When a patient receives a positive test result from a diagnostic test they assume they have the disease. However, the positive predictive value (PPV), ie the probability that they have the disease given a positive test result, is rarely equal to one. To assist their patients, doctors must explain the chance that they do in fact have the disease. However, physicians frequently miscalculate the PPV as the sensitivity and/or misinterpret the PPV, which results in increased anxiety in patients and generates unnecessary tests and consultations. The reasons for this miscalculation as well as three ways to calculate the PPV are reviewed here.

Background

Evaluation of tumor-tissue images stained with hematoxylin and eosin (H&E) is pivotal in diagnosis, yet only a fraction of the rich phenotypic information is considered for clinical care. Here, we propose a survival deep learning (SDL) framework to extract this information to predict glioma survival.

Methods

Digitized whole slide images were downloaded from The Cancer Genome Atlas (TCGA) for 766 diffuse glioma patients, including isocitrate dehydrogenase (IDH)-mutant/1p19q-codeleted oligodendroglioma, IDH-mutant/1p19q-intact astrocytoma, and IDH-wildtype astrocytoma/glioblastoma. Our SDL framework employs a residual convolutional neural network with a survival model to predict patient risk from H&E-stained whole-slide images. We used statistical sampling techniques and randomized the transformation of images to address challenges in learning from histology images. The SDL risk score was evaluated in traditional and recursive partitioning (RPA) survival models.

Results

The SDL risk score demonstrated substantial univariate prognostic power (median concordance index of 0.79 [se: 0.01]). After adjusting for age and World Health Organization 2016 subtype, the SDL risk score was significantly associated with overall survival (OS; hazard ratio = 2.45; 95% CI: 2.01 to 3.00). Four distinct survival risk groups were characterized by RPA based on SDL risk score, IDH status, and age with markedly different median OS ranging from 1.03 years to 14.14 years.

Conclusions

The present study highlights the independent prognostic power of the SDL risk score for objective and accurate prediction of glioma outcomes. Further, we show that the RPA delineation of patient-specific risk scores and clinical prognostic factors can successfully demarcate the OS of glioma patients.

Malignant glioma represents a diverse set of molecularly heterogeneous diseases. Few therapeutic agents have been approved despite decades of clinical trials research and pre-clinical investigation. Attempts to refine neuroimaging criteria and recent discovery of the genomic profiles linking tumor subsets to survival outcomes have spurred discussion on a variety of new approaches in clinical trial design and relevant endpoints. Here we focus on those endpoints in clinical trial design for patients with primary glioma and related issues still to be resolved.

The use of cumulative incidence functions for characterizing the risk of one type of event in the presence of others has become increasingly popular over the past two decades. The problems of modeling, estimation and inference have been treated using parametric, nonparametric and semi-parametric methods. Efforts to develop suitable extensions of machine learning methods, such as regression trees and ensemble methods, have begun comparatively recently. In this paper, we propose a novel approach to estimating cumulative incidence curves in a competing risks setting using regression trees and associated ensemble estimators. The proposed methods use augmented estimators of the Brier score risk as the primary basis for building and pruning trees, and lead to methods that are easily implemented using existing R packages. Data from the Radiation Therapy Oncology Group (trial 9410) is used to illustrate these new methods.

BACKGROUND: Focused ultrasound (FUS-T) and stereotactic radiosurgery thalamotomy (SRS-T) targeting the ventral intermediate nucleus are effective incisionless surgeries for essential tremor (ET). However, their efficacy for tremor reduction and, importantly, adverse event incidence have not been directly compared. OBJECTIVE: To present a comprehensive systematic review with network meta-analysis examining both efficacy and adverse events (AEs) of FUS-T vs SRS-T for treating medically refractory ET. METHODS: We conducted a systematic review and network meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, using the PubMed and Embase databases. We included all primary FUS-T/SRS-T studies with approximately 1-year follow-up, with unilateral Fahn-Tolosa-Marin Tremor Rating Scale or Clinical Rating Scale for Tremor scores prethalamotomy/post-thalamotomy and/or AEs. The primary efficacy outcome was Fahn-Tolosa-Marin Tremor Rating Scale A+B score reduction. AEs were reported as an estimated incidence. RESULTS: Fifteen studies of 464 patients and 3 studies of 62 patients met inclusion criteria for FUS-T/SRS-T efficacy comparison, respectively. Network meta-analysis demonstrated similar tremor reduction between modalities (absolute tremor reduction: FUS-T: -11.6 (95% CI: -13.3, -9.9); SRS-T: -10.3 (95% CI: -14.2, -6.0). FUS-T had a greater 1-year adverse event rate, particularly imbalance and gait disturbances (10.5%) and sensory disturbances (8.3%). Contralateral hemiparesis (2.7%) often accompanied by speech impairment (2.4%) were most common after SRS-T. There was no correlation between efficacy and lesion volume. CONCLUSION: Our systematic review found similar efficacy between FUS-T and SRS-T for ET, with trend toward higher efficacy yet greater adverse event incidence with FUS-T. Smaller lesion volumes could mitigate FUS-T off-target effects for greater safety.

