Although need more studies to confirm, this nomogram indicates further treatment may be applied to the high risk subgroup. Legal entity responsible for the study Zefen Xiao Funding N/A Disclosure All authors have declared no conflicts of interest.

Genetic variations in the PI3K-PTEN-AKT-mTOR pathway are associated with distant metastasis in nasopharyngeal carcinoma patients treated with intensity-modulated radiation therapy.

Distant metastasis is the primary failure pattern of nasopharyngeal carcinoma(NPC) in intensity-modulated radiation therapy(IMRT) era. This study was conducted to find the impact of genetic variations in the phosphatidylinositol 3-kinase(PI3K)/phosphatase and tensin homologue(PTEN)/v-akt murine thymoma viral oncogene homologue(AKT)/mammalian target of rapamycin(mTOR) pathway on the risk of distant metastasis in NPC. We genotyped 16 single-nucleotide polymorphisms(SNPs) in five core genes in this pathway from 496 patients treated by IMRT with or without chemotherapy. The relationships between genetic polymorphisms and distant progression were evaluated. We observed that two loci in the AKT1 gene(rs3803300 and rs2494738 alone or combined) were associated with prognosis, with patients carrying at least one variant allele had significantly reduced risk of distant failure, especially in N2-3 group. In addition, we found that genetic variation may had some joint effect with N classification in recursive-partitioning analysis(RPA) analysis, with which patients were stratified into four different risk subgroups (RPA model): RPA1(low risk), RPA2(moderate risk), RPA3(high risk) and RPA4(highest risk). Our findings suggested that genetic variations within the PI3K signaling pathway modulate the development and invasion of NPC patients. Further research is needed to replicate the study in other centers and races, and to unravel the functional significance of these polymorphisms.

Hierarchic Interaction of Factors Associated With Liver Decompensation After Resection for Hepatocellular Carcinoma

Liver resection is the treatment of choice for hepatocellular carcinoma (HCC) in well-compensated liver cirrhosis. Postoperative liver decompensation (LD) is the most representative and least predictable cause of morbidity and mortality. To determine the hierarchy and interaction of factors associated with the risk for LD and to define applicable risk classes among surgical candidates. This retrospective review collected data from 543 patients with chronic liver disease who underwent hepatic resection for HCC from January 1, 2000, through December 31, 2013, in a tertiary comprehensive cancer center. Final follow-up was completed on January 31, 2015, and data were assessed from February 1 to 28, 2015. Preoperative prognostic factors and risk stratification for postoperative LD. Multivariate logistic regression was performed, and the independent risk factors for LD were included in a recursive partitioning analysis model. Results were validated by means of 10-fold cross-validation. The analysis included 543 patients, of whom 411 (75.7%) were male, 132 (24.3%) were female, and the median age was 68 (interquartile range, 62-73) years. An independent association with LD was found for major hepatectomy (odds ratio [OR], 2.41; 95% CI, 1.17-4.30; P = .01), portal hypertension (OR, 2.20; 95% CI, 1.13-4.30; P = .01), and Model for End-Stage Liver Disease (MELD) score greater than 9 (OR, 2.26; 95% CI, 1.10-4.58; P = .02). Recursive partitioning analysis confirmed portal hypertension as the most important factor (OR, 2.99; 95% CI, 1.93-4.62; P < .001), followed by extension of hepatectomy with (OR, 2.76; 95% CI, 1.85-4.77; P = .03) and without (OR, 2.98; 95% CI, 1.97-4.52; P < .001) portal hypertension, and MELD score (OR, 1.79; 95% CI, 1.23-2.13; P < .001). Low-risk patients (LD rate, 4.9% [11 of 226]) without portal hypertension underwent minor resection with a MELD score of 9 or less; intermediate-risk patients (LD rate, 28.6% [85 of 297]) had no portal hypertension and underwent major resections or, in case of minor resections, had portal hypertension or a MELD score greater than 9; and high-risk patients (LD rate, 60.0% [12 of 20]) underwent major resection with portal hypertension. Risk-class progression paralleled median length of stay (7, 8, and 11 days, respectively; P < .001) and liver-related mortality (4.4% [10 of 226], 9.0% [27 of 297], and 25.0% [5 of 20], respectively; P = .001). A 10-fold cross-validation of the model resulted in a C index of 0.78 (95% CI, 0.74-0.82) and an overall error rate of 0.06. The risk for postoperative LD after resection for HCC in chronic liver disease is associated with preoperative hierarchic interaction of portal hypertension, planned extension of hepatectomy, and the MELD score.

Recursive partition analysis of peritoneal and systemic recurrence in patients with gastric cancer who underwent D2 gastrectomy: Implications for neoadjuvant therapy consideration.

To classify patients with nonmetastatic advanced gastric cancer who underwent D2-gastrectomy into prognostic groups based on peritoneal and systemic recurrence risks. Between 2004 and 2007, 1,090 patients with T3-4 or N+ gastric cancer were identified from our registry. Recurrence rates were estimated using a competing-risk analysis. Different prognostic groups were defined using recursive partitioning analysis (RPA). Median follow-up was 7 years. In the RPA-model for peritoneal recurrence risk, the initial node was split by T stage, indicating that differences between patients with T1-3 and T4 cancer were the greatest. The 5-year peritoneal recurrence rates for patients with T4 (n = 627) and T1-3 (n = 463) disease were 34.3% and 9.1%, respectively. N stage and neural invasion had an additive impact on high-risk patients. The RPA model for systemic relapse incorporated N stage alone and gave two terminal nodes: N0-2 (n = 721) and N3 (n = 369). The 5-year cumulative incidences were 7.7% and 24.5%, respectively. We proposed risk stratification models of peritoneal and systemic recurrence in patients undergoing D2-gastrectomy. This classification could be used for stratification protocols in future studies evaluating adjuvant therapies such as preoperative chemoradiotherapy. J. Surg. Oncol. 2016;114:859-864. © 2016 2016 Wiley Periodicals, Inc.

Nomogram and recursive partitioning analysis to predict overall survival in patients with stage IIB-III thoracic esophageal squamous cell carcinoma after esophagectomy.

We have developed statistical models for predicting survival in patients with stage IIB–III thoracic esophageal squamous cell carcinoma (ESCC) and assessing the efficacy of adjuvant treatment. From a retrospective review of 3,636 patients, we created a database of 1,004 patients with stage IIB–III thoracic ESCC who underwent esophagectomy with or without postoperative radiation. Using a multivariate Cox regression model, we assessed the prognostic impact of clinical and histological factors on overall survival (OS). Logistic analysis was performed to identify factors to include in a recursive partitioning analysis (RPA) to predict 5-year OS. The nomogram was evaluated internally based on the concordance index (C-index) and a calibration plot. The median survival time in the training dataset was 30.9 months, and the 5-year survival rate was 33.9%. T stage, differentiated grade, adjuvant treatment, tumor location, lymph node metastatic ratio (LNMR), and the presence of vascular carcinomatous thrombi were statistically significant predictors of 5-year OS. The C-index of the nomogram was 0.70 (95% CI 0.67–0.73). RPA resulted in a three-class stratification: class 1, LNMR ≤ 0.15 with adjuvant treatment; class 2, LNMR ≤ 0.15 without adjuvant treatment and LNMR > 0.15 with adjuvant treatment; and class 3, LNMR > 0.15 without adjuvant treatment. The three classes were statistically significant for OS (P < 0.001). Thus, the nomogram and RPA models predicted the prognosis of stage IIB–III ESCC patients and could be used in decision-making and clinical trials.