Using simulation modeling to guide policy to reduce disparities and achieve equity in cancer outcomes: state of the science and a road map for the future

Cancer mortality in the United States has steadily decreased over time, and this trend has largely been attributed to the discovery and implementation of effective cancer prevention, early detection, and systemic therapies (1-6). These benefits, however, have not been realized equally across the population, with African American or Black populations (hereinafter referred to as “Black”) having higher a incidence of several cancer types, such as cervical, lung, and prostate cancer, and disproportionate mortality rates for all the major cancer types (6). In this issue of the Journal, teams of researchers collaborated to develop and use population simulation models to quantify the contributions of cancer risk, early detection, treatment and survival, and noncancer mortality to cancer disparities between the Black population and the overall US population. In a series of papers, mortality was modeled for cancer sites where the Black population has higher mortality rates than the overall population, even when they experienced lower incidence (breast, cervical, colorectal, lung, and prostate cancer). Other articles in this issue highlight future research needs prerequisite to expanding modeling approaches, including contextualizing modeling in a conceptual framework of the effects of systemic racism, where systemic racism is defined broadly as policies and practices that perpetuate unfair treatment of people of color (7). Greater uniformity in measuring systemic racism and collection of data linking specific aspects of systemic racism to downstream impacts on model input parameters will be necessary to implement future modeling within this context. As noted in the commentary by Winn and colleagues (8), the lack of such data reflect a “data divide” and can create data deserts, where there are limited data for an entire population of people based on race and other characteristics. This data divide means that many people may not completely benefit from innovative, personalized cancer care that is increasingly data driven, perpetuating cancer disparities (8). Overall, the body of research in this issue is intended to inform policymakers and researchers about how modeling can be part of the solutions by synthesizing the best-quality existing data, highlighting data gaps and illustrating interventions that are most likely to reduce disparities and achieve racial equity in cancer burden. Population simulation modeling is 1 research method recommended by the National Academy of Sciences and others to evaluate the comparative effectiveness of different interventions (9). The Cancer Intervention and Surveillance Modeling Network (CISNET) has been continuously funded by the National Cancer Institute since the year 2000 to advance modeling science, guide public health research and priorities, and support the development of optimal cancer control strategies. Although past efforts have had a limited focus on incorporating race into simulation models, many aspects of the CISNET modeling approach readily lend themselves to addressing questions about strategies that could achieve equity in cancer outcomes though equal opportunity to access high-quality prevention, early detection, and treatment (10). The CISNET models provide a virtual laboratory to synthesize the best available data on specific components of the cancer control process, from prevention and early detection to diagnosis, treatment, and survivorship care, and their effects on population trends in overall incidence and mortality over the life course of populations (11). By 2023, CISNET included modeling of 10 cancer sites—bladder, breast, cervical, colorectal, endometrial, esophageal, gastric, lung, prostate cancer, and multiple myeloma—with bladder, endometrial, and gastric cancer as well as multiple myeloma funded most recently. Within each cancer group, 2 or more independent teams collaborate to use common data inputs and evaluate how results vary logically by model based on specific assumptions and how well the models replicate observed cancer trends over time or predict results in clinical trials (12-14). In this issue, the CISNET modeling teams from the breast, cervical, colorectal, lung, and prostate cancer groups report on results of collaboration to estimate, and the bladder, endometrial, and gastric cancer as well as the multiple myeloma groups develop plans to estimate the contributions of cancer risk, screening, timely diagnosis, treatment, and survival to differences in cancer mortality between the US Black population and the overall US population. The models used or plan to use racial group–specific data from clinical trials, national registries, nationally representative surveys, and observational studies to develop input parameters for model components, including risk factor exposure, access to screening and treatment, and variation in tumor biology and response to therapy; ancestry was not considered (Figure 1). Use of clinical trial data was possible only when Black individuals were included in sufficient numbers to estimate effects separately by racial group (15). Cancer Intervention and Surveillance Modeling Network model components After development and incorporation of input parameters by racial group, the breast, cervical, colorectal, lung, and prostate cancer models estimated life events for each individual from each respective population in the absence of any intervention, such as prevention, screening, or treatment. Model analyses were then replicated for each individual, representing a counterfactual version of a person’s life, depicting changes in incidence, stage, and mortality in the presence of interventions through prevention of incident cases, detection of smaller tumors with screening, or changes in survival based on treatment. To quantify the impact of each intervention on cancer mortality, most of the teams started with a model of cancer mortality in the overall population and sequentially replaced model parameters 1 by 1 with parameters for Black populations. Results for each cancer type highlighted both similar and different patterns based on the features of each cancer site. Breast cancer is unique because despite