Martingale Residuals Research Articles

Sequential Stratification Prognostic Score for Patients with Intermediate-Risk Myelodysplastic Syndrome

Background: The International Prognostic Scoring System-Revised (IPSS-R) is one gold standard for myelodysplastic syndrome (MDS) risk stratification. An accurate estimate of prognosis allows clinicians to make appropriate management decisions. The IPSS-R intermediate-risk (int-risk) group however, shows marked heterogeneity with respect to disease outcomes, and contributes to uncertainty about patient management. A more refined yet simple sequential scoring model is therefore proposed to improve the classification of int-risk patients to allow for appropriate therapeutic decisions.Methods: A retrospective patient-centered analysis was conducted at MD Anderson Cancer Center on patients diagnosed with primary MDS between 2000 and 2015. All patients were scored according to the IPSS-R criteria and int-risk patients were included. The primary outcome was overall survival (OS), and covariate characteristics included: hemoglobin, ferritin, white blood cell (WBC) count, platelet count, absolute neutrophil count (ANC), percentage of peripheral blood (PB) and bone marrow (BM) blasts at diagnosis, number of cytopenias, performance status, RBC and platelet transfusion history, and type of treatment received. Survival differences between groups were compared and univariate Cox proportional hazards regressionwas used to evaluate association between each covariate and OS. Classification and regression tree (CART) analysis was used to find the best cut-off points to differentiate OS for some continuous variables. CART analysis for failure time data used the Martingale residuals of a Cox model to approximate chi-square values for all possible cut points. A multivariate Cox proportional hazards model was then fitted for OS by including all statistically significant covariates from univariate Cox models. A backward stepwise selection was performed using a threshold significance of 0.05 for covariates to stay in the final model. Internal validation testing of the final stratification score was performed using the bootstrap method, and external validation was performed using an independent external cohort from Cleveland Clinic.Results: Out of 3167 patients, a total of 298 were identified with IPSS-R int-risk MDS and retrospectively analyzed in this study. Age of 66 years or greater (2 points), peripheral blood blasts of 2% or more (1 point), and history of RBC transfusion (1 point) were significantly associated with inferior survivaland were included in the final OS prognostic model. The score divided MDS patients who have IPSS-R intermediate-risk into two prognostic risk groups: an int-favorable group with a score of 0-1, and an int-adverse group with a score of 2-4. Overall survival was significantly different between the two risk groups of patients, with median survival of approximately 6 and 2 years for int-favorable and int-adverse, respectively (p <0.001, Figure 1A). Transfusions per patient year were significantly higher in int-adverse patients for both red blood cells and platelets (p<0.001, Figure 1B). The stratification prognostic score was robust in the classification of patients with IPSS-R intermediate-risk myelodysplastic syndrome by internal validation, and analysis of 111 IPSS-R int-risk MDS patients from a separate institution demonstrated significantly different OS between int-favorable and int-adverse patients, with median survival of approximately 4 and 2 years respectively (p=0.04).Conclusions: Sequential prognostication provides a clinically practical framework to aid clinicians in determining appropriate treatment approaches for the heterogeneous group of int-risk MDS patients. [Display omitted] DisclosuresJabbour:Bristol-Myers Squibb: Consultancy. Komrokji:Celgene: Honoraria; Novartis: Honoraria, Speakers Bureau. Steensma:Pfizer: Consultancy; Incyte: Equity Ownership; Janssen: Consultancy, Research Funding; H3 Biosciences: Consultancy; Celgene: Consultancy; Pfizer: Consultancy, Membership on an entity's Board of Directors or advisory committees; Takeda: Consultancy; Onconova: Consultancy; Amgen: Consultancy, Membership on an entity's Board of Directors or advisory committees; Novartis: Consultancy, Membership on an entity's Board of Directors or advisory committees. Roboz:Cellectis: Research Funding; AbbVie, Agios, Amgen, Amphivena, Array Biopharma Inc., Astex, AstraZeneca, Celator, Celgene, Clovis Oncology, CTI BioPharma, Genoptix, Immune Pharmaceuticals, Janssen Pharmaceuticals, Juno, MedImmune, MEI Pharma, Novartis, Onconova, Pfizer, Roche Pharmace: Consultancy. Sekeres:Celgene: Membership on an entity's Board of Directors or advisory committees. Kantarjian:ARIAD: Research Funding; Pfizer: Research Funding; Delta-Fly Pharma: Research Funding; Bristol-Meyers Squibb: Research Funding; Amgen: Research Funding; Novartis: Research Funding.

