Tobacco smoking and risk of 36 cardiovascular disease subtypes: fatal and non-fatal outcomes in a large prospective Australian study

Cardiovascular disease (CVD) is the leading cause of mortality globally and encompasses a broad and diverse range of subtypes, including ischaemic heart disease (IHD), cardiac dysrhythmias, cerebrovascular disease, peripheral arterial disease and heart failure [1].

Smoking increases the risk of CVD [2–6]. However, the magnitude of this increase in risk varies substantively according to a range of factors. Smoking-related CVD risks are highest in current and recent smokers, compared to never smokers and those who have quit in the more distant past [5]. Risk also increases with increasing duration of use and with greater intensity of smoking, as measured by the number of cigarettes smoked per day [5]. The data that are available indicate variation according to CVD subtype [5–7]. For example, while mortality from both IHD and abdominal aortic aneurysm is elevated in current smokers, compared to those who have never smoked, contemporary US cohort study data show considerable variation in the relative risks, from 2.6 (95%CI 2.4–2.7) for IHD mortality to 7.5 (5.8–9.7) for mortality from abdominal aortic aneurysm for men [6]. In general, studies to date have focused on fatal CVD outcomes, with limited evidence on non-fatal outcomes. As with most risk factors for CVD, excess relative risks attributable to current smoking tend to decline with age [5], and absolute excess risks increase [8]. While data on variation in risk according to other factors are limited, there is some evidence of a greater impact of smoking on IHD in women than in men [9]. Relative and absolute excess CVD risks attributable to smoking will vary according to the absolute CVD risk in never smokers [8].

Hence, smoking-attributable risks vary according to the CVD outcome under examination and according to the pattern of smoking in a population, which, in turn, relate to the stage of the tobacco epidemic and will vary over time and between countries [10]. It follows that a comprehensive contemporary picture of the relationship of smoking to CVD, necessary for appropriate policy and practice responses, requires examination of the full spectrum of CVD, including fatal and non-fatal outcomes, according to smoking intensity and recency. However, our search of the published evidence to date did not locate any studies that included all of these features (see the “Discussion” section and Additional file 1: Brief systematic review of smoking and cardiovascular disease). Moreover, estimates of the proportion of CVD events attributable to smoking in the population often present summaries for the whole population [11], which tend to be weighted towards older ages, rather than providing age-specific evidence and evidence relating to premature CVD, which is more informative for tobacco control.

This study aimed to quantify the relation of tobacco smoking, including measures of intensity and recency, to the risk of CVD and CVD subtypes, including acute myocardial infarction, cerebrovascular disease, heart failure, peripheral arterial disease and dysrhythmias, considering fatal and non-fatal outcomes, where appropriate. The relationship of smoking to outcomes in different population subgroups was also considered, along with the proportion and number of CVD deaths attributable to smoking, by age.

In common with North America and much of Europe, Australia is considered to have a mature smoking epidemic, with historically high prevalences of smoking, high intensity and very long durations of smoking among smokers in the population, similar prevalences of smoking among men and women and low and decreasing current prevalence of smoking [10]. The findings of this study are therefore likely to be informative internationally, as well as nationally.

The Sax Institute’s 45 and Up Study is a cohort study of 267,153 men and women aged 45 and over at baseline, randomly sampled from the general population of New South Wales (NSW), Australia, using the Department of Human Services enrolment database. Individuals joined the study by completing a postal questionnaire from January 2006 to December 2009 and are followed up through repeated data collection and through data linkage. The general study methods are described in detail elsewhere [12].

Baseline questionnaire data include information on socio-demographic factors, health behaviours, height and body weight, medical and surgical history, functional capacity and physical activity. The study questionnaire is available at https://www.saxinstitute.org.au/our-work/45-up-study/questionnaires/.

Questionnaire data from study participants were linked probabilistically to the data from the NSW Admitted Patient Data Collection, which is a complete census of all public and private hospital admissions in NSW, the NSW Register of Births, Deaths and Marriages (to provide fact of death); the Australian Bureau of Statistics Cause of Death unit record files (to provide cause of death); and the National Death Index (to provide fact and cause of death), up to the end of December 2015. Combined, these death registration data capture all deaths in New South Wales and Australia. The linked Admitted Patient Data Collection data that were used contained details of admissions in participants from the year 2001 up to the end of December 2015, including the primary reason for admission using the International Classification of Diseases 10th revision–Australian Modification (ICD-10-AM) [13] and up to 50 additional clinical diagnoses. Probabilistic linkage is conducted by the NSW Centre for Health Record Linkage and is known to be highly accurate (false-positive and false-negative rates < 0.4%) [14].

