An evaluation of penalised survival methods for developing prognostic models with rare events

The effect of neoadjuvant chemotherapy and chemoradiotherapy on exercise capacity and outcome following upper gastrointestinal cancer surgery: an observational cohort study

Background

In 2014 approximately 21,200 patients were diagnosed with oesophageal and gastric cancer in England and Wales, of whom 37 % underwent planned curative treatments. Potentially curative surgical resection is associated with significant morbidity and mortality. For operable locally advanced disease, neoadjuvant chemotherapy (NAC) improves survival over surgery alone. However, NAC carries the risk of toxicity and is associated with a decrease in physical fitness, which may in turn influence subsequent clinical outcome. Lower levels of physical fitness are associated with worse outcome following major surgery in general and Upper Gastrointestinal Surgery (UGI) surgery in particular. Cardiopulmonary exercise testing (CPET) provides an objective assessment of physical fitness. The aim of this study is to test the hypothesis that NAC prior to upper gastrointestinal cancer surgery is associated with a decrease in physical fitness and that the magnitude of the change in physical fitness will predict mortality 1 year following surgery.

Methods

This study is a multi-centre, prospective, blinded, observational cohort study of participants with oesophageal and gastric cancer scheduled for neoadjuvant cancer treatment (chemo- and chemoradiotherapy) and surgery. The primary endpoints are physical fitness (oxygen uptake at lactate threshold measured using CPET) and 1-year mortality following surgery; secondary endpoints include post-operative morbidity (Post-Operative Morbidity Survey (POMS)) 5 days after surgery and patient related quality of life (EQ-5D-5 L).

Discussion

The principal benefits of this study, if the underlying hypothesis is correct, will be to facilitate better selection of treatments (e.g. NAC, Surgery) in patients with oesophageal or gastric cancer. It may also be possible to develop new treatments to reduce the effects of neoadjuvant cancer treatment on physical fitness. These results will contribute to the design of a large, multi-centre trial to determine whether an in-hospital exercise-training programme that increases physical fitness leads to improved overall survival.

Trial registration

ClinicalTrials.gov NCT01325883 - 29^th March 2011.

A computational approach to compare regression modelling strategies in prediction research

Background

It is often unclear which approach to fit, assess and adjust a model will yield the most accurate prediction model. We present an extension of an approach for comparing modelling strategies in linear regression to the setting of logistic regression and demonstrate its application in clinical prediction research.

Methods

A framework for comparing logistic regression modelling strategies by their likelihoods was formulated using a wrapper approach. Five different strategies for modelling, including simple shrinkage methods, were compared in four empirical data sets to illustrate the concept of a priori strategy comparison. Simulations were performed in both randomly generated data and empirical data to investigate the influence of data characteristics on strategy performance. We applied the comparison framework in a case study setting. Optimal strategies were selected based on the results of a priori comparisons in a clinical data set and the performance of models built according to each strategy was assessed using the Brier score and calibration plots.

Results

The performance of modelling strategies was highly dependent on the characteristics of the development data in both linear and logistic regression settings. A priori comparisons in four empirical data sets found that no strategy consistently outperformed the others. The percentage of times that a model adjustment strategy outperformed a logistic model ranged from 3.9 to 94.9 %, depending on the strategy and data set. However, in our case study setting the a priori selection of optimal methods did not result in detectable improvement in model performance when assessed in an external data set.

Conclusion

The performance of prediction modelling strategies is a data-dependent process and can be highly variable between data sets within the same clinical domain. A priori strategy comparison can be used to determine an optimal logistic regression modelling strategy for a given data set before selecting a final modelling approach.

Electronic supplementary material

The online version of this article (doi:10.1186/s12874-016-0209-0) contains supplementary material, which is available to authorized users.

Adequate sample size for developing prediction models is not simply related to events per variable

OBJECTIVES

The choice of an adequate sample size for a Cox regression analysis is generally based on the rule of thumb derived from simulation studies of a minimum of 10 events per variable (EPV). One simulation study suggested scenarios in which the 10 EPV rule can be relaxed. The effect of a range of binary predictors with varying prevalence, reflecting clinical practice, has not yet been fully investigated.