A strategy for optimal fitting of multiplicative and additive hazards regression models

BackgroundIn survival analysis, data can be modeled using either a multiplicative hazards regression model (such as the Cox model) or an additive hazards regression model (such as Lin’s or Aalen’s model). While several diagnostic tools are available to check the assumptions underpinning each type of model, there is no defined procedure to fit these models optimally. Moreover, the two types of models are rarely combined in survival analysis. Here, we propose a strategy for optimal fitting of multiplicative and additive hazards regression models in survival analysis.MethodsThis section details our proposed strategy for optimal fitting of multiplicative and additive hazards regression models, with a focus on the assumptions underpinning each type of model, the diagnostic tools used to check these assumptions, and the steps followed to fit the data. The proposed strategy draws on classical diagnostic tools (Schoenfeld and martingale residuals) and less common tools (pseudo-observations, martingale residual processes, and Arjas plots).ResultsThe proposed strategy is applied to a dataset of patients with myocardial infarction (TRACE data frame). The effects of 5 covariates (age, sex, diabetes, ventricular fibrillation, and clinical heart failure) on the hazard of death are analyzed using multiplicative and additive hazards regression models. The proposed strategy is shown to fit the data optimally.ConclusionsSurvival analysis is improved by using multiplicative and additive hazards regression models together, but specific steps must be followed to fit the data optimally. By providing different measures of the same effect, our proposed strategy allows for better interpretation of the data.

A comparative study on the unified model based multifactor dimensionality reduction methods for identifying gene-gene interactions associated with the survival phenotype

BackgroundFor gene-gene interaction analysis, the multifactor dimensionality reduction (MDR) method has been widely employed to reduce multi-levels of gene-gene interactions into high- or low-risk groups using a binary attribute. For the survival phenotype, the Cox-MDR method has been proposed using a martingale residual of a Cox model since Surv-MDR was first proposed using a log-rank test statistic. Recently, the KM-MDR method was proposed using the Kaplan-Meier median survival time as a classifier. All three methods used the cross-validation procedure to identify single nucleotide polymorphism (SNP) using SNP interactions among all possible SNP pairs. Furthermore, these methods require the permutation test to verify the significance of the selected SNP pairs. However, the unified model-based multifactor dimensionality reduction method (UM-MDR) overcomes this shortcoming of MDR by unifying the significance testing with the MDR algorithm within the framework of the regression model. Neither cross-validation nor permutation testing is required to identify SNP by SNP interactions in the UM-MDR method. The UM-MDR method comprises two steps: in the first step, multi-level genotypes are classified into high- or low-risk groups, and an indicator variable for the high-risk group is defined. In the second step, the significance of the indicator variable of the high-risk group is tested in the regression model included with other adjusting covariates. The Cox-UMMDR method was recently proposed by combining Cox-MDR with UM-MDR to identify gene-gene interactions associated with the survival phenotype. In this study, we propose two simple methods either by combining KM-MDR with UM-MDR, called KM-UMMDR or by modifying Cox-UMMDR by adjusting for the covariate effect in step 1, rather than in step 2, a process called Cox2-UMMDR. The KM-UMMDR method allows the covariate effect to be adjusted for in the regression model of step 2, although KM-MDR cannot adjust for the covariate effect in the classification procedure of step 1. In contrast, Cox2-UMMDR differs from Cox-UMMDR in the sense that the martingale residuals are obtained from a Cox model by adjusting for the covariate effect in step 1 of Cox2-UMMDR whereas Cox-UMMDR adjusts for the covariate effect in the regression model in step 2. We performed simulation studies to compare the power of several methods such as KM-UMMDR, Cox-UMMDR, Cox2-UMMDR, Cox-MDR, and KM-MDR by considering the effect of covariates and the marginal effect of SNPs. We also analyzed a real example of Korean leukemia patient data for illustration and a short discussion is provided.ResultsIn the simulation study, two different scenarios are considered: the first scenario compares the power of the cases with and without the covariate effect. The second scenario is to compare the power of cases with the main effect of SNPs versus without the main effect of SNPs. From the simulation results, Cox-UMMDR performs the best across all scenarios among KM-UMMDR, Cox2-UMMDR, Cox-MDR and KM-MDR. As expected, both Cox-UMMDR and Cox-MDR perform better than KM-UMMDR and KM-MDR when a covariate effect exists because the former adjusts for the covariate effect but the latter cannot. However, Cox2-UMMDR behaves similarly to KM-UMMDR and KM-MDR even though there is a covariate effect. This implies that the covariate effect would be more efficiently adjusted for in the regression model of the second step rather than under the classification procedure of the first step. When there is a main effect of any SNP, Cox-UMMDR, Cox2-UMMDR and KM-UMMDR perform better than Cox-MDR and KM-MDR if the main effects of SNPs are properly adjusted for in the regression model. From the simulation results of two different scenarios, Cox-UMMDR seems to be the most robust when there is either any covariate effect adjusting for or any SNP that has a main effect on the survival phenotype. In addition, the power of all methods decreased as the censoring fraction increased from 0.1 to 0.3, as heritability increased. The power of all methods seems to be greater under MAF = 0.2 than under MAF = 0.4. For illustration, both KM-UMMDR and Cox2-UMMDR were applied to identify SNP by SNP interactions with the survival phenotype to a real dataset of Korean leukemia patients.ConclusionBoth KM-UMMDR and Cox2-UMMDR were easily implemented by combining KM-MDR and Cox-MDR with UM-MDR, respectively, to detect significant gene-gene interactions associated with survival time without cross-validation and permutation testing. The simulation results demonstrate the utility of KM-UMMDR, Cox2-UMMDR and Cox-UMMDR, which outperforms Cox-MDR and KM-MDR when some SNPs with only marginal effects might mask the detection of causal epistasis. In addition, Cox-UMMDR, Cox2-UMMDR and Cox-MDR performed better than KM-UMMDR and KM-MDR when there were potentially confounding covariate effects.

