Comparing tumour staging and grading systems: a case study and a review of the issues, using thymoma as a model

Pan-cancer analysis implicates novel insights of lactate metabolism into immunotherapy response prediction and survival prognostication

BACKGROUND

Immunotherapy has emerged as a potent clinical approach for cancer treatment, but only subsets of cancer patients can benefit from it. Targeting lactate metabolism (LM) in tumor cells as a method to potentiate anti-tumor immune responses represents a promising therapeutic strategy.

METHODS

Public single-cell RNA-Seq (scRNA-seq) cohorts collected from patients who received immunotherapy were systematically gathered and scrutinized to delineate the association between LM and the immunotherapy response. A novel LM-related signature (LM.SIG) was formulated through an extensive examination of 40 pan-cancer scRNA-seq cohorts. Then, multiple machine learning (ML) algorithms were employed to validate the capacity of LM.SIG for immunotherapy response prediction and survival prognostication based on 8 immunotherapy transcriptomic cohorts and 30 The Cancer Genome Atlas (TCGA) pan-cancer datasets. Moreover, potential targets for immunotherapy were identified based on 17 CRISPR datasets and validated via in vivo and in vitro experiments.

RESULTS

The assessment of LM was confirmed to possess a substantial relationship with immunotherapy resistance in 2 immunotherapy scRNA-seq cohorts. Based on large-scale pan-cancer data, there exists a notably adverse correlation between LM.SIG and anti-tumor immunity as well as imbalance infiltration of immune cells, whereas a positive association was observed between LM.SIG and pro-tumorigenic signaling. Utilizing this signature, the ML model predicted immunotherapy response and prognosis with an AUC of 0.73/0.80 in validation sets and 0.70/0.87 in testing sets respectively. Notably, LM.SIG exhibited superior predictive performance across various cancers compared to published signatures. Subsequently, CRISPR screening identified LDHA as a pan-cancer biomarker for estimating immunotherapy response and survival probability which was further validated using immunohistochemistry (IHC) and spatial transcriptomics (ST) datasets. Furthermore, experiments demonstrated that LDHA deficiency in pancreatic cancer elevated the CD8+ T cell antitumor immunity and improved macrophage antitumoral polarization, which in turn enhanced the efficacy of immunotherapy.

CONCLUSIONS

We unveiled the tight correlation between LM and resistance to immunotherapy and further established the pan-cancer LM.SIG, holds the potential to emerge as a competitive instrument for the selection of patients suitable for immunotherapy.

Evaluating Dynamic Discrimination Performance of Risk Prediction Models for Survival Outcomes

Risk prediction models for survival outcomes are widely applied in medical research to predict future risk for the occurrence of the event. In many clinical studies, the biomarker data are measured repeatedly over time. To facilitate timely disease prognosis and decision making, many dynamic prediction models have been developed and generate predictions on a real-time basis. As a dynamic prediction model updates an individual's risk prediction over time based on new measurements, it is often important to examine how well the model performs at different measurement times and prediction times. In this article, we propose a two-dimensional area under curve (AUC) measure for dynamic prediction models and develop associated estimation and inference procedures. The estimation procedures are discussed under two types of biomarker measurement schedules: regular visits and irregular visits. The model parameters are estimated effectively by maximizing a pseudo-partial likelihood function. We apply the proposed method to a renal transplantation study to evaluate the discrimination performance of dynamic prediction models based on longitudinal biomarkers for graft failure.

Discovery of a biomarker for β-Thalassemia by HPLC-MS and improvement from Proton Transfer Reaction - Parallel Ion Parking

β-thalassemia is a quantitative hemoglobin (Hb) disorder resulting in reduced production of Hb A and increased levels of Hb A₂. Diagnosis of β-thalassemia can be problematic when combined with other structural Hb variants, so that the separation approaches in routine clinical centers are not sufficiently decisive to obtain accurate results. Here, we separate the intact Hb subunits by high-performance liquid chromatography, followed by top-down tandem mass spectrometry of intact subunits to distinguish Hb variants. Proton transfer reaction-parallel ion parking (PTR-PIP), in which a radical anion removes protons from multiply charged precursor ions and produces charge-reduced ions spanning a limited m/z range, was used to increase the signal-to-noise ratio of the subunits of interest. We demonstrate that the δ/β ratio can act as a biomarker to identify β-thalassemia in normal electrospray ionization MS1 and PTR-PIP MS1. The application of PTR-PIP significantly increases the sensitivity and specificity of the HPLC-MS method to identify δ/β ratio as a thalassemia biomarker.

SurvBenchmark: comprehensive benchmarking study of survival analysis methods using both omics data and clinical data

Survival analysis is a branch of statistics that deals with both the tracking of time and the survival status simultaneously as the dependent response. Current comparisons of survival model performance mostly center on clinical data with classic statistical survival models, with prediction accuracy often serving as the sole metric of model performance. Moreover, survival analysis approaches for censored omics data have not been thoroughly investigated. The common approach is to binarize the survival time and perform a classification analysis.

Here, we develop a benchmarking design, SurvBenchmark, that evaluates a diverse collection of survival models for both clinical and omics data sets. SurvBenchmark not only focuses on classical approaches such as the Cox model but also evaluates state-of-the-art machine learning survival models. All approaches were assessed using multiple performance metrics; these include model predictability, stability, flexibility, and computational issues. Our systematic comparison design with 320 comparisons (20 methods over 16 data sets) shows that the performances of survival models vary in practice over real-world data sets and over the choice of the evaluation metric. In particular, we highlight that using multiple performance metrics is critical in providing a balanced assessment of various models. The results in our study will provide practical guidelines for translational scientists and clinicians, as well as define possible areas of investigation in both survival technique and benchmarking strategies.

Estimation and inference of predictive discrimination for survival outcome risk prediction models

Accurate risk prediction has been the central goal in many studies of survival outcomes. In the presence of multiple risk factors, a censored regression model can be employed to estimate a risk prediction rule. Before the prediction tool can be popularized for practical use, it is crucial to rigorously assess its prediction performance. In our motivating example, researchers are interested in developing and validating a risk prediction tool to identify future lung cancer cases by integrating demographic information, disease characteristics and smoking-related data. Considering the long latency period of cancer, it is desirable for a prediction tool to achieve discriminative performance that does not weaken over time. We propose estimation and inferential procedures to comprehensively assess both the overall predictive discrimination and the temporal pattern of an estimated prediction rule. The proposed methods readily accommodate commonly used censored regression models, including the Cox proportional hazards model and the accelerated failure time model. The estimators are consistent and asymptotically normal, and reliable variance estimators are also developed. The proposed methods offer an informative tool for inferring time-dependent predictive discrimination, as well as for comparing the discrimination performance between candidate models. Applications of the proposed methods demonstrate enduring performance of the risk prediction tool in the PLCO study and detected decaying performance in a study of liver disease.