At baseline, 8% of the 188,167 participants reported being current smokers and 34% were past smokers. The prevalence of smoking was similar in men and women. Overall, 94.9% (n = 74,141) of all current and past smokers reported smoking cigarettes only, 1.8% (n = 1379) smoked cigarettes and cigars or pipes and 1.8% (n = 1407) smoked only cigars or pipes.

During a total follow-up time of 1.35 million person-years (mean 7.4 years), there were 27,511 major CVD events—fatal and non-fatal. These included the following: 11,778 IHD, including 4548 AMI; 3991 cerebrovascular disease; 3874 heart failure; and 2311 peripheral arterial disease events. There were 2540 deaths from major CVD: 1101 from IHD including 558 from AMI, 697 from cerebrovascular disease and 135 from peripheral arterial disease.

Rates of hospitalisation with or death from acute myocardial infarction, cerebrovascular disease, heart failure, peripheral arterial disease and major CVD combined were elevated in current versus never smokers, for men and women (Fig. 1a, b; Additional file 2: Table S1). Former smokers showed an intermediate level of risk. Compared to never smokers, current smokers were at a significantly increased risk of hospitalisation with or mortality from IHD (adjusted RR 1.65, 95%CI 1.54–1.77), including AMI (2.45, 2.22–2.70); cerebrovascular disease (2.16, 1.93–2.42); heart failure (2.23, 1.96–2.53); peripheral arterial disease (5.06, 4.47–5.74); and any major CVD (1.63, 1.56–1.71) (Fig. 1a). These estimates did not change materially when the first year of follow-up was excluded or when recent quitters were re-classified as current smokers (Additional file 2: Tables S2 and S3). The RRs of non-fatal events for these outcomes are given in Additional file 2: Table S4. The RR of dying from AMI in current versus never smokers was 2.79 (2.04–3.80); the corresponding result for mortality from cerebrovascular disease was 2.26 (1.65–3.10) (Fig. 1b). The RR of dying from any major CVD in current versus never smokers was 2.75 (2.37–3.19) (Fig. 1b).

When considered in finer pathological and clinical categories (single level 3 ICD-10-AM codes), significantly elevated risks in current versus never smokers were observed for 29 of the 36 subtypes investigated (Fig. 2). Details of the finer CVD subtypes contributing to the numbers of fatal and non-fatal events, separately, and age-standardised rates of disease according to smoking status are given in Additional file 2: Tables S5 and S6. The markedly elevated RRs for current versus never smokers of combined fatal and non-fatal outcomes relating to peripheral arterial disease were observed for subcategories of that disease, including for abdominal aortic aneurysm (ICD-10-AM I71) and atherosclerosis (I70) (Fig. 2). Significantly increased risks of hospitalisation/mortality for vascular disorders of the intestine (K55) and subcategories of cerebrovascular disease and IHD were observed, although the latter showed significant heterogeneity, with higher RR for AMI (I21) and lesser elevations in risk for chronic ischaemic heart disease (I25) and angina (I20). Compared to people who had never smoked, current smokers had significantly elevated risks of heart failure (I50); cardiomyopathy (I42); venous thromboembolism, including pulmonary embolism (I26) and deep vein thrombosis (I80); dysrhythmias, including atrial fibrillation and flutter (I48) and paroxysmal tachycardia (I47); and essential hypertension (I10). The RR of supraventricular tachycardia (I47.1) in current versus never smokers was 1.35 (1.07-1.72) and that of ventricular tachycardia (I47.2) was 1.82 (1.35-2.42). Combined hospitalisation/mortality outcomes for various valvular disorders (I34, I35) and atrioventricular and left bundle-branch block and other cardiac arrhythmias (I44, I49) did not show significant associations with current smoking (Fig. 2).

