Approaches to developing de novo cancer population models to examine questions about cancer and race in bladder, gastric, and endometrial cancer and multiple myeloma: the Cancer Intervention and Surveillance Modeling Network incubator program.

BACKGROUND: We are developing 10 de novo population-level mathematical models in 4 malignancies (multiple myeloma and bladder, gastric, and uterine cancers). Each of these sites has documented disparities in outcome that are believed to be downstream effects of systemic racism. METHODS: Ten models are being independently developed as part of the Cancer Intervention and Surveillance Modeling Network incubator program. These models simulate trends in cancer incidence, early diagnosis, treatment, and mortality for the general population and are stratified by racial subgroup. Model inputs are based on large population datasets, clinical trials, and observational studies. Some core parameters are shared, and other parameters are model specific. All models are microsimulation models that use self-reported race to stratify model inputs. They can simulate the distribution of relevant risk factors (eg, smoking, obesity) and insurance status (for multiple myeloma and uterine cancer) in US birth cohorts and population. DISCUSSION: The models aim to refine approaches in prevention, detection, and management of 4 cancers given uncertainties and constraints. They will help explore whether the observed racial disparities are explainable by inequities, assess the effects of existing and potential cancer prevention and control policies on health equity and disparities, and identify policies that balance efficiency and fairness in decreasing cancer mortality.

BayCANN: Streamlining Bayesian Calibration With Artificial Neural Network Metamodeling.

Purpose: Bayesian calibration is generally superior to standard direct-search algorithms in that it estimates the full joint posterior distribution of the calibrated parameters. However, there are many barriers to using Bayesian calibration in health decision sciences stemming from the need to program complex models in probabilistic programming languages and the associated computational burden of applying Bayesian calibration. In this paper, we propose to use artificial neural networks (ANN) as one practical solution to these challenges. Methods: Bayesian Calibration using Artificial Neural Networks (BayCANN) involves (1) training an ANN metamodel on a sample of model inputs and outputs, and (2) then calibrating the trained ANN metamodel instead of the full model in a probabilistic programming language to obtain the posterior joint distribution of the calibrated parameters. We illustrate BayCANN using a colorectal cancer natural history model. We conduct a confirmatory simulation analysis by first obtaining parameter estimates from the literature and then using them to generate adenoma prevalence and cancer incidence targets. We compare the performance of BayCANN in recovering these "true" parameter values against performing a Bayesian calibration directly on the simulation model using an incremental mixture importance sampling (IMIS) algorithm. Results: We were able to apply BayCANN using only a dataset of the model inputs and outputs and minor modification of BayCANN's code. In this example, BayCANN was slightly more accurate in recovering the true posterior parameter estimates compared to IMIS. Obtaining the dataset of samples, and running BayCANN took 15 min compared to the IMIS which took 80 min. In applications involving computationally more expensive simulations (e.g., microsimulations), BayCANN may offer higher relative speed gains. Conclusions: BayCANN only uses a dataset of model inputs and outputs to obtain the calibrated joint parameter distributions. Thus, it can be adapted to models of various levels of complexity with minor or no change to its structure. In addition, BayCANN's efficiency can be especially useful in computationally expensive models. To facilitate BayCANN's wider adoption, we provide BayCANN's open-source implementation in R and Stan.

Microsimulation Model to Compare Enteral and Parenteral Iron Supplementation in Children With Intestinal Failure.

BACKGROUND: Children with chronic intestinal failure have a high prevalence of anemia, commonly from iron deficiency, leading to frequent blood transfusions. No current guideline exists for iron supplementation in these children. In this analysis, we evaluate the effectiveness and the cost-effectiveness of using parenteral, enteral, and no iron supplementation to reduce blood transfusions. METHODS: We created a microsimulation model of pediatric intestinal failure over a 1-year time horizon. Model outcomes included cost (US dollars), blood transfusions received, and hemoglobin trend. Strategies tested included no supplementation, daily enteral supplementation, and monthly parenteral supplementation. We estimated parameters for the model using an institutional cohort of 55 patients. Model parameters updated each 1-month cycle using 2 regressions. A multivariate mixed-effects linear regression estimated hemoglobin values at the next month based on data from the prior month. A mixed-effects logistic regression on hemoglobin predicted the probability of receiving a blood transfusion in a given month. RESULTS: Compared with no supplementation, both enteral and parenteral iron supplementation reduced blood transfusions required per patient by 0.3 and 0.5 transfusions per year, respectively. Enteral iron cost $34 per avoided blood transfusion. Parenteral iron cost an additional $6600 per avoided blood transfusion compared with enteral iron. CONCLUSIONS: We found both parenteral and enteral iron to be effective at reducing blood transfusions. The cost of enteral iron makes it the desired choice in patients who can tolerate it. Future work should aim to identify which subpopulations of patients may benefit most from one strategy over the other.

