Oguzhan Alagoz

1. Breast cancer remains the most commonly diagnosed cancer and the second leading cause of cancer death among women worldwide, with approximately 2.3 million new cases and 670,000 deaths estimated...

Federally mandated breast density notifications motivate consideration of supplemental breast magnetic resonance imaging (MRI). To evaluate supplemental breast MRI strategies. Simulation of women at average to 4 times higher-than-average relative risk (RR) for breast cancer incidence undergoing screening digital breast tomosynthesis (DBT) with or without supplemental MRI. Breast Cancer Surveillance Consortium and literature. Women aged 40 years or older. Lifetime. U.S. federal payer. Screening with DBT with or without breast density-targeted MRI by starting age (40, 45, or 50 years) and interval (annual or biennial). Breast cancer deaths averted, false-positive biopsy recommendations, harm-benefit ratios, and incremental cost-effectiveness ratios (ICERs). Across all starting ages and intervals, DBT averted 7.4 to 10.5 breast cancer deaths per 1000 average-risk women screened and 23.2 to 33.6 per 1000 women with 4 times higher-than-average risk. Across all RR levels, DBT with supplemental MRI for women with extremely dense breasts (DBT+MRId) averted 0.1 to 0.8 additional breast cancer deaths and resulted in 22 to 186 additional false-positive biopsy recommendations. False-positive biopsies per breast cancer death averted for biennial DBT+MRId for women with 2 times higher-than-average risk were similar to those associated with DBT in average-risk women. For all risk groups, biennial DBT+MRId starting at age 50 years was more effective but less cost-effective than DBT starting at age 45 years. The ICERs were sensitive to cancer risk, MRI costs, and false-positive biopsy rates. Subgroups considered risk and breast density only. Supplemental MRI for women aged 40 years or older with extremely dense breasts and higher-than-average risk (RR ≥2.0) had harm-benefit ratios similar to biennial DBT alone and could be cost-effective if MRI costs and false-positive biopsy rates are reduced. National Cancer Institute.

Introduction: The Coronavirus 2019 (COVID-19) pandemic disrupted healthcare delivery across the United States, leading to a sharp decline in outpatient visits. Given that thyroid nodule evaluation and diagnosis typically occur in outpatient settings, we hypothesized that these disruptions would reduce thyroid cancer incidence. This study evaluates trends in thyroid cancer diagnosis, treatment, and outcomes at a U.S. comprehensive cancer center from 2017 to 2023. Methods: We included adults with pathologically confirmed thyroid cancer diagnosis between January 1, 2017, and December 31, 2023. Primary analyses assessed (1) annual percent change (APC) in thyroid cancer incidence from 2017–2023 and (2) incidence trends across three periods: pre-pandemic (Jan 2017–Mar 2020), pandemic (Apr 2020–Mar 2021), and COVID-years (Apr 2021–Dec 2023). Annual incidence rates were calculated using the total number of individuals cared for within our health system each year as the denominator. Secondary outcomes included tumor size, extent and timing of surgery, and 1-year all-cause mortality across time periods. Results: Among 1,031 patients (68% women, 92% White, mean age 52 ± 15.5), age- and sex-adjusted thyroid cancer incidence declined from 34 to 15.6 per 100,000 persons between 2017 and 2022, with an average APC of –14.4%. Diagnoses rose slightly in 2023 (21 per 100,000) compared to 2022 but remained 38% below 2017 levels. No significant differences were observed in tumor size, surgical treatment, gender ratio, or 1-year all-cause mortality across timeframes. Conclusion: Age- and sex-adjusted annual incidence of thyroid cancer declined steadily from 2017 to 2022. The annual decrease in thyroid cancer continued through the pandemic years. Tumor size and mortality remained stable across all years evaluated. The increase in diagnoses noted in 2023 may be a correction due to missed cases during the years of the pandemic or could represent a new steady state incidence after years of decline.