Timing of diagnosis of complex lower urinary tract injury in the 30-day postoperative period following benign hysterectomy

Complex lower urinary tract injury resulting from hysterectomy is a rare but highly morbid complication. Although intraoperative recognition reduces the risk of serious sequelae, observational studies have shown that most complex lower urinary tract injuries are recognized in the postoperative period. To date, limited research exists describing the timing of diagnosis of complex lower urinary tract injury or risk factors associated with complex lower urinary tract injury diagnosed in the postoperative period. This analysis aimed to describe the time to diagnosis of complex lower urinary tract injury among women undergoing benign hysterectomy. We also aimed to identify the intraoperative risk factors for differences in type and timing of complex lower urinary tract injury in the 30-day postoperative period using a large prospective national surgical database. This was a retrospective analysis using the National Surgical Quality Improvement Program hysterectomy data set from 2014 to 2018. All benign hysterectomies were included. Sociodemographic factors, health status, surgeon type, and other operative characteristics were extracted. A complex lower urinary tract injury was defined as at least 1 ureteral obstruction, ureteral fistula, or bladder fistula diagnosed within the first 30 days following surgery. Bivariate and multivariate logistic regression and cox proportional hazards assessed differences in odds of and time until diagnosis of complex lower urinary tract injury. Proportional hazard assumptions were evaluated with martingale residuals and supremum tests. Significance thresholds were 0.05 for all analyses. In this study, 100,823 women met the inclusion criteria. Median time to diagnosis of complex lower urinary tract injury was 10 days (interquartile range, 3-19) and varied significantly based on type of injury (P<.01) with ureteral obstruction (6; interquartile range, 2-16) recognized earlier than ureteral fistula (12; interquartile range, 7-21) and bladder fistula (14; interquartile range, 4-23). In addition, 8.65% of complex lower urinary tract injury were diagnosed on the day of surgery. Total laparoscopic hysterectomy had the lowest rate of complex lower urinary tract injury in unadjusted and adjusted analysis, with abdominal hysterectomy (adjusted odds ratio, 2.02; 95% confidence interval, 1.21-3.36) and vaginal hysterectomy (adjusted odds ratio, 2.05; 95% confidence interval, 1.16-3.62) having greater odds of ureteral obstruction, whereas laparoscopic assisted vaginal hysterectomy had the greatest odds of fistula (adjusted odds ratio, 2.10; 95% confidence interval, 1.26-3.48). Concomitant apical suspension was associated with a 6-day reduction in median time to diagnosis (P=.01), and surgery with a gynecologic oncologist was associated with a 9.5-day increase in median time to diagnosis (P=.01). Cox proportional hazards analysis confirmed these findings when controlling for confounders. Greater than 91% of complex lower urinary tract injury diagnoses in the National Surgical Quality Improvement Program hysterectomy database were diagnosed after the day of surgery. Route of hysterectomy, concomitant apical suspension, and primary surgeon specialty are associated with differences in both type of injury and time until diagnosis. These intraoperative risk factors should be considered when assessing for complex lower urinary tract injury in the 30-day postoperative period.