An imaging signature to predict outcome in metastatic colorectal cancer using routine computed tomography scans

BACKGROUND & AIMS

Quantitative analysis of computed tomography (CT) scans of patients with metastatic colorectal cancer (mCRC) can identify imaging signatures that predict overall survival (OS).

METHODS

We retrospectively analysed CT images from 1584 mCRC patients on two phase III trials evaluating FOLFOX ± panitumumab (n = 331, 350) and FOLFIRI ± aflibercept (n = 437, 466). In the training set (n = 720), an algorithm was trained to predict OS landmarked from month 2; the output was a signature value on a scale from 0 to 1 (most to least favourable predicted OS). In the validation set (n = 864), hazard ratios (HRs) evaluated the association of the signature with OS using RECIST1.1 as a benchmark of comparison.

RESULTS

In the training set, the selected signature combined three features - change in tumour volume, change in tumour spatial heterogeneity, and tumour volume - to predict OS. In the validation set, RECIST1.1 classified patients in three categories: response (n = 166, 19.2%), stable disease (n = 636, 73.6%), and progression (n = 62, 7.2%). The HR was 3.93 (2.79-5.54). Using the same distribution for the signature, the HR was 21.04 (14.88-30.58), showing an incremental prognostic separation. Stable disease by RECIST1.1 was reclassified by the signature along a continuum where patients belonging to the most and least favourable signature quartiles had a median OS of 40.73 (28.49 to NA) months (n = 94) and 7.03 (5.66-7.89) months (n = 166), respectively.

CONCLUSIONS

A signature combining three imaging features provides early prognostic information that can improve treatment decisions for individual patients and clinical trial analyses.

On the time-varying predictive performance of longitudinal biomarkers: Measure and estimation

In many biomedical studies, participants are monitored at periodic visits until the occurrence of the failure event. Biomarkers are often measured repeatedly during these visits, and such measurements can facilitate updated disease prediction. In this work, we propose a two-dimensional incident dynamic area under curve (AUC), to capture the variability due to both the biomarker assessment time and the prediction time to comprehensively quantify the predictive performance of a longitudinal biomarker. We propose a pseudo partial-likelihood to achieve consistent estimation of the AUC under two realistic scenarios of visit schedules. Variance estimation methods are designed to facilitate inferential procedures. We examine the finite-sample performance of our method through extensive simulations. The methods are applied to a study of chronic myeloid leukemia to evaluate the predictive performance of longitudinally collected gene expression levels.

Investigating the appropriateness of different concordance measures in a time-to-event setting

PURPOSE

Prediction models that assess a patient's risk of an event are used to inform treatment options and confirm screening tests. The concordance (c) statistic is one measure to validate the accuracy of these models, but has many extensions when applied to censored data. The purpose was to determine which c-statistic is most accurate at different rates of censoring.

METHODS

A simulation study was conducted for n = 750, and censoring rates of 20%, 50%, and 80%. The mean of three different concordance definitions were compared as well as the mean of three different c-statistics, including one, parametric c-statistic for exponentially distributed data, developed by the authors. The SE was also calculated but was of secondary interest.

RESULTS

The c-statistic developed by the authors yielded the a mean closest to the gold standard concordance measure when censoring is present in data, even when the exponentially distributed parametric assumptions do not hold. Similar results were found for SE.

CONCLUSIONS

The c-statistic developed by the authors appears to be the most robust to censored data. Thus, it is recommended to use this c-statistic to validate prediction models applied to censored data. This will improve the reliability and comparability across future time-to-event studies.

Re-excision or "wait and watch"-a prediction model in breast phyllodes tumors after surgery

Background

The prognosis of breast phyllodes tumors (PTs) largely depending on the pathological grading, which lacks objectivity. This study aimed to develop a nomogram based on clinicopathological features to evaluate the recurrence probability of PTs following surgery.

Methods

Data from 334 patients with breast PTs, who underwent surgical treatment at Sun Yat-sen Memorial Hospital from January 2005 to December 2014, were used to develop a prediction model. Additionally, data of 36 patients from Peking University Shenzhen Hospital (cohort 1) and data of 140 patients from Sun Yat-sen University Cancer Center (cohort 2) during the same period were used to validate the model. The medical records and tumor slides were retrospectively reviewed. The log-rank and Cox regression tests were used to develop a clinical prediction model of breast PTs. All statistical analyses were performed using R and STATA.

Results

Of all 334 patients included in the primary cohort, 224 had benign, 91 had borderline, and 19 had malignant tumors. The 1-, 3-, and 5-year recurrence-free survival was 98.5%, 97.9%, and 96.8%, respectively. Ultrasound-guided vacuum-assisted biopsy (UGVAB) is a non-inferior treatment application in benign PTs compared with open surgery [hazard ratio (HR), 2.38; 95% confidence interval (CI), 0.59–9.58]. Width of surgical margin, mitoses, and tumor border were identified as independent risk factors for breast PTs. A nomogram was developed based on these three variables. The C-index of internal and external validation was 0.71, 0.67 (cohort 1) and 0.73 (cohort 2), respectively.

Conclusions

The study model presented more concise and objective variables to evaluate the recurrence-free survival of patients after surgery, which can help deciding whether to do a re-excision or “wait and watch”.

Assessing predictive accuracy of survival regressions subject to nonindependent censoring

Survival regression is commonly applied in biomedical studies or clinical trials, and evaluating their predictive performance plays an essential role for model diagnosis and selection. The presence of censored data, particularly if informative, may pose more challenges for the assessment of predictive accuracy. Existing literature mainly focuses on prediction for survival probabilities with limitation work for survival time. In this work, we focus on accuracy measures of predicted survival times adjusted for a potentially informative censoring mechanism (ie, coarsening at random (CAR); non-CAR) by adopting the technique of inverse probability of censoring weighting. Our proposed predictive metric can be adaptive to various survival regression frameworks including but not limited to accelerated failure time models and proportional hazards models. Moreover, we provide the asymptotic properties of the inverse probability of censoring weighting estimators under CAR. We consider the settings of high-dimensional data under CAR or non-CAR for extensions. The performance of the proposed method is evaluated through extensive simulation studies and analysis of real data from the Critical Assessment of Microarray Data Analysis.

Clinical Validation of the Prognostic Stage Groups of the Eighth-Edition TNM Staging for Medullary Thyroid Carcinoma

CONTEXT

Despite advances in thyroid cancer staging systems, considerable controversy about the current staging system for medullary thyroid carcinoma (MTC) continues.