The RRs for the main CVD outcomes increased with increasing intensity of smoking, among current smokers relative to never smokers (Fig. 3a, b; p_trend < 0.0001 for all outcomes). There was a doubling in the rate of major CVD events (fatal and non-fatal) in current smokers of 25 or more cigarettes per day, compared with never smokers (2.12, 1.96–2.29); at least a tripling in the rate of AMI (3.34, 2.85–3.92), cerebrovascular disease (3.22, 2.66–3.89) or heart failure (3.91, 3.17–4.83); and a roughly sevenfold risk of peripheral arterial disease (7.26, 5.95–8.88). In current smokers of 25 or more cigarettes per day, compared with never smokers, RRs were 4.90 (3.79–6.34) for mortality from major CVD and 5.50 (3.89–7.79) for mortality from IHD (Fig. 3b). Considering finer increments in smoking intensity, the risk of any fatal or non-fatal major CVD was significantly elevated from a current smoking level of 7–9 cigarettes per day (1.35, 1.10–1.66) and of CVD mortality from 4 to 6 cigarettes per day (1.92, 1.11–3.32), compared to never smokers (Fig. 4a, b; p_trend < 0.0001 for both).

For the common CVD subtypes, the risks for past smokers quitting prior to the age of 35 did not differ significantly from those of never smokers (Fig. 5a, b). Persistent elevations in the risk of peripheral arterial disease following cessation of smoking were observed, with a small elevation in the composite major CVD outcome among those quitting at age 35–44 or 45–54 years. No significant elevations in mortality from IHD and from all major CVD were seen in past smokers quitting by age 55, compared to never smokers (Fig. 5b).

Compared to never smokers, current smokers had a significantly higher risk of AMI in all of the subgroups examined (Fig. 6). The absolute rates of CVD were higher in older versus younger age groups and in men compared to women. Current smoking was more strongly related to AMI in younger compared to older people (Fig. 6), with RRs of 2.93 (2.51–3.41) for age 45–64 and 1.65 (1.25–2.17) for age ≥ 80 years (p_interaction = 0.0001). Comparing smoking-related RR across grouped CVD outcomes, significantly greater RRs were observed for women versus men for fatal and non-fatal outcomes combined for IHD (p_interaction = 0.02), including AMI (p_interaction = 0.007) and non-AMI IHD (p_interaction = 0.02) but not for stroke, heart failure, peripheral arterial disease, other CVD and any major CVD, nor was any significant sex-related variation apparent for fatal outcomes (Fig. 5, Additional file 2: Figure S8). There was no significant variation in the smoking-related RR according to any of the other factors examined, including according to education, physical activity, diabetes status or alcohol consumption. A greater RR of CVD for current versus never smokers in younger compared to older age groups was observed for chronic ischaemic heart disease and CVD mortality. RRs for CVD mortality in current (versus never) smokers were 3.47 (2.76–4.37) and 2.12 (1.70–2.63) in those aged 45–74 years and ≥ 75 years, respectively; RRs for chronic ischaemic heart disease in current (versus never smokers) were 1.73 (1.60–1.86) and 1.27 (1.08–1.50) in those aged 45–74 years and ≥ 75 years, respectively (data not shown).

This study demonstrates the damage that tobacco smoking causes across the entire cardiovascular system. Current tobacco smokers have at least double the risk of developing most significant types of CVD, including AMI, cerebrovascular disease and heart failure, and over five times the risk of developing peripheral arterial disease, compared to people who have never smoked. Mortality from CVD was almost tripled in current versus never smokers. This means that, on average, for current smokers experiencing any of these events, including an AMI, cerebrovascular disease or death from CVD, the balance of probabilities is that it was more likely than not to have been caused by smoking. The relative risks and attributable fractions are greater at younger ages. Around 36% of CVD deaths prior to age 65 and 15% of all CVD deaths in Australia (around 6400 deaths annually)—a country with a relatively low prevalence of current smoking—can be attributed to smoking.

Risks of CVD increased with increasing numbers of cigarettes smoked per day and were greatly diminished among those quitting smoking. Our study showed that what would be generally be regarded as “light smoking” was accompanied by substantial increases in the risk of CVD; smokers of 4–6 cigarettes per day had around double the risk of dying of CVD than people who had never smoked, adding to the emerging evidence on the CVD harms of light smoking [25]. Ex-smokers had substantially lower risks of CVD compared with those continuing to smoke, with those stopping by age 35–44 avoiding around 90% of the excess smoking-attributable risk of CVD conditions such as AMI and stroke.