STUDY DESIGN AND SETTING

We conducted an extended resampling study using a large general-practice data set, comprising over 2 million anonymized patient records, to examine the EPV requirements for prediction models with low-prevalence binary predictors developed using Cox regression. The performance of the models was then evaluated using an independent external validation data set. We investigated both fully specified models and models derived using variable selection.

RESULTS

Our results indicated that an EPV rule of thumb should be data driven and that EPV ≥ 20 ​ generally eliminates bias in regression coefficients when many low-prevalence predictors are included in a Cox model.

CONCLUSION

Higher EPV is needed when low-prevalence predictors are present in a model to eliminate bias in regression coefficients and improve predictive accuracy.

Survival analysis for high-dimensional, heterogeneous medical data: Exploring feature extraction as an alternative to feature selection

BACKGROUND

In clinical research, the primary interest is often the time until occurrence of an adverse event, i.e., survival analysis. Its application to electronic health records is challenging for two main reasons: (1) patient records are comprised of high-dimensional feature vectors, and (2) feature vectors are a mix of categorical and real-valued features, which implies varying statistical properties among features. To learn from high-dimensional data, researchers can choose from a wide range of methods in the fields of feature selection and feature extraction. Whereas feature selection is well studied, little work focused on utilizing feature extraction techniques for survival analysis.

RESULTS

We investigate how well feature extraction methods can deal with features having varying statistical properties. In particular, we consider multiview spectral embedding algorithms, which specifically have been developed for these situations. We propose to use random survival forests to accurately determine local neighborhood relations from right censored survival data. We evaluated 10 combinations of feature extraction methods and 6 survival models with and without intrinsic feature selection in the context of survival analysis on 3 clinical datasets. Our results demonstrate that for small sample sizes - less than 500 patients - models with built-in feature selection (Cox model with ℓ1 penalty, random survival forest, and gradient boosted models) outperform feature extraction methods by a median margin of 6.3% in concordance index (inter-quartile range: [-1.2%;14.6%]).

CONCLUSIONS

If the number of samples is insufficient, feature extraction methods are unable to reliably identify the underlying manifold, which makes them of limited use in these situations. For large sample sizes - in our experiments, 2500 samples or more - feature extraction methods perform as well as feature selection methods.

Comparison of Cox Model Methods in A Low-dimensional Setting with Few Events

Prognostic models based on survival data frequently make use of the Cox proportional hazards model. Developing reliable Cox models with few events relative to the number of predictors can be challenging, even in low-dimensional datasets, with a much larger number of observations than variables. In such a setting we examined the performance of methods used to estimate a Cox model, including (i) full model using all available predictors and estimated by standard techniques, (ii) backward elimination (BE), (iii) ridge regression, (iv) least absolute shrinkage and selection operator (lasso), and (v) elastic net. Based on a prospective cohort of patients with manifest coronary artery disease (CAD), we performed a simulation study to compare the predictive accuracy, calibration, and discrimination of these approaches. Candidate predictors for incident cardiovascular events we used included clinical variables, biomarkers, and a selection of genetic variants associated with CAD. The penalized methods, i.e., ridge, lasso, and elastic net, showed a comparable performance, in terms of predictive accuracy, calibration, and discrimination, and outperformed BE and the full model. Excessive shrinkage was observed in some cases for the penalized methods, mostly on the simulation scenarios having the lowest ratio of a number of events to the number of variables. We conclude that in similar settings, these three penalized methods can be used interchangeably. The full model and backward elimination are not recommended in rare event scenarios.

