Effectiveness of tobacco control policy interventions on tobacco use behaviours and health outcomes based on ITS research methodology: a scoping review

Abstract

Objective

To provide an overview of the effectiveness of global tobacco control policies using Interrupted Time Series (ITS) methodology and analyse key health outcome indicators and methodological differences in ITS studies.

Data sources

PubMed, Web of Science and Embase were searched for ITS studies on tobacco control policies from database inception to 1 July 2024.

Study selection

55 English-language studies were selected based on strict inclusion criteria.

Data extraction

Three researchers extracted data on study characteristics, design, conclusions and limitations.

Data synthesis

Of the studies, 28 used monthly time intervals, 40 focused on a single time break, 47 used non-Autoregressive Integrated Moving Average models, 40 considered seasonality, 34 controlled for confounding factors and only 13 included a control group. The implementation and evaluation of tobacco control policies demonstrated considerable geographical imbalance, with Europe reporting the largest number of evaluations (n = 23), followed by Asia (n = 12) and North America (n = 10), whereas South America (n = 4), multiple regions (n = 4) and Oceania (n = 2) were markedly under-represented. Overall, policies were associated with beneficial effects across multiple health outcomes, including reductions in emergency visit rate, hospitalisation rates, secondhand smoke exposure rate and smoking prevalence, although certain outcome measures exhibited notable heterogeneity.

Conclusions

ITS applications differ in time-interval selection, modelling and control of seasonality and confounding, contributing to result heterogeneity. Tobacco control policies are effective across multiple health indicators, though variability warrants further study. Future work should enhance reporting and methodological standardisation, particularly in control group use and statistical modelling, and strengthen policy implementation and evaluation in low- and middle-income countries to promote equitable global progress.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study provides a globally scoped systematic review focused specifically on the application of Interrupted Time Series (ITS) methods to evaluate tobacco control policies, offering strong methodological specificity and an international perspective.

• It encompasses a variety of policy types and health outcome indicators, highlighting key methodological differences in ITS applications, thereby informing improvements in future policy evaluation designs.

• The inclusion criterion limited to English-language studies employing ITS methods may underestimate evaluations conducted using other methodologies or those from non-English-speaking regions.

Introduction

According to the WHO,¹ over 1 billion people were smokers in 2019, with 78% being men and 22% women. Annually, about 8 million people die due to smoking or secondhand smoke exposure, 15% of whom are non-smokers. Smoking and secondhand smoke are major risk factors for various chronic non-communicable diseases globally.² The Global Burden of Disease 2021 report³ states that smoking causes over 40% of deaths from chronic obstructive pulmonary disease and more than 80% of lung cancer deaths worldwide.

In 2003, the WHO adopted the Framework Convention on Tobacco Control (FCTC), the first international treaty focused on reducing tobacco use and its health risks. Over 180 countries have since ratified the treaty, resulting in a significant global reduction in smoking prevalence,⁴ which has dropped to 19.2% among individuals aged 15 and older.⁵ Countries have implemented various tobacco control measures, including indoor smoking bans, higher tobacco taxes, restrictions on advertising and mandatory health warnings on packaging.

In addition, Interrupted Time Series (ITS) has been widely used internationally to assess the impact of tobacco control policies on population health and behaviour. As a quasi-experimental design, ITS does not require randomised control conditions and can identify both the immediate effects and changes in trends attributable to policy interventions by comparing time trends and level changes before and after the intervention. In particular, in the context of large-scale interventions such as national-level tobacco control policies, which cannot be randomly allocated, ITS is regarded as one of the most appropriate and highly explanatory evaluation methods currently available.⁶ For example, an ITS study in Canada⁷ found that graphic health warnings led to a 12.1% reduction in smoking prevalence within the first year and 19.6% in the second year.

Therefore, this study aimed to systematically review global research employing the ITS method to evaluate tobacco control policies, exploring the design and application of ITS in tobacco policy research and revealing the actual impact of different policies on public health across countries. This provides policymakers with a scientific basis and offers recommendations for optimising future tobacco control policies.