In common with many other CVD risk factors, relative risks from current smoking are greater at younger ages, with around a tripling in the risk of AMI in current versus never smokers aged 45–64 and a relative risk of 1.65 at age 80 and over. However, because the incidence of CVD is much greater at older ages, the absolute excess CVD incidence attributable to smoking is often greater for older versus younger age groups. For example, the AMI risk difference between current and never smokers is 2.57/1000 person-years at age 45–64 and 7.58/1000 person-years at age 80 years and over. Similarly, while women had lower absolute rates of AMI and non-AMI IHD, the relative risks of these conditions in female current versus never smokers were significantly greater than those seen in their male counterparts. This is in keeping with the published evidence [9]. We did not find any significant variation in the relation of smoking to fatal IHD outcomes, or to CVD outcomes other than IHD, according to sex; we are not aware of other publications examining this. There was no significant variation in the current smoking-AMI relationship according to any of the other population characteristics examined. It is important to bear the absolute risks of CVD in mind when interpreting the findings on relative risks. The absolute number of CVD events attributable to smoking will tend to be greater in groups with higher absolute risks of CVD, including men and those with pre-existing CVD and/or diabetes. Moreover, while the RR in current versus never smokers are greater for peripheral arterial disease than for AMI, the absolute differences in risk between current and never smokers are greater for AMI.

Considering the available published evidence (based on formal searches outlined in the Additional file 1: Brief systematic review of smoking and cardiovascular disease), we conclude that our study presents the most systematic and comprehensive analyses of the relation of smoking to detailed CVD subtypes to date worldwide. It also provides the only contemporary estimates, to our knowledge, from Australia. The study combines detailed prospective data on smoking behaviour and potential confounding factors with virtually complete and independent follow-up of participants for hospitalisation and death, through data linkage. CVD mortality, reported as the underlying cause of death on the death certificate, is therefore captured using this design, as are hospitalisations including CVD as either a primary or secondary diagnosis. For conditions that generally result in hospitalisation and/or death and are well recorded in routine data, such as AMI [26] and cerebrovascular disease [27], combined hospitalisation and mortality data provide sound estimates of incidence. Other conditions, such as angina, chronic IHD, heart failure and dysrhythmias, do not consistently result in hospitalisation or death—and in the case of angina and heart failure, are not generally regarded as legitimate underlying causes of death [28]. In these cases, the rates presented should be considered to reflect hospitalisation with or death from these conditions and not necessarily incidence.

Our study finds, for the first time, a significant increase in the risk of incident paroxysmal tachycardia hospitalisation or death, including for supraventricular tachycardia and ventricular tachycardia, in current versus never smokers; our searches did not locate any comparable studies reporting this,worldwide. These findings are consistent with limited smaller-scale clinical data demonstrating elevated risks in current versus never smokers of ventricular tachyarrhythmias in trial patients with IHD and reduced ejection fraction receiving an implantable cardioverter defibrillator [29], ventricular tachycardia among patients with systolic heart failure [30], and complex arrhythmias in men with existing CVD [31]. Paroxysmal tachycardia (ICD10-AM I47), particularly ventricular tachycardia (I47.2), is potentially life threatening [13, 32]. We also found a relative risk of 1.88 (1.46–2.43) for hospitalisation with or death from cardiomyopathy in current versus never smokers and were unable to locate any directly comparable studies. However, a 1994 publication from the Multiple Risk Factor Intervention Trial found a relative risk of death from idiopathic dilated cardiomyopathy of 1.39 (1.18–1.63) per additional pack of cigarettes smoked per day, using a reference group of men who did not smoke at baseline [33], and a 1993 publication showed statistically compatible findings [34].