Importance: General mammography screening guidelines target women at average risk within a specified age range (age based) and do not consider absolute risk of individual women at a given age (risk based). Objective: To compare outcomes of mammography screening strategies that vary by 5-year risk of invasive breast cancer vs age-based strategies. Design, Setting, and Participants: This decision analytical model used 2 established Cancer Intervention and Surveillance Modeling Network (CISNET) breast cancer models and simulated US women born in 1980 who were aged 40 years or older without a prior history of breast cancer. Modeling analyses were conducted from April 2023 to April 2025. Intervention: Digital breast tomosynthesis delivered via 50 screening strategies (3 age based and 47 risk based) vs a no-screening scenario. Five-year absolute invasive breast cancer risk was based on the validated Breast Cancer Surveillance Consortium, version 3 calculator. Women’s 5-year breast cancer risk was categorized as low, average, intermediate, or high. Main Outcomes and Measures: Primary outcomes included lifetime number of breast cancer deaths averted and false-positive screening recalls. Lifetime outcomes were averaged across models and expressed per 1000 women screened. Results: Nine risk-based screening strategies were associated with a comparable or greater number of deaths averted than biennial age-based screening from ages 40 to 74 years (B40-74) (range across strategies for mean model estimates, 6.8-7.5 per 1000 women vs 6.8 per 1000 women) as well as reduced false-positive recalls by 8% to 23% (1050-1257 per 1000 women for risk-based screening strategies vs 1365 per 1000 women for B40-74). For example, a risk-based approach using a combination of biennial screening (for women at low risk aged 55-74 years, at average risk aged 50-59 years, at intermediate risk aged 45-54 years, and at high risk aged 40-49 years) and annual screening (for women at average risk aged 60-74 years, at intermediate risk aged 55-74 years, and at high risk aged 50-74 years) would be associated with 6% more breast cancer deaths averted than B40-74 (7.2 vs 6.8 per 1000 women) and 13% fewer false-positive recalls (1190 vs 1365 per 1000 women). Results were consistent across the 2 CISNET models, and the relative difference in breast cancer deaths averted between B40-74 and risk-based screening strategies was more pronounced than for life-years gained. Conclusions and Relevance: In this decision analytical modeling study of breast cancer screening, population risk-based screening using 5-year invasive breast cancer risk was associated with similar or greater benefits than age-based screening as well as reduced false-positive recalls. As personalized medicine advances, risk-based screening is poised to become a cornerstone of breast cancer prevention, offering a more nuanced and tailored approach to patient care.

Introduction: Subgroup analyses are vital components of health technology assessments, but randomized controlled trials (RCTs) do not commonly report survival distributions for subgroups. This study developed an analytical framework to elicit unreported subgroup-specific survival curves from aggregate RCT data. Methods: Assuming exponentially distributed subgroup survival durations, we developed an optimization model that approximates the restricted mean survival time (RMST) for the overall population via the weighted average of the RMSTs of 2 subgroups in each arm. Reported hazard ratios from the forest plots between the arms were used to enforce the relationship among subgroups’ hazard rates in the model. The performance of the model was tested in a real-life test set of 8 RCTs in advanced-stage gastrointestinal tumors, which also reported KM curves for overall survival (OS) for 40 subgroups as well as in 42 synthetic test cases with 168 subgroups as a benchmark. For each subgroup, predicted median survival, OS rates, and the RMSTs were compared against their actual counterparts as well as their 95% confidence intervals (CIs). Results: Predicted median survivals and RMSTs were within the 95% CIs of the reported values in 32 (80%) and 34 (85%) of 40 subgroups in real-life test cases and in 163 (97%) and 146 (87%) of 168 subgroups in synthetic test cases, respectively. Across all cases, on average, the predicted survival curves laid within the 95% CIs of reported KM curves 71% and 97% of the time in real-life and synthetic test cases, respectively. Discussion: Our study offers a useful and scalable method for extracting subgroup-specific survival from aggregate RCT data to enable subgroup-specific indirect comparisons, and cost-utility and meta-analyses. Higlights: Most randomized controlled trials report survival curves for the overall patient population but do not provide subgroup-specific survival curves, which are crucial for cost-effectiveness analyses and meta-analyses focusing on these subgroups. This study developed an optimization modeling approach to elicit unreported subgroup-specific survival curves from aggregate trial data. The proposed modeling approach accurately predicted the reported subgroup-specific survival curves in 42 simulated test cases with 168 subgroups overall, in which each subgroup-specific survival curve was assumed to followed an exponential distribution. The performance of the proposed modeling approach was sensitive to the assumptions when it was tested using a real-life test set of 8 oncology trials, which also reported survival curves for a total of 40 subgroups.