Methods

Search strategy

This study systematically searched PubMed, Web of Science and Embase databases for ITS studies on tobacco control policies from the date of creation of the databases to 1 July 2024. The search terms consisted of “Tobacco Control”, “Smoke-Free Policy”, “Smoking Regulation”, “Smoking Bans”, “Smoking Prohibition”, “Interrupted Time Series Analysis”, “ITS Study”, “Health benefit”, “Health Impact”, “Health outcome” and other combinations of subject terms and free words. A complete list of search terms is provided in the online supplemental tables S1–S3.

Inclusion and exclusion criteria

This study placed no restrictions on population, country or time frame. Included interventions were various tobacco control policies implemented by governments or international organisations, such as smoke-free legislation, increased tobacco taxes, bans on tobacco advertising or promotion and mandatory health warnings or graphic labels on packaging. Outcome measures included: disease-related indicators (eg, morbidity, mortality, hospitalisation rates); maternal and child health outcomes (perinatal mortality, preterm birth and the incidence of small for gestational age (SGA)); public health and preventive measures (suicide mortality, secondhand smoke exposure and smoking cessation rates) and smoking prevalence and tobacco consumption. Eligible studies had at least two segments (pre-intervention and post-intervention) separated by a clear intervention or exposure, with repeated observations on a group (deg, community, hospital).

Exclusion criteria were as follows: (1) studies without outcome measures or with unclear data (eg, only qualitative descriptions, no quantitative time-series data or no defined pre-intervention and post-intervention periods); (2) interventions unrelated to tobacco control, such as medications or counselling; (3) discussed ITS methodology without applying it to tobacco control policy; (4) systematic reviews, meta-analyses and literature reviews; (5) non-English language, duplicate publications or conference abstracts and (6) lack of full-text availability.

Data extraction and review

In this study, three researchers extracted data from the 55 included studies and cross-checked them. There was suitable inter-rater reliability across each category (mean Cohen’s kappa=0.90). The extracted data included the title of the studies, authors’ names, year of publication, country/region, study population, intervention (policy), time frame of the study, time interval, pre-intervention points, post-intervention points, type of model, statistical methodology, whether there was adjustment for seasonality, whether there was control for confounders, outcome indicators, control indicators and indicator effect (online supplemental tables S4–S6).

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Results

Overview of findings

As of 1 July 2024, 234 relevant studies were retrieved for this study, of which 72 were retrieved by PubMed, 48 by Web of Science, 108 by Embase and 6 by other sources. After excluding 58 duplicate studies, 176 studies were finally obtained. 120 studies were excluded by evaluating study titles and abstracts, 1 study was excluded by reading the full text on this basis. 55 studies⁸,⁶² were finally included in the analysis. and the study screening process and results are shown in figure 1.

Figure 1. Flowchart of global tobacco control policy intervention effectiveness literature search and screening based on ITS research methodology. ITS, Interrupted Time Series.

The publication time ranges from 2007 to 2024. A large number of studies were published from 2016 onwards, with a total of 42 studies⁸,^{1315 16 18} published. 2023 has the highest number of studies published in a single year, with a total of nine studies.⁸,¹⁶ Overall, the number of publications in the literature showed an increasing trend from year to year (figure 2).

Figure 2. Statistics on the year of publication of the literature on the effectiveness of global tobacco control policy interventions based on the ITS research methodology. ITS, Interrupted Time Series.

ITS methodology

It was found that monthly data were used most (28 studies,^{1113 15 17},⁴¹ 50.91%), followed by yearly data (11 studies,⁸,^{1012 42} 20.00%). Seasonal data were used in six studies¹⁶⁴⁹,⁵³ (10.91%), and four studies¹⁴⁵⁴,⁵⁶ were not reported (table 1).

Table 1. Characteristics of the design of included studies on the effects of global tobacco control policy interventions based on the ITS research methodology.

