Cox PH Model Research Articles

Delays from neoadjuvant chemotherapy to interval debulking surgery and survival in ovarian cancer.

5552 Background: Prolonged time from primary surgery to chemotherapy is associated with worse survival in ovarian cancer (OC); however, the impact of prolonged time from neoadjuvant chemotherapy (NACT) to interval debulking surgery (IDS) is unknown. Given increasing utilization of NACT, we seek to evaluate the role of delays from NACT to IDS (TIDS) on survival. Methods: At a single center, we prospectively identified 224 women with newly diagnosed stage III/IV OC given NACT from 7/1/15 to 12/1/17. Clinical characteristics were abstracted by two independent reviewers. Delays in TIDS were defined as time from last preoperative carboplatin to IDS > 6 weeks. Fisher’s exact/Wilcoxon rank sum tests were used to compare clinical characteristics by delay in TIDS. Kaplan Meier method was used to estimate progression-free (PFS) and overall survival (OS) from date of IDS. Log-rank test/multivariate CoxPH models were used to examine differences by delay groups, adjusting for covariates. Results: Of the 224 women, 159 underwent IDS, and 34 (21%) experienced TIDS delays. These women were older (median 68 vs. 65 years, p = 0.05) and had more preoperative NACT cycles (median 6 vs. 4, p = 0.003). Patients with delays in TIDS also had a longer interval from pathological diagnosis to start of NACT (TNACT), median 22 vs. 17 days, p = 0.01, and interval from IDS to postoperative chemotherapy (TPOC), median 37 vs. 30 days, p = 0.01; however, neither TNACT nor TPOC predicted survival, p > 0.05. On univariate analysis, delays in TIDS were significantly associated with worse OS (HR 2.4 95% CI 1.2-4.8, p = 0.01); however, this was attenuated in multivariate models (HR 1.66 95% CI 0.8-3.4, p = 0.17), adjusting for age, stage and complete gross resection (CGR). On univariate analysis, delays in TIDS were not associated with PFS (HR 1.55 95% CI 0.97-2.5, p = 0.062), and in multivariate models, increase in number of preoperative NACT cycles (p = 0.005) and lack of CGR (p < 0.001) were the only variables predictive of worse PFS. Conclusions: Delays in TIDS are associated with OS, but not after adjustment for age, stage and CGR, suggesting a need to maximize cytoreduction regardless of delays in NACT. The role of preoperative NACT cycles on survival should be further studied.

Development of an artificial intelligence model to predict survival at specific time intervals for lung cancer patients.

6556 Background: Survival prediction models for lung cancer patients could help guide their care and therapy decisions. The objectives of this study were to predict probability of survival beyond 90, 180 and 360 days from any point in a lung cancer patient’s journey. Methods: We developed a Gradient Boosting model (XGBoost) using data from 55k lung cancer patients in the ASCO CancerLinQ database that used 3958 unique variables including Dx and Rx codes, biomarkers, surgeries and lab tests from ≤1 year prior to the prediction point, which was chosen at random for each patient. We used 40% data for training, 25% for hyper-parameter tuning, 20% for testing and 15% for holdout validation. Death date available in the Electronic Health Record was cross checked by linkage to death registries. Results: The model was validated on the holdout set of 8,468 patients. The Area Under the Curve (AUC) for the model was 0.79. The precision and recall for predicting survival beyond the three time points were between 0.7-0.8 and 0.8-0.9 respectively (see table). This compares favourably to other lung cancer survival models created using different machine learning techniques (Jochems 2017, Dekker 2009). A Cox-PH model created using the top 20 variables also had a significantly lower performance (see table). Analysis of input variables yielded distinctive patterns for patient subgroups and time points. Tumor status, medications, lab values and functional status were found to be significant in patient sub cohorts. Conclusions: An AI model to predict survival of lung cancer patients built using a large real world dataset yielded high accuracy. This general model can further be used to predict survival of sub cohorts stratified by variables such as stage or various treatment effects. Such a model could be useful for assessing patient risk and treatment options, evaluating cost and quality of care or determining clinical trial eligibility. [Table: see text]

Epithelial cell death markers in bronchoalveolar lavage correlate with chronic lung allograft dysfunction subtypes and survival in lung transplant recipients-a single-center retrospective cohort study.

