Introduction Smoking is still the most preventable cause of cancer, and a leading cause of premature mortality and health inequalities in the UK. This study modelled the health and economic impacts of achieving a ‘tobacco-free’ ambition (TFA) where, by 2035, less than 5% of the population smoke tobacco across all socioeconomic groups.
Methods A non-linear multivariate regression model was fitted to cross-sectional smoking data to create projections to 2035. These projections were used to predict the future incidence and costs of 17 smoking-related diseases using a microsimulation approach. The health and economic impacts of achieving a TFA were evaluated against a predicted baseline scenario, where current smoking trends continue.
Results If trends continue, the prevalence of smoking in the UK was projected to be 10% by 2035—well above a TFA. If this ambition were achieved by 2035, it could mean 97 300 +/- 5 300 new cases of smoking-related diseases are avoided by 2035 (tobacco-related cancers: 35 900+/- 4 100; chronic obstructive pulmonary disease: 29 000 +/- 2 700; stroke: 24 900 +/- 2 700; coronary heart disease: 7600 +/- 2 700), including around 12 350 diseases avoided in 2035 alone. The consequence of this health improvement is predicted to avoid £67 +/- 8 million in direct National Health Service and social care costs, and £548 million in non-health costs, in 2035 alone.
Conclusion These findings strengthen the case to set bold targets on long-term declines in smoking prevalence to achieve a tobacco ‘endgame’. Results demonstrate the health and economic benefits that meeting a TFA can achieve over just 20 years. Effective ambitions and policy interventions are needed to reduce the disease and economic burden of smoking.
- End game
- tobacco microsimulation
- disease burden
- economic burden
Statistics from Altmetric.com
Tobacco remains the leading cause of preventable cancer and death in developed countries and is estimated to kill up to two in three long-term users.1 In the UK, an estimated 19% of cancer cases each year are linked to exposure to tobacco smoke.2 It is also a major cause of coronary heart disease (CHD), stroke and chronic obstructive pulmonary disease (COPD).3
The growing burden of non-communicable diseases (NCDs) is evident among individuals, families, health systems and wider society, causing increased pressures on economic productivity, health and welfare resources.4 Despite smoking prevalence having halved over the past 35 years to 16.9% in England in 2015, it is clear more is required to reduce the disease burden caused by tobacco.5 The pervasive and costly nature of preventable NCDs6 7 means policy makers urgently need to establish preventative public health interventions that are both effective and cost-effective.
To address this urgency, many in the tobacco control community support a ‘tobacco-free’ ambition (TFA), in which a 5% smoking prevalence is achieved across all income or socioeconomic groups.8 Countries such as New Zealand, Finland, Scotland and Ireland have adopted this ambition but not the UK.9–12 There is interest among many in the UK tobacco control community, notably the Smokefree Action Coalition, in the UK Government establishing a similar TFA achieving a 5% smoking prevalence across all socioeconomic groups by 2035.
Recent policy commitments, such as the creation of smoking cessation services, the introduction of smoke-free workplaces and implementation of standardised packaging for tobacco products, are effective interventions designed to reduce smoking prevalence by promoting cessation and reducing uptake.13 14 However, these interventions alone cannot drive the UK smoking prevalence to reach 5% by 2035.
Understanding how smoking prevalence will change over time and the resultant impact it has on tobacco-related disease and economic burden is necessary for policy planning. Simulation models can test the impact of policy interventions, demonstrating their effects, such as number of disease cases and costs incurred by the public purse, by changing key parameters. Recent examples of advancing policy debates in tobacco control through simulation modelling include: predicting the impact on disease outcomes or smoking prevalence of changing levels of tobacco taxation15 and the availability of tobacco retail outlets,16 17 as well as appraising progress in countries where a TFA has been established, such as New Zealand.18–20
These models can predict likely consequences of current trends continuing, strengthening the justification for policy interventions.21 A recent systematic review on mathematical modelling in tobacco control notes that while individual models differ, there are benefits in taking a ‘trend and policy’ approach as well as seeking to understand consequential impacts on smoking prevalence, tobacco-related morbidity and mortality, and economic burdens.22
The model presented here projected trends in UK smoking prevalence from 2015 to 2035 as baseline. This baseline was compared with a TFA scenario of 5% smoking prevalence by 2035 in the UK. It evaluated the impact of these changing trends in smoking on the future UK burden of NCDs—namely a range of cancers, CHD, stroke and COPD—and resultant economic consequences as a result of achieving a TFA of 5% smoking prevalence in the total population.