Design characteristics	n	% or n
Time interval type
Monthly	28	50.91%
Annually	11	20.00%
Quarterly	6	10.91%
Weekly	5	9.09%
Two weekly	1	1.82%
NR	4	7.27%
Number of exposure/intervention periods
1	40	72.73%
2	7	12.73%
3 or more	6	10.91%
NR	2	3.64%
Statistical methods
Non-ARIMA	47
Linear regression		18
Poisson		14
Negative binomial		8
Logistic		5
GAM (generaliseded additive model)		2
GEE (generalised estimating equations)		1
Other		1
Joinpoint regression		1
ARIMA	4
SARIMA	3
NR	1
Whether there was adjustment for seasonality
Yes	40	72.73%
No	15	27.27%
Whether there was control for confounders
Yes	34	61.82%
No	19	34.55%
NR	2	3.64%
Consider a control group
No	42	76.36%
Yes	13	23.64%

Three studies involving various statistical methods (Polus et al,²⁵ Glantz and Gibbs,³³ Alpert et al⁵⁶).

ARIMA, Autoregressive Integrated Moving Average; ITS, Interrupted Time Series; NR, not reported; SARIMA, Seasonal Autoregressive Integrated Moving Average.

Most (40 studies, 72.73%) studies⁸¹¹,21 23 24 26 27 29 focused on the intervention effects of tobacco control policies at one time break, 13 (23.6%) studies^{11 13 16 17 23 30 32 38 43 46 48 57 58} two and more time breaks and 2 studies^{8 56} were not explicitly reported.

Among the statistical methods used, Autoregressive Integrated Moving Average (ARIMA) model was applied in 4 studies,^{19 20 41 49} while the Seasonal Autoregressive Integrated Moving Average (SARIMA) model was used in 3 studies.^{11 17 58} There were also 47 studies in which researchers chose non-ARIMA models to analyse the data. These models include linear regression (18 studies912 23 24 26 30 31 33 34 36 38 42,44 50 51 55 57), Poisson regression (14 studies^{10 15 16 18 25 27 29 35 39 40 45 52 59 60}), negative binomial regression (8 studies^{21 22 25 28 32 33 47 61}), logistic regression (5 studies^{37 46 53 54 56}), generalised additive model (2 studies^{13 62}), generalised estimating equations (1 study¹⁴).

Most studies considered seasonality (40 studies,^{810 11 13},^{15 17 19} 72.73%), but less than half of the studies controlled for confounding factors (34 studies,1112 15 19 20 22 29 31,^{41 44 46} 61.82%).

Only 13 studies^{10 14 17 24 29 40 41 45 47 50 53 58 61} (23.64%) had a control group, and the majority had no control group. In terms of the type of control group, it can be divided into two main categories, non-tobacco-related health outcomes and different geographically located populations.

Tobacco control policies by region-based on the ITS methodology

Globally, ITS-based studies of tobacco control policies cover all continents, but the focus of policy content varies (table 2). Europe is the most represented region in ITS-based tobacco control studies (41.82%, primarily from the UK, the Netherlands and Germany). European studies predominantly addressed smoke-free legislation (10 studies^{17 18 22 23 26 28 35 48 54 60} across seven countries) and smoke-free environments (10 studies^{10 16 25 27 36 37 39 55 57 58} across five countries), with the remaining 2 studies^{46 47} multicomponent smoke-free policy for the Netherlands and 1 study⁵³ health warnings for Norway.

Table 2. Types and distribution of global tobacco control policies assessed via ITS.

Tobacco control policy classification	Europe	Asia	North America	Multiple	South America	Oceania	Subtotal
Smoking-free legislation	10	5	3	1	3		22
Smoke-free environment	10	3	3	3	1		20
Multicomponent smoke-free policy	2	4				1	7
Health warnings	1		2				3
Raise taxes on tobacco			1			1	2
Advertising ban			1				1
Total	23	12	10	4	4	2	55

ITS, Interrupted Time Series.