OBJECTIVE

We aimed to evaluate the prognostic performance of the current eighth edition of the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control TNM staging system (TNM-8) and the alternative proposed prognostic stage groups based on recursive partitioning analysis (TNM-RPA).

DESIGN, SETTING, AND PATIENTS

We retrospectively analyzed 182 patients with MTC treated at a single tertiary Korean hospital between 1995 and 2015.

INTERVENTIONS AND MAIN OUTCOME MEASURES

Survival analysis was conducted according to TNM-8 and TNM-RPA. The area under the receiver-operating characteristic curve (AUC), the proportion of variation explained (PVE), and the Harrell concordance index (C-index) were used to evaluate predictive performance.

RESULTS

Under TNM-8, only two (1.1%) patients were downstaged compared with the seventh edition of the AJCC TNM staging system (TNM-7). The AUC at 10 years, PVE, and C-index were 0.679, 8.7%, and 0.744 for TNM-7 and 0.681, 8.9%, and 0.747 for TNM-8, respectively. Under TNM-RPA, 104 (57.14%) patients were downstaged compared with TNM-8. TNM-RPA had better prognostic performance with respect to cancer-specific survival (AUC at 10 years, 0.750; PVE, 20.9%; C-index, 0.881).

CONCLUSIONS

The predictive performance of the revised TNM-8 in patients with MTC has not changed despite its modification from TNM-7. The proposed changes in TNM-RPA were statistically valid and may present a more reproducible system that better estimates cancer-specific survival of individual patients.

A Modified Prognostic Score for Patients with Curatively Resected Gastric Cancer

Aims and background

Gastric cancer is the second leading cause of cancer death worldwide; the risk of dying depends on several patient and disease characteristics. An existing prognostic score predicts survival in gastric cancer patients undergoing curative resection based on patient age, tumor site, extent of wall invasion and nodal status, categorized as simply as negative or positive.

Methods

Our aim was to modify the original prognostic score by incorporating information on nodal stage according to the latest TNM classification (number of involved nodes), based on a retrospective series of 610 chemotherapy-naïve gastric cancer patients recruited to a surgical clinical trial. We then tested the modified score on an independent series of 136 gastric cancer patients.

Results

Nodal stage added significant prognostic information to the nodal status classification (P <0.001), and was therefore included in the modified score. With the latter, we were able to identify three risk groups with overall five-year survival varying from more than 70% to less than 30%. The prognostic performance of the modified score was better than that achieved with the AJCC-UICC TNM staging.

Conclusions

The modified score, based on established prognostic factors, is proposed as a simple tool for prognostic grouping of gastric cancer patients undergoing curative surgery.

Germline genetic variation in JAK2 as a prognostic marker in castration-resistant prostate cancer

OBJECTIVES

To evaluate the prognostic significance of germline variation in candidate genes in patients with castration-resistant prostate cancer (CRPC).

METHODS

Germline DNA was extracted from peripheral blood mononuclear cells of patients with CRPC enrolled in a clinically annotated registry. Fourteen candidate genes implicated in either initiation or progression of prostate cancer were tagged using single nucleotide polymorphisms (SNPs) from HapMap with a minor allele frequency of >5%. The primary endpoint was overall survival (OS), defined as time from development of CRPC to death. Principal component analysis was used for gene levels tests of significance. For SNP-level results the per allele hazard ratios (HRs) and 95% confidence intervals (CIs) under the additive allele model were estimated using Cox regression, adjusted for age at CRPC and Gleason score (GS).

RESULTS

A total of 240 patients with CRPC were genotyped (14 genes; 84 SNPs). The median (range) age of the cohort was 69 (43-93) years. The GS distribution was 55% with GS ≥8, 32% with GS = 7 and 13% with GS <7 or unknown. The median (interquartile range) time from castration resistance to death for the cohort was 2.67 (1.6-4.07) years (144 deaths). At the gene level, a single gene, JAK2 was associated with OS (P < 0.01), and 11 of 18 JAK2 SNPs were individually associated with OS after adjustment for age and GS. A multivariate model consisting of age, GS, rs2149556 (HR 0.67; 95% CI 0.38-1.18) and rs4372063 (HR 2.17; 95% CI 1.25-3.76) was constructed to predict survival in patients with CRPC (concordance of 0.69, P < 3.2 × 10^-9 ).

CONCLUSIONS

Germline variation in the JAK2 gene was associated with survival in patients with CRPC and warrants further validation as a potential prognostic biomarker.

Impact of Polymorphic Variations of Gemcitabine Metabolism, DNA Damage Repair, and Drug-Resistance Genes on the Effect of High-Dose Chemotherapy for Relapsed or Refractory Lymphoid Malignancies

The goal of this study was to determine whether single nucleotide polymorphisms (SNPs) in genes involved in gemcitabine metabolism, DNA damage repair, multidrug resistance, and alkylator detoxification influence the clinical outcome of patients with refractory/relapsed lymphoid malignancies receiving high-dose gemcitabine/busulfan/melphalan (Gem/Bu/Mel) with autologous stem cell support. We evaluated 21 germline SNPs of the gemcitabine metabolism genes CDA, deoxycytidine kinase, and hCNT3; DNA damage repair genes RECQL, X-ray repair complementing 1, RAD54L, ATM, ATR, MLH1, MSH2, MSH3, TREX1, EXO1, and TP73; and multidrug-resistance genes MRP2 and MRP5; as well as glutathione-S-transferase GSTP1 in 153 patients with relapsed or refractory lymphoma or myeloma receiving Gem/Bu/Mel. We studied the association of genotypes with overall survival (OS), progression-free survival (PFS), and nonhematological grade 3 or 4 toxicity. CDA C111T and TREX1 Ex14-460C>T genotypes had a significant effect on OS (P = .007 and P = .005, respectively), and CDA C111T, ATR C340T, and EXO1 P757L genotypes were significant predictors for severe toxicity (P = .037, P = .024, and P = .025, respectively) in multivariable models that adjusted for clinical variables. The multi-SNP risk score analysis identified the combined genotypes of TREX1 Ex14-460 TT and hCNT3 Ex5 +25A>G AA as significant predictors for OS and the combination of MRP2 Ex10 + 40GG/GA and MLH1 IVS12-169 TT as significant predictor for PFS. Polymorphic variants of certain genes involved in gemcitabine metabolism and DNA damage repair pathways may be potential biomarkers for clinical outcome in patients with refractory/relapsed lymphoid tumors receiving Gem/Bu/Mel.