A new efficient method to detect genetic interactions for lung cancer GWAS

BackgroundGenome-wide association studies (GWAS) have proven successful in predicting genetic risk of disease using single-locus models; however, identifying single nucleotide polymorphism (SNP) interactions at the genome-wide scale is limited due to computational and statistical challenges. We addressed the computational burden encountered when detecting SNP interactions for survival analysis, such as age of disease-onset. To confront this problem, we developed a novel algorithm, called the Efficient Survival Multifactor Dimensionality Reduction (ES-MDR) method, which used Martingale Residuals as the outcome parameter to estimate survival outcomes, and implemented the Quantitative Multifactor Dimensionality Reduction method to identify significant interactions associated with age of disease-onset.MethodsTo demonstrate efficacy, we evaluated this method on two simulation data sets to estimate the type I error rate and power. Simulations showed that ES-MDR identified interactions using less computational workload and allowed for adjustment of covariates. We applied ES-MDR on the OncoArray-TRICL Consortium data with 14,935 cases and 12,787 controls for lung cancer (SNPs = 108,254) to search over all two-way interactions to identify genetic interactions associated with lung cancer age-of-onset. We tested the best model in an independent data set from the OncoArray-TRICL data.ResultsOur experiment on the OncoArray-TRICL data identified many one-way and two-way models with a single-base deletion in the noncoding region of BRCA1 (HR 1.24, P = 3.15 × 10–15), as the top marker to predict age of lung cancer onset.ConclusionsFrom the results of our extensive simulations and analysis of a large GWAS study, we demonstrated that our method is an efficient algorithm that identified genetic interactions to include in our models to predict survival outcomes.

A new linear regression-like residual for survival analysis, with application to genome wide association studies of time-to-event data.

In linear regression, a residual measures how far a subject's observation is from expectation; in survival analysis, a subject's Martingale or deviance residual is sometimes interpreted similarly. Here we consider ways in which a linear regression-like interpretation is not appropriate for Martingale and deviance residuals, and we develop a novel time-to-event residual which does have a linear regression-like interpretation. We illustrate the utility of this new residual via simulation of a time-to-event genome-wide association study, motivated by a real study seeking genetic modifiers of Duchenne Muscular Dystrophy. By virtue of its linear regression-like characteristics, our new residual may prove useful in other contexts as well.

Does facility volume influence survival in patients with primary malignant bone tumors of the vertebral column? A comparative cohort study

Facility volume has been correlated with survival in many cancers. This relationship has not been established in primary malignant bone tumors of the vertebral column (BTVC). To investigate whether facility patient volume is associated with overall survival in patients with primary malignant BTVCs. Retrospective comparative cohort. Adult patients with chordomas, chondrosarcomas, or osteosarcomas of the mobile spine. Five-year survival. We retrospectively analyzed 733 patients with primary malignant BTVCs in the national cancer database from 2004 through 2015. Univariate and multivariate analyses were used to correlate specific outcome measures with facility volume. Volume was stratified based on cumulative martingale residuals to determine the inflection point of negative to positive impact on survival based on the patient cohort. Long-term survival was compared between patients treated at high and low volume using the Kaplan-Meier method. Only patients with malignant primary tumors were considered eligible for inclusion; patients with incomplete treatment data or benign tumors were excluded. Patients treated at high-volume centers (HVCs) were younger (p=.0003) and more likely to be insured (p<.0001). There were no significant differences in tumor characteristics. Patients treated at high-volume facilities had improved 5-year survival of 71% versus 58% at low-volume centers (p<.0001). Patients treated at HVCs were more likely to receive surgical treatment (91% vs. 80%, p<.0001); if surgery was performed, they were more likely to undergo an en bloc resection (48% vs. 30%, p<.0001). However, there were no differences in margin status or utilization of radiotherapy or chemotherapy between HVCs and low-volume centers. In a multivariate analysis, facility volume was independently associated with improved survival overall (HR 0.75 [0.58-0.97], p=.03). Primary malignant BTVCs are rare, even for HVCs. Despite this, patient survival was significantly improved when treatment was performed at HVCs.

Clustering of Largely Right-Censored Oropharyngeal Head and Neck Cancer Patients for Discriminative Groupings to Improve Outcome Prediction

Clustering is the task of identifying groups of similar subjects according to certain criteria. The AJCC staging system can be thought as a clustering mechanism that groups patients based on their disease stage. This grouping drives prognosis and influences treatment. The goal of this work is to evaluate the efficacy of machine learning algorithms to cluster the patients into discriminative groups to improve prognosis for overall survival (OS) and relapse free survival (RFS) outcomes. We apply clustering over a retrospectively collected data from 644 head and neck cancer patients including both clinical and radiomic features. In order to incorporate outcome information into the clustering process and deal with the large proportion of censored samples, the feature space was scaled using the regression coefficients fitted using a proxy dependent variable, martingale residuals, instead of follow-up time. Two clusters were identified and evaluated using cross validation. The Kaplan Meier (KM) curves between the two clusters differ significantly for OS and RFS (p-value < 0.0001). Moreover, there was a relative predictive improvement when using the cluster label in addition to the clinical features compared to using only clinical features where AUC increased by 5.7% and 13.0% for OS and RFS, respectively.

