David T. Levy, Ph.D. Lombardi Comprehensive Cancer Center
Computational Models Simulation models/computational models are used in other fields, but are increasingly common in public health, especially in the fields of tobacco control and obesity Models are especially useful where there are dynamic systems with many stages (e.g., policy -> environment -> behaviors -> health outcomes) and where the effects unfold over time. Models attempt to make the connections between stages across stages and over time explicit, focusing on the movement of whole system rather than an isolated part
Characteristics of Modeling Generally combine data and parameters from different sources Provides structure by developing a framework and making assumptions explicit Incorporates the effects that are difficult to distinguish empirically in statistical studies Non-linear relationships Interdependencies Dynamic processes Feedback loops
Types of Model Macro-simulations: groups of individuals (e.g., current, former and never smokers) Uni-directional causality Systems dynamic (feedback loops) Micro-simulations: individuals in proportion to their composition in the population Monte-Carlo Agent-based and network models; make explicit assumptions about behaviors
Tobacco Control and Smoking Tobacco control policies provide an example of one the greatest public health success stories important to study what type of policies work in tobacco control and lessons for other public health risks Smoking is a behavioral risk factor with clearest link to cancer- can study the role of dose, duration, and age; and the interaction with other non-cancer chronic diseases
What is SimSmoke? SimSmoke simulates the dynamics of smoking rates and smoking-attributed deaths in a State or Nation, and the effects of policies on those outcomes. Compartmental (macro) model with smokers, exsmokers and never smokers evolving through time by age and gender. Focus on tobacco control policies Effects vary by: depending on the way the policy is implemented, by age and gender the length of time that the policy is in effect Nonlinear and interactive effects of policies
SimSmoke: Basic Approach Policy Changes Taxes Clean air laws Media Camp. Marketing Bans Warning labels Cessation Tx Youth Access Norms, Attitudes, Opportunities Cigarette Use Former and current smokers, relative risks Smoking- Attributable Deaths Total Mortality and by type: Lung cancer Other cancers Heart disease Stroke COPD MCH Outcomes
Relationship between policies and smoking rates based on: Evidence from tobacco and other risky behavior literature, Theories (Economics, Sociology, Psychology, Epidemiology, etc), and Advice by a multidisciplinary expert panel
Policies based on FCTC/MPOWER Cigarette excise taxes: Through prices Smoke-Free Air Laws: Worksites, restaurant and bars, other public places Tobacco control/media campaigns Marketing/Advertising Bans Health Warnings Cessation Treatment: Availability of pharmacotherapy, cessation treatment (financial access, quitlines and web-based treatment Youth access (minimum purchase age): enforcement and vending and self-service bans
Past vs. Future Tracking Period- starts from year where requisite data available, e.g., 1993 for most US models, and continues to the current recent year. The tracking period is used to: Calibrate the model- adjust the parameters Validate the model- test how well it predicts Examine the role of past policies Future Projection- examine the effect of policies from current year forward, e.g., the effect of a cigarette tax increase or the ability to reach the Healthy People 2020 smoking prevalence goal of 12%
Models built for: 32 Countries: Albania*, Argentina*, Bangladesh, Brazil,* China, Czech Republic,* Egypt, Finland,* France,* Germany,* Great Britain,* India, Indonesia, Ireland,* Italy,* Japan,* Korea*, Malaysia, Mexico, Netherlands*, Pakistan, Poland, Philippines, Taiwan*, Russia, Spain, Sweden, Thailand,* Turkey, Ukraine, US,* Vietnam* 6 States: Arizona*, California*, Kentucky*, Massachusetts, Minnesota,* NY * Paper published
Policymakers have used models for: ADVOCACY: Justification by forecasting future tobacco use and health outcomes and showing the effect of past policies PLANNING: Estimate the likely impact of alternative interventions in specific situations and on specific populations Assess and rank strategies for reaching goals prior to commitment of resources Develop more systematic surveillance and evaluation networks HEURISTIC: Understanding the complex network of policies surrounding tobacco use and health outcomes at research and policy-making levels.
Counterfactuals: Brazil Past Policies 1989-2010 To consider the effect of all policies implemented since 1989, we first set policies through 2010 to their 1989 levels to obtain the counterfactual smoking rates in the absence of post-1989 policies. The difference between the smoking prevalence with polices at 1989 levels and the smoking rate with actual policies implemented yields the net effect of policies implemented since 1989. For the role of single policies, we compared the scenario with only that policy implemented to the counterfactual policy scenario. The impact of policies on deaths was estimated by subtracting the number of SADs with policies implemented from their number with policies kept at 1989 levels.