A unified Bayesian semiparametric approach to assess discrimination ability in survival analysis

The discriminatory ability of a marker for censored survival data is routinely assessed by the time-dependent ROC curve and the c-index. The time-dependent ROC curve evaluates the ability of a biomarker to predict whether a patient lives past a particular time t. The c-index measures the global concordance of the marker and the survival time regardless of the time point. We propose a Bayesian semiparametric approach to estimate these two measures. The proposed estimators are based on the conditional distribution of the survival time given the biomarker and the empirical biomarker distribution. The conditional distribution is estimated by a linear-dependent Dirichlet process mixture model. The resulting ROC curve is smooth as it is estimated by a mixture of parametric functions. The proposed c-index estimator is shown to be more efficient than the commonly used Harrell's c-index since it uses all pairs of data rather than only informative pairs. The proposed estimators are evaluated through simulations and illustrated using a lung cancer dataset.

Comparing two correlated C indices with right-censored survival outcome: a one-shot nonparametric approach

The area under the receiver operating characteristic curve is often used as a summary index of the diagnostic ability in evaluating biomarkers when the clinical outcome (truth) is binary. When the clinical outcome is right-censored survival time, the C index, motivated as an extension of area under the receiver operating characteristic curve, has been proposed by Harrell as a measure of concordance between a predictive biomarker and the right-censored survival outcome. In this work, we investigate methods for statistical comparison of two diagnostic or predictive systems, of which they could either be two biomarkers or two fixed algorithms, in terms of their C indices. We adopt a U-statistics-based C estimator that is asymptotically normal and develop a nonparametric analytical approach to estimate the variance of the C estimator and the covariance of two C estimators. A z-score test is then constructed to compare the two C indices. We validate our one-shot nonparametric method via simulation studies in terms of the type I error rate and power. We also compare our one-shot method with resampling methods including the jackknife and the bootstrap. Simulation results show that the proposed one-shot method provides almost unbiased variance estimations and has satisfactory type I error control and power. Finally, we illustrate the use of the proposed method with an example from the Framingham Heart Study.

Staging of thymic epithelial neoplasms: thymoma and thymic carcinoma

Thymic epithelial neoplasms are uncommon tumors that have been the subject of interest in the last few decades with regard to their histogenesis, histopathologic classification, treatment and prognosis. These tumors are a group of heterogenous neoplasms that are often difficult to subtype and the value of such subclassification with regard to prognosis remains obscure. One factor, however, that appears strongly associated with clinical behavior is tumor staging. The focus of this review will be an overview of the different staging systems for thymic epithelial neoplasms that have been presented in the literature over the years. Particular emphasis is paid to the latest developments in this context.

Development of Hong Kong Liver Cancer staging system with treatment stratification for patients with hepatocellular carcinoma

BACKGROUND & AIMS

We aimed to develop a prognostic classification scheme with treatment guidance for Asian patients with hepatocellular carcinoma (HCC).

METHODS

We collected data from 3856 patients with HCC predominantly related to hepatitis B treated at Queen Mary Hospital in Hong Kong from January 1995 through December 2008. Data on patient performance status, Child-Pugh grade, tumor status (size, number of nodules, and presence of intrahepatic vascular invasion), and presence of extrahepatic vascular invasion or metastasis were included, and randomly separated into training and test sets for analysis. Cox regression and classification and regression tree analyses were used to account for the relative effects of factors in predicting overall survival times and to classify disparate treatment decision rules, respectively; the staging system and treatment recommendation then were constructed by integration of clinical judgments. The Hong Kong Liver Cancer (HKLC) classification was compared with the Barcelona Clinic Liver Cancer (BCLC) classification in terms of discriminatory ability and effectiveness of treatment recommendation.

RESULTS

The HKLC system had significantly better ability than the BCLC system to distinguish between patients with specific overall survival times (area under the receiver operating characteristic curve values, approximately 0.84 vs 0.80; concordance index, 0.74 vs 0.70). More importantly, HKLC identified subsets of BCLC intermediate- and advanced-stage patients for more aggressive treatments than what were recommended by the BCLC system, which improved survival outcomes. Of BCLC-B patients classified as HKLC-II in our system, the survival benefit of radical therapies, compared with transarterial chemoembolization, was substantial (5-year survival probability, 52.1% vs 18.7%; P < .0001). In BCLC-C patients classified as HKLC-II, the survival benefit of radical therapies compared with systemic therapy was even more pronounced (5-year survival probability, 48.6% vs 0.0%; P < .0001).

CONCLUSIONS

We collected data from patients with HCC in Hong Kong to create a system to identify patients who are suitable for more aggressive treatment than the currently used BCLC system. The HKLC system should be validated in non-Asian patient populations and in patients with different etiologies of HCC.

Radiologic response to transcatheter hepatic arterial chemoembolization and clinical outcomes in patients with hepatocellular carcinoma

BACKGROUND & AIMS

The current study analysed the association between radiologic tumour response and survival times of patients with hepatocellular carcinoma (HCC) who were treated with transcatheter hepatic arterial chemoembolization (TACE).

METHODS

Among 493 consecutive patients presenting to our institution between July 2002 and June 2010 with radiologically (n = 398) or histologically (n = 95) confirmed HCC, 368 patients who met inclusion criteria, underwent TACE and had confirmed survival data were retrospectively reviewed. The radiologic response was assessed using RECIST 1.1, EASL and mRECIST criteria at 1 month after the initial TACE.

RESULTS

By univariate analysis, higher Child-Turcotte-Pugh (CTP) score, bilobar and multifocal distribution of tumours, larger tumour size (>5 cm), higher serum alpha-foetoprotein (AFP) level (>200 ng/ml), no subsequent radiofrequency ablation, advanced ECOG, UNOS and BCLC staging, absence of complete necrosis and non-responder (SD or PD) in RECIST 1.1, EASL and mRECIST response assessment were significantly associated with shorter overall survival times. By Cox proportional hazards model, advanced age, presence of ascites, higher MELD score, advanced BCLC staging, absence of complete necrosis and non-responder by RECIST 1.1, EASL and mRECIST criteria were independent and significant prognosticators for overall survival times in patients with HCC who underwent TACE. By time-dependent ROC curve analysis, mRECIST response criteria showed greatest accuracy in predicting survival (AUROC = 0.8676), followed by EASL (AUROC = 0.8471) and RECIST 1.1 (AUROC = 0.7986).

CONCLUSION

mRECIST and EASL criteria for assessing radiologic response 1 month after initial TACE more consistently predict the differences in overall survival between responders and non-responders than conventional RECIST 1.1 criteria.