Marginal variable screening for survival endpoints.

When performing survival analysis in very high dimensions, it is often required to reduce the number of covariates using preliminary screening. During the last years, a large number of variable screening methods for the survival context have been developed. However, guidance is missing for choosing an appropriate method in practice. The aim of this work is to provide an overview of marginal variable screening methods for survival and develop recommendations for their use. For this purpose, a literature review is given, offering a comprehensive and structured introduction to the topic. In addition, a novel screening procedure based on distance correlation and martingale residuals is proposed, which is particularly useful in detecting nonmonotone associations. For evaluating the performance of the discussed approaches, a simulation study is conducted, comparing the true positive rates of competing variable screening methods in different settings. A real data example on mantle cell lymphoma isprovided.

Quantifying the relationship between age at diagnosis and breast cancer-specific mortality.

The relationship between age at diagnosis and breast cancer-specific mortality (BCSM) is unclear. The aim of this study was to examine the nature of this relationship using rigorous statistical methodology. A historical cohort study of adult women with invasive breast cancer in the SEER database from 2000 to 2015 was conducted. Multivariable Cox's cause-specific hazards model was used to evaluate the association of age at diagnosis with risk of BCSM. Functional relationship of age was assessed using cumulative sums of Martingale residuals and the Kolmogorov-type supremum test. A total of 206,332 women were eligible for study. Mean age at diagnosis was 59.7 ± 13.8years. Median follow-up was 80months. During the study period, 21,771 women (10.6%) died from breast cancer and 18,566 (9.0%) died from other causes. Cumulative incidence of BCSM at 120months post-diagnosis was 14.4% (95% CI 14.2-14.6%). Age was found to be quadratically related to the risk of BCSM (p < 0.001), with a nadir at 45years of age. The final Cox model suggests that a 30-year-old woman has approximately the same adjusted BCSM risk (HR 1.187, 95% CI 1.187-1.188) as a 60-year-old woman (HR 1.174, 95% CI 1.174-1.175). Women diagnosed with breast cancer at the extremes of age suffer disproportionate rates of cancer-specific mortality. The relationship between age at diagnosis and adjusted risk of BCSM is complex, consistent with a quadratic function. With the growing appreciation for breast cancer as a heterogeneous disease, it is essential to accurately address age as a prognostic risk factor in predictive models.

A goodness-of-fit test for the stratified proportional hazards model for survival data

Goodness-of-fit testing is addressed in the stratified proportional hazards model for survival data. A test statistic based on within-strata cumulative sums of martingale residuals over covariates is proposed and its asymptotic distribution is derived under the null hypothesis of model adequacy. A Monte Carlo procedure is proposed to approximate the critical value of the test. Simulation studies are conducted to examine finite-sample performance of the proposed statistic.

Validation of lymphocyte to monocyte ratio in resected small bowel adenocarcinoma as a predictor of survival.

e15799 Background: Lymphocyte to monocyte ratio (LMR) has been described as a prognostic factor in many solid tumors including colorectal adenocarcinoma. LMR has not been investigated as a prognostic factor in small bowel cancers. In this study, we aimed to evaluate prognostic factors in resected small bowel adenocarcinoma including LMR. Methods: Two hundred forty-one patients who underwent resection for stage I-III small bowel adenocarcinoma were retrospectively identified utilizing the pathology database at a single tertiary referral institution from 1994 to 2015. All patients had complete follow up data and were included in the survival analysis. One hundred sixty-nine patients had preoperative peripheral blood counts available for analysis. Plot of martingale residuals against LMR were used to establish best cutoff points for LMR. A training set for LMR included consecutively identified patients from 2006 to 2015, and a validation cohort including patients identified from 1994 to 2005 was used. Overall survival was performed utilizing Kaplan-Meier method, and Wilcoxon tests were used for statistical comparisons. Cox proportional hazards were performed and all tests were two sided. P value of &lt; 0.05 was considered significant. Results: Median overall survival for the entire group was 54.5 months (95% CI: 37.2-81.2 months) with 5- and 10-year overall survival of 48% and 35%. The training set for LMR included 81 patients, and the validation set included 88 patients. The cutoff of 1.56 was chosen based on most significant p value (p = 0.002). When combined, the overall area under the curve (AUC) for LMR was 0.63, p &lt; 0.01, (specificity 37.3%, sensitivity 90.1%, positive predictive value 33.1%, and negative predictive value 92.2%). There were 126 patients with LMR &gt; 1.56 and 43 patients with LMR &lt; 1.56 in the entire cohort. In multivariate analysis, LMR under 1.56 was a negative prognostic factor, HR = 2.20 (95% CI: 1.27-3.84, p &lt; 0.01). In addition to LMR, age &gt; 60 years and advanced T stage were independently negative predictors of overall survival in all patients. Conclusions: Lymphocyte-to-monocyte ratio &lt; 1.56 is a validated negative prognostic factor in resected small bowel adenocarcinoma.