Chronic lung allograft dysfunction (CLAD) remains the leading cause of late death after lung transplantation. Epithelial injury is thought to be a key event in the pathogenesis of CLAD. M30 and M65 are fragments of cytokeratin-18 released specifically during epithelial cell apoptosis and total cell death, respectively. We investigated whether M30 and M65 levels in bronchoalveolar lavage (BAL) correlate with CLAD subtypes: restrictive allograft syndrome (RAS) versus bronchiolitis obliterans syndrome (BOS). BALs were obtained from 26 patients with established CLAD (10 RAS, 16 BOS) and 19 long-term CLAD-free controls. Samples with concurrent infection or acute rejection were excluded. Protein levels were measured by ELISA. Variables were compared using Kruskal-Wallis, Mann-Whitney U test and Chi-squared tests. Association of M30 and M65 levels with post-CLAD survival was assessed using a Cox PH models. M65 levels were significantly higher in RAS compared to BOS and long-term CLAD-free controls and correlated with worse post-CLAD survival. Lung epithelial cell death is enhanced in patients with RAS. Detection of BAL M65 may be used to differentiate CLAD subtypes and as a prognostic marker in patients with established CLAD. Understanding the role of epithelial cell death in CLAD pathogenesis may help identify new therapeutic targets to improve outcome.

OR14-4 Early Glomerular Hyperfiltration and Long Term Kidney Outcomes in Type 1 Diabetes: The DCCT/EDIC Experience

INTRODUCTION Glomerular hyperfiltration has long been considered to be a major contributing factor to the development of diabetic kidney disease but most studies assessed increased albumin excretion rather than reduced GFR as an outcome. To address whether early glomerular hyperfiltration results in subsequent increased risk of clinically significant loss of GFR, namely Stage 3 CKD (eGFR <60 ml/min/1.73m2), we analyzed 30 year GFR follow-up data in participants undergoing 125I-iothalamate clearance on entry into the DCCT/EDIC Study. METHODS125I-iothalamate clearance was added to the DCCT protocol in 1986 and was assessed at DCCT baseline in 446 participants. This analysis reports long-term eGFR (CKD-EPI equation) follow-up data on these participants. The association of baseline hyperfiltration (primary cutpoint of ≥140 ml/min/1.73m2) with the risk of developing Stage 3 CKD (eGFR < 60 ml/min/1.73m2) was analyzed using Cox proportional hazards models. RESULTS Of the 446 participants, 178 had iothalamate GFR levels ≥ 130 mL/min/1.73m2 and of these, 106 had levels ≥ 140 mL/min/1.73m2 and 55 had levels ≥ 150 mL/min/1.73m2 upon entry into the DCCT. Among these 446 participants, 53 developed an eGFR < 60 mL/min/1.73m2 events over a median follow-up time of 28 years (rate of 4.69 events per 1000 individuals at risk for one year), and 34 developed a sustained (i.e., two consecutive visits) eGFR <60 mL/min/1.73m2 events over a median follow-up time of 28 years (rate of 2.98 sustained events per 1000 individuals at risk for one year) in DCCT/EDIC. The proportion maintaining an eGFR ≥ 60 mL/min/1.73m2 was not decreased and was actually somewhat greater in the hyperfiltration group (95/106 = 89.6% vs. 298/340 = 87.6%) using the cutoff of 140 mL/min/1.73m2. The cumulative incidences of developing an eGFR < 60 mL/min/1.73m2 were again similar in the two hyperfiltration groups (≥140 vs. < 140 mL/min/1.73m2 - 4.1% vs. 5.9% after 20 years and 11% vs. 12.8% after 28 years). Hyperfiltration as assessed by iothalamate GFR ≥140 mL/min/1.73m2 was not associated with subsequent risk of developing an eGFR < 60 mL/min/1.73m2 in an unadjusted Cox PH model (HR= 0.83, 95%CI [0.43, 1.62]) nor in the adjusted model (HR= 0.77, 95%CI [0.38, 1.54]). Similar results were obtained for the developing of a sustained eGFR < 60 mL/min/1.73m2. Hyperfiltration cut-offs of 130 and 150 ml/min/1.73m2 showed similar findings. CONCLUSIONS Early hyperfiltration in patients with type 1 diabetes was not associated with any long-term decrease in kidney function. Though it is known with certainty that long-term improved glycemic control reduces the development of microalbuminuria, macroalbuminuria and Stage 3 CKD, the notion that early hyperfiltration is a marker of poor long term renal outcome is not supported by these findings.