45.0% Brazil Counterfactuals: Smoking prevalence 1989-2010 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Status quo Counterfactual price only
Percent of the Reduction in 2010 Smoking Prevalence* Due to Individual Policies Implemented Since 1989 9.8% 0.3% 7.8% 13.7% 48.4% 6.3% 13.6% Price only Smoke-Free Air only Media only Advertising only Health warnings only Cessation tx only Youth Access only
Effect of Policies Implemented: 1989-2010 Policy Implementation Year SMOKING PREVALENCE 1989 2000 2010 2010 Lower Bound a 2010 Upper Bound a 2050 Counterfactual: all policies at 1989 level 35.4% 32.6% 31.0% 24.9% All policies implemented 35.4% 23.7% 16.8% 22.2% 10.5% 10.3% Percent reduction in smoking prevalence from policy change a All policies 27.4% 45.9% 27.8% 66.4% 59.1% SMOKING ATTRIBUTABLE DEATHS 1989 2010 Cumulative 2010 2010 Lower Bound a 2010 Upper Bound a Cumulative 2050 Counterfactual: all policies at 1989 level 181,957 283,048 4,998,024 20,401,516 All policies implemented 181,957 225,048 4,578,810 4,739,196 4,282,963 13,471,388 Deaths averted from policy change All policies 58,000 419,214 258,828 715,061 6,930,128 Low birth weight babies avoided from policy changes 1989-2010 With ;policies implemented 14,827 704,976
Advocacy: Other successes due to tobacco policies Percent reduction in smoking prevalence (18 and above): > 30% reduction Brazil (almost 50% reduction due to policies) California At least 25% Reduction United Kingdom Minnesota Thailand 20% Reduction Arizona Korea Ireland NYS Netherlands
Advocacy: There may be limits to current policies: We may need more than traditional policies to reduce smoking by more than 50% Those with the weakest current policies (e.g., Russia and China) show the potential for largest reductions in smoking prevalence, with forecasts of about a 50% reduction in smoking prevalence in going from very limited policies to fully FCTC-consistent policies How can we surpass a 50% reduction? Improved cessation treatments, e.g. better and more tailored interventions with follow-up and integrated services May need to alter the tobacco products available, e.g., reduce nicotine and other addictive constituents or disallow current cigarettes in favor of safer forms of tobacco
Planning: Male Smoking Prevalence: SimSmoke Predictions vs. Surveys, Minnesota 28.0% 26.0% 24.0% 22.0% 20.0% 18.0% 16.0% 14.0% 12.0% 10.0% 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 SimSmoke CPS-TUS MATS
Ireland Male Smoking Prevalence,1998-2010 data, data, data
Mexico: Many surveys ENA, ENAULT, GATS Ask different questions, may have important implications for some day smokers vs every day smokers
Planning: Ranking the effect of future policies Brazil SimSmoke smoking prevalence Policies/Year Status quo 16.8% All FCTC policies implemented 16.8% 2010 2015 2050 Smoking Prevalence 15.5% 10.3% Lower Bound Upper Bound 11.9% 6.3% 7.3% 4.7% Cumulative 2011 2050 Lower Bound 2011 2050 Smoking Attributable Deaths Upper Bound 2011 2050 8,892,578 9,513,874 8,749,842 Reduction in Smoking Prevalence 7,563,664 8,657,395 6,783,055 Independent policy effects Tax at 75% of retail price 10.2% 16.7% 13.0% 21.5% Well-enforced smoke-free air laws 4.5% 6.4% 3.1% 9.5% Well-enforced marketing ban 3.0% 4.8% 2.4% 7.2% High-intensity media campaign 4.8% 7.4% 3.6% 10.9% Cessation treatment programs 2.3% 4.6% 6.9% 9.3% 469,463 365,730 565,492 268,042 135,972 396,336 171,180 86,231 254,867 305,436 157,126 459,018 198,382 100,530 489,257 Well-enforced youth access restrictions 0.8% 5.1% 0.0% 10.1% 28,491 0 42,734 With all policies implemented 23.5% 38.5% 29.0% 54.0% 1,328,914 856,474 1,966,787
Planning: Health Effects Delayed SimSmoke Projections Smoking-Attributable Deaths Status Quo vs. All FCTC Policies for Finland More immediate impact on heart disease and maternal and child health
Planning: There may be limits to current policies: We may need more than traditional policies to reduce smoking by more than 50% Those with the weakest current policies (e.g., Russia and China) show the potential for largest reductions in smoking prevalence, with forecasts of about a 50% reduction in smoking prevalence in going from very limited policies to fully FCTC-consistent policies How can we surpass a 50% reduction? Improved cessation treatments, e.g. better and more tailored interventions with follow-up and integrated services May need to alter the tobacco products available, e.g., reduce nicotine and other addictive constituents or disallow current cigarettes in favor of safer forms of tobacco
FDA Public health standard Public health standard calls for the review of the scientific evidence regarding 1. Risks and benefits of the tobacco product standard to the population as a whole, including both users and non-users of tobacco products; 2. Whether there is an increased or decreased likelihood that existing users of tobacco products will stop using such products; and 3. Whether there is an increased or decreased likelihood that those who do not currently use tobacco products, most notably youth, will start to use tobacco products Example: Mandatory product standards that would limit the allowable levels of ingredients in tobacco products (menthol, nicotine, etc) 25
Planning: Modeling the effects of a ban on menthol cigarettes Possible effects of a ban: Menthol smokers switch to non-menthol brand. Menthol smokers quit at differential rate than if non-menthol smoker. Some individuals who would have initiated smoking with menthol cigarettes never start. Scenarios investigated: 1. 10% of the former menthol smokers quit and 10% of those who would have initiated as menthol smokers never smoke; 2. 20% quit and 20% do not initiate, and; 3. 30% quit and 30% do not initiate