Neutrophil-to-lymphocyte ratio as a prognostic marker in patients with metastatic gallbladder cancer.

442 Background: Neutrophil to lymphocyte ratio (NLR) has been used as an inflammation-based prognostic marker for various malignancies. The aim of our study was to determine whether NLR can independently predict the overall survival in patients with metastatic gallbladder cancer (GBC). Methods: We identified patients diagnosed with GBC who were treated at Mayo Clinic between the years 2000 and 2016. Patients who had nonmetastatic GBC were excluded along with the patients who did not have data for neutrophils and lymphocytes. Optimal cutoff point for NLR was identified by plotting martingale residuals against NLR and patients were divided into two groups, ≥ 5 or &lt; 5. Demographic, follow-up data and outcomes were collected by retrospective review of electronic medical records. Fisher’s exact test was used to compare categorical variables, while The Mann- Whitney U test was used to compare continuous variables. Kaplan-Meier curves were plotted for NLR ≥ 5 and NLR &lt; 5 and overall survival (OS) between the two groups were compared using log rank test. Multivariate survival analysis was performed using Cox-proportional hazard regression. Results: A total of 231 patients met our inclusion criteria, of which, 138 (60%) had NLR &lt; 5 and 93 (40%) had NLR ≥ 5. Patients with NLR ≥ 5 were more likely to be older and have poor performance score, lower albumin level, higher alkaline phosphatase level, and higher platelet count. There were no significant differences noted in gender, race and administration of chemotherapy between the two groups. In univariate analysis, patients with NLR ≥ 5 at presentation had a significantly worse OS compared to those with NLR &lt; 5 (Median survival: 3.6 vs. 8.8 months, p &lt; 0.001). In multivariate analysis, adjusting for age, ECOG status, albumin, ALP, AST, ALT, bilirubin, platelet count and administration of chemotherapy, NLR of ≥ 5 was associated with a worse OS compared to NLR &lt; 5 (HR: 1.70, 95% CI:1.20 – 2.39. p &lt; 0.05). Conclusions: Our study demonstrates that NLR ≥ 5 is an independent predictor of poor prognosis in patients with metastatic gallbladder cancer.

Evaluating the Effect of Right-Censored End Point Transformation for Radiomic Feature Selection of Data From Patients With Oropharyngeal Cancer

To evaluate the effect of transforming a right-censored outcome into binary, continuous, and censored-aware representations on radiomics feature selection and subsequent prediction of overall survival (OS) and relapse-free survival (RFS) of patients with oropharyngeal cancer. Different feature selection techniques were applied using a binary outcome indicating event occurrence before median follow-up time, a continuous outcome using the Martingale residuals from a proportional hazards model, and the raw right-censored time-to-event outcome. Radiomic signatures combined with clinical variables were used for risk prediction. Three metrics for accuracy and calibration were used to evaluate eight feature selectors and six predictive models. Feature selection across 529 patients on more than 3,800 radiomic features resulted in increases ranging from 0.01 to 0.11 in C-index and area under the curve (AUC) scores compared with clinical features alone. The ensemble model yielded the best scores for AUC and C-index (often > 0.7) and calibration (Nam-D'Agostino test statistic often < 15.5 with 8 df). The random forest feature selectors achieved the best performance considering all metrics. Random regression forest performed the best in OS prediction with the ensemble model (AUC, 0.75; C-index, 0.76; calibration, 8.7). Random survival forest performed the best in RFS prediction with the ensemble model (AUC, 0.71; C-index, 0.68; calibration, 19.1). Including a radiomic signature results in better prediction than using only clinical data. Signatures generated randomly or without considering the outcome result in poor calibration scores. The random forest feature selectors for each of the three transformations typically selected the greatest number of features and produced the best predictions at acceptable calibration levels. In particular, random regression forest and random survival forest performed best for OS and RFS, respectively.