somewhat lower incidence rates, Black women have higher mortality rates than White women (16). In this issue of the Journal (17), 3 CISNET breast cancer modeling teams used a common set of racial group–specific input parameters for incidence, screening, therapy, and competing mortality to estimate the portion of mortality disparities each set of parameters explains. The treatment parameter included both the initiation of systemic therapy and the collective impact of disparities in delays between diagnosis and treatment initiation, dose reductions and incomplete cycles, and nonstandard regimens to estimate actual effectiveness vs efficacy observed in clinical trials (18). After confirming that modeled incidence and mortality rates closely matched observed Surveillance, Epidemiology, and End Results program data, the model results indicated that in 2019, racial differences incidence and competing mortality would have reduced mortality disparities by a net ‒1% because Black women have lower age-adjusted incidence. Changes in tumor subtype and stage distributions accounted for a mean of 20% (range across models = 13%-24%), and screening accounted for a mean of 3% (range = 3%-4%) of the modeled mortality disparities between the Black and the overall population (Table 1). Treatment parameters accounted for the majority of modeled mortality disparities for Black women, with a mean of 17% (range = 16%-19%) for treatment initiation and a mean of 61% (range = 57%-63%) for effectiveness (Table 1). Summary of the percentage of modeled cancer mortality in 2019 among Black people, explained by parameters representing different phases of the Cancer Control Continuuma,b The data in this table were generated using several steps. First, mortality was modeled for each population using racial group–specific data. Next, the overall difference in modeled mortality between the Black and overall populations was calculated. Then, starting with the overall US population, parameters for the Black population were sequentially substituted into the overall population model. At each step in the substitution, the difference between the modeled mortality in that step and the mortality in the overall population model was calculated. That latter difference was divided by the overall difference in modeled mortality between the Black and overall populations to generate the percentages summarized in the table. NA = the component was not included in the model. All percentages total 100% within each column. Adjusted for differences in smoking rates because Black people have higher incidence despite lower pack-year histories; results are for the 1960 birth cohort. The breast analyses are for women. Because the breast analysis was conducted by 3 groups, the mean results are presented. Note that breast cancer incidence is lower in Black vs all women, leading to a 7% lower mortality (‒7%), but competing noncancer mortality accounted for 6% of the modeled mortality in Black women, for a net impact of this component of ‒1%. This component considers competing noncancer mortality and hysterectomy rates. The estimate for prostate cancer incidence includes racial differences in the population age distribution and differences in noncancer mortality. Age-standardized rates. The rates for lung cancer are among people born in 1960 and aged 59 years in 2019; all other rates are among multiple birth cohorts. Summary of the percentage of modeled cancer mortality in 2019 among Black people, explained by parameters representing different phases of the Cancer Control Continuuma,b The data in this table were generated using several steps. First, mortality was modeled for each population using racial group–specific data. Next, the overall difference in modeled mortality between the Black and overall populations was calculated. Then, starting with the overall US population, parameters for the Black population were sequentially substituted into the overall population model. At each step in the substitution, the difference between the modeled mortality in that step and the mortality in the overall population model was calculated. That latter difference was divided by the overall difference in modeled mortality between the Black and overall populations to generate the percentages summarized in the table. NA = the component was not included in the model. All percentages total 100% within each column. Adjusted for differences in smoking rates because Black people have higher incidence despite lower pack-year histories; results are for the 1960 birth cohort. The breast analyses are for women. Because the breast analysis was conducted by 3 groups, the mean results are presented. Note that breast cancer incidence is lower in Black vs all women, leading to a 7% lower mortality (‒7%), but competing noncancer mortality accounted for 6% of the modeled mortality in Black women, for a net impact of this component of ‒1%. This component considers competing noncancer mortality and hysterectomy rates. The estimate for prostate cancer incidence includes racial differences in the population age distribution and differences in noncancer mortality. Age-standardized rates. The rates for lung cancer are among people born in 1960 and aged 59 years in 2019; all other rates are among multiple birth cohorts. Cervical cancer is another cancer with known racial mortality disparities and has an intervention (colposcopy) that can remove precancerous lesions and avoid invasive cancer incidence and mortality (19). To understand factors that contribute to racial disparities, Spencer and colleagues (20) adapted an established microsimulation model of human papillomavirus (HPV) infection and cervical cancer to reflect demographic, screening, and survival data for populations of Black women and compared results to a model reflecting data for all US women. Input parameters for all women were sequentially replaced with Black race specific for all-cause mortality, hysterectomy rates, screening frequency, screening modality (Papanicolaou smear vs HPV plus Papanicolaou smear co-testing), diagnostic follow-up, and cancer treatment and survival. They found that lower use of co-testing relative to Papanicolaou-only testing among populations of Black women increased disparities in both incidence and mortality. Differences in screening rates, timeliness