A systematic selection method for the development of cancer staging systems

The tumor-node-metastasis (TNM) staging system has been the anchor of cancer diagnosis, treatment, and prognosis for many years. For meaningful clinical use, an orderly, progressive condensation of the T and N categories into an overall staging system needs to be defined, usually with respect to a time-to-event outcome. This can be considered as a cutpoint selection problem for a censored response partitioned with respect to two ordered categorical covariates and their interaction. The aim is to select the best grouping of the TN categories. A novel bootstrap cutpoint/model selection method is proposed for this task by maximizing bootstrap estimates of the chosen statistical criteria. The criteria are based on prognostic ability including a landmark measure of the explained variation, the area under the receiver operating characteristic (ROC) curve, and a concordance probability generalized from Harrell's c-index. We illustrate the utility of our method by applying it to the staging of colorectal cancer.

Staging of primary mediastinal tumors

Primary mediastinal tumors, with the exception of lymphoma, are generally rare neoplasms. The diagnosis, classification, and treatment of these tumors still cause some degree of difficulty due to their low incidence and morphologic heterogeneity. This is particularly true for thymic epithelial neoplasms, that is, thymoma and thymic carcinoma, and also applies to primary neuroendocrine carcinomas of the thymus and primary mediastinal germ cell tumors. The appropriate staging of these tumors, likewise, has been a matter of debate over the years and numerous proposals for the staging of thymic epithelial neoplasms have been put forward in the last few decades. Unfortunately, variations of such proposals have been used in some instances to stage other tumors such as thymic neuroendocrine carcinomas and germ cell tumors, which has led to the often inappropriate use of a single staging system for different types of tumors with different biological behavior. This review will provide an overview of the staging of primary mediastinal tumors with special emphasis on more recent assessments in this particular area.

Lasso tree for cancer staging with survival data

The tumor-node-metastasis staging system has been the lynchpin of cancer diagnosis, treatment, and prognosis for many years. For meaningful clinical use, an orderly grouping of the T and N categories into a staging system needs to be defined, usually with respect to a time-to-event outcome. This can be reframed as a model selection problem with respect to features arranged on a partially ordered two-way grid, and a penalized regression method is proposed for selecting the optimal grouping. Instead of penalizing the L1-norm of the coefficients like lasso, in order to enforce the stage grouping, we place L1 constraints on the differences between neighboring coefficients. The underlying mechanism is the sparsity-enforcing property of the L1 penalty, which forces some estimated coefficients to be the same and hence leads to stage grouping. Partial ordering constraints is also required as both the T and N categories are ordinal. A series of optimal groupings with different numbers of stages can be obtained by varying the tuning parameter, which gives a tree-like structure offering a visual aid on how the groupings are progressively made. We hence call the proposed method the lasso tree. We illustrate the utility of our method by applying it to the staging of colorectal cancer using survival outcomes. Simulation studies are carried out to examine the finite sample performance of the selection procedure. We demonstrate that the lasso tree is able to give the right grouping with moderate sample size, is stable with regard to changes in the data, and is not affected by random censoring.

A comparison of estimators to evaluate the discriminatory power of time-to-event models

Discrimination measures for continuous time-to-event outcomes have become an important tool in medical decision making. The idea behind discrimination measures is to evaluate the performance of a prediction model by measuring its ability to distinguish between observations having an event and those having no event. Researchers proposed a variety of approaches to estimate discrimination measures from a set of right-censored data. These approaches rely on different regularity assumptions that are needed to ensure consistency of the respective estimators. Typical examples of regularity assumptions include the proportional hazards assumption in Cox regression and the random censoring assumption. Because regularity assumptions are often violated in practice, conducting a sensitivity analysis of the estimators is of considerable interest. The aim of the paper is to analyze and to compare the most popular estimators of discrimination measures for event time outcomes. On the basis of the results of an extensive simulation study and the analysis of molecular data, we investigate the behavior of the estimators in situations where the underlying regularity assumptions do not hold. We show that violations of the regularity assumptions may induce a nonignorable bias and may therefore result in biased medical decision making.

The New American Joint Committee on Cancer/International Union Against Cancer staging system for adenocarcinoma of the stomach: increased complexity without clear improvement in predictive accuracy

Purpose

To evaluate the changes in the 7th edition American Joint Committee on Cancer (AJCC) staging system for stomach cancer compared to the 6th edition; to compare the predictive accuracy of the two staging systems.

Methods

In a combined database containing 2,196 patients who underwent an R0 resection for gastric adenocarcinoma, differences between the two staging systems were evaluated and stage-specific survival estimates compared. Concordance probability and Brier scores were estimated for both systems to examine the predictive accuracy.

Results

Nodal status cutoff values were changed, leading to a more even distribution for the redefined N1, N2, and N3 group. AJCC 6th edition stage II reflected a highly heterogeneous population, which is now adequately subdivided in the AJCC 7th edition into stages IIA, IIB, and IIIA. The predictive accuracy of N classification improved significantly as measured by concordance. Despite increased complexity, the predictive accuracy of AJCC 7th stage grouping was significantly worse than that of the AJCC 6th edition.

Discussion

The increased complexity of the 7th edition staging system is accompanied by improvements in the predictive value of nodal staging as compared to the 6th edition, but it was no better in overall stage-specific predictive accuracy. Future refinements of the tumor, node, metastasis staging system should consider whether increased complexity is balanced by improved prognostic accuracy.

Value of the first post-transplant biopsy for predicting long-term cardiac allograft vasculopathy (CAV) and graft failure in heart transplant patients

Background

Cardiac allograft vasculopathy (CAV) is the principal cause of long-term graft failure following heart transplantation. Early identification of patients at risk of CAV is essential to target invasive follow-up procedures more effectively and to establish appropriate therapies. We evaluated the prognostic value of the first heart biopsy (median: 9 days post-transplant) versus all biopsies obtained within the first three months for the prediction of CAV and graft failure due to CAV.

Methods and Findings

In a prospective cohort study, we developed multivariate regression models evaluating markers of atherothrombosis (fibrin, antithrombin and tissue plasminogen activator [tPA]) and endothelial activation (intercellular adhesion molecule-1) in serial biopsies obtained during the first three months post-transplantation from 172 patients (median follow-up = 6.3 years; min = 0.37 years, max = 16.3 years). Presence of fibrin was the dominant predictor in first-biopsy models (Odds Ratio [OR] for one- and 10-year graft failure due to CAV = 38.70, p = 0.002, 95% CI = 4.00–374.77; and 3.99, p = 0.005, 95% CI = 1.53–10.40) and loss of tPA was predominant in three-month models (OR for one- and 10-year graft failure due to CAV = 1.81, p = 0.025, 95% CI = 1.08–3.03; and 1.31, p = 0.001, 95% CI = 1.12–1.55). First-biopsy and three-month models had similar predictive and discriminative accuracy and were comparable in their capacities to correctly classify patient outcomes, with the exception of 10-year graft failure due to CAV in which the three-month model was more predictive. Both models had particularly high negative predictive values (e.g., First-biopsy vs. three-month models: 99% vs. 100% at 1-year and 96% vs. 95% at 10-years).