Introduction. Calibration, a critical step in the development of simulation models, involves adjusting unobservable parameters to ensure that the outcomes of the model closely align with observed target data. This process is particularly vital in cancer simulation models with a natural history component, where direct data to inform natural history parameters are rarely available. Methods. We conducted a scoping review of studies published from 1980 to August 11, 2024, using keyword searches in PubMed and Web of Science. Eligible studies included cancer simulation models with a natural history component that used calibration methods for parameter estimation. Results. A total of 117 studies met the inclusion criteria. Nearly all studies ( n = 115) specified calibration targets, while most studies ( n = 91) described the parameter search algorithms used. Goodness-of-fit metrics ( n = 87), acceptance criteria ( n = 53), and stopping rule ( n = 46) were reported less frequently. The most commonly used calibration targets were incidence, mortality, and prevalence, typically drawn from cancer registries and observational studies. Mean squared error was the most commonly used goodness-of-fit measure. Random search was the predominant method for parameter search, followed by the Bayesian approach and the Nelder–Mead method. Discussion. Despite recent advances in machine learning, such algorithms remain underutilized in the calibration of cancer simulation models. Further research is needed to compare the efficiency of different parameter search algorithms used for calibration. Highlights: This work reviewed the literature of cancer simulation models with a natural history component and identified the calibration approaches used in these models with respect to the following attributes: cancer type, calibration target data source, calibration target type, goodness-of-fit metrics, search algorithms, acceptance criteria, stopping rule, computational time, modeling approach, and model stochasticity. Random search has been the predominant method for parameter search, followed by Bayesian approach and Nelder–Mead method. Machine learning–based algorithms, despite their fast advancement in the recent decade, have been underutilized in the cancer simulation models. Furthermore, more research is needed to compare different parameter search algorithms used for calibration.

Background: Cancer simulation models can answer research and policy questions when prospective evidence is incomplete or not feasible. However, such models require incorporating unmeasureable inputs for which there is often not strong evidence, and model utility is limited if assumptions lack face validity or if the model is not clinically relevant. We systematically incorporated formal advisory input to mitigate these challenges as we developed a microsimulation model of papillary thyroid cancer (PApillary Thyroid CArcinoma Microsimulation model [PATCAM]). Methods: We used a participatory action research approach incorporating focus group techniques and using principles of bidirectional learning. Results: We assembled a formal standing advisory group with representation by perspective (medical, patient, and payor), geography, and local practice culture to understand current and historical clinical beliefs and practices about thyroid cancer diagnosis and treatment. The group provided input on critical modeling assumptions and decisions: 1) the role of nodule size in biopsy decisions, 2) trends in provider biopsy behavior, 3) specialty propensity to biopsy, 4) population prevalence of thyroid cancer over time, 5) proportion of malignant tumors showing regression, and 6) cancer epidemiology and diagnostic practices by sex and age. Advisory group questions and concerns about model development will inform future research questions and strategies to communicate and disseminate model results. Conclusions: We successfully used our advisory group to provide critical inputs on unmeasurable assumptions, increasing the face validity of our model. The use of a standing advisory group improved model transparency and contributed to future research plans and dissemination of model results so they can have maximum impact when guiding clinical decisions and policy. Highlights: Unfamiliarity with simulation modeling poses a threat to its acceptability and adoption. The effectiveness of these models is contingent on end-users’ willingness to accept and adopt model results. The effectiveness of the models is further limited if they lack face validity to potential users or do not have clinical relevance. Several approaches to overcoming validity challenges have been advanced, such as collaborative modeling, which involves developing multiple models independently using common data sources. However, when only a single model exists, another approach is needed. We used an Advisory Group and “participatory modeling,” which has been used in other settings but has not been previously reported in cancer modeling. We describe the methods used for and results of incorporating a formal advisory group into the development of a cancer microsimulation model. The use of a formal, standing advisory group (as opposed to one-off focus groups or interviews) strengthened our model by rigorously vetting modeling assumptions and model inputs with subject matter experts. The formal, ongoing structure promoted transparency. The group education in cancer modeling improved participant ability to provide useful input and may help with dissemination. The advisory group also provided critical feedback about how to effectively communicate model results and informed planned future research questions.