Education, experience, social network and firm survival: the case of the electrical fittings cluster in Sargodha, Pakistan

In modern era of industrialization, firm survival is widely acknowledged and geared through industrial cluster in developing countries. The present study unfolds the impact of education, experience and social network on firm death by using two rounds of electrical fittings cluster survey from 2008 to 2017. The empirical outcomes from Cox PH model spotlight that entrepreneurs are relatively more educated, rich in experience of production and marketing, have low hazard rate in their business. The empirical analysis also revealed that firm survival can be achieved through strong social network of entrepreneur. The study findings infer that Small and Medium Enterprises Development Authority and United Nations Industrial Development Organization training programs are needed to focus on cluster entrepreneurs and sustained firm survival.

Statistical Analysis of Risk factors of Malaria re-infection among Outpatients in DR, Congo: A Comparison Approach of AFT and Cox PH Models

Background: Malaria is an infectious disease caused by a Plasmodium parasite and is one of the highest causes of mortality globally. This study aims to determine models to detect the effect of risk factors on malaria re-infection of patients survival. Methods: The study includes 109 malaria outpatients in Lubumbashi Congo Hospital, who had re-infection status after six months follow-up. The survival status of the re-infected patients was based on the effect of various factors. The best model was selected through Akaike Information Criterion (AIC) and Cox-Snell Residuals using SAS and R packages. Results: The results from the analysis showed that Gamma model (AIC=147.092) was better in the analysis compared to accelerated failure-time models. Conclusion: Although, many researchers prefer proportional hazard model in analysing a survival data but accelerated failure-time model is a good alternative method as they do not require proportionality of hazards as key assumption.

Decreased Club Cell Secretory Protein in Bronchoalveolar Lavage is Associated with Chronic Lung Allograft Dysfunction (CLAD) but Not with CLAD Phenotypes

Purpose Chronic lung allograft dysfunction (CLAD) remains the main impediment to long-term survival after lung transplantation. Club cell secretory protein (CCSP), is an abundant protein secreted from non-ciliated bronchial epithelial cells with both antioxidative and anti-inflammatory properties. BAL CCSP levels were previously shown to be lower in patients who later developed CLAD. We aimed to examine whether levels of CCSP in BAL obtained early after CLAD diagnosis are associated with the major CLAD subtypes: bronchiolitis obliterans syndrome (BOS) and restrictive allograft syndrome (RAS). Methods Among all adult, first, bilateral, lung transplants, between 2004-2013, CLAD status was retrospectively determined as a sustained irreversible drop in FEV1, and patients were classified into BOS and RAS (based on 10% reduction in TLC), and long-term (>4 years) CLAD-free. First BAL available after CLAD diagnosis not associated with concurrent infection or acute rejection was chosen for this study. Protein levels were measured by ELISA. Log transformed CCSP levels were compared using linear regression models. Association of CCSP levels with post-CLAD survival was assessed using Cox PH models. Results BALs were obtained from 26 patients with established CLAD (10 RAS, 16 BOS) and 19 CLAD-free controls. CCSP levels were significantly lower in CLAD compared to CLAD-free controls (P=0.019) but did not differ between CLAD subgroups RAS and BOS (P=0.57). CCSP level was not associated with post-CLAD survival (HR 1.05, CI (0.72, 1.56), P=0.78). Conclusion We confirm that BAL CCSP levels are reduced in lung transplant recipients with CLAD. However, CCSP levels do not differ between BOS and RAS, possibly since airways are similarly affected in both conditions. Further investigation of other types of epithelial injury may provide insight into the pathogenesis of CLAD and its subtypes.

Bronchoalveolar Lavage Pentraxin 3 Levels during Clinically-Stable Minimal Acute Rejection are Associated with CLAD and Death Risk