Conclusions

Patients with absence of fibrin in the first biopsy and persistence of normal tPA in subsequent biopsies rarely develop CAV or graft failure during the next 10 years and potentially could be monitored less invasively. Presence of early risk markers in the transplanted heart may be secondary to ischemia/reperfusion injury, a potentially modifiable factor.

Time-dependent Predictive Values of Prognostic Biomarkers with Failure Time Outcome

In a prospective cohort study, information on clinical parameters, tests and molecular markers is often collected. Such information is useful to predict patient prognosis and to select patients for targeted therapy. We propose a new graphical approach, the positive predictive value (PPV) curve, to quantify the predictive accuracy of prognostic markers measured on a continuous scale with censored failure time outcome. The proposed method highlights the need to consider both predictive values and the marker distribution in the population when evaluating a marker, and it provides a common scale for comparing different markers. We consider both semiparametric and nonparametric based estimating procedures. In addition, we provide asymptotic distribution theory and resampling based procedures for making statistical inference. We illustrate our approach with numerical studies and datasets from the Seattle Heart Failure Study.

Use of nomograms for predictions of outcome in patients with advanced bladder cancer

INTRODUCTION

Accurate estimates of risk are essential for physicians if they are to recommend a specific management to patients with bladder cancer. In this review, we discuss the criteria for the evaluation of nomograms and review current available nomograms for advanced bladder cancer.

METHODS

A retrospective review of the Pubmed database between 2002 and 2008 was performed using the keywords 'nomogram' and 'bladder'. We limited the articles to advanced bladder cancer. We recorded input variables, prediction form, number of patients used to develop the prediction tools, the outcome being predicted, prediction tool-specific features, predictive accuracy, and whether validation was performed.

RESULTS

We discuss the characteristics needed to evaluate nomograms such as predictive accuracy, calibration, generalizability, level of complexity, effect of competing risks, conditional probabilities, and head-to-head comparison with other prediction methods. The predictive accuracies of the pre-cystectomy tools (n = 2) range from ∼65-75% and that of the post-cystectomy tools (n = 5) range from ∼75-80%. While some of these nomograms are well-calibrated and outperform AJCC staging, none has been externally validated. To date, four studies demonstrated a statistically significant improvement in predictive accuracy of nomograms by including biomarkers.

CONCLUSIONS

Nomograms provide accurate individualized estimates of outcomes. They currently represent the most accurate and discriminatory decision-making aids tools for predicting outcomes in patients with bladder cancer. Use of current nomograms could improve current selection of patients for standard therapy and investigational trial design by ensuring homogeneous groups. The addition of biological markers to the currently available nomograms using clinical and pathologic data holds the promise of improving prediction and refining management of patients with bladder cancer.

Survival prediction for terminally ill cancer patients: revision of the palliative prognostic score with incorporation of delirium

PURPOSE

An existing and validated palliative prognostic (PaP) score predicts survival in terminally ill cancer patients based on dyspnea, anorexia, Karnofsky performance status score, clinical prediction of survival, total WBC, and lymphocyte percentage. The PaP score assigns patients to three different risk groups according to a 30-day survival probability--group A, >70%; group B, 30%-70%; group C, <30%. The impact of delirium is known but was not incorporated into the PaP score.

MATERIALS AND METHODS

Our aim was to incorporate information on delirium into the PaP score based on a retrospective series of 361 terminally ill cancer patients. We followed the approach of "validation by calibration," proposed by van Houwelingen and later adapted by Miceli for achieving score revision with inclusion of a new variable. The discriminating performance of the scores was estimated using the K statistic.

RESULTS

The prognostic contribution of delirium was confirmed as statistically significant (p < .001) and the variable was accordingly incorporated into the PaP score (D-PaP score). Following this revision, 30-day survival estimates in groups A, B, and C were 83%, 50%, and 9% for the D-PaP score and 87%, 51%, and 16% for the PaP score, respectively. The overall performance of the D-PaP score was better than that of the PaP score.

CONCLUSION

The revision of the PaP score was carried out by modifying the cutoff values used for prognostic grouping without, however, affecting the partial scores of the original tool. The performance of the D-PaP score was better than that of the PaP score and its key feature of simplicity was maintained.

Risk Prediction for Prostate Cancer Recurrence Through Regularized Estimation with Simultaneous Adjustment for Nonlinear Clinical Effects

In biomedical studies, it is of substantial interest to develop risk prediction scores using high-dimensional data such as gene expression data for clinical endpoints that are subject to censoring. In the presence of well-established clinical risk factors, investigators often prefer a procedure that also adjusts for these clinical variables. While accelerated failure time (AFT) models are a useful tool for the analysis of censored outcome data, it assumes that covariate effects on the logarithm of time-to-event are linear, which is often unrealistic in practice. We propose to build risk prediction scores through regularized rank estimation in partly linear AFT models, where high-dimensional data such as gene expression data are modeled linearly and important clinical variables are modeled nonlinearly using penalized regression splines. We show through simulation studies that our model has better operating characteristics compared to several existing models. In particular, we show that there is a non-negligible effect on prediction as well as feature selection when nonlinear clinical effects are misspecified as linear. This work is motivated by a recent prostate cancer study, where investigators collected gene expression data along with established prognostic clinical variables and the primary endpoint is time to prostate cancer recurrence. We analyzed the prostate cancer data and evaluated prediction performance of several models based on the extended c statistic for censored data, showing that 1) the relationship between the clinical variable, prostate specific antigen, and the prostate cancer recurrence is likely nonlinear, i.e., the time to recurrence decreases as PSA increases and it starts to level off when PSA becomes greater than 11; 2) correct specification of this nonlinear effect improves performance in prediction and feature selection; and 3) addition of gene expression data does not seem to further improve the performance of the resultant risk prediction scores.

Anterior craniofacial resection for malignant paranasal tumors: a monoinstitutional experience of 366 cases

BACKGROUND

The purpose of this study was to evaluate the results of a mono-institutional series of patients treated with anterior craniofacial resection for malignant paranasal sinus tumors.

METHODS

We analyzed all patients with malignant paranasal sinus tumors treated with anterior craniofacial resection at our institution between 1987 and 2007. All tumors were classified according to both the American Joint Committee on Cancer (AJCC)-2002 and the Istituto Nazionale Tumori (INT) classifications.

RESULTS

The sample included 366 patients. There was intraorbital spread in 108 cases. The skull base was eroded in 127 patients, with dura or brain involvement in 93 patients. The 10-year disease-specific survival was 53.1%. Histologic subtype, INT stage, surgical margins, and postsurgical radiotherapy were significant, independent predictors for both local relapse and disease-specific survival (DSS). The AJCC-2002 classification was not significant when tested in place of INT stage.