Where comparison is possible and bearing in mind the general variation in smoking-related CVD risks according to age, the relative risks observed in this study are broadly consistent with contemporary estimates from countries with similar historical smoking profiles. For example, our finding of a relative risk of IHD mortality of 3.07 (2.48–3.79) in current versus never smokers is consistent with recent US estimates of IHD mortality of 3.0 (2.8–3.2) and 3.5 (2.7–4.6) in women and 2.0 (1.8–2.2) and 3.2 (2.5–4.1) in men [6, 35], and somewhat lower than estimates in the UK women of 4.47 (4.19–4.77) [5]. It is also similar to that seen in the INTERHEART study in 52 countries from 2000 to 2002, which found a relative risk of non-fatal AMI of 2.95 (95%CI 2.77–3.14) in current versus never smokers [36]. Our finding of a relative risk of fatal cerebrovascular disease of 2.26 (1.65–3.10) in current versus never smokers compares to relative risks of 2.1 (1.8–2.3) and 3.2 (2.2–4.7) in women and 1.9 (1.7–2.2) and 1.7 (1.0–2.8) in men from the USA [6, 35] and 3.06 (2.83–3.31) in women from the UK [5]. Our findings of RRs in current versus never smokers for ischaemic stroke (I63) of 2.01 (1.69–2.40) and of intracerebral haemorrhage (I61) of 1.84 (1.33–2.54) are comparable to the figures of 2.17 (2.06–2.28) and 1.77 (1.60–1.95), respectively, from European and North American studies included in a recent meta-analysis [37]. The large > 6-fold smoking-related increases in the risk of abdominal aortic aneurysm and peripheral arterial disease are also consistent with those observed elsewhere for hospitalisation/incidence [7] and mortality [5, 6, 35].

Broad consistency with the published evidence—characterised by overlapping confidence intervals between our estimates and those of the bulk of the published estimates—was observed for our findings in current versus never smokers of 30–40% increases in the risk of atrial fibrillation (Additional file 1: references 14–23) and pulmonary embolism and other venous thrombosis (phlebitis and thrombophlebitis) (Additional file 1: references 3, 24–41). We found a relative risk of 2.23 (1.96–2.53) for heart failure in current versus never smokers; a number of other studies report at least a doubling in risk (Additional file 1: references 42–47), while others report lesser increases in relative risk (Additional file 1: references 48–53). Previous studies consistently show significantly elevated risks of non-rheumatic aortic and mitral valve disorders in current versus never smokers (Additional file 1: references 54–60), with confidence intervals overlapping substantively those of our null results.

Although hypertensive kidney disease and hypertensive heart disease were not significantly associated with smoking in our study, the confidence intervals around these estimates are relatively wide and our point estimates are compatible with findings from a previous study showing significantly elevated mortality from these conditions in current versus never smokers [6].

Composite endpoints, such as “any major CVD” and “IHD”, will necessarily include a different combination of CVD subtypes for fatal and non-fatal outcomes; fatal outcomes will tend to be dominated by mortality from AMI, stroke and heart failure whereas hospitalisation for angina is a common non-fatal outcome (Additional file 2: Table S5). Hence, composite outcomes based on fatal versus non-fatal events are not directly comparable. It follows that studies that investigate smoking-related risks combining multiple CVD subtypes as endpoints (e.g. IHD, combining AMI and angina) are, in effect, summarising potentially heterogeneous RR. Moreover, consistent with previous evidence, our study finds generally higher RR for current versus never smokers for outcomes captured virtually completely by hospitalisation and death data, such as AMI, compared to outcomes which are not, such as angina. The extent to which these differences reflect real differences in the relationship of smoking to disease incidence across subtypes, or under-ascertainment of incidence in those outcomes not well captured, cannot be determined from this study. We therefore emphasise the analyses based on specific subtypes, and the fact that findings relate to hospitalisation and mortality, and, when considering the relationship of smoking to disease incidence, those well captured.

The risks of 29 CVD subtypes were significantly elevated in current compared to never smokers in our study, contributing evidence that few, if any parts of the cardiovascular system remain untouched by smoking, as measured by the occurrence of disease serious enough to cause hospitalisation or death. The strength and consistency of the observed relationship with smoking, the finding of a dose-response relationship and attenuation of risk with quitting are all consistent with the accepted fact that smoking causes CVD. While CVD subtypes tend to be closely related to one another, the underlying pathophysiological mechanisms of these conditions cover the diversity of causes, including atherosclerotic disease, thrombosis, dysrhythmia, hypertension, heart failure and cardiomyopathy.