A Gaussian Approximation Approach for Value of Information Analysis.

Most decisions are associated with uncertainty. Value of information (VOI) analysis quantifies the opportunity loss associated with choosing a suboptimal intervention based on current imperfect information. VOI can inform the value of collecting additional information, resource allocation, research prioritization, and future research designs. However, in practice, VOI remains underused due to many conceptual and computational challenges associated with its application. Expected value of sample information (EVSI) is rooted in Bayesian statistical decision theory and measures the value of information from a finite sample. The past few years have witnessed a dramatic growth in computationally efficient methods to calculate EVSI, including metamodeling. However, little research has been done to simplify the experimental data collection step inherent to all EVSI computations, especially for correlated model parameters. This article proposes a general Gaussian approximation (GA) of the traditional Bayesian updating approach based on the original work by Raiffa and Schlaifer to compute EVSI. The proposed approach uses a single probabilistic sensitivity analysis (PSA) data set and involves 2 steps: 1) a linear metamodel step to compute the EVSI on the preposterior distributions and 2) a GA step to compute the preposterior distribution of the parameters of interest. The proposed approach is efficient and can be applied for a wide range of data collection designs involving multiple non-Gaussian parameters and unbalanced study designs. Our approach is particularly useful when the parameters of an economic evaluation are correlated or interact.

Prioritizing Future Research on Allopurinol and Febuxostat for the Management of Gout: Value of Information Analysis.

OBJECTIVES: The aim of this study was to quantify the value of conducting additional research and reducing uncertainty regarding the cost effectiveness of allopurinol and febuxostat for the management of gout. METHODS: We used a previously developed Markov model that evaluated the cost effectiveness of nine urate-lowering strategies: no treatment, allopurinol-only fixed dose (300 mg), allopurinol-only dose escalation (up to 800 mg), febuxostat-only fixed dose (80 mg), febuxostat-only dose escalation (up to 120 mg), allopurinol-febuxostat sequential therapy fixed dose, allopurinol-febuxostat sequential therapy dose escalation, febuxostat-allopurinol sequential therapy fixed dose, and febuxostat-allopurinol sequential therapy dose escalation. Each strategy was evaluated over the lifetime of a hypothetical gout patient. We calculated population expected value of perfect information (EVPI). We used a linear regression meta-modeling approach to calculate population expected value of partial perfect information (EVPPI), and a Gaussian approximation to calculate the population expected value of sample information for parameters (EVSI) and the expected net benefit of sampling (ENBS) for four potential study designs: (1) an allopurinol efficacy trial; (2) a febuxostat efficacy trial; (3) a prospective observational study evaluating health utilities; and (4) a comprehensive study evaluating the efficacy of allopurinol and febuxostat and health utilities. A 5-year decision time horizon was used in the base-case analysis. RESULTS: EVPI varied by a decision maker's willingness-to-pay (WTP) per quality-adjusted life-year (QALY) and was $US900 million for WTP of $US60,000 per QALY. Population EVPPI was highest across all WTP values for study design #4. For study design #4 and a WTP of $US60,000 per QALY, the optimal sample size was 735 patients per study arm. CONCLUSIONS: Future studies are needed to evaluate the effectiveness of allopurinol and febuxostat dose escalation.

Some Health States Are Better Than Others: Using Health State Rank Order to Improve Probabilistic Analyses.

