The research aims to project state-specific cigarette consumption trends in the US until 2035, assess the feasibility of reaching ideal targets (approximating 5% smoking prevalence), and establish realistic state-level targets. Previous projections often relied on simplistic linear extrapolations of national trends, neglecting the considerable state-level variation in smoking prevalence and the inherently asymptotic nature of behavioral change. While a national target of zero prevalence might be ideal, it's considered unrealistic by many experts. The existing disparity in smoking prevalence across states (ranging from 7.9% in Utah to 23.8% in West Virginia in 2020) underscores the need for a state-by-state analysis. Various jurisdictions have established different thresholds for tobacco use reduction, ranging from less than 5% prevalence in some to more aggressive targets like less than 2% in Finland. Although Healthy People 2030 has set targets for the US, a quantitative, state-level assessment of achieving those targets using long-term data was lacking. This study addresses this gap by utilizing a comprehensive dataset and advanced statistical modeling to provide a more nuanced and realistic projection of future cigarette consumption.

The paper reviews existing literature on tobacco control, highlighting the limitations of simple linear projections for predicting smoking behavior. It discusses the importance of considering state-specific trends and the need for more realistic targets. The literature reviewed encompasses various national and international targets for reducing tobacco use, emphasizing the range of ambitions and the varying levels of success seen across different countries and regions. It highlights the existing research on successful tobacco control programs and their impact on cigarette consumption and lung cancer rates.

The study utilized 70 years (1950-2020) of annual state-specific estimates of per capita cigarette consumption (packs per capita or 'ppc') from the Tax Burden on Tobacco reports. The data, representing tax-paid cigarette sales, were sourced from tobacco tax administrators in each state and the US Department of Treasury. Trends within each state were analyzed using linear regression models, while the variation across states was assessed using the Gini coefficient. Autoregressive Integrated Moving Average (ARIMA) models were used to forecast state-specific ppc from 2021 to 2035. To define a consumption level approximating the 5% smoking prevalence target, a random intercepts model was used to regress state-level ppc on adult smoking prevalence from the 2011-2020 Behavioral Risk Factor Surveillance System (BRFSS) survey. This resulted in a population-average consumption of approximately 13 ppc as a reasonable approximation of the 5% prevalence target. ARIMA models were selected using the Hyndman-Khandakar algorithm, with model accuracy evaluated using the Mean Absolute Error (MAE), comparing ARIMA against exponential smoothing models. The final ARIMA models were trained on 25 years of data (1980-2005) to account for the historical trends in cigarette consumption. Forecasts were generated using the forecast function, with residuals resampled to simulate 1000 possible future paths, accounting for increasing uncertainty over time. The Gini coefficient was calculated annually to assess inequity in cigarette consumption. The proportion of simulated forecasts below the 13 ppc threshold was used to estimate the likelihood of reaching the target. Realistic (50% chance) and aggressive (25% chance) targets for each state were determined based on the 50th and 25th percentiles of the simulated forecasts, respectively.

Since 1980, the average annual decline in US per capita cigarette consumption was 3.3%, but rates varied considerably across states (SD = 1.1%). The Gini coefficient, measuring inequity, showed a steady increase from 1985 to 2020 and is projected to continue rising. ARIMA models suggest that by 2035, only 12 states have a >50% chance of reaching the 13 ppc target (approximating 5% prevalence). However, all states show potential for progress. The forecasts indicate a widening gap in consumption levels between states, with California projected to have significantly lower consumption than states like Michigan and Missouri by 2035. Specific probabilities of reaching the 13 ppc target by 2030 and 2035 were provided for each state. Realistic and aggressive target thresholds for 2030 and 2035 were also calculated for each state based on the simulated forecast distributions. For example, California had realistic targets of below 9 ppc in 2030 and 7 ppc in 2035, while Michigan’s were 29 ppc and 25 ppc, and Missouri’s were 57 ppc and 51 ppc, respectively. Aggressive targets were also presented for each state.

The findings highlight the significant disparities in cigarette consumption reduction rates across US states. The increasing inequity, as shown by the rising Gini coefficient, necessitates state-specific interventions rather than a uniform national approach. The success of comprehensive tobacco control programs in states like California, New York, and Washington, which are projected to meet the target, underscores the effectiveness of multi-pronged strategies including tax increases, smoke-free policies, and public awareness campaigns. Expansion of these comprehensive programs to other states is crucial for reducing consumption. Even successful programs can be enhanced by continuing to implement new policy initiatives. The role of local initiatives in complementing state and national efforts was highlighted. While the study provides state-specific targets, it acknowledges the limitations of relying on historical trends, which might not accurately reflect all influencing factors such as illicit trade.

The study demonstrates that while cigarette consumption is declining nationally, the rate of decline varies substantially across US states, leading to increasing inequity. Reaching national aspirational targets may be unrealistic for most states without significant expansion of comprehensive tobacco control programs. State-specific, realistic, and aggressive targets are provided to guide future interventions. Future research should incorporate the potential impact of policy changes and explore alternative forecasting methods.

The accuracy of the “ideal target” relies on the continued correlation between prevalence and consumption observed in the 2011-2020 data. Model uncertainty might lead to less certain target attainment; for example, Maryland's 50% chance contrasts with Colorado's 11% chance despite similar median projections. The prediction intervals might be too narrow due to not fully accounting for uncertainty in autocorrelation estimates and model order. Illicit cigarette trade and changes in population structure might also affect the reliability of sales trends for some states.