The UK Health Forum (UKHF) microsimulation model
A dual-module modelling process written in C++ software, initially developed by the UKHF as part of the Foresight working group23 and described in more detail in online supplementary appendix 1, was utilised and further developed for this study. In this model, smoking status was included as a single, categorical, risk factor to determine future disease burden. Module 1 uses a non-linear multivariate, categorical regression model fitted to cross-sectional age–sex risk factor data to create longitudinal smoking prevalence projections to 2035, using annual health survey data from 2000 to 2012 obtained from the General Lifestyle Survey (GLS)/General Household Survey (GHS) , and including time as a covariate. Sample weights are included to adjust for categorical data, but we are unable to account for births that are subsequently included in module 2. Each year, an individual has a probability of initiating or quitting smoking based on probabilities provided by these predictions. Given the modular nature of the model, different risk factors can be run through the microsimulation relatively easily by selecting different input parameters, including smoking-related relative risks. Module 2 uses these smoking prevalence trends within a Monte Carlo microsimulation to predict the future incidence, prevalence and costs of 17 smoking-related diseases within a virtual UK population (specified by current and projected Office of National Statistics (ONS) population statistics). The health and economic impacts of a hypothetical policy scenario, a TFA where smoking prevalence gradually drops to 5% from current prevalence across both sexes and all age groups by 2035, were modelled against a predicted baseline scenario based on future projections of current and historical smoking prevalence obtained from the GLS/GHS. This scenario was created by producing a trend to 2035 using a regression model through existing data with the addition of a 5% smoking prevalence data point in 2035. This resulted in a gradual decline in smoking prevalence to 5% by 2035. Decay rates were applied to the relative risks of smokers to determine relative risk of ex-smokers.24
Input data used for the model are displayed in box 1. Smoking prevalence data and socioeconomic status data for Great Britain were extracted from the GHS/GLS, following issue by special licence from the UK Data Service for years 2000 to 2012.25 These data were extrapolated to fit current and future UK population estimates by using ONS National Population Projections.26
UK population distribution data, stratified by age and sex, were used in conjunction with UK mortality distribution data. This was obtained from the ONS along with principal projections data and input into the microsimulation.27 Mortality distributions were used to compute the probability of death for the diseases of interest as well as other unspecified causes of death. Migration, total fertility rates—stratified by mothers’ ages—and mortality rates affected the population size in the simulation over time.
The smoking-related diseases included in the study were 14 different tobacco-related cancers classified by the International Agency for Research on Cancer28 as caused by tobacco smoking (colorectal, laryngeal, liver, oesophageal, oral, bladder, cervical, acute myelogenous leukaemia (AML), chronic myelogenous leukaemia, gallbladder, ovarian, kidney, lung and pancreatic), plus CHD, COPD and stroke. International Classification of Diseases version 10 codes were used to define diseases and are provided in online supplementary appendix 2.
Incidence, prevalence, mortality, survival and relative risks were extracted from the literature and online national databases. Online supplementary appendix 3 provides a reference table of the data inputs used in the model and online supplementary appendix 4 provides data input tables for each of the parameters. Where data were not available, they were calculated using available parameters based on DISMOD equations.29 Calculated data were used to estimate available data by way of validation and to ensure coherence. Online supplementary appendix 1, ‘Survival Statistics CRUK 2010/11’, provides more detail.
