Emulator-Based Bayesian Calibration of the CISNET Colorectal Cancer Models.

PURPOSE: To calibrate Cancer Intervention and Surveillance Modeling Network (CISNET)'s SimCRC, MISCAN-Colon, and CRC-SPIN simulation models of the natural history colorectal cancer (CRC) with an emulator-based Bayesian algorithm and internally validate the model-predicted outcomes to calibration targets. METHODS: We used Latin hypercube sampling to sample up to 50,000 parameter sets for each CISNET-CRC model and generated the corresponding outputs. We trained multilayer perceptron artificial neural networks (ANNs) as emulators using the input and output samples for each CISNET-CRC model. We selected ANN structures with corresponding hyperparameters (i.e., number of hidden layers, nodes, activation functions, epochs, and optimizer) that minimize the predicted mean square error on the validation sample. We implemented the ANN emulators in a probabilistic programming language and calibrated the input parameters with Hamiltonian Monte Carlo-based algorithms to obtain the joint posterior distributions of the CISNET-CRC models' parameters. We internally validated each calibrated emulator by comparing the model-predicted posterior outputs against the calibration targets. RESULTS: The optimal ANN for SimCRC had 4 hidden layers and 360 hidden nodes, MISCAN-Colon had 4 hidden layers and 114 hidden nodes, and CRC-SPIN had 1 hidden layer and 140 hidden nodes. The total time for training and calibrating the emulators was 7.3, 4.0, and 0.66 h for SimCRC, MISCAN-Colon, and CRC-SPIN, respectively. The mean of the model-predicted outputs fell within the 95% confidence intervals of the calibration targets in 98 of 110 for SimCRC, 65 of 93 for MISCAN, and 31 of 41 targets for CRC-SPIN. CONCLUSIONS: Using ANN emulators is a practical solution to reduce the computational burden and complexity for Bayesian calibration of individual-level simulation models used for policy analysis, such as the CISNET CRC models. In this work, we present a step-by-step guide to constructing emulators for calibrating 3 realistic CRC individual-level models using a Bayesian approach. HIGHLIGHTS: We use artificial neural networks (ANNs) to build emulators that surrogate complex individual-based models to reduce the computational burden in the Bayesian calibration process.ANNs showed good performance in emulating the CISNET-CRC microsimulation models, despite having many input parameters and outputs.Using ANN emulators is a practical solution to reduce the computational burden and complexity for Bayesian calibration of individual-level simulation models used for policy analysis.This work aims to support health decision scientists who want to quantify the uncertainty of calibrated parameters of computationally intensive simulation models under a Bayesian framework.

Emulator-based Bayesian calibration of the CISNET colorectal cancer models.

Purpose: To calibrate Cancer Intervention and Surveillance Modeling Network (CISNET) 's SimCRC, MISCAN-Colon, and CRC-SPIN simulation models of the natural history colorectal cancer (CRC) with an emulator-based Bayesian algorithm and internally validate the model-predicted outcomes to calibration targets. Methods: We used Latin hypercube sampling to sample up to 50,000 parameter sets for each CISNET-CRC model and generated the corresponding outputs. We trained multilayer perceptron artificial neural networks (ANN) as emulators using the input and output samples for each CISNET-CRC model. We selected ANN structures with corresponding hyperparameters (i.e., number of hidden layers, nodes, activation functions, epochs, and optimizer) that minimize the predicted mean square error on the validation sample. We implemented the ANN emulators in a probabilistic programming language and calibrated the input parameters with Hamiltonian Monte Carlo-based algorithms to obtain the joint posterior distributions of the CISNET-CRC models' parameters. We internally validated each calibrated emulator by comparing the model-predicted posterior outputs against the calibration targets. Results: The optimal ANN for SimCRC had four hidden layers and 360 hidden nodes, MISCAN-Colon had 4 hidden layers and 114 hidden nodes, and CRC-SPIN had one hidden layer and 140 hidden nodes. The total time for training and calibrating the emulators was 7.3, 4.0, and 0.66 hours for SimCRC, MISCAN-Colon, and CRC-SPIN, respectively. The mean of the model-predicted outputs fell within the 95% confidence intervals of the calibration targets in 98 of 110 for SimCRC, 65 of 93 for MISCAN, and 31 of 41 targets for CRC-SPIN. Conclusions: Using ANN emulators is a practical solution to reduce the computational burden and complexity for Bayesian calibration of individual-level simulation models used for policy analysis, like the CISNET CRC models. In this work, we present a step-by-step guide to constructing emulators for calibrating three realistic CRC individual-level models using a Bayesian approach.

Cost-Effectiveness Analysis of Screening for Pancreatic Cancer Among High-Risk Populations.