Purpose Acute cellular rejection (ACR) remains the most significant risk factor for chronic lung allograft dysfunction (CLAD). While clinically-unstable and higher-grade ≥A2 ACR is generally treated with augmented immunosuppression (aIS), management of clinically-stable A1 ACR (CSA1R) remains controversial. At our center, patients with CSA1R are routinely not treated with aIS and constitute a unique population that enabled us to identify sub-phenotypes with differential risks. Our aim was to determine if bronchoalveolar lavage (BAL) biomarkers at the time of first CSA1R (fCSA1R) are predictive of subsequent CLAD or death. Methods Among all adult, first, bilateral lung transplants, performed 2010-2016, post-transplant transbronchial biopsies obtained within the first year, without concurrent infection, were categorized as clinically unstable or stable based on ≥10% concurrent drop in FEV1. fCSA1R episodes were defined as stable biopsy-FEV1 pairs not preceded by another ACR. 21 proteins previously associated with ACR or CLAD were measured in the BAL using a multiplex bead assay. Association between protein levels and subsequent CLAD or death was assessed using Cox PH models, adjusted for relevant peri-transplant clinical covariates. Results We identified 75 patients with fCSA1R at a median time of 98 days (range 48.5-197) post-transplant. Median time from transplant to CLAD or death was 1247 (756.5-1921.5) and 1641 days (1024.5-2326.5), respectively. In multivariable models, levels of IL6, S100A8, IL10, TNF-receptor 1, IL-1α and pentraxin 3 (PTX3) were associated with both CLAD development and death (P Conclusion Our data indicate that a focused BAL protein signature, with PTX3 having the strongest association, may be useful in determining a subset of CSA1R patients who are at increased risk and may benefit from a more aggressive management strategy.

Abstract P378: A Novel Risk Prediction Score for Incident Heart Failure Among Patients With Diabetes

Introduction: Diabetes is a well-established risk factor for heart failure (HF). However, risk prediction models for incident HF in patients with diabetes are lacking. Accordingly, we aim to develop a novel risk-prediction model for HF in a diabetes cohort using a hybrid machine learning method for variable selection and CoxPH approach for risk prediction. Methods: We used data from the ACCORD trial that included patients with prevalent diabetes at baseline and had adjudicated HF events on follow up. The data was divided in to 80/20% training and testing (n = 6,896/1,725, respectively) and 152 baseline variables were included. A random survival forest (RSF) of 1000 trees was constructed to identify the top 20 predictors. Finally, a 5-year incident HF risk score was developed using the 9 variables selected by backward CoxPH regression. Results: The study included 8,621 participants (Mean age: 63 y, 62% male) without prevalent HF. The RSF model for incident HF with top 20 predictors derived from the training cohort had superior discrimination than the CoxPH model in the testing cohort ( C -index 0.80 vs. 0.78; P &lt; 0.001). The top 9 variables identified included age, waist circumference, BP, serum creatinine, HDL, prior MI, prior CABG, and EKG QTC duration and T-axis. The weighted risk score based on the nine clinical variables was used to create a composite HF risk score ranging from -6 to 36. A 1-unit increment in the risk score was associated with a 27% higher risk of HF at 5 years. A graded increment in the 5-year HF incidence was noted across quintiles of the HF risk score ranging from 1.2% in the lowest to 9.8% in the highest quintile (Figure). The HF risk score demonstrated good discrimination with an overall C -index of 75.2. Conclusions: In conclusion, an RSF model in a clinical diabetes cohort outperformed conventional CoxPH in the identification of important risk factors of incident HF. Additionally, the diabetes-HF prediction model using common clinical variables demonstrated good discrimination and risk stratification in the overall cohort.

Abiraterone acetate plus prednisone/prednisolone compared with enzalutamide in metastatic castration resistant prostate cancer before or after chemotherapy: A retrospective study of real-world data (ACES).

303 Background: Abiraterone acetate (a prodrug of abiraterone, which is a selective inhibitor of androgen biosynthesis) combined with prednisone/prednisolone (AA+P) and enzalutamide (ENZ) (an androgen-receptor–signalling inhibitor) have proven survival benefit in men with metastatic castration resistant prostate cancer (mCRPC) in chemo naïve and prior chemo patients. There have been no studies directly comparing the effectiveness of ENZ to AA+P in mCRPC patients. Methods: A retrospective, survival analysis study of 143 real world mCRPC patients (90 in AA+P and 53 in ENZ group) was conducted. Patients who started their treatment between 1st February 2012 and 31st May 2016 were included. The primary endpoint was biochemical progression free survival (PFS). Secondary end points were radiographic progression free survival (rPFS) and overall survival (OS). Data was analysed using Cox proportional hazards models, adjusting for covariates: prior radical or palliative treatment; Gleason score; baseline PSA; age; and chemo naïve or not. Results: After median follow up of 15 months (IQR 7 to 23) 112 events of biochemical progression were observed (71 in AA+P and 41 in ENZ). 41%in AA+P group and 30% patients in ENZ group received prior chemo. The chance of biochemical progression was significantly lower among ENZ patients than AA+P patients, when adjusting for all covariates in the Cox PH model (Hazard Ratio 0.54, 95% CI 0.35 to 0.82, p=0.004. There was a trend implying the chance of rPFS could be higher among ENZ patients than AA+P patients (HR 1.24, 95% CI 0.76 to 2.02, p=0.4). OS is lower among ENZ patients than AA+P patients, when adjusting for all covariates in the Cox PH model (HR 0.91, 95% CI 0.59 to 1.41, p=0.7). 38% of ENZ patients reported fatigue compared to 16% of AA+P patients while hypertension was reported slightly more in AA+P patients than in ENZ patients. Conclusions: This study showed a statistically significant difference in biochemical progression-free survival, favouring ENZ, but no significant difference in radiographic progression-free survival or overall survival.