Review and evaluation of penalised regression methods for risk prediction in low-dimensional data with few events

Risk prediction models are used to predict a clinical outcome for patients using a set of predictors. We focus on predicting low-dimensional binary outcomes typically arising in epidemiology, health services and public health research where logistic regression is commonly used. When the number of events is small compared with the number of regression coefficients, model overfitting can be a serious problem. An overfitted model tends to demonstrate poor predictive accuracy when applied to new data. We review frequentist and Bayesian shrinkage methods that may alleviate overfitting by shrinking the regression coefficients towards zero (some methods can also provide more parsimonious models by omitting some predictors). We evaluated their predictive performance in comparison with maximum likelihood estimation using real and simulated data. The simulation study showed that maximum likelihood estimation tends to produce overfitted models with poor predictive performance in scenarios with few events, and penalised methods can offer improvement. Ridge regression performed well, except in scenarios with many noise predictors. Lasso performed better than ridge in scenarios with many noise predictors and worse in the presence of correlated predictors. Elastic net, a hybrid of the two, performed well in all scenarios. Adaptive lasso and smoothly clipped absolute deviation performed best in scenarios with many noise predictors; in other scenarios, their performance was inferior to that of ridge and lasso. Bayesian approaches performed well when the hyperparameters for the priors were chosen carefully. Their use may aid variable selection, and they can be easily extended to clustered-data settings and to incorporate external information.

A calibration hierarchy for risk models was defined: from utopia to empirical data

OBJECTIVE

Calibrated risk models are vital for valid decision support. We define four levels of calibration and describe implications for model development and external validation of predictions.

STUDY DESIGN AND SETTING

We present results based on simulated data sets.

RESULTS

A common definition of calibration is "having an event rate of R% among patients with a predicted risk of R%," which we refer to as "moderate calibration." Weaker forms of calibration only require the average predicted risk (mean calibration) or the average prediction effects (weak calibration) to be correct. "Strong calibration" requires that the event rate equals the predicted risk for every covariate pattern. This implies that the model is fully correct for the validation setting. We argue that this is unrealistic: the model type may be incorrect, the linear predictor is only asymptotically unbiased, and all nonlinear and interaction effects should be correctly modeled. In addition, we prove that moderate calibration guarantees nonharmful decision making. Finally, results indicate that a flexible assessment of calibration in small validation data sets is problematic.

CONCLUSION

Strong calibration is desirable for individualized decision support but unrealistic and counter productive by stimulating the development of overly complex models. Model development and external validation should focus on moderate calibration.

Clinical Features and Outcomes in Patients With Membranous Nephropathy and Crescent Formation

Cases of membranous nephropathy (MN) with crescent formation, in the absence of lupus, hepatitis B virus infection, anti-glomerular basement membrane (GBM) nephritis, or antineutrophil cytoplasmic antibody (ANCA), are on record. Clinical presentation and treatment outcomes in these patients are unclear. All patients with biopsy-proven MN diagnosed between years 2008 and 2014 and followed up were enrolled retrospectively. Patients with ANCA, anti-GBM antibodies, lupus, hepatitis B virus infection, or malignance were excluded. Clinical features and outcomes were compared between MN patients with and without crescent. Out of 401 consecutive patients with idiopathic MN, 28 (6.9%) showed crescent formation in 4.9% (2.2%-16.7%) of glomeruli. Mean age of these patients was 50.1 ± 11.1 years, and they presented with heavy proteinuria (6.5 ± 4.8  g/24  h) and hematuria; 21.4% of these patients had declined estimated glomerular filtration rate (<60  mL/min/1.73  m2) on biopsy. Anti-phospholipase A2 receptor antibody was detectable in 79.7% of these patients. These clinical features were comparable to the MN patients without crescent (P > 0.05). Twelve (42.9%) patients received steroids plus immunosuppressive therapy similar to that in patients without crescent (41.3%). Fewer patients with crescents achieved remission (67.9% vs 86.7%, P = 0.029). Crescent formation was a risk factor for no response to the treatments (odds ratio [OR] = 3.1, P = 0.033). Higher percentage of crescents predicted more risk for no remission (OR = 1.2, P = 0.038). Patients with crescents presented more frequencies of abnormal serum creatinine during follow-up (10.7% vs 1.3%, P = 0.031). Crescent formation was also a risk factor for worse renal outcome (relative risk = 10.2, P = 0.046). MN patients with crescents showed unfavorable therapeutic response and tended to have worse renal outcomes. More aggressive treatments and renal protection might be considered to improve the outcomes.

