Background: Circulating free DNA (cfDNA) is an exciting novel method to diagnose, monitor, and predict resistance and response to cancer therapies, with the potential to radically alter the management of cancer patients. To fulfill its potential, greater knowledge about preanalytical variables is required to optimize and standardize the collection process, and maximize the yield and utility of the small quantities of cfDNA extracted. Methods: To this end, we have compared the cfDNA extraction efficiency of three different protocols, including a protocol developed in house (Jewish General Hospital). We evaluated the impact on cfDNA levels of preanalytical variables including speed and timing of the second centrifugation and the use of k-EDTA and CTAD blood collection tubes. Finally, we analyzed the impact on fractional abundance of targeted pre-amplification and whole genome amplification on tumor and circulating tumor DNA (ctDNA) from patients with breast cancer. Results: Making use of a novel protocol for cfDNA extraction we increased cfDNA quantities, up to double that of commercial kits. We found that a second centrifugation at 3,000 g on frozen plasma is as efficient as a high-speed (16,000 g) centrifugation on fresh plasma and does not affect cfDNA levels. Conclusions: These results allow for the implementation of protocols more suitable to the clinical setting. Finally, we found that, unlike targeted gene amplification, whole genome amplification resulted in altered fractional abundance of selected ctDNA variants. Impact: Our study of the preanalytical variables affecting cfDNA recovery and testing will significantly enhance the quality and application of ctDNA testing in clinical oncology.

No abstract

Background: Prognostic models are of high relevance in many medical application domains. However, many common machine learning methods have not been developed for direct applicability to right-censored outcome data. Recently there have been adaptations of these methods to make predictions based on only structured data (such as clinical data). Pseudo-observations has been suggested as a data pre-processing step to address right-censoring in deep neural network. There is a theoretical backing for the use of pseudo-observations to replace the right-censored response outcome, and this allows for algorithms and loss functions designed for continuous, non-censored data to be used. Medical images have been used to predict time-to-event outcomes applying deep convolutional neural network (CNN) methods using a Cox partial likelihood loss function under the assumption of proportional hazard. We propose a method to predict the cumulative incidence from images and structured clinical data by integrating (or combining) pseudo-observations and convolutional neural networks.Results: The performance of the proposed method is assessed in simulation studies and a real data example in breast cancer from The Cancer Genome Atlas (TCGA). The results are compared to the existing convolutional neural network with Cox loss. Our simulation results show that our proposed method performs similar to or even outperforms the comparator, particularly in settings where both the dependent censoring and the survival time do not follow proportional hazards in large sample sizes. The results found in the application in the TCGA data are consistent with the results found in the simulation for small sample settings, where both methods perform similarly. Conclusions: The proposed method facilitates the application of deep CNN methods to time-to-event data and allows for the use of simple and easy to modify loss functions thus contributing to modern image-based precision medicine.

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

Background: Administrative health records (AHRs), which are generated primarily for management and billing purposes, are now widely used in drug safety and comparative effectiveness studies. The development of analytic methods for multi-site studies can benefit from the availability of simulated data, which do not require ethical approvals and data access permissions. We simulated AHRs using both the Observational Medical Dataset Simulator II (OSIM2) proposed by the Observational Medical Outcomes Partnership, and a modified OSIM (ModOSIM) method developed by the Canadian Network for Observational Drug Effect Studies (CNODES). Our objective was to compare the simulated data to real-world AHR data to assess the representativeness of the simulated data.Methods: The real-world data comprised prescription drug records for all individuals with healthcare coverage at any point in a 10-year period (2008 – 2017) from the Manitoba Population Research Data Repository (MPRDR) in the province of Manitoba, Canada. OSIM2 and ModOSIM, which are empirical simulation models for longitudinal patient data, were used to simulate AHRs. The data were described using frequencies and percentages. We estimated agreement of prescription drug use measures in MPRDR, OSIM2 and ModOSIM using the concordance coefficient.Results: The MPRDR cohort included 169,586,633 drug records and 1,395 drug types for 1,604,734 individuals. Data for 50,000 individuals were simulated using OSIM2 and ModOSIM. Sex and age group distributions were similar in the real-world and simulated data. There were significant differences in the total number of drug records and number of unique drugs for OSIM2 and ModOSIM when compared with MPRDR; the median number of unique drugs in MPRDR, OSIM2 and ModOSIM was 9.0, 6.0 and 10.0, respectively. For average number of days of drug use, concordance was 16% (95% confidence interval [CI]: 12% – 19%) for MPRDR and OSIM2 and 88% (95% CI: 87%-90%) for MPRDR and ModOSIM.Conclusions: ModOSIM data were more similar to MPRDR than OSIM2 data on many measures of prescription drug use. Simulated AHRs that are consistent with those found in real-world settings can be generated using ModOSIM; these simulated data will benefit methodological studies and data analyst training.