SURVIVAL REGRESSION MODELS FOR SINGLE EVENTS AND COMPETING RISKS BASED ON PSEUDO-OBSERVATIONS

Abstract Survival data is a special type of data that measures the time to an event of interest. The most important feature of survival data is the presence of censored observations. An observation is said to be right-censored if the time of the observation is, for some reason, shorter than the time to the event. If no censoring occurs in the data, standard statistical models can be used to analyse the data. Pseudo-observations can replace censored observations and thereby allow standard statistical models to be used. In this paper, a pseudo-observation approach was applied to single-event and competing-risks analysis, with special attention paid to the properties of the pseudo-observations. In the empirical part of the study, the use of regression models based on pseudo-observations in credit-risk assessment was investigated. Default, defined as a delay in payment, was considered to be the event of interest, while prepayment of credit was treated as a possible competing risk. Credits that neither default nor are prepaid during the follow-up were censored observations. Typical application characteristics of the credit and creditor were the covariates in the regression model. In a sample of retail credits provided by a Polish financial institution, regression models based on pseudo-observations were built for the single-event and competing-risks approaches. Estimates and discriminatory power of these models were compared to the Cox PH and Fine-Gray models.

Clinical prognostic model for older patients with advanced non-small cell lung cancer.

Older patients with non-small cell lung cancer (NSCLC) are often not prescribed standard therapy. It is important to know which older patients would be candidates for aggressive therapy based on their prognosis, and to develop a model that can help determine prognosis. Data on older patients (≥70 years) enrolled on 38 NCI cooperative group trials of advanced NSCLC from 1991 to 2011 were analyzed. Multivariable Cox PH model was built with a stepwise selection. We derived a prognostic score using the estimated Cox PH regression coefficient. We then calculated the area under receiver operating characteristic (ROC) curve of survival in the testing set. The final analysis included 1467 patients, who were randomly divided into a training (n = 963) and a testing set (n = 504). The prognostic risk score was calculated as: 3 (if male) + 3 (if PS = 1) + 8 (if PS = 2) + 11 (if initial stage = IV) + 4 (if weight loss). Patients were classified into two prognostic groups: good (0-8) and poor (≥9). The median survival in the two groups in the testing set were 13.15 (95% CI, 10.82-15.91) and 8.52 months (95% CI, 7.5-9.63), respectively. The model had area under the 1-year and 2-year ROCs (0.6 and 0.65, respectively) that were higher than existing models. Male gender, poor performance status, distant metastases and recent weight loss predict for poor overall survival (OS) in older patients with advanced NSCLC. This study proposes a simple prognostic model for older adults with advanced NSCLC.

A model-based concordance-type index for evaluating the added predictive ability of novel risk factors and markers in the logistic regression models

ABSTRACTThe Concordance statistic (C-statistic) is commonly used to assess the predictive performance (discriminatory ability) of logistic regression model. Although there are several approaches for the C-statistic, their performance in quantifying the subsequent improvement in predictive accuracy due to inclusion of novel risk factors or biomarkers in the model has been extremely criticized in literature. This paper proposed a model-based concordance-type index, CK, for use with logistic regression model. The CK and its asymptotic sampling distribution is derived following Gonen and Heller's approach for Cox PH model for survival data but taking necessary modifications for use with binary data. Unlike the existing C-statistics for logistic model, it quantifies the concordance probability by taking the difference in the predicted risks between two subjects in a pair rather than ranking them and hence is able to quantify the equivalent incremental value from the new risk factor or marker. The simulation study revealed that the CK performed well when the model parameters are correctly estimated for large sample and showed greater improvement in quantifying the additional predictive value from the new risk factor or marker than the existing C-statistics. Furthermore, the illustration using three datasets supports the findings from simulation study.