BACKGROUND: Probabilistic sensitivity analyses (PSA) may lead policy makers to take nonoptimal actions due to misestimates of decision uncertainty caused by ignoring correlations. We developed a method to establish joint uncertainty distributions of quality-of-life (QoL) weights exploiting ordinal preferences over health states. METHODS: Our method takes as inputs independent, univariate marginal distributions for each QoL weight and a preference ordering. It establishes a correlation matrix between QoL weights intended to preserve the ordering. It samples QoL weight values from their distributions, ordering them with the correlation matrix. It calculates the proportion of samples violating the ordering, iteratively adjusting the correlation matrix until this proportion is below an arbitrarily small threshold. We compare our method with the uncorrelated method and other methods for preserving rank ordering in terms of violation proportions and fidelity to the specified marginal distributions along with PSA and expected value of partial perfect information (EVPPI) estimates, using 2 models: 1) a decision tree with 2 decision alternatives and 2) a chronic hepatitis C virus (HCV) Markov model with 3 alternatives. RESULTS: All methods make tradeoffs between violating preference orderings and altering marginal distributions. For both models, our method simultaneously performed best, with largest performance advantages when distributions reflected wider uncertainty. For PSA, larger changes to the marginal distributions induced by existing methods resulted in differing conclusions about which strategy was most likely optimal. For EVPPI, both preference order violations and altered marginal distributions caused existing methods to misestimate the maximum value of seeking additional information, sometimes concluding that there was no value. CONCLUSIONS: Analysts can characterize the joint uncertainty in QoL weights to improve PSA and value-of-information estimates using Open Source implementations of our method.

Linear regression metamodeling as a tool to summarize and present simulation model results.

BACKGROUND / OBJECTIVE: Modelers lack a tool to systematically and clearly present complex model results, including those from sensitivity analyses. The objective was to propose linear regression metamodeling as a tool to increase transparency of decision analytic models and better communicate their results. METHODS: We used a simplified cancer cure model to demonstrate our approach. The model computed the lifetime cost and benefit of 3 treatment options for cancer patients. We simulated 10,000 cohorts in a probabilistic sensitivity analysis (PSA) and regressed the model outcomes on the standardized input parameter values in a set of regression analyses. We used the regression coefficients to describe measures of sensitivity analyses, including threshold and parameter sensitivity analyses. We also compared the results of the PSA to deterministic full-factorial and one-factor-at-a-time designs. RESULTS: The regression intercept represented the estimated base-case outcome, and the other coefficients described the relative parameter uncertainty in the model. We defined simple relationships that compute the average and incremental net benefit of each intervention. Metamodeling produced outputs similar to traditional deterministic 1-way or 2-way sensitivity analyses but was more reliable since it used all parameter values. CONCLUSIONS: Linear regression metamodeling is a simple, yet powerful, tool that can assist modelers in communicating model characteristics and sensitivity analyses.

Network meta-analysis of margin threshold for women with ductal carcinoma in situ.

BACKGROUND: Negative margins are associated with reduced risk of ipsilateral breast tumor recurrence (IBTR) for women with ductal carcinoma in situ (DCIS) treated with breast-conserving surgery (BCS). However, there is no consensus about the best minimum margin width. METHODS: We searched the PubMed database for studies of DCIS published in English between January 1970 and July 2010 and examined the relationship between IBTR and margin status after BCS for DCIS. Women with DCIS were stratified into two groups, BCS with or without radiotherapy. We used frequentist and Bayesian approaches to estimate the odds ratios (OR) of IBTR for groups with negative margins and positive margins. We further examined specific margin thresholds using mixed treatment comparisons and meta-regression techniques. All statistical tests were two-sided. RESULTS: We identified 21 studies published in 24 articles. A total of 1066 IBTR events occurred in 7564 patients, including BCS alone (565 IBTR events in 3098 patients) and BCS with radiotherapy (501 IBTR events in 4466 patients). Compared with positive margins, negative margins were associated with reduced risk of IBTR in patients with radiotherapy (OR = 0.46, 95% credible interval [CrI] = 0.35 to 0.59), and in patients without radiotherapy (OR = 0.34, 95% CrI = 0.24 to 0.47). Compared with patients with positive margins, the risk of IBTR for patients with negative margins was smaller (negative margin >0 mm, OR = 0.45, 95% CrI = 0.38 to 0.53; >2 mm, OR = 0.38, 95% CrI = 0.28 to 0.51; >5 mm, OR = 0.55, 95% CrI = 0.15 to 1.30; and >10 mm, OR = 0.17, 95% CrI = 0.12 to 0.24). Compared with a negative margin greater than 2 mm, a negative margin of at least 10 mm was associated with a lower risk of IBTR (OR = 0.46, 95% CrI = 0.29 to 0.69). We found a probability of .96 that a negative margin threshold greater than 10 mm is the best option compared with other margin thresholds. CONCLUSIONS: Negative surgical margins should be obtained for DCIS patients after BCS regardless of radiotherapy. Within cosmetic constraint, surgeons should attempt to achieve negative margins as wide as possible in their first attempt. More studies are needed to understand whether margin thresholds greater than 10 mm are warranted.