Asia followed with 21.82% of studies, primarily from China and South Korea. Key policies included smoke-free legislation (5 studies^{9 11 40 59 61} across four countries), smoke-free environments (3 studies^{12 13 62} across two countries) and multicomponent smoke-free policy packages (4 studies^{15 34 40 45} across two countries). Of these, China accounted for a total of 7 studies.^{13 15 34 45 59 61 62}

North America (18.18%, focused on the USA and Canada) showed the most diverse range of tobacco control policies, with nearly all multicomponent smoke-free policies addressed. Among them, there are 2 studies^{20 51} of smoke-free legislation in Canada and 1 study³³ in the USA, 3 studies^{41 49 50} of smoke-free environment in the USA, 1 study¹⁴ of health warnings in Canada and 1 study⁵⁶ in the USA, 1 study⁵² of advertising ban in the USA and 1 study³¹ of raising tobacco tax in the USA.

In contrast, relatively few studies were reported from South America (4 studies^{19 21 24 29}), multiple regions (4 studies^{19 21 24 29}) and Oceania (2 studies,^{30 38} both from Australia).

Health outcomes based on the ITS methodology

Tobacco control policies positively impacted health outcomes, including disease-related indicators, maternal and child health, public health and prevention measures, smoking prevalence and tobacco consumption, although individual outcomes showed variability (table 3).

Table 3. Summary of the effects of tobacco control policy outcome indicators based on the ITS method.

Outcome measures	Decrease (n)	Increase (n)	No effect (n)
Disease-related indicators	60	3	27
Mortality rate	20	0	6
Hospitalisation rate	19	0	7
Incidence rate	7	1	7
Hospital admissions	6	0	4
Consultation rate	4	1	2
Emergency visit rate	3	0	0
Dispensing rates of medications	1	1	1
Maternal and child Health	4	0	4
Preterm birth	2	0	1
Perinatal mortality	1	0	1
Small-for-gestational age (SGA)	1	0	1
Pregnancy outcomes	0	0	1
Public health and prevention measures	3	2	2
Secondhand smoke exposure rate	2	0	0
Suicide mortality rate	1	0	0
Cessation rate	0	2	2
Smoking prevalence and tobacco consumption	17	0	6
Smoking rate	10	0	4
Cigarette consumption	6	0	1
Numbers of smokers (below the age of 25 years)	1	0	0
Market share changes	0	0	1

ITS, Interrupted Time Series.

Disease-related indicators

Of the multiple disease-related indicators analysed, 60 results⁹,¹⁵¹⁷ showed a decrease in tobacco-induced disease-related indicators, only 3 results^{28 47 58} showed an increase and another 27 results^{16 18 20 27 31 32 35 37 45 49 58 61 62} showed no significant effect. For emergency visits, three results^{17 21 33} consistently reported reductions in asthma, bronchospasm and emergency medical services visits. Outcomes for admissions, hospitalisation, morbidity and mortality varied, particularly for acute myocardial infarction and asthma. Whereas complete heterogeneity was found for the other two outcome indicators, specifically for consultation rates (four decreases,^{20 47} two no effect,²⁰ one increase⁴⁷) and medication dispensing rates (one decrease,⁵⁸ one no effect,⁵⁸ one increase⁵⁸), no heterogeneity was found for consultation or medication dispensing rates for the same disease when subdivided into specific diseases.

Maternal and child health

There was heterogeneity in the effects of tobacco control policies on perinatal mortality, preterm birth and SGA. Outcomes showed either a reduction or no effect, with no reports of increases. Specifically, perinatal mortality (one decrease,⁴⁶ one no effect⁴⁶), preterm birth (two decrease,^{25 46} one no effect⁴⁶) and SGA (one decrease,⁴⁶ one no effect⁴⁶). One additional study²⁵ indicated no effect on pregnancy outcome.

Public health and prevention measures

Tobacco control policies have also had a positive impact on the effectiveness of public health and preventive measures, particularly in terms of suicide mortality rates, secondhand smoke exposure rates and smoking cessation rates. Two results^{9 12} and one result,⁴⁰ respectively, showed that secondhand smoke exposure and suicide mortality rates decreased significantly after the implementation of tobacco control policies. However, cessation rates showed mixed results, with two studies^{14 44} reporting increases and two others^{55 56} showing no effect.