A prediction-driven mixture cure model and its application in credit scoring

In the credit market, assessment of a borrower's default risk over time is essential to enabling timely risk management, since borrowers’ exposure to risk and the losses that result from defaults are strongly related to the time when they default. Mixture cure models, with their ability to predict not only whether borrowers will default but also when they are likely to default, have been applied to credit scoring. We propose a prediction-driven mixture cure model, which sacrifices interpretability for potentially better prediction performance, and apply it to credit scoring. In the incidence part of the mixture cure model, we substitute the typical statistical incidence model (i.e., logistic regression) with a more flexible, and hopefully more accurate, classification method (i.e., random forests). For the latency part, we propose a survival analysis model, named Time-Dependent Hazards, which accommodates a direct relationship between failure times and covariates and can potentially better predict the probability of default over time than the standard Cox PH model. Empirical evaluation using real-world data from a major P2P lending institution in China shows that both extensions contributed to performance improvement in both discrimination and calibration.

Nrf2 metagene as a prognostic biomarker across all stages of colorectal cancer (CRC).

557 Background: Keap1/Nrf2 is an important intracellular canonical stress response pathway, with Nrf2 acting as a potent transcription factor. Lung cancer is known to acquire constitutive activation of the pathway thus promoting survival, resisting chemo-radiation and dysregulating metabolism. Mechanisms of pathway activation include methylation of KEAP1, somatic mutation and direct activation by KRAS, BRAF and MYC. Its role in CRC is unknown. Due to its role as a transcription factor, activated in many ways, we hypothesise a metagene of Nrf2 regulated genes would act as a prognostic biomarker in CRC. Methods: Using a candidate gene approach and publicly available data (GSE17536), we derived and trained a 36 gene metagene to aggregate Nrf2 pathway expression, using principal component analysis and cox proportional hazard models. GSE14333 was used for validation in stage I/II/III CRC. The first line metastatic FOCUS trial was used for validation in Stage IV disease. Results: 601 patients are included in the validation analysis. Nrf2 metagene expression is associated with worse DFS outcomes in stage I/II/III disease in Cox PH model comparisons (LRT, p = 0.00175) and worse OS in stage IV disease (LRT, p = 0.00574). On multivariate analysis, Nrf2 expression remained significant when adjusted for known prognostic factors of adjuvant chemotherapy and Duke’s stage in stage I/II/III disease (ANOVA, p = 0.03), BRAF V600E and sidedness in stage IV disease (ANOVA, p = 0.00255). Using the median cut point, high expression has a HR 2.36 (C.I 1.27-4.38) in stage I/II/III disease and HR 1.39 (C.I 1.12-1.72) in metastatic disease. Conclusions: Nrf2 expression is a novel, robust prognostic biomarker of poor prognosis across all stages of colorectal cancer. Greater understanding of its mechanistic role could lead to targeted strategies to improve outcomes in CRC. [Table: see text]

Demonstrating the Performance of Accelerated Failure Time Model Over Cox-PH Model of Survival Data Analysis with Application to HIV-Infected Patients Under HAART

Human Immunodeficiency Virus (HIV) is a virus that kills CD4 cells. These CD4 cells are white blood cells that fight infection. CD4 count is like a snapshot of how well our immune system is functioning. Studying the way of CD4+ count over time provides an insight to the disease evolution. This study was considering the data of HIV/AIDS patients who were undergoing Antiretroviral Therapy in the ART clinic of Menellik II Referral Hospital, Addis Ababa, Ethiopia, during the period 1<i>^st</i> January 2014 to 31<i>^st</i> December 2017. For separate survival model log-logistic model is more appropriate for the survival data than other parametric models. Therefore; functional status and regimen class are significant covariates in determining the hazard function patients. Log rank and Wilcoxon tests showed that the significant difference in survival situation among the categorical variables selected for this study sex, marital status, functional status, WHO-clinical stages and regimen class subgroups. But, there was no significant difference in the time-to-event between subgroups of sex, Marital Status and WHO clinical Stage, while, Regimen Class and Functional Status there was a significant difference in the time-to-event between subgroups.