Direct NHS health and social care costs were based on healthcare expenditure data obtained from the NHS England Programme Budget Database.30 Non-health societal costs resulting from losses of economic productivity due to premature morbidity and mortality were estimated from a range of data sources including the ONS, the Annual Survey of Hours and Earnings and the Labour Force Survey.31–33 The number of days off work for a given disease was obtained using modelled outputs from a previous health economic modelling project.34 35
Risk factor data
Historical and current prevalence of smoker status (never smoker, ex-smoker and smoker) broken down by age, sex and income quintile.
Most recent incidence, mortality and survival of the diseases of interest, broken down by age and sex.
Relative risk of acquiring the diseases of interest, broken down by age and sex.
Most recent UK population, broken down by age and sex.
Most recent mortality and fertility rates of the UK population.
Health economic data
Mean utility weights of the diseases of interest without medical intervention.
Most recent direct healthcare (National Health Service) costs associated with the diseases of interest.
Most recent non-healthcare costs (non-health societal costs) associated with the diseases of interest.
By 2035, the prevalence of UK smokers is predicted to reach 10% in the baseline scenario. This percentage varies when split by income quintile (figures 1 and 2). For men, 2.4% of those in quintile 1 are predicted to smoke in 2035 compared with 15.7% in the lowest income quintile. A similar pattern is observed for women (2.6% vs 14.3%, respectively). Online supplementary appendix 5 presents the smoking projections by equivalised income quintiles for men and women.
Despite the short time horizon of the study, the TFA scenarios delivered an impact on incidence of diseases relative to the baseline. The incidence of a number of the modelled diseases such as COPD, stroke and lung cancer is predicted to decrease from 2015 to 2035 in the TFA scenario relative to baseline. All results are presented in table 1.
Relative to the baseline scenario, the TFA scenario is predicted to avoid around 12 353 (2.8% of total) new cases of disease in the UK in the year 2035 alone. Note that total diseases are the sum of non-rounded figures. The majority of these cases were tobacco-related cancers (5087; 1.4%), followed by COPD (3633), stroke (2907) and CHD (727) in 2035 alone. Decreased cases of lung cancers are predicted to be the biggest beneficiary out of the cancers, avoiding 2907 cases (4.9%) in the year 2035 alone. Certain diseases, such as COPD, stroke and lung cancer, have yielded a smaller CI, whereas there are higher degrees of uncertainty for certain cancer types, namely renal, gastric, laryngeal, colorectal and hepatic cancers. Note the uncertainty that accompanies the output data represents the accuracy of the microsimulation as opposed to the confidence of the input data itself. Errors around the input data were not available (see online supplementary appendix 1 for more detail).
Achieving a TFA over this 20-year period is predicted to avoid around 97 300 cumulative incident cases of disease in the UK (1% of total expected new cases to 2035). As presented in table 1, these cases are predominantly tobacco-related cancers (35 901) followed by stroke (24 854), COPD (28 997) and CHD (7594). Again, lung cancer is the biggest site-specific cancer to benefit, with 19 331 cases avoided (1.4%) over this time period, with higher degrees of uncertainty observed for renal, gastric, laryngeal, colorectal and hepatic cancers. Online supplementary appendix 6 presents the incidence and cumulative incidence cases every 5 years of the simulation.
As a consequence of delivering improved health outcomes, the TFA scenario is also predicted to avoid £67 million in direct NHS and social care costs in the year 2035 alone. Most of these costs are avoided as a result of fewer cancer cases (£32 million), followed by other diseases. Stroke is a relatively more expensive disease, so despite being predicted to result in relatively fewer cases of disease than COPD, it will result in higher costs avoided: £15 million in the year 2035 alone compared with COPD at £13 million. CHD produced the smallest cost avoided for non-cancers at £7 million. Unlike for disease incident cases, the biggest cancer costs savings were seen in AML (£13 million), though with a higher degree of uncertainty, followed by lung cancer (£11 million), with small cost savings expected for other cancer sites.