Smoking prevalence and tobacco consumption

Tobacco control policies significantly reduced smoking prevalence and tobacco consumption, though some heterogeneity existed. Smoking prevalence declined in 10 studies,²⁹,3138 43 50 52 53 55 56 and the decline in smoking among young people under 25 was particularly significant after the implementation of tobacco control policies, but four results^{30 31 52 56} reported no effect. There was also heterogeneity in tobacco consumption (six decrease,^{8 23 43 51 56 57} one no effect⁴²).

Discussion

Current status of ITS methodology

The ITS approach has shown considerable strengths in evaluating tobacco control policies, but significant heterogeneity across studies can lead to variability and limit comparability.

Time intervals

Differences in time intervals can result in inconsistent outcomes. Monthly data is most commonly used, providing a detailed view of short-term policy impacts. While annual and quarterly data capture long-term trends, they may obscure short-term changes. Conversely, monthly data, while sensitive, can overemphasise short-term fluctuations and complicate the evaluation of long-term effects.^{63 64}

Policy observation nodes

Most studies focus on a single policy intervention point, with fewer exploring the cumulative effects of multiple or phased policies. This simplification may limit a comprehensive understanding of policy impacts.⁶⁴

Statistical methods

The 55 reviewed studies employed various statistical methods, each with its own strengths and limitations. ARIMA and SARIMA are effective for handling autocorrelation and seasonality, but may struggle with non-stationary or structurally changing data.⁶⁵ Non-ARIMA models offer alternative analytical perspectives, handling simple trends to complex multivariate relationships and accommodating count, ratio and binary data. Linear regression, the most used method, is suited for trend data but is sensitive to outliers, while negative binomial regression is better for over-dispersed data. Each method has specific trade-offs.^{63 66}

Seasonality

Seasonality is a crucial factor, particularly in regions with seasonal variations in tobacco consumption. Nearly 30% of studies failed to account for seasonality. Among those that did, methods and depth of consideration varied, potentially affecting the reliability and comparability of findings.^{63 66}

Confounding factors

Controlling for confounding factors is critical, especially in observational studies lacking randomised controls. Fewer than half of the studies considered factors such as socioeconomic status or other health policies, which may bias estimates of tobacco control effects.^{63 64}

Control group

Including a control group in ITS analyses helps distinguish changes attributable to the policy intervention from those arising from concurrent factors, such as socioeconomic trends or other public health initiatives, thereby enhancing internal validity. However, most included studies did not incorporate a control group, potentially limiting the ability to rule out confounding effects from co-interventions or other events occurring before or after the intervention.^{64 67 68}

Review of global tobacco control policies

Since the beginning of the 21st century, the adoption and implementation of tobacco control policies have increased significantly worldwide, especially following the 2003 adoption of the WHO FCTC. Countries have successively introduced various measures, including the establishment of smoke-free environments, strengthened health warning labels, tobacco advertising bans and tobacco taxation policies. Some countries, like Australia, revised policies based on implementation outcomes and social needs, such as increasing tobacco taxes in 2010 and 2013⁵² (online supplemental table S5).

However, there are notable regional disparities in implementation and evaluation. Europe leads in tobacco control, with countries like the UK and the Netherlands enforcing wide-reaching smoke-free legislation across public places, workplaces and outdoor areas.^{69 70} Notably, they are often accompanied by ITS-based evaluations, reflecting a strong emphasis on data-driven policy improvement. In Asia, where China has recently introduced a range of measures including public smoking bans and packaging regulations, alongside increased use of ITS methods for evaluation. Several Chinese provinces and cities have conducted ITS studies to assess policy impacts on smoking prevalence and hospitalisation rates.^{15 45 61 62} In contrast, countries like South Korea and Singapore, despite implementing policies, have fewer published ITS evaluations, possibly due to limited data availability or evaluation capacity.^{12 40}