of follow-up of abnormal screening test results, and stage-specific treatment-related survival also contributed to disparities. Overall, 70% of mortality disparities were due to stage-specific survival, followed by timeliness of screening follow-up and screening frequency (Table 1). Differences in HPV rates did not explain disparities and suggest that access to timely diagnosis and recommended treatment could improve equity in cervical cancer mortality. Notably, this was the only cancer site to model timely follow-up after an abnormal cancer screening test. As noted in the commentary by Doubeni and colleagues (21), considering follow-up will become increasingly important in future modeling of all cancer screening and has begun to be addressed in US Preventive Services Task Force recommendations and coverage decisions. Colorectal cancer, like cervical cancer, has known precursor lesions. Detection and removal of precursor lesions at colonoscopy prevents their progression to invasive cancer, so that access to colonoscopy can strongly influence colorectal cancer incidence and mortality. In this issue, Rutter and colleagues (22) estimated screening effectiveness in Black vs White individuals (in contrast to the other models in this issue that compared Black populations with the overall population). They first updated their model to include recent changes in incidence rates for Black and White individuals aged 20 to 44 years, including race-specific data on tumor location. They then compared screening effectiveness in Black and White individuals born in 1970 (single birth cohort), considering racial group–specific competing mortality, tumor location, and survival after treatment. The results suggested that Black and White individuals with access to the same screening quality and modality would have similar benefits from screening but that the benefit of screening will be reduced when colonoscopy quality is reduced (ie, receipt of colonoscopy from a gastroenterologist with low adenoma detection rates). The screening benefits also depended on treatment-related survival, with greater benefits when the most effective therapy was delivered. Lung cancer, the leading cause of cancer death in the United States, has substantial racial disparities in mortality, despite Black populations having lower smoking levels than the overall population (23). In this issue, Skolnick and colleagues adapted an established CISNET lung cancer model to evaluate how different portions of the cancer control process, from risk factors to screening and treatment, affected mortality disparities among people born in 1950 or 1960 (single-cohort analyses) (24). Given the relative historical paucity of screening and curative therapies for lung cancer, the racial differences in incidence adjusted for smoking differences accounted for about 90% of the mortality disparities for both men and women (Table 1). The authors stress, however, that with increasing use of lung cancer screening and a growing number of effective therapies, ensuring equity in these aspects of care will be important to avoid exacerbating inequities. Prostate cancer deaths in the United States remain significantly higher in the Black populations than in the overall populations of men (19). In the paper by Gulati and colleagues (25), 1 of the CISNET prostate models was adapted to include data for Black men. Similar to the approach that the breast, cervical, and lung models used, the prostate analysis began with the model for the overall US population, and then sequentially replaced parameters with data for Black men. The authors found that differences in underlying disease risk explained 36% of the modeled mortality, while more aggressive tumor features in Black men explained 34% and worse cancer-specific survival explained 31% of mortality (Table 1). Contributions from differences in historical screening (‒1%) were small, but this result may reflect the assumption that the efficacy of prostate-specific antigen screening seen in randomized clinical trials did not vary by race. Taken together, the results of the modeling analyses included several notable patterns (Figure 2). In cancers with large racial incidence disparities and widely used screening procedures (eg, cervix), access to and the quality of screening played a larger role than in cancers with smaller racial group differences in incidence (eg, breast). The results also suggested opportunities to reduce disparities through primary prevention using HPV vaccination for cervical cancer or smoking cessation to reduce lung cancer. Percentage of modeled mortality disparities in 2019 between the Black population and the overall population explained by model components, by cancer and gender For cervix and colorectal cancers, where screening can detect and remove precancerous lesions, improving high-quality screening use can have a substantial impact on mortality disparities. In other cancers, such as breast cancer, where screening rates and quality are high and fairly comparable in Black and all people, this portion of the cancer control process was much less impactful in terms of mortality disparities than the quality of treatment. Other common findings were that the portion of mortality differences explained by incidence-related vs survival-related parameters depended on the availability of effective treatments, especially in lung cancer, where many cases are diagnosed in advanced stages. With current shifts to earlier stages because of increasing use of lung screening and development of more effective treatments, such as and these components of care will become increasing in trends in disparities, especially therapies are not equally (24). For cancer sites such as breast cancer, has effective therapies for the most common access to and the effectiveness of treatment explained the majority of the disparities. group differences in access to any future treatment may also be important for modeling mortality disparities in the CISNET sites and gastric cancer as well as multiple are in earlier stages of their and colleagues have an approach in this issue to test how different policies could be to access to the mortality for each cancer data to achieve public health (Figure For an intervention that 20% of a of in over an intervention that of a of 10 in Overall, by the