Unified Cox model based multifactor dimensionality reduction method for gene-gene interaction analysis of the survival phenotype

BackgroundOne strategy for addressing missing heritability in genome-wide association study is gene-gene interaction analysis, which, unlike a single gene approach, involves high-dimensionality. The multifactor dimensionality reduction method (MDR) has been widely applied to reduce multi-levels of genotypes into high or low risk groups. The Cox-MDR method has been proposed to detect gene-gene interactions associated with the survival phenotype by using the martingale residuals from a Cox model. However, this method requires a cross-validation procedure to find the best SNP pair among all possible pairs and the permutation procedure should be followed for the significance of gene-gene interactions. Recently, the unified model based multifactor dimensionality reduction method (UM-MDR) has been proposed to unify the significance testing with the MDR algorithm within the regression model framework, in which neither cross-validation nor permutation testing are needed. In this paper, we proposed a simple approach, called Cox UM-MDR, which combines Cox-MDR with the key procedure of UM-MDR to identify gene-gene interactions associated with the survival phenotype.ResultsThe simulation study was performed to compare Cox UM-MDR with Cox-MDR with and without the marginal effects of SNPs. We found that Cox UM-MDR has similar power to Cox-MDR without marginal effects, whereas it outperforms Cox-MDR with marginal effects and more robust to heavy censoring. We also applied Cox UM-MDR to a dataset of leukemia patients and detected gene-gene interactions with regard to the survival time.ConclusionCox UM-MDR is easily implemented by combining Cox-MDR with UM-MDR to detect the significant gene-gene interactions associated with the survival time without cross-validation and permutation testing. The simulation results are shown to demonstrate the utility of the proposed method, which achieves at least the same power as Cox-MDR in most scenarios, and outperforms Cox-MDR when some SNPs having only marginal effects might mask the detection of the causal epistasis.

The number of resected lymph nodes is associated with the long-term survival outcome in patients with T2 N0 non-small cell lung cancer.

ObjectiveFor the patients with pathologic T2 N0 non-small cell lung cancer (NSCLC), the extent of lymph node (LN) removal required for survival is controversial. We aimed to explore the prognostic significance of examined LNs and to identify how many nodes should be examined.MethodsWe reviewed 549 patients who underwent pulmonary or pneumonectomy surgery or plus lymphadenectomy who were confirmed as T2 stage and LN negative by postoperative pathological diagnosis. According to Martingale residuals of the Cox model, the patients were classified into four groups by the number of examined LNs (1–2 LNs, 3–7 LNs, 8–11 LNs, and ≥12 LNs). Kaplan–Meier analysis and Cox regression analysis were used to evaluate the association between survival and the number of examined LNs.ResultCompared with the 1–2 LNs, 3–7 LNs, and 8–11 LNs groups, the survival was significantly better in the ≥12 LNs group. The 5-year cancer-specific survival rate was 60.5% for patients with 1–2 negative LNs, compared with 68.7%, 72.6%, and 78.4% for those with 3–7, 8–11, and >11 LNs examined, respectively. The 7-year cancer-specific survival rate was 52.9% for patients with 1–2 negative LNs, compared with 63.7%, 63.8%, and 70.8% for those with 3–7, 8–11, and >11 LNs examined, respectively (P=0.045). There was a significant drop in mortality risk with the examination of more LNs. The lowest mortality risk occurred in those with 32 or more LNs examined. Multivariate analysis showed that age and the number of examined LNs were strong independent predictors of survival.ConclusionThe number of examined LNs is a strong independent prognostic factor. Our study demonstrates that patients with T2 N0 NSCLC should have at least 12 LNs examined and that the results of this study may provide information for the optimal number of resected LNs in surgery.

A New Log-location Regression Model with Influence Diagnostics and Residual Analysis

A new four-parameter lifetime model called OddLog-Logistic Burr XII distribution, is defined and investigated. Some of itsmathematical properties are derived. Some useful characterization resultsbased on \ the ratio of two truncated moments, based on the hazard functionas well as on the conditional expectation of certain functions of the randomvariable are presented. The maximum likelihood method is used to estimatethe model parameters by means of a graphical Monte Carlo simulation study.Moreover, we introduce a new log-location regression model based on theproposed distribution. The Jackknife estimation method as an alternativemethod is used to estimate the unknown parameters of new regression model. Thegeneralized cook distance and likelihood distance measures are used todetect the possible influential observations. The martingale and modifieddeviance residuals are defined to detect outliers and evaluate the modelassumptions. The potentiality of the new regression model is illustrated bymeans of a real data set.

Skeletal Muscle Attenuation (Sarcopenia) Predicts Reduced Overall Survival in Patients with Advanced Epithelial Ovarian Cancer Undergoing Primary Debulking Surgery.