The National Cancer Institute–funded Cancer Intervention and Surveillance Modeling Network (CISNET) breast cancer mathematical models have been increasingly utilized by policymakers to address breast cancer screening policy decisions and influence clinical practice. These well-established and validated models have a successful track record of use in collaborations spanning over 2 decades. While mathematical modeling is a valuable approach to translate short-term screening performance data into long-term breast cancer outcomes, it is inherently complex and requires numerous inputs to approximate the impacts of breast cancer screening. This review article describes the 6 independently developed CISNET breast cancer models, with a particular focus on how they represent breast cancer screening and estimate the contribution of screening to breast cancer mortality reduction and improvements in life expectancy. We also describe differences in structures and assumptions across the models and how variation in model results can highlight areas of uncertainty. Finally, we offer insight into how the results generated by the models can be used to aid decision-making regarding breast cancer screening policy.

Calibration, a critical step in the development of simulation models, involves adjusting unobservable parameters to ensure that the outcomes of the model closely align with observed target data. This process is particularly vital in cancer simulation models with a natural history component where direct data to inform natural history parameters are rarely available. This work reviews the literature of cancer simulation models with a natural history component and identifies the calibration approaches used in these models with respect to the following attributes: calibration target, goodness-of-fit (GOF) measure, parameter search algorithm, acceptance criteria, and stopping rules. After a comprehensive search of the PubMed database from 1981 to June 2023, 68 studies were included in the review. Nearly all (n=66) articles specified the calibration targets, and most articles (n=56) specified the parameter search algorithms they used, whereas goodness-of-fit metric (n=51) and acceptance criteria/stopping rule (n=45) were reported for fewer times. The most frequently used calibration targets were incidence, mortality, and prevalence, whose data sources primarily come from cancer registries and observational studies. The most used goodness-of-fit measure was weighted mean squared error. Random search has been the predominant method for parameter search, followed by grid search and Nelder-mead method. Machine learning-based algorithms, despite their fast advancement in the recent decade, has been underutilized in the cancer simulation models. More research is needed to compare different parameter search algorithms used for calibration. Key points: This work reviewed the literature of cancer simulation models with a natural history component and identified the calibration approaches used in these models with respect to the following attributes: calibration target, goodness-of-fit (GOF) measure, parameter search algorithm, acceptance criteria, and stopping rules. Random search has been the predominant method for parameter search, followed by grid search and Nelder-mead method. Machine learning-based algorithms, despite their fast advancement in the recent decade, has been underutilized in the cancer simulation models. Furthermore, more research is needed to compare different parameter search algorithms used for calibration.

Objective: To quantify the resource use of revising breast cancer screening guidelines to include average-risk women aged 40–49 years across Canada from 2024 to 2043 using a validated microsimulation model. Setting: OncoSim-Breast microsimulation platform was used to simulate the entire Canadian population in 2015–2051. Methods: We compared resource use between current screening guidelines (biennial screening ages 50–74) and alternate screening scenarios, which included annual and biennial screening for ages 40–49 and ages 45–49, followed by biennial screening ages 50–74. We estimated absolute and relative differences in number of screens, abnormal screening recalls without cancer, total and negative biopsies, screen-detected cancers, stage of diagnosis, and breast cancer deaths averted. Results: Compared with current guidelines in Canada, the most intensive screening scenario (annual screening ages 40–49) would result in 13.3% increases in the number of screens and abnormal screening recalls without cancer whereas the least intensive scenario (biennial screening ages 45–49) would result in a 3.4% increase in number of screens and 3.8% increase in number of abnormal screening recalls without cancer. More intensive screening would be associated with fewer stage II, III, and IV diagnoses, and more breast cancer deaths averted. Conclusions: Revising breast cancer screening in Canada to include average-risk women aged 40–49 would detect cancers earlier leading to fewer breast cancer deaths. To realize this potential clinical benefit, a considerable increase in screening resources would be required in terms of number of screens and screen follow-ups. Further economic analyses are required to fully understand cost and budget implications.