CONCLUSION

Our data indicated that craniofacial resection and postsurgical radiotherapy remain the primary option for malignant tumors involving the anterior skull base.

Statistical consideration for clinical biomarker research in bladder cancer

OBJECTIVE

To critically review and illustrate current methodological and statistical considerations for bladder cancer biomarker discovery and evaluation.

METHODS

Original, review, and methodological articles, and editorials were reviewed and summarized.

RESULTS

Biomarkers may be useful at multiple stages of bladder cancer management: early detection, diagnosis, staging, prognosis, and treatment; however, few novel biomarkers are currently used in clinical practice. The reasons for this disjunction are many and reflect the long and difficult pathway from candidate biomarker discovery to clinical assay, and the lack of coherent and comprehensive processes (pipelines) for biomarker development. Conceptually, the development of new biomarkers should be a process that is similar to therapeutic drug evaluation-a highly regulated process with carefully regulated phases from discovery to human applications. In a further effort to address the pervasive problem of inadequacies in the design, analysis, and reporting of biomarker prognostic studies, a set of reporting recommendations are discussed. For example, biomarkers should provide unique information that adds to known clinical and pathologic information. Conventional multivariable analyses are not sufficient to demonstrate improved prediction of outcomes. Predictive models, including or excluding any new putative biomarker, need to show clinically significant improvement of performance in order to claim any real benefit. Towards this end, proper model building, avoidance of overfitting, and external validation are crucial. In addition, it is important to choose appropriate performance measures dependent on outcome and prediction type and to avoid the use of cutpoints. Biomarkers providing a continuous score provide potentially more useful information than cutpoints since risk fits a continuum model. Combination of complementary and independent biomarkers is likely to better capture the biological potential of a tumor than any single biomarker. Finally, methods that incorporate clinical consequences such as decision curve analysis are crucial to the evaluation of biomarkers.

CONCLUSIONS

Attention to sound design and statistical practice should be delivered as early as possible and will help maximize the promise of biomarkers for patient care. Studies should include a measure of predictive accuracy and clinical decision-analysis. External validation using data from an independent cohort provides the strongest evidence that a model is valid. There is a need for adequately assessed clinical biomarkers in bladder cancer.

New concepts of biobanks--strategic chance for uro-oncology

Cancer, as well as other common diseases, is a complex condition that not only causes a major threat to human health, but also represents a huge burden to society in terms of healthcare cost and loss of economic productivity. Treatment improvements remain elusive, since the causes of cancer are due to a huge number of small and possibly additive effects arising from genetic susceptibility, lifestyle, and environmental conditions. Thus, progress in translational cancer research investigating these changes and their complex interaction is highly dependent on large series of cases (affected and unaffected individuals) including high quality samples and their associated data. Therefore, large and well-organized biobanks have been established, are underway, or are planned in many countries and institutions. The integration of these resources with powerful molecular and "omics" approaches, integrated bioinformatic tools hold the promise to further advance our knowledge of disease development, thus leading to better prevention and treatment strategies. However, these valuable and irreplaceable collections typically suffer from underutilization, due to fragmentation of the collections and their accessibility, lack of common management strategies, including consensus on standard operating procedures, unique policies of utilization, and distribution as well as missing input on a broad basis reflecting research needs on an interdisciplinary, multi-institutional fashion beyond project-driven interest. The uro-oncologic community has not yet contributed to these efforts to its full potential, and broad knowledge on the contemporary developments in the field of biobanking and input into these efforts are still missing. This review presents an overview on biobanking and may serve as an update to be integrated into future discussions on managing biobanks involving uro-oncology. It is based on the discussions at the last meeting of the International Bladder Cancer Network in Barcelona (Spain) in fall 2008 and has been also largely influenced by the works and discussions of the Marble Arch International Working Group on Biobanking for Biomedical Research.

Reporting performance of prognostic models in cancer: a review

BACKGROUND

Appropriate choice and use of prognostic models in clinical practice require the use of good methods for both model development, and for developing prognostic indices and risk groups from the models. In order to assess reliability and generalizability for use, models need to have been validated and measures of model performance reported. We reviewed published articles to assess the methods and reporting used to develop and evaluate performance of prognostic indices and risk groups from prognostic models.

METHODS

We developed a systematic search string and identified articles from PubMed. Forty-seven articles were included that satisfied the following inclusion criteria: published in 2005; aiming to predict patient outcome; presenting new prognostic models in cancer with outcome time to an event and including a combination of at least two separate variables; and analysing data using multivariable analysis suitable for time to event data.

RESULTS

In 47 studies, Cox models were used in 94% (44), but the coefficients or hazard ratios for the variables in the final model were reported in only 72% (34). The reproducibility of the derived model was assessed in only 11% (5) of the articles. A prognostic index was developed from the model in 81% (38) of the articles, but researchers derived the prognostic index from the final prognostic model in only 34% (13) of the studies; different coefficients or variables from those in the final model were used in 50% (19) of models and the methods used were unclear in 16% (6) of the articles. Methods used to derive prognostic groups were also poor, with researchers not reporting the methods used in 39% (14 of 36) of the studies and data derived methods likely to bias estimates of differences between risk groups being used in 28% (10) of the studies. Validation of their models was reported in only 34% (16) of the studies. In 15 studies validation used data from the same population and in five studies from a different population. Including reports of validation with external data from publications up to four years following model development, external validation was attempted for only 21% (10) of models. Insufficient information was provided on the performance of models in terms of discrimination and calibration.

CONCLUSIONS

Many published prognostic models have been developed using poor methods and many with poor reporting, both of which compromise the reliability and clinical relevance of models, prognostic indices and risk groups derived from them.

Critical review of prostate cancer predictive tools

Prostate cancer is a very complex disease, and the decision-making process requires the clinician to balance clinical benefits, life expectancy, comorbidities and potential treatment-related side effects. Accurate prediction of clinical outcomes may help in the difficult process of making decisions related to prostate cancer. In this review, we discuss attributes of predictive tools and systematically review those available for prostate cancer. Types of tools include probability formulas, look-up and propensity scoring tables, risk-class stratification prediction tools, classification and regression tree analysis, nomograms and artificial neural networks. Criteria to evaluate tools include discrimination, calibration, generalizability, level of complexity, decision analysis and ability to account for competing risks and conditional probabilities. The available predictive tools and their features, with a focus on nomograms, are described. While some tools are well-calibrated, few have been externally validated or directly compared with other tools. In addition, the clinical consequences of applying predictive tools need thorough assessment. Nevertheless, predictive tools can facilitate medical decision-making by showing patients tailored predictions of their outcomes with various alternatives. Additionally, accurate tools may improve clinical trial design.