Sarcopenia was reported as a prognostic factor in cancer patients. Using computed tomography (CT), we analyzed the impact of sarcopenia on overall survival (OS) in patients with advanced epithelial ovarian cancer (EOC) after primary debulking surgery (PDS). Preoperative CT scans of consecutive EOC patients (n = 323) were retrospectively assessed for skeletal muscle index (SMI) and muscle attenuation (MA; Hounsfield units [HU]). The optimal cut-off point for MA (32HU) was calculated using the Martingale residuals method, and previously reported cut-offs for SMI were used. Logistic regression was used to determine univariate and multivariate factors associated with OS. Sarcopenia defined as SMI < 38.5, < 39, and 41cm2/m2 was detected in 29.4, 33.7, and 47.1% of patients, respectively; however, none of these SMI cut-off levels were associated with OS. MA < 32HU was present in 21.1% (68/323) of the total cohort. Significant differences between patients with MA < 32 and ≥ 32HU were detected for median age (67 vs. 57years), Eastern Cooperative Oncology Group (ECOG) > 0 (13.2 vs. 3.1%), comorbidity (age-adjusted Charlson Comorbidity Index [ACCI] ≥ 4; 36.8 vs. 13.3%), median body mass index (BMI; 27 vs. 24kg/m2), International Federation of Gynecology and Obstetrics (FIGO) stage, histology (high-grade serous 95.6 vs. 84.7%), and complete resection (38.2 vs. 68.2%). MA < 32HU remained a significant prognostic factor for OS in multivariate Cox regression analysis (hazard ratio 1.79, p = 0.003). Median OS in patients with MA < 32HU versus MA ≥ 32HU was 28 versus 56months (p < 0.001). Furthermore, MA < 32HU was significantly associated with OS in the prognostically poor population of patients with residual tumor (p = 0.015). Low MA was significantly associated with poor survival, especially in patients with residual tumor after PDS. MA assessment could be used for risk stratification after PDS.

Asthma is not associated with the need for surgery in Crohn's disease when controlling for smoking status: a population-based cohort study.

PurposeGrowing evidence suggests asthma and Crohn’s disease commonly cooccur. However, the impact of asthma on the prognosis of Crohn’s disease is unknown. The aim of our study was to assess the effect of asthma on the need for intestinal resection in patients with Crohn’s disease while adjusting for smoking status, imputed from a smaller, secondary data set.Patients and methodsUsing health administrative data from a universally funded healthcare plan in Alberta, Canada, we conducted a cohort study to assess the effect of asthma on the need for surgery in patients with Crohn’s disease diagnosed between 2002 and 2008 (N=2,113). Validated algorithms were used to identify incident cases of Crohn’s disease, cooccurring asthma, and intestinal resection. The association between asthma and intestinal resection was estimated using multivariable Cox proportional hazards regression. Smoking status was imputed using a novel method using martingale residuals, derived from a data set of 485 patients enrolled in the Alberta Inflammatory Bowel Disease Consortium (2007 to 2014) who completed environmental questionnaires. All analyses were adjusted for age, sex, rural/urban status, and mean neighborhood income quintile.ResultsAsthma did not increase the risk of surgery in the health administrative data when not adjusting for smoking status (HR 1.03, 95% CI 0.81 to 1.29). The association remained nonsignificant after imputing smoking status in the health administrative data (HR 1.03, 95% CI 0.81 to 1.29).ConclusionAlthough asthma is associated with an increased risk of Crohn’s disease, co-occurring asthma is not associated with the risk of surgery in these patients. This null association persisted after adjusting for smoking status. This study described a novel method to adjust for confounding (smoking status) in time-to-event analyses, even when the confounding variable is unmeasured in health administrative data.

Testing Calibration of Cox Survival Models at Extremes of Event Risk

Risk prediction models can translate genetic association findings for clinical decision-making. Most models are evaluated on their ability to discriminate, and the calibration of risk-prediction models is largely overlooked in applications. Models that demonstrate good discrimination in training datasets, if not properly calibrated to produce unbiased estimates of risk, can perform poorly in new patient populations. Poorly calibrated models arise due to missing covariates, such as genetic interactions that may be unknown or not measured. We demonstrate that models omitting interactions can lead to increased bias in predicted risk for patients at the tails of the risk distribution; i.e., those patients who are most likely to be affected by clinical decision making. We propose a new calibration test for Cox risk-prediction models that aggregates martingale residuals for subjects from extreme high and low risk groups with a test statistic maximum chosen by varying which risk groups are included in the extremes. To estimate the empirical significance of our test statistic, we simulate from a Gaussian distribution using the covariance matrix for the grouped sums of martingale residuals. Simulation shows the new test maintains control of type 1 error with improved power over a conventional goodness-of-fit test when risk prediction deviates at the tails of the risk distribution. We apply our method in the development of a prediction model for risk of cystic fibrosis-related diabetes. Our study highlights the importance of assessing calibration and discrimination in predictive modeling, and provides a complementary tool in the assessment of risk model calibration.