An efficient phase I trial is a crucial step in developing a new drug in a safe and timely manner. The main objective of a phase I trial is to determine the maximum tolerated dose in order to recommend the dose for a phase II trial. There are many designs that are implemented in phase I trials. Rule-based designs such as the traditional 3 + 3 method and rolling six design are easy to implement and assess for safety using a conservative approach. Model-based designs such as the continual reassessment method and the time-to-event continual reassessment method use mathematical models to increase the precision of dose estimation. The advantages and shortcomings of these designs, along with other designs, are reviewed.

Given the lack of beneficial treatments in glioma, there is a need for prognostic models for therapeutic decision making and life planning. Recently several studies defining subtypes of glioma have been published. Here, we review the statistical considerations of how to build and validate prognostic models, explain the models presented in the current glioma literature, and discuss advantages and disadvantages of each model. The 3 statistical considerations to establishing clinically useful prognostic models are: study design, model building, and validation. Careful study design helps to ensure that the model is unbiased and generalizable to the population of interest. During model building, a discovery cohort of patients can be used to choose variables, construct models, and estimate prediction performance via internal validation. Via external validation, an independent dataset can assess how well the model performs. It is imperative that published models properly detail the study design and methods for both model building and validation. This provides readers the information necessary to assess the bias in a study, compare other published models, and determine the model's clinical usefulness. As editors, reviewers, and readers of the relevant literature, we should be cognizant of the needed statistical considerations and insist on their use.

Estimating a patient's mortality risk is important in making treatment decisions. Survival trees are a useful tool and employ recursive partitioning to separate patients into different risk groups. Existing 'loss based' recursive partitioning procedures that would be used in the absence of censoring have previously been extended to the setting of right censored outcomes using inverse probability censoring weighted estimators of loss functions. In this paper, we propose new 'doubly robust' extensions of these loss estimators motivated by semiparametric efficiency theory for missing data that better utilize available data. Simulations and a data analysis demonstrate strong performance of the doubly robust survival trees compared with previously used methods. Copyright © 2016 John Wiley & Sons, Ltd.

Among women diagnosed with ductal carcinoma in situ (DCIS), we identified factors associated with local invasive cancer (LIC) and regional/metastatic invasive cancer (RMIC) and provide 10-year risks based on clinically relevant factors. We created a retrospective, population-based cohort of 1492 women with an initial diagnosis of DCIS (1983–1996) treated by lumpectomy alone. Histological and molecular markers (Ki67, ER, PR, COX-2, p16, ERBB2) were collected on DCIS cases with a subsequent tumor (DCIS, LIC, or RMIC) and a subsample of frequency-matched controls without subsequent tumors. Competing risks methods were used to identify factors associated with LIC and RMIC and cumulative incidence methods to estimate 10-year risks for combinations of factors. Median follow-up time was 12.6 years (range 0.5–29.5 years). The overall 10-year risk of LIC (11.9 %) was higher than for RMIC (3.8 %). About half of women with initial DCIS lesions are detected by mammography and p16 negative and have a 10-year risk of LIC of 6.2 % (95 % CI 5.8–6.8 %) and RMIC of 1.2 % (95 % CI 1.1–1.3 %). Premenopausal women whose DCIS lesion was p16 positive or p16 negative and detected by palpation had high 10-year risk of LIC of 23.0 % (95 % CI 19.3–27.4 %). Ten-year risk of RMIC was highest at 22.5 % (95 % CI 13.8–48.1 %) for those positive for p16, COX-2, and ERRB2, and negative for ER, but prevalence of this group is low at 3 %. Ten-year risk of LIC and RMIC is low for the majority diagnosed with DCIS. Combinations of molecular markers and method of detection of initial DCIS lesion can differentiate women at low and high risk of LIC and RMIC.

Among women diagnosed with ductal carcinoma in situ (DCIS), we identified factors associated with local invasive cancer (LIC) and regional/metastatic invasive cancer (RMIC) and provide 10-year risks based on clinically relevant factors. We created a retrospective, population-based cohort of 1492 women with an initial diagnosis of DCIS (1983-1996) treated by lumpectomy alone. Histological and molecular markers (Ki67, ER, PR, COX-2, p16, ERBB2) were collected on DCIS cases with a subsequent tumor (DCIS, LIC, or RMIC) and a subsample of frequency-matched controls without subsequent tumors. Competing risks methods were used to identify factors associated with LIC and RMIC and cumulative incidence methods to estimate 10-year risks for combinations of factors. Median follow-up time was 12.6 years (range 0.5-29.5 years). The overall 10-year risk of LIC (11.9 %) was higher than for RMIC (3.8 %). About half of women with initial DCIS lesions are detected by mammography and p16 negative and have a 10-year risk of LIC of 6.2 % (95 % CI 5.8-6.8 %) and RMIC of 1.2 % (95 % CI 1.1-1.3 %). Premenopausal women whose DCIS lesion was p16 positive or p16 negative and detected by palpation had high 10-year risk of LIC of 23.0 % (95 % CI 19.3-27.4 %). Ten-year risk of RMIC was highest at 22.5 % (95 % CI 13.8-48.1 %) for those positive for p16, COX-2, and ERRB2, and negative for ER, but prevalence of this group is low at 3 %. Ten-year risk of LIC and RMIC is low for the majority diagnosed with DCIS. Combinations of molecular markers and method of detection of initial DCIS lesion can differentiate women at low and high risk of LIC and RMIC.