Application of Penalized Regression Techniques in Modelling Insulin Sensitivity by Correlated Metabolic Parameters

This paper aims to introduce penalized estimation techniques in clinical investigations of diabetes, as well as to assess their possible advantages and limitations. Data from a previous study was used to carry out the simulations to assess: a) which procedure results in the lowest prediction error of the final model in the setting of a large number of predictor variables with high multicollinearity (of importance if insulin sensitivity should be predicted) and b) which procedure achieves the most accurate estimate of regression coefficients in the setting of fewer predictors with small unidirectional effects and moderate correlation between explanatory variables (of importance if the specific relation between an independent variable and insulin sensitivity should be examined). Moreover a special focus is on the correct direction of estimated parameter effects, a non-negligible source of error and misinterpretation of study results. The simulations were performed for varying sample size to evaluate the performance of LASSO, Ridge as well as different algorithms for Elastic Net. These methods were also compared with automatic variable selection procedures (i.e. optimizing AIC or BIC).We were not able to identify one method achieving superior performance in all situations. However, the improved accuracy of estimated effects underlines the importance of using penalized regression techniques in our example (e.g. if a researcher aims to compare relations of several correlated parameters with insulin sensitivity). However, the decision which procedure should be used depends on the specific context of a study (accuracy versus complexity) and moreover should involve clinical prior knowledge.

Discrimination-based sample size calculations for multivariable prognostic models for time-to-event data

Background

Prognostic studies of time-to-event data, where researchers aim to develop or validate multivariable prognostic models in order to predict survival, are commonly seen in the medical literature; however, most are performed retrospectively and few consider sample size prior to analysis. Events per variable rules are sometimes cited, but these are based on bias and coverage of confidence intervals for model terms, which are not of primary interest when developing a model to predict outcome. In this paper we aim to develop sample size recommendations for multivariable models of time-to-event data, based on their prognostic ability.

Methods

We derive formulae for determining the sample size required for multivariable prognostic models in time-to-event data, based on a measure of discrimination, D, developed by Royston and Sauerbrei. These formulae fall into two categories: either based on the significance of the value of D in a new study compared to a previous estimate, or based on the precision of the estimate of D in a new study in terms of confidence interval width. Using simulation we show that they give the desired power and type I error and are not affected by random censoring. Additionally, we conduct a literature review to collate published values of D in different disease areas.

Results

We illustrate our methods using parameters from a published prognostic study in liver cancer. The resulting sample sizes can be large, and we suggest controlling study size by expressing the desired accuracy in the new study as a relative value as well as an absolute value. To improve usability we use the values of D obtained from the literature review to develop an equation to approximately convert the commonly reported Harrell’s c-index to D. A flow chart is provided to aid decision making when using these methods.

Conclusion

We have developed a suite of sample size calculations based on the prognostic ability of a survival model, rather than the magnitude or significance of model coefficients. We have taken care to develop the practical utility of the calculations and give recommendations for their use in contemporary clinical research.

How to develop a more accurate risk prediction model when there are few events

When the number of events is low relative to the number of predictors, standard regression could produce overfitted risk models that make inaccurate predictions. Use of penalised regression may improve the accuracy of risk prediction

The Clinical Relevance of Microbleeds in Stroke study (CROMIS-2): rationale, design, and methods

BACKGROUND AND RATIONALE

The increasing use of oral anticoagulants, mostly to prevent ischemic stroke due to atrial fibrillation in an ageing population, has been associated with a fivefold increased incidence of oral anticoagulant-associated intracerebral hemorrhage: a rare, serious, and unpredictable complication. We hypothesize that cerebral microbleeds and other markers of cerebral small vessel disease on magnetic resonance imaging, and genetic polymorphisms (e.g. influencing cerebral small vessel integrity or anticoagulation stability), are associated with an increased risk of oral anticoagulant-associated intracerebral hemorrhage, with potential to improve risk prediction.

AIMS

(1) To determine the incidence, clinical, radiological, and genetic associations of oral anticoagulant-associated intracerebral hemorrhage in a prospective, multicentre cohort study of patients with atrial fibrillation-related ischemic stroke or transient ischemic attack started on oral anticoagulants; (2) To investigate characteristics of oral anticoagulant-associated intracerebral hemorrhage compared with non-oral anticoagulants related intracerebral hemorrhage in a prospective study.