Large-margin learning of Cox proportional hazard models for survival analysis

Machine learning approaches have been recently attempted to tackle the prediction tasks in survival analysis. However, most existing methods aim to learn the prognostic function directly via linear regression or ranking models, unable to exploit the underlying density family, notably the famous CoxPH model. In this paper we propose a novel estimator for the CoxPH model based on the margin maximization principle, which was proven to achieve superb generalization performance in standard classification problems in machine learning. The censored data are effectively handled by incorporating cost-sensitive margin violation loss. We demonstrate the improved prediction performance on several survival datasets.

Effect Modification of the Association of Long-Term PM2.5 Exposure and Cause-Specific Mortality: An Analysis of 64 Million U.S. Medicare Beneficiaries

Introduction. Long-term exposure to fine particulate matter (PM2.5) has been consistently associated with mortality; however, our understanding of how these associations are modified by certain characteristics, such as urbanicity and race, is more limited.Methods. We considered demographic and mortality data for over 64 million Medicare enrollees across all 40 thousand ZIP codes in the conterminous United States from 2000 to 2008. We linked these data to ZIP code and month-specific PM2.5 exposures estimated using GIS-based spatio-temporal models and fit the data using Cox proportional hazard model that were modified using data aggregation and limited-memory BFGS optimization. This modification allowed our Cox PH model to analyze data simultaneously for all 64 million Medicare beneficiaries within 10 minutes (Intel Xeon E5-2680, 56 cores). We used these models to estimate the association of PM2.5 on all cause, cardiovascular (CVD), respiratory disease and cancers (including specifics: ischemic heart disease, cerebrovascular disease, congestive heart failure, COPD, pneumonia, upper respiratory infection (URI), lung cancer), with strata for age, race, gender, and ZIP code of residence. We examined effect modification by age, race, gender, land use, and region and assessed the linearity of the dose-response relationship for each cause and by modifier.Results. We found significant positive associations of PM2.5 and all causes of death, except URI, with mortality risk ratios (RR) for non-accidental, CVD, respiratory, and cancer mortality equaling 1.244 (95% CI: 1.238, 1.251), 1.683 (95% CI: 1.669, 1.696), 1.241 (95% CI: 1.223, 1.260) and 1.160 (95% CI: 1.147, 1.172) per 10 ug/m3 increase in PM2.5, respectively. We showed risk of death to be higher for beneficiaries living in urban as compared to non-urban areas, for men as compared to women, for younger ages, and individuals living in the northeastern US. PM2.5 associated RRs were similar for whites and non-whites.

MASAN: a novel staging system for prognosis of patients with oesophageal squamous cell carcinoma

BackgroundOesophageal squamous cell carcinoma (ESCC) is one of the most malignant cancers worldwide. Treatment of ESCC is in progress through accurate staging and risk assessment of patients. The emergence of potential molecular markers inspired us to construct novel staging systems with better accuracy by incorporating molecular markers.MethodsWe measured H scores of 23 protein markers and analysed eight clinical factors of 77 ESCC patients in a training set, from which we identified an optimal MASAN (MYC, ANO1, SLC52A3, Age and N-stage) signature. We constructed MASAN models using Cox PH models, and created MASAN-staging systems based on k-means clustering and minimum-distance classifier. MASAN was validated in a test set (n = 77) and an independent validation set (n = 150).ResultsMASAN possessed high predictive accuracies and stratified ESCC patients into three prognostic groups that were more accurate than the current pTNM-staging system for both overall survival and disease-free survival. To facilitate clinical utilisation, we also constructed MASAN-SI staging systems based on staining indices (SI) of protein markers, which possessed similar prognostic performance as MASAN.ConclusionMASAN provides a good alternative staging system for ESCC prognosis with a high precision using a simple model.

Under-five child mortality in Bangladesh: Classical and Bayesian approaches to Cox proportional hazard model

This study examines the determinants of under-five child mortality in Bangladesh using the data extracted from the Bangladesh Demographic and Health Survey (BDHS), 2011 and 2014. Product-Limit method and Log-Rank test have been used for bivariate analysis. Cox proportional hazard model has been employed under both classical and Bayesian approaches. Cox regression analysis reveals that region (Barisal and Sylhet), maternal education (higher education), mother’s membership of NGO have significant impact on child mortality. The results obtained using Bayesian Cox PH model are almost similar except one key finding. Under Bayesian analysis, child’s size at birth appeared as potential determinant of under-five mortality whereas it has insignificant effect on child survival when classical Cox model has been applied.Bangladesh J. Sci. Res. 30(1&amp;2): 45-54, December-2017