As well as the impact on NHS and social care costs, the economic benefits of achieving a TFA are reaped more widely. As a result of delivering savings in morbidity and mortality costs, this scenario is predicted to avoid £548 million in non-health societal costs to the UK population in the year 2035 alone.
In their call to action, Beaglehole et al 36 state that ‘the time has come for the world to acknowledge the unacceptability of the damage being done by the tobacco industry’ and that achieving tobacco-free ambitions are ‘socially desirable, technically feasible, and could become politically practical’. This research would yield even greater benefits by capturing a fuller range of diseases avoided and extending the gains past the year 2035. As a result, these ambitions may also be epidemiologically and financially necessary.
Strengthening the case for achieving a TFA is benefited by expanding the ‘endgame’ narrative. As McDaniel et al 37 note, endgame strategies could encourage governments to move beyond ‘controlling’ the tobacco epidemic towards ending it, with policies that support users, the supply chain and/or institutional structures. They also note these strategies can be shaped through different lenses that require innovative policy approaches and require political leadership in the face of vociferous tobacco industry interests. The findings from this research demonstrate the value of such leadership. It is intended that quantifying the benefits of achieving a TFA can provide a supportive framework for impactful policy action.
Findings from this study demonstrate that (A) there are clear health benefits delivered through achieving a TFA by 2035 and (B) maintaining current trends in smoking prevalence is still likely to produce a sizeable disease and economic burden.
Smoking rates only decline with action. The UK has a history of implementing comprehensive tobacco control measures, including the standardised packaging of tobacco products.38 Establishing a TFA is an important policy endeavour to improve the health outcomes of the UK and a predicted baseline scenario of policy interventions to rid the country from the lethal grip of smoking.
A comprehensive policy response to accelerate progress in reducing the disease and economic burden of tobacco is also vital to alleviate NHS England’s warning that ‘the health of millions of children, the future sustainability of the NHS and the economic prosperity of Britain all now depend on a radical upgrade in prevention and public health’.39 This warning has remained largely unaddressed, and findings from this research demonstrate that sizeable costs to NHS and social care could be avoided by achieving a TFA, in addition to avoiding the substantial societal costs incurred from premature mortality and morbidity.
As TFA begin to be introduced by countries across the world, there is widespread support in the tobacco control community for the UK Government to establish such an ambition.8 A comprehensive response to achieve it will continue to need a suite of policies that learn from other endgame strategies and the emerging literature and measures grounded in the ‘MPOWER’ framework.40 Sustained increases in tobacco taxation, supported with measured to address illicit tobacco trade, are necessary to mitigate industry pricing strategies and deliver a pro-equity benefit by encouraging low-income smokers to stop. We also note discussions about the role of electronic nicotine delivery systems to reduce smoking prevalence. With research suggesting that e-cigarette use in England is positively associated with the success rate of quit attempts, but not the total rate,41 more research is needed to continue to understand their impact on smoking cessation.
Practically, achieving sustained reductions in smoking prevalence in the UK will also now require sustained resources to support smokers to stop, with a particular focus on engaging hard-to-reach groups. With a substantial number of local authorities in England cutting back on their budgets for smoking cessation activity and mass media campaigns,42 a sustainable approach to public health funding should be sought in the immediacy.
Limitations and future work
These data complement the small but valuable evidence base using modelling to quantify the benefits of decreasing smoking prevalence on disease and economic outcomes both internationally15 43–47 and in the UK48 49 using a microsimulation method that simulates 100 million individuals. However, assessing a 20-year period from 2015 to 2035 is unlikely to sufficiently convey the full harms of tobacco and impact of smoking-related diseases experienced across the life-course of a population. For example, the findings exclude smoking-related diseases that would be developed in 2040, despite being attributable to continuous exposure of the risk factor for all or part of the studied time period. As such, measuring beyond 2035 is likely to yield a greater disease and economic burden attributable to smoking.