Our study shows, for the first time to our knowledge, that current smoking confers similarly elevated risks of fatal and non-fatal disease, at least for the outcomes where both incidence and mortality are captured well using data linkage, namely AMI and cerebrovascular disease. Where non-fatal outcomes are unlikely to represent incidence, and smoking-related RRs appear to differ between fatal and non-fatal outcomes, those from the fatal outcomes are likely to be more reliable estimates of the true effect of smoking on disease incidence, as it is not possible in this situation to distinguish between genuinely different RR and differences resulting from issues capturing incidence.

Our findings add to the worldwide evidence that smoking-related risks vary according to CVD subtype, as well as providing independent international confirmation of the magnitude of relative and absolute risks. That such variation is seen within fatal outcomes indicates that it is unlikely to be the result of differential ascertainment or other methodological issues [5, 6].

This study is large and population based, with prospective detailed ascertainment of smoking status from questionnaire items that have been used by the Million Women Study, supporting direct international comparison of results [5]. Since many people now survive a CVD event, a key strength of this study is the inclusion of both fatal and non-fatal events, which provides a more complete and accurate picture of the health effects of smoking. We excluded people with cardiovascular disease and with cancer at baseline, to minimise the impact of smoking cessation due to ill health and to focus on the likely causal effects of smoking. Because of this tendency for smokers—especially older smokers—to quit due to illness, we could not reliably quantify the relationship of smoking cessation after age 55 to the risk of CVD. Overall, 32% of current smokers at baseline who were part of a sample resurveyed 3 years later had quit [16]. This means that the findings presented here are likely to be conservative estimates of the true effect of smoking, apart from the estimates of the relation of CVD to the number of cigarettes smoked each day, where the risks relate more closely to the usual number of cigarettes smoked during follow-up. At the time of the baseline survey, around 12% of individuals aged 45 and over in New South Wales were estimated to be current smokers [38], and following exclusions, around 8% of the 45 and Up Study cohort included in this study were current smokers. Although the 45 and Up Study is, like the vast majority of cohort studies, not designed to be strictly representative of the general population, RR estimates based on internal comparisons within the cohort, such as those presented here, remain valid [38, 39].

The evidence presented here is likely to be of use to the community, policymakers and practitioners in appreciating the magnitude of the harms of smoking across the spectrum of CVD, and the likely benefits of cessation, at the individual and population level; it is especially useful in providing information from a context with low current smoking prevalence and high historical prevalences. CVD is a major cause of morbidity, disability and reduced quality of life, as well as being responsible for large healthcare costs. Our findings strengthen the established evidence on the impact of smoking on CVD mortality and conditions such as ischaemic heart disease, stroke and peripheral vascular disease. They expand the breadth of the known CVD harms of smoking, to include greater evidence on non-fatal disease, smoking relatively few cigarettes per day and on common, high morbidity and mortality conditions such as heart failure, cardiac dysrhythmias and cardiomyopathy. They indicate that quitting smoking and other tobacco control measures are likely to prevent these conditions, reduce healthcare burden and morbidity and improve quality of life. These findings are informative for population-level tobacco control efforts to prevent CVD events and are also likely to be of use clinically, particularly in terms of encouraging quitting in those affected by the CVD conditions identified here as being smoking-related, as well as informing early detection of CVD in current and past smokers.

Because CVD is common and smoking substantially increases the risk of CVD, smoking is responsible for large numbers of cases of CVD, even in countries with extensive tobacco control measures and large declines in smoking prevalence, such as Australia. The prevalence of current adult daily smoking in Australia in 2016 was estimated at 13%, amounting to around 2.7 million current daily smokers; 2% are current smokers who do not smoke on a daily basis and 24% have smoked in the past [40]. Despite the low prevalence, smoking is ranked as the leading or second leading risk factor for burden of disease in Australia [2, 11]; 12% of the CVD burden in 2011 was attributable to smoking [11]. Particularly motivating findings for individuals considering quitting include the observed CVD benefits of quitting at any age, with those quitting by age 45 avoiding almost all of the excess risk, as well as the fact that the balance of probabilities means that a CVD death, myocardial infarction, stroke, hospitalisation with heart failure or aortic aneurysm in a smoker is more likely than not to have been caused by smoking.

In conclusion, smoking increases the risk of virtually all CVD subtypes, including that of the newly identified risk of paroxysmal tachycardia. Quitting reduces the risk. Smoking accounts for a substantial proportion of premature CVD events in the setting of an established smoking epidemic with low current smoking prevalence.