DESIGN AND METHODS

Study 1: Prospective, multicentre, inception cohort study of 1425 adults started on oral anticoagulants (including vitamin K antagonists and the nonvitamin K oral anticoagulants) after recent ischemic stroke and concurrent atrial fibrillation. Participants will have standardized brain magnetic resonance imaging (including a T2*-weighted gradient-recalled echo sequence) and DNA sample collection at baseline, with two-year follow-up by postal questionnaire and medical records surveillance for symptomatic intracranial hemorrhage, other serious vascular events, and death. We will compare the rates of symptomatic intracranial hemorrhage (primary outcome; subclassified as intracerebral, subdural, extradural, intraventricular), other vascular events, and death (secondary outcomes) in participants with one or more cerebral microbleeds to the rates in those without cerebral microbleeds.

STUDY

Prospective case-control study of oral anticoagulant-associated intracerebral hemorrhage compared with non-oral anticoagulant-associated intracerebral hemorrhage to investigate genetic, clinical, and radiological associations with oral anticoagulant-associated intracerebral hemorrhage. In participants with intracerebral hemorrhage (including at least 300 with oral anticoagulant-associated intracerebral hemorrhage), we will collect a DNA sample, standardized clinical data and routine brain imaging (magnetic resonance imaging or computed tomography), and information on functional outcome.

EXPECTED OUTCOMES

We will identify the factors associated with increased intracranial hemorrhage risk after oral anticoagulants for secondary prevention after ischemic stroke due to atrial fibrillation. We will determine clinical, radiological and genetic factors, and clinical outcomes associated with oral anticoagulant-associated intracerebral hemorrhage.

Hypertrophic Cardiomyopathy Registry: The rationale and design of an international, observational study of hypertrophic cardiomyopathy

Hypertrophic cardiomyopathy (HCM) is the most common monogenic heart disease with a frequency as high as 1 in 200. In many cases, HCM is caused by mutations in genes encoding the different components of the sarcomere apparatus. Hypertrophic cardiomyopathy is characterized by unexplained left ventricular hypertrophy, myofibrillar disarray, and myocardial fibrosis. The phenotypic expression is quite variable. Although most patients with HCM are asymptomatic, serious consequences are experienced in a subset of affected individuals who present initially with sudden cardiac death or progress to refractory heart failure. The Hypertrophic Cardiomyopathy Registry study is a National Heart, Lung, and Blood Institute-sponsored 2,750-patient, 44-site, international registry and natural history study designed to address limitations in extant evidence to improve prognostication in HCM (NCT01915615). In addition to the collection of standard demographic, clinical, and echocardiographic variables, patients will undergo state-of-the-art cardiac magnetic resonance for assessment of left ventricular mass and volumes as well as replacement scarring and interstitial fibrosis. In addition, genetic and biomarker analyses will be performed. The Hypertrophic Cardiomyopathy Registry has the potential to change the paradigm of risk stratification in HCM, using novel markers to identify those at higher risk.

Prediction of thrombo-embolic risk in patients with hypertrophic cardiomyopathy (HCM Risk-CVA)

Aims

Atrial fibrillation (AF) and thrombo‐embolism (TE) are associated with reduced survival in hypertrophic cardiomyopathy (HCM), but the absolute risk of TE in patients with and without AF is unclear. The primary aim of this study was to derive and validate a model for estimating the risk of TE in HCM. Exploratory analyses were performed to determine predictors of TE, the performance of the CHA₂DS₂‐VASc score, and outcome with vitamin K antagonists (VKAs).