PURPOSE: We evaluated the potential cost-effectiveness of combined magnetic resonance imaging (MRI) and endoscopic ultrasound (EUS) screening for pancreatic ductal adenocarcinoma (PDAC) among populations at high risk for the disease. METHODS: We used a microsimulation model of the natural history of PDAC to estimate the lifetime health benefits, costs, and cost-effectiveness of PDAC screening among populations with specific genetic risk factors for PDAC, including BRCA1 and BRCA2, PALB2, ATM, Lynch syndrome, TP53, CDKN2A, and STK11. For each high-risk population, we simulated 29 screening strategies, defined by starting age and frequency. Screening included MRI with follow-up EUS in a subset of patients. Costs of tests were based on Medicare reimbursement for MRI, EUS, fine-needle aspiration biopsy, and pancreatectomy. Cancer-related cost by stage of disease and phase of treatment was based on the literature. For each high-risk population, we performed an incremental cost-effectiveness analysis, assuming a willingness-to-pay (WTP) threshold of $100,000 US dollars (USD) per quality-adjusted life year (QALY) gained. RESULTS: For men with relative risk (RR) 12.33 (CDKN2A) and RR 28 (STK11), annual screening was cost-effective, starting at age 55 and 40 years, respectively. For women, screening was only cost-effective for those with RR 28 (STK11), with annual screening starting at age 45 years. CONCLUSION: Combined MRI/EUS screening may be a cost-effective approach for the highest-risk populations (among mutations considered, those with RR >12). However, for those with moderate risk (RR, 5-12), screening would only be cost-effective at higher WTP thresholds (eg, $200K USD/QALY) or with once-only screening.

Ultrasound Surveillance of Small, Incidentally Detected Gallbladder Polyps: Projected Benefits by Sex, Age, and Comorbidity Level.

OBJECTIVE: Incidentally detected gallbladder polyps are commonly encountered when performing upper abdominal ultrasound. Our purpose was to estimate the life expectancy (LE) benefit of ultrasound-based gallbladder surveillance in patients with small (6-7 to <10 mm), incidentally detected gallbladder polyps, accounting for patient sex, age, and comorbidity level. METHODS: We developed a decision-analytic Markov model to evaluate hypothetical cohorts of women and men with small gallbladder polyps, with varying age (66-80 years) and comorbidity level (none, mild, moderate, severe). Drawing from current evidence, in the base case, we assumed no increased risk of gallbladder cancer in patients with small gallbladder polyps. To estimate maximal possible LE gains from surveillance, we assumed perfect cancer control consequent to 5 years of surveillance. We varied key assumptions including cancer risk and test performance characteristics in sensitivity analysis. RESULTS: Projected LE gains from surveillance were <3 days across most cohorts and scenarios evaluated. For 66- and 80-year-olds with no comorbidities, LE gains were 1.46 and 1.45 days, respectively, for women, and 0.67 and 0.75 days for men. With 10 years of surveillance, LE gains increased to 2.94 days for 66-year-old women with no comorbidities (men: 1.35 days). If we assumed a 10% increase in gallbladder cancer risk among individuals with polyps, LE gains increased slightly to 1.60 days for 66-year-old women with no comorbidities (men: 0.74 days). Results were sensitive to test performance and surgical mortality. DISCUSSION: Even under unrealistic, optimistic assumptions of cancer control, ultrasound surveillance of incidentally detected small gallbladder polyps provided limited benefit.

Iodide-doped precious metal nanoparticles: measuring oxidative stress in vivo via photoacoustic imaging.

Accumulation of reactive oxygen and nitrogen species (RONS) can induce cell damage and even cell death. RONS are short-lived species, which makes direct, precise, and real-time measurement difficult. Biologically-relevant RONS levels are in the nM-µM scale; hence, there is a need for highly sensitive RONS probes. We previously used hybrid gold-core silver-shell nanoparticles with mM sensitivity to H2O2. These particles reported the presence of RONS via spectral shifts which could easily be quantified via photoacoustic imaging. Here, we used halide doping to tune the electrochemical properties of these materials to better match the oxidation potential of RONS. This work describes the synthesis, characterization, and application of these AgI-coated gold nanorods (AgI/AuNR). The I : Ag molar ratio, pH, and initial Ag shell thickness were optimized for good RONS detection limits. Halide doping lowers the reduction potential of Ag from to resulting in a 1000-fold increase in H2O2 and 100 000-fold increase in ONOO- sensitivity. The AgI/AuNR system also etches 45-times faster than undoped Ag/AuNR. The AgI/AuNR easily reported the endogenously produced RONS in established cells lines as well as murine models.