Clinical benefits of a multivariate prediction model for bladder cancer: a decision analytic approach

BACKGROUND

It has been demonstrated that multivariate prediction models predict cancer outcomes more accurately than cancer stage; however, the effects of these models on clinical management are unclear. The objective of the current study was to determine whether a previously published multivariate prediction model for bladder cancer ("bladder nomogram") improved medical decision making when referral for adjuvant chemotherapy was used as a model.

METHODS

Data were analyzed from an international cohort study of 4462 patients who underwent cystectomy without chemotherapy from 1969 to 2004. The number of patients eligible for chemotherapy was determined using pathologic stage criteria (lymph node-positive disease or pathologic T3 [pT3] or pT4 tumor classification) and for 3 cutoff levels on the bladder nomogram (10%, 25%, and 70% risk of recurrence with surgery alone). The number of recurrences was calculated by applying a relative risk reduction to the baseline risk among eligible patients. Clinical net benefit was then calculated by combining recurrences and treatments and weighting the latter by a factor related to drug tolerability.

RESULTS

A nomogram cutoff outperformed pathologic stage for chemotherapy in every scenario of drug effectiveness and tolerability. For a drug with a relative risk of 0.80, with which clinicians would treat <or=20 patients to prevent 1 recurrence, use of the nomogram was equivalent to a strategy that resulted in 60 fewer chemotherapy treatments per 1000 patients without any increase in recurrence rates.

CONCLUSIONS

The authors concluded that referring patients who undergo cystectomy to adjuvant chemotherapy on the basis of a multivariate model is likely to lead to better patient outcomes than the use of pathologic stage. Further research is warranted to evaluate the clinical effects of multivariate prediction models.

Location, extent and number of positive surgical margins do not improve accuracy of predicting prostate cancer recurrence after radical prostatectomy

PURPOSE

Positive surgical margins increase the risk of biochemical recurrence after radical prostatectomy by 2 to 4-fold. The risk of biochemical recurrence may be influenced by the anatomical location and extent of positive surgical margins. In a multicenter study we analyzed the predictive usefulness of several subclassifications of positive surgical margins.

MATERIALS AND METHODS

The clinical information and followup data of 7,160 patients treated with radical prostatectomy alone at 1 of 3 institutions between 1995 and 2006 were modeled using Cox proportional hazards regression analysis for biochemical recurrence. Positive surgical margins were analyzed as solitary vs multiple, focal vs extensive and apical location vs other. The usefulness of these subclassifications was assessed by the improvement in predictive accuracy of nomograms containing these parameters compared to one in which the surgical margin was modeled simply as positive vs negative.

RESULTS

The 7-year progression-free probability was 60% in patients with positive surgical margins. A positive surgical margin was significantly associated with biochemical recurrence (HR 2.3, p <0.001) after adjusting for age, prostate specific antigen, pathological Gleason score, pathological stage and year of surgery. An increased risk of biochemical recurrence was associated with multiple vs solitary positive surgical margins (adjusted HR 1.4, p = 0.002) and extensive vs focal positive surgical margins (adjusted HR 1.3, p = 0.004) on multivariable analysis. However, neither parameter improved the predictive accuracy of a nomogram compared to one in which surgical margin status was modeled as positive vs negative (concordance index 0.851 vs 0.850 vs 0.850).

CONCLUSIONS

The number and extent of positive surgical margin significantly influence the risk of biochemical recurrence after radical prostatectomy. However, the empirical prognostic usefulness of subclassifications of positive surgical margins is limited.

Pre-treatment biomarker levels improve the accuracy of post-prostatectomy nomogram for prediction of biochemical recurrence

PURPOSE

We tested the ability of several pre-operative blood-based biomarkers to enhance the accuracy of standard post-operative features for the prediction of biochemical recurrence (BCR) after radical prostatectomy (RP).

METHODS

Pre-operative plasma levels of Endoglin, interleukin-6 (IL-6), interleukin-6 soluble receptor (IL-6sR), transforming growth factor-beta1 (TGF-beta1), urokinase plasminogen activator (uPA), urokinase plasminogen inhibitor-1 (PAI-1), urokinase plasminogen receptor (uPAR), vascular cell adhesion molecule-1 (VCAM1), and vascular endothelial growth factor (VEGF) were measured using commercially available enzyme immunoassays in 423 consecutive patients treated with RP for clinically localized prostate cancer. Standard post-operative features consisted of surgical margin status, extracapsular extension, seminal vesicle invasion, lymph node involvement, and pathologic Gleason sum. Multivariable modeling was used to explore the gain in the predictive accuracy. The accuracy was quantified by the c-index statistic and was internally validated with 200 bootstrap resamples.

RESULTS

Plasma IL-6 (P = 0.03), IL-6sR (P < 0.001), TGF-beta1 (P = 0.005), and V-CAM1 (P = 0.01) achieved independent predictor status after adjusting for the effects of standard post-operative features. After stepwise backward variable elimination, a model relying on RP Gleason sum, IL-6sR, TGF-beta1, VCAM1, and uPA improved the predictive accuracy of the standard post-operative model by 4% (86.1% vs. 82.1%, P < 0.001).

CONCLUSIONS

Pre-operative plasma biomarkers improved the accuracy of established post-operative prognostic factors of BCR by a significant margin. Incorporation of these biomarkers into standard predictive models may allow more accurate identification of patients who are likely to fail RP thereby allowing more efficient delivery of adjuvant therapy.

Extensions to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers

Background

Decision curve analysis is a novel method for evaluating diagnostic tests, prediction models and molecular markers. It combines the mathematical simplicity of accuracy measures, such as sensitivity and specificity, with the clinical applicability of decision analytic approaches. Most critically, decision curve analysis can be applied directly to a data set, and does not require the sort of external data on costs, benefits and preferences typically required by traditional decision analytic techniques.

Methods

In this paper we present several extensions to decision curve analysis including correction for overfit, confidence intervals, application to censored data (including competing risk) and calculation of decision curves directly from predicted probabilities. All of these extensions are based on straightforward methods that have previously been described in the literature for application to analogous statistical techniques.

Results

Simulation studies showed that repeated 10-fold crossvalidation provided the best method for correcting a decision curve for overfit. The method for applying decision curves to censored data had little bias and coverage was excellent; for competing risk, decision curves were appropriately affected by the incidence of the competing risk and the association between the competing risk and the predictor of interest. Calculation of decision curves directly from predicted probabilities led to a smoothing of the decision curve.

Conclusion

Decision curve analysis can be easily extended to many of the applications common to performance measures for prediction models. Software to implement decision curve analysis is provided.