Methods and results

A retrospective, longitudinal cohort of seven institutions was used to develop multivariable Cox regression models fitted with pre‐selected predictors. Bootstrapping was used for validation. Of 4821 HCM patients recruited between 1986 and 2008, 172 (3.6%) reached the primary endpoint of cerebrovascular accident (CVA), transient ischaemic attack (TIA), or systemic peripheral embolus within 10 years. A total of 27.5% of patients had a CHA₂DS₂‐VASc score of 0, of whom 9.8% developed TE during follow‐up. Cox regression revealed an association between TE and age, AF, the interaction between age and AF, TE prior to first evaluation, NYHA class, left atrial (LA) diameter, vascular disease, and maximal LV wall thickness. There was a curvilinear relationship between LA size and TE risk. The model predicted TE with a C‐index of 0.75 [95% confidence interval (CI) 0.70–0.80] and the D‐statistic was 1.30 (95% CI 1.05–1.56). VKA treatment was associated with a 54.8% (95% CI 31–97%, P = 0.037) relative risk reduction in HCM patients with AF.

Conclusions

The study shows that the risk of TE in HCM patients can be identified using a small number of simple clinical features. LA size, in particular, should be monitored closely, and the assessment and treatment of conventional vascular risk factors should be routine practice in older patients. Exploratory analyses show for the first time evidence for a reduction of TE with VKA treatment. The CHA₂DS₂‐VASc score does not appear to correlate well with the clinical outcome in patients with HCM and should not be used to assess TE risk in this population.

Evaluating hospital performance based on excess cause-specific incidence

Formal evaluation of hospital performance in specific types of care is becoming an indispensable tool for quality assurance in the health care system. When the prime concern lies in reducing the risk of a cause-specific event, we propose to evaluate performance in terms of an average excess cumulative incidence, referring to the center's observed patient mix. Its intuitive interpretation helps give meaning to the evaluation results and facilitates the determination of important benchmarks for hospital performance. We apply it to the evaluation of cerebrovascular deaths after stroke in Swedish stroke centers, using data from Riksstroke, the Swedish stroke registry. © 2015 The Authors. Statistics in Medicine Published by John Wiley & Sons Ltd.

Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration

The TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) Statement includes a 22-item checklist, which aims to improve the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. This explanation and elaboration document describes the rationale; clarifies the meaning of each item; and discusses why transparent reporting is important, with a view to assessing risk of bias and clinical usefulness of the prediction model. Each checklist item of the TRIPOD Statement is explained in detail and accompanied by published examples of good reporting. The document also provides a valuable reference of issues to consider when designing, conducting, and analyzing prediction model studies. To aid the editorial process and help peer reviewers and, ultimately, readers and systematic reviewers of prediction model studies, it is recommended that authors include a completed checklist in their submission. The TRIPOD checklist can also be downloaded from www.tripod-statement.org.

A genomic classifier improves prediction of metastatic disease within 5 years after surgery in node-negative high-risk prostate cancer patients managed by radical prostatectomy without adjuvant therapy

BACKGROUND

Surgery is a standard first-line therapy for men with intermediate- or high-risk prostate cancer. Clinical factors such as tumor grade, stage, and prostate-specific antigen (PSA) are currently used to identify those who are at risk of recurrence and who may benefit from adjuvant therapy, but novel biomarkers that improve risk stratification and that distinguish local from systemic recurrence are needed.

OBJECTIVE

To determine whether adding the Decipher genomic classifier, a validated metastasis risk-prediction model, to standard risk-stratification tools (CAPRA-S and Stephenson nomogram) improves accuracy in predicting metastatic disease within 5 yr after surgery (rapid metastasis [RM]) in an independent cohort of men with adverse pathologic features after radical prostatectomy (RP).

DESIGN, SETTING, AND PARTICIPANTS

The study population consisted of 169 patients selected from 2641 men who underwent RP at the Cleveland Clinic between 1987 and 2008 who met the following criteria: (1) preoperative PSA>20 ng/ml, stage pT3 or margin positive, or Gleason score≥8; (2) pathologic node negative; (3) undetectable post-RP PSA; (4) no neoadjuvant or adjuvant therapy; and (5) minimum of 5-yr follow-up for controls. The final study cohort consisted of 15 RM patients and 154 patients as non-RM controls.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

The performance of Decipher was evaluated individually and in combination with clinical risk factors using concordance index (c-index), decision curve analysis, and logistic regression for prediction of RM.