In addition, the diseases modelled do not cover the full range of tobacco-related conditions in the UK, meaning additional harms are not included. This extends to maternal and child health outcomes related to smoking-related conditions, such as low birth weights and preterm births, which research suggests are often overlooked in tobacco simulation modelling50 as well as conditions for which smoking may be both a risk factor for disease and an aggravating factor, such as type 2 diabetes.51
In further work, it would also be beneficial to simulate the contribution of individual policies and ‘endgame strategies’.52 This would be a valuable next step in simulation modelling tobacco control interventions. Understanding how these policies vary by income group would also be valuable.
Delivering this assessment of specific policies would require improved data in the following recent UK-specific domains: cross-price elasticity figures of various legal tobacco products; cross-price elasticity figures of illicit tobacco products; elasticity figures for tobacco products stratified by socio-economic class; and pass-on rates for tobacco products.
Further work should develop the existing model to include passive smoking and take account of upward trends in hand-rolled tobacco use or other non-cigarette tobacco products like shisha, cigars and pipes. However, relative risk data and reliable exposure data for these specific types of tobacco use are necessary. Further developments should also consider more addicted smokers for whom it is harder to quit and specifically consider those of low income (who currently make up a large percentage of the smokers).
Additional data in other areas would build on the current estimates, such as more precise relative risks for disease incidence data. Data by age, sex and income group are necessary to test the long-term health impacts within the microsimulation. While these data do exist, the sample size in most groups is too small to extract meaningful results. However, this model can easily be refined and updated when new data become available.
Cumulative costs avoided over a 20-year period have not been included in this paper to avoid producing a large figure that may be difficult to contextualise. In addition, these calculations only explored non-health costs as those resulting from lost productivity due to premature morbidity and mortality. As such, they do not capture the full range of non-health societal harms, such as wider social costs, passive smoking, domestic fires or litter, as captured in models produced by Action on Smoking and Health53 and Brunel University.54
Although data are intensive, the microsimulation method has been deemed the most suitable for risk factor and chronic disease modelling in a review by the Organisation for Economic Co-Operation and Development (OECD).55 The present model is highly flexible and has been implemented in over 70 countries.56–60 The model includes a number of different risk factors as well as functionality to run multiple risks.
There are a number of interesting areas of exploration in the endgame narrative that can inform, and be informed by, future simulation modelling. These include building on international evidence exploring public support for TFA,43 modelling the impact of tobacco control policies on groups with disproportionately high smoking prevalence, such as measuring the impact of tobacco taxation on smoking prevalence in low income groups,61 and suggesting ambitions may not result in a core of smokers more addicted and less able to quit.44
What this paper adds
Understanding the impact of changes to smoking prevalence through simulation models is important for policy planning and understanding the costs of inaction.
This paper strengthens the currently small evidence base, both in the UK and internationally, that uses microsimulation modelling to understand the impact of reduced smoking prevalence on disease and economic outcomes.
Strategies that achieve a tobacco ‘endgame’ could deliver substantially improved health outcomes. However, such ambitions have not yet been readily adopted internationally and the evidence base on the benefits of achieving such ambitions is an emergent area of research.
For the first time to our knowledge, this paper communicates the benefits that a ‘tobacco-free’ ambition could bring on lessening the disease and economic burden caused by smoking in the UK.
The authors would like to express gratitude to those organisations and individuals who supported this research, in particular Seb Hinde and Anna Gilmore. The authors would also like to thank Mark Sculpher, Gavin Roberts and John Brazier for allowing the authors to make use of their wider societal benefits approach to estimating non-health costs.
Contributors All authors were involved in the design of the study. AKT, AB and LW provided information on the study methodology and data inputs and outputs, and DH wrote the introduction, discussion and policy components of this paper. AKT, DH, KB, AJ, LR and LW have contributed to manuscript revisions. AJ, LR and MB developed the model methodology, including development of algorithms and model assumptions.
Funding The research was commissioned by Cancer Research UK.
Competing interests AKT and AB worked at the UK Health Forum when this research was undertaken. No other interests are declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.