P l a n n i n g M o d e l i n g a M e n t h o l B a n U s i n g S i m S m o k e 27
Heuristic: Youth Access Policy Past literature suggests youth access policies lead to increased retail compliance. Effects on actual smoking rates are unclear. Two potential reasons Role of non-retail sources of cigarettes (parents older friends theft) Level and extent of policies
Heuristic: Policy Components Affecting Enforcement Publicity/ Education Compliance Checks Per Year Penalties Multiplicative relationship Anti-tobacco Norms Compliance S-shaped curve, subject to substitution into other sources Reduced Smoking Originally applied to youth access, but applies to marketing restrictions and smoke-free air laws
Heuristic: The Decision to Quit Current Smoker Attempts to Quit Self Quit Rx Pharm. NRT OTC Behavioral Treatment Success Fail Success Fail Success Fail Success Fail No quit attempt Continues Smoking Behavioral & Rx Pharm Behavioral & NRT OTC Framework used to show effects for specific policies Success Fail Success Fail
Heuristic: Cessation Treatment Policies AVAILABILITY: Ability to obtain NRT, Buproprion and Varenecline by Rx or over-the counter FINANCIAL ACCESS: payment or mandatory coverage for cessation treatments Prescription or OTC pharmacotherapies alone Behavioral treatment alone Pharmacotherapies and behavioral QUITLINES: delivered by government and coordinated through health care system BRIEF INTERVENTIONS: delivered by health care providers Web-based treatment: supervised and used by health care agencies of provider Follow-up of Care: health care providers, quitlines, web Each of the above affects quit attempts and treatment use with potential interactions (synergies among policies)
Heuristic: Smokeless as Harm Reduction Harm reduction: As a substitute for cigarettes (provides the nicotine fix), it has been suggested that use of at least some smokeless can reduce overall harm, because of lower health risk, similar to methadone for heroine addicts. Smokeless risks less than cigarettes (which are not inhaled into lung), but depends on contents, also no second hand smoke. Potentially harm increasing, if: If smokeless leads to increased youth initiation and acts as a gateway to cigarettes Encourages dual use with cigarettes instead of cessation from cigarettes
Heuristic: Health effects and polytobacco use: simple example with only cigarettes and smokeless Initiation cigarette use Initiation smokeless use Sole cigarette use (habit) Dual cigarette & smokeless habit Sole smokeless us (habit) Cigarette only attributable death Dual use attributable death Smokeless only attributable death Need to know relative risks for those who continue to use and for former users
Tobacco Use in Sweden, Males, 2004-2020 18.0% 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% Declines in cigarette use accompanied by constant rates of sole and dual use of snus, suggesting that users are shifting from single to dual use 0.0% 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 Male Cigarette Use (alone) Male Snus Use (alone) Male Combined Snus and Cigarette Use 34
New Tobacco Products Will be important to consider whether smokers become new users or dual users Whether youth use these products instead of cigarettes and whether they eventually use cigarettes Whether former smokers use these products and then become smokers 35
Heuristic: Tobacco control is complex: Modeling provides a framework Industry behavior Tobacco, retail Tobacco Control Policy Taxes, laws, regulations Environment Attitudes, norms, opportunities (economic, other) Physiology Genetics, diet, other Risky behaviors: Using cigarettes, cigars, and smokeless and other non-combustibles Health Outcomes Death, disease, dollars Limited evidence for many of these linkages, models provide guidance on areas for future research
Heuristic: Future challenges for Sim- Smoke and tobacco control modeling Better understanding of the initiation and cessation process Constantly changing market with new products and dual uses for cigarettes, smokeless, cigars, and pipes; transitions in the use of the different products is unlikely to be stable Difficult to anticipate industry reactions to policies both in consumer markets and in the political arena Need to consider the heterogeneity of nations and individuals; tobacco users are increasingly low SES in MICs and general population in LMICs with economic growth
Collaborative Modeling Since different models will highlight different aspects of the problem, information from the different models will need to be combined in a systematic manner An example is NCI s CISNET program: The models consider common research questions using a natural history of disease framework The models use a common data sources to help identify reasons for any differences results The results are compared to provide a reasonable range of outcomes for decision-makers Models are well documented using publicly available model profiler Georgetown University is home for smoking/lung group Levy and coordinating center for the breast cancer group
Para la elaboración de constancias, favor de enviar lista de participantes presenciales con: Profra. Berta Luz Téllez btellez@insp.mx Videos y presentaciones anteriores en: http://www.inspvirtual.mx -Videoconferencias https://www.facebook.com/videoconferenciasinsp