RESULTS AND LIMITATIONS

RM patients developed metastasis at a median of 2.3 yr (interquartile range: 1.7-3.3). In multivariable analysis, Decipher was a significant predictor of RM (odds ratio: 1.48; p=0.018) after adjusting for clinical risk factors. Decipher had the highest c-index, 0.77, compared with the Stephenson model (c-index: 0.75) and CAPRA-S (c-index: 0.72) as well as with a panel of previously reported prostate cancer biomarkers unrelated to Decipher. Integration of Decipher into the Stephenson nomogram increased the c-index from 0.75 (95% confidence interval [CI], 0.65-0.85) to 0.79 (95% CI, 0.68-0.89).

CONCLUSIONS

Decipher was independently validated as a genomic metastasis signature for predicting metastatic disease within 5 yr after surgery in a cohort of high-risk men treated with RP and managed conservatively without any adjuvant therapy. Integration of Decipher into clinical nomograms increased prediction of RM. Decipher may allow identification of men most at risk for metastatic progression who should be considered for multimodal therapy or inclusion in clinical trials.

PATIENT SUMMARY

Use of Decipher in addition to standard clinical information more accurately identified men who developed metastatic disease within 5 yr after surgery. The results suggest that Decipher allows improved identification of the men who should consider secondary therapy from among the majority that may be managed conservatively after surgery.

ASXL1 but not TET2 mutations adversely impact overall survival of patients suffering systemic mastocytosis with associated clonal hematologic non-mast-cell diseases

Systemic mastocytosis with associated hematologic clonal non-mast cell disease (SM-AHNMD) is a rare and heterogeneous subtype of SM and few studies on this specific entity have been reported. Sixty two patients with Systemic mastocytosis with associated hematologic clonal non-mast cell disease (SM-AHNMD) were presented. Myeloid AHNMD was the most frequent (82%) cases. This subset of patients were older, had more cutaneous lesions, splenomegaly, liver enlargement, ascites; lower bone mineral density and hemoglobin levels and higher tryptase level than lymphoid AHNMD. Defects in KIT, TET2, ASXL1 and CBL were positive in 87%, 27%, 14%, and 11% of cases respectively. The overall survival of patients with SM-AHNMD was 85.2 months. Within the myeloid group, SM-MPN fared better than SM-MDS or SM-AML (p = 0.044,). In univariate analysis, the presence of C-findings, the AHNMD subtypes (SM-MDS/CMML/AML versus SM-MPN/hypereosinophilia) (p = 0.044), Neutropenia (p = 0.015), high monocyte level (p = 0.015) and the presence of ASXL1 mutation had detrimental effects on OS (p = 0.007). In multivariate analysis and penalized Cox model, only the presence of ASXL1 mutation remained an independent prognostic factor that negatively affected OS (p = 0.035). SM-AHNMD is heterogeneous with variable prognosis according to the type of the AHNMD. ASXL1 is mutated in a subset of myeloid AHNMD and adversely impact on OS.

A novel clinical risk prediction model for sudden cardiac death in hypertrophic cardiomyopathy (HCM risk-SCD)

AIMS

Hypertrophic cardiomyopathy (HCM) is a leading cause of sudden cardiac death (SCD) in young adults. Current risk algorithms provide only a crude estimate of risk and fail to account for the different effect size of individual risk factors. The aim of this study was to develop and validate a new SCD risk prediction model that provides individualized risk estimates.

METHODS AND RESULTS

The prognostic model was derived from a retrospective, multi-centre longitudinal cohort study. The model was developed from the entire data set using the Cox proportional hazards model and internally validated using bootstrapping. The cohort consisted of 3675 consecutive patients from six centres. During a follow-up period of 24 313 patient-years (median 5.7 years), 198 patients (5%) died suddenly or had an appropriate implantable cardioverter defibrillator (ICD) shock. Of eight pre-specified predictors, age, maximal left ventricular wall thickness, left atrial diameter, left ventricular outflow tract gradient, family history of SCD, non-sustained ventricular tachycardia, and unexplained syncope were associated with SCD/appropriate ICD shock at the 15% significance level. These predictors were included in the final model to estimate individual probabilities of SCD at 5 years. The calibration slope was 0.91 (95% CI: 0.74, 1.08), C-index was 0.70 (95% CI: 0.68, 0.72), and D-statistic was 1.07 (95% CI: 0.81, 1.32). For every 16 ICDs implanted in patients with ≥4% 5-year SCD risk, potentially 1 patient will be saved from SCD at 5 years. A second model with the data set split into independent development and validation cohorts had very similar estimates of coefficients and performance when externally validated.