In North America, the USA and Canada have enforced various policies, particularly focusing on health warnings and advertising bans. Meanwhile, both the USA and Canada have conducted ITS evaluations across various tobacco control policies. US studies encompass a wide range of measures, including taxation, advertising bans and packaging regulations, reflecting the breadth of its policy implementation^{41 49 50 52 56}; in contrast, Canadian ITS research tends to focus on the assessment of specific regional regulations.^{20 51} Chile and Australia are strong enforcers in South America and Oceania, respectively, while other countries in these regions show limited ITS evaluation and weaker policy coverage.^{71 72} Factors such as economic interests, weak enforcement and low public health awareness are likely contributing to these shortcomings.⁷³

Health benefits of tobacco control policies

This study found that tobacco control policies significantly improved several health-related outcomes, though there was some variability in their effectiveness. Beyond policy-specific factors—such as type, approach, target population and tobacco product—variations also stem from sociocultural context, enforcement and healthcare accessibility.^{74 75}

Disease-related indicators

Most tobacco control policies effectively reduced smoking-related emergency visits, hospitalisations and mortality, particularly for conditions like asthma and cardiovascular disease.^{17 21 33} However, outcome variability suggests that factors such as differences in target populations, regional health infrastructure, public health awareness and policy enforcement may influence the success of these policies, leading to ineffectiveness in some areas.⁷⁶

Maternal and child health

While tobacco control policies generally reduced perinatal mortality and preterm births, variability in outcomes may be linked to differences in policy type, intervention approach (eg, whether accompanied by smoking cessation support for pregnant women), baseline maternal smoking prevalence and healthcare accessibility.^{77 78} Improved targeting and broader coverage, especially for high-risk groups, could further enhance these health benefits.

Public health and preventive measures

Policies significantly reduced secondhand smoke exposure and suicide mortality, but their impact on smoking cessation rates was more inconsistent. This may be due to behavioural habits, cultural influences and the availability of cessation support services. In contexts lacking cultural or structural support for cessation, legislation alone rarely achieves high quit rates.⁷⁹ Strengthening cessation programmes and public education could help improve cessation rates.⁸⁰

Smoking prevalence and tobacco consumption

Tobacco control policies consistently reduced smoking prevalence and tobacco use, especially among youth, highlighting their effectiveness in preventing youth smoking. However, tobacco control policies have shown limited impact on market share,^{81 82} as the industry may adapt by modifying product types or marketing strategies.⁸³

Conclusion

This review offers a cross-country analysis of tobacco control policies, highlighting their significant role in improving health outcomes. ITS methodology effectively captures changes before and after policy implementation, but heterogeneity in time intervals, models and confounder control leads to inconsistent results and limits comparability. Future research should explicitly justify time interval and statistical model selections, improve control of seasonality and confounders, and use proper control groups to standardise evaluations across diverse settings.

Since the implementation of the FCTC, global tobacco control efforts have strengthened, but enforcement and research are stronger in high-income regions like Europe and North America, while low- and middle-income countries lag behind. This underscores the need to strengthen tobacco control efforts and research investment in these settings.

Tobacco control policies effectively reduce emergency visits, hospitalisations, secondhand smoke exposure and smoking prevalence, though outcome variability persists due to differences in policy type, implementation, population and healthcare infrastructure. Future research should focus on clearer policy classification and exploring this heterogeneity.

In summary, future research and policy development should focus on improving the quality of policy implementation, increasing public participation and assessing long-term impacts. Collaboration between governments and international organisations will be crucial in advancing global tobacco control efforts.

Research limitations

The object of this study is a global overview of tobacco control policies based on literature applying ITS methods, which may lead to a biased description of the global application profile of tobacco control policies. Studies that do not include the use of other quantitative or qualitative research methods may limit our understanding of the comprehensiveness of the impact of tobacco control policies. In addition, the study’s reliance on published findings, its failure to cover unpublished grey literature, and its lack of attention to tobacco control policy research in developing countries and elsewhere may have contributed to the incomplete nature of the analysis.

The opinions expressed are those of the authors alone.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.