cancer care disease available data on Black people, and equity the commentary by Doubeni and colleagues that the results of this CISNET population simulation modeling can be used to strategies with the for the National Cancer and the Cancer of the cancer death by within cancer mortality rates in 2019 among the overall and Black populations and the mortality in the Black population in this issue, and colleagues conducted a of the on racial disparities in US cancer mortality. articles between and the authors variation in studies mortality disparities in or groups, with only a on American or included cancer registries, or data to from the US or historical data. The authors found different across multiple of disparities. The most used on and with racial the most was by several including and racial and the policies in was with cancer mortality disparities among the US Black population compared with the overall population. of the studies included in the considered multiple disparities or their effects on to cancer risk or treatment The authors recommended that future modeling studies use data that several of racial disparities, including effects of disparities on (eg, stress, that contribute to cancer incidence, and mortality. They also suggested that modeling teams include to the of the modeling approach to The modeling analyses in this issue highlight in data that can be used to other populations with and of disparities to cancer risk, screening and survival. In the by and colleagues these data gaps were and several strategies were to support development of future model in these strategies is of cancer data collection to include factors such as gender and racial group, especially for and The authors also noted that modeling could more reflect the impact of cancer control interventions on cancer disparities cancer and included risk factor data to for cancer screening (eg, of method of detection and the of or diagnostic Although such data collection could in or registries, collection of these data at the national would be important to generate data on the most representative populations. The Surveillance, and has updated data collection to estimate population and racial and group is 1 for such data Other by and colleagues include use of data and of cancer data with data to disparities (eg, and their In the commentary by Winn and data were highlighted as for future modeling because the where has on They also highlighted the of using such as where the model results could provide data on specific to the policymakers for (8). data will not individuals at risk for but they could highlight for and suggest interventions to with the to such as and for screening and diagnostic in the same in to that cancer are within with high cancer mortality rates, and or other for of to or of Overall, data on multiple disparities will be to advance modeling and the of population modeling to guide efforts to cancer As noted in the commentary by Doubeni and the of modeling to racism is a and In the paper in this issue, and colleagues a conceptual framework to guide future data collection and modeling analyses addressing In that systemic racism was as a cause of racism is defined in this framework as the racism in or and practices and that and perpetuate unfair treatment of people of color through of benefits, health care, and (7). These and have health including to and lack of access to cancer care leading to in cancer incidence, and death rates In the conceptual model by and colleagues systemic racism includes racism and of racism include and and policies that access to care, including impact on screening, diagnosis and treatment initiation, and receipt of therapy and survivorship care racism can in of that care, such as assumptions that Black individuals will not treatment, so is not necessary to all to the or that a lack of the effects of systemic racism, including to of racism and and lack of access to care, can leading to an in the incidence of such as and each of can risk of cancer and treatment the efficacy of cancer treatment, and overall mortality (10). As in and the in components of systemic racism could be into future CISNET models through use of data linking of systemic racism to specific model of effects of systemic racism on Cancer Intervention and Surveillance Modeling Network model components that could be used in future analyses of strategies to improve cancer equity The racial group–specific modeling in this issue is the first from the entire CISNET to a research to that could be to equity in cancer among Black populations. The body of research how population simulation modeling can be to public health The studies that high-quality therapy could have a major impact on mortality disparities between the Black and overall especially as and more effective The findings also that solutions will to be for each cancer site because cancers, such as cervical, lung, and prostate cancer, have large incidence by race and available primary prevention strategies. Cervical and colorectal cancer screening also has the unique opportunity to remove precancerous lesions and lower mortality. This body of also data gaps and and conceptual that will to be addressed to more inform efforts to achieve cancer among these is the to representative national data using and linking the effects of systemic racism and to cancer among population groups, including people from racial and and gender groups as well as with Overall, the model studies in this provide a framework for other cancer and disease to their to health This also a to for the modeling to use and to contribute to national and efforts to achieve health by highlighting of modeling to studies and guide of important cancer disparities, this and data to understand the benefits of modeling analyses as an important to the population impact of policies and interventions to achieve cancer This was by the National of National Cancer Institute and This research was also in part by the National Cancer Institute and the National Institute on to and National Cancer Institute to The had role in The and in this are the of the authors and not the of the This as part of the to Cancer Modeling to and by the National Cancer National of Modeling of Breast Cancer Control the 3 The authors of

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.