CONCLUSION

This is the first validated SCD risk prediction model for patients with HCM and provides accurate individualized estimates for the probability of SCD using readily collected clinical parameters.

Risk stratification for advanced proximal colon neoplasm and individualized endoscopic screening for colorectal cancer by a risk-scoring model

BACKGROUND

Only 30% to 40% of patients with advanced proximal neoplasms (APN) have distal colon neoplasms.

OBJECTIVE

To develop a risk score model for APN and propose an individualized screening protocol for colorectal cancer.

DESIGN

Retrospective cohort study.

SETTING

Tertiary-care center.

PATIENTS

Derivation cohort (6200 adults) and validation cohort (1389 adults).

INTERVENTION

Screening colonoscopy.

MAIN OUTCOME MEASUREMENTS

An APN risk score model was developed from the derivation cohort (6200 adults) and was tested in the validation cohort (1389 adults), who underwent screening colonoscopy.

RESULTS

Age, male sex, and smoking were clinical risk factors for APN. The presence of a distal neoplasm was a sigmoidoscopic risk factor for APN. We calculated APN risk scores (0-8) based on these variables and classified patients as low risk (0-2) or high risk (3-8). In the validation cohort, the relative risk of APN was 3.5-fold higher in the high-risk group than in the low-risk group. Our model suggests that colonoscopy should be performed as an initial screening test in patients with a high clinical risk for APN. Sigmoidoscopy should be performed initially in patients with low clinical risk for APN followed by supplementary colonoscopy in those with high APN risk scores based on both clinical and sigmoidoscopic risk factors. This protocol detected APN in 22 of 34 APN+ patients (64.7%) with little increase in the endoscopy burden, whereas only 16 of 34 APN+ patients (47.1%) would be identified by initial sigmoidoscopy followed by colonoscopy only in cases with distal neoplasms.

LIMITATIONS

Retrospective design.

CONCLUSION

Our APN risk score model provides an algorithm for efficient screening of colorectal cancer by sigmoidoscopy and colonoscopy.

Nomogram prediction for overall survival of patients diagnosed with cervical cancer

Background:

Nomograms are predictive tools that are widely used for estimating cancer prognosis. The aim of this study was to develop a nomogram for the prediction of overall survival (OS) in patients diagnosed with cervical cancer.

Methods:

Cervical cancer databases of two large institutions were analysed. Overall survival was defined as the clinical endpoint and OS probabilities were estimated using the Kaplan–Meier method. Based on the results of survival analyses and previous studies, relevant covariates were identified, a nomogram was constructed and validated using bootstrap cross-validation. Discrimination of the nomogram was quantified with the concordance probability.

Results:

In total, 528 consecutive patients with invasive cervical cancer, who had all nomogram variables available, were identified. Mean 5-year OS rates for patients with International Federation of Gynecologists and Obstetricians (FIGO) stage IA, IB, II, III, and IV were 99.0%, 88.6%, 65.8%, 58.7%, and 41.5%, respectively. Seventy-six cancer-related deaths were observed during the follow-up period. FIGO stage, tumour size, age, histologic subtype, lymph node ratio, and parametrial involvement were selected as nomogram covariates. The prognostic performance of the model exceeded that of FIGO stage alone and the model’s estimated optimism-corrected concordance probability was 0.723, indicating accurate prediction of OS. We present the prediction model as nomogram and provide a web-based risk calculator (http://www.ccc.ac.at/gcu).

Conclusion:

Based on six easily available parameters, a novel statistical model to predict OS of patients diagnosed with cervical cancer was constructed and validated. The model was implemented in a nomogram and provides accurate prediction of individual patients’ prognosis useful for patient counselling and deciding on follow-up strategies.