Global smoking trends in inflammatory bowel disease: A systematic review of inception cohorts

Background and aims The effect of smoking on the risk of developing inflammatory bowel diseases (IBD) may be heterogeneous across ethnicity and geography. Although trends in smoking for the general population are well described, it is unknown whether these can be extrapolated to the IBD cohort. Smoking prevalence trends specific to the global IBD cohort over time have not been previously reported. This is a systematic review of smoking prevalence specific to the IBD cohort across geography. Methods A systematic literature search was conducted on Medline and Embase from January 1st 1946 to April 5th 2018 to identify population-based studies assessing the prevalence of smoking at diagnosis in inception cohorts of Crohn’s disease(CD) or ulcerative colitis(UC). Studies that did not report smoking data from time of diagnosis or the year of IBD diagnosis were excluded. Prevalence of smoking in IBD was stratified by geography and across time. Results We identified 56 studies that were eligible for inclusion. Smoking prevalence data at diagnosis of CD and UC was collected from twenty and twenty-five countries respectively. Never-smokers in the newly diagnosed CD population in the West has increased over the last two decades, especially in the United Kingdom and Sweden; +26.6% and +11.2% respectively. Never-smokers at CD diagnosis in newly industrialised nations have decreased over the 1990s and 2000s; China (-19.36%). Never-smokers at UC diagnosis also decreased in China; -15.4%. The former-smoker population at UC diagnosis in China is expanding; 11%(1990–2006) to 34%(2011–2013). Conclusion There has been a reduction in the prevalence of smoking in the IBD cohort in the West. This is not consistent globally. Although, smoking prevalence has decreased in the general population of newly industrialised nations, this remains an important risk factor with longer term outcomes awaiting translation in both UC and CD.


Introduction
Our group has extensively reported that inflammatory bowel diseases (IBD) have become a global challenge in the 21 st century. [1][2][3][4][5] The rapidly accelerating incidence of both Crohn's disease (CD) and ulcerative colitis (UC) in the newly industrialized countries in the East mirrors epidemiological patterns of IBD in the West more than 75 years ago. [2] The evolving epidemiology of IBD is thought to be associated with the industrialisation of society. The rise of IBD incidence in newly industrialised nations combined with reports of comparable rates of IBD in migrant and native populations in the West[6] support the theory that environmental triggers and Western lifestyle have an integral role in the pathogenesis of IBD. [3,4] The dichotomous relationship between smoking and the development of IBD has been the subject of intense scrutiny and is a complex interplay of genetics, immunology and environment. In the West, smoking has been consistently reported as a risk factor for developing CD and adversely affects disease course [7][8][9], whereas former smokers and non-smokers are at increased risk of developing UC in comparison to current smokers. [10,11] In contrast, studies in non-Western populations have been unable to replicate this association between CD and smoking. [12] The interaction between smoking and the NOD-2 gene and their effect on the risk of CD has been postulated to be specific to the 1007 fs mutation and a negative association between NOD-2 mutation and smoking could be explained by their inverse relationship. [13] An understanding of global smoking prevalence trends specific to the IBD cohort is required as the foundation for further investigation of the heterogeneous influence of this risk factor in IBD pathogenesis and disease course across different regions. In addition, this is increasingly important in light of the identification of smoking as a key risk factor for nonresponse to anti-TNF agents in patients with CD. [14] However, the global prevalence of smoking associated with IBD have not been collated and reported. We conducted a systematic review to assess the prevalence of smoking in all population based IBD inception cohort studies. We examined smoking prevalence specific to individual IBD cohorts across time and geography.

Search strategy and selection criteria
This systematic review was conducted according to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines. [15] A systematic literature search (S1 Table) was conducted on Medline (01 January 1946 to April 5th 2018) and Embase (01 January 1947 to April 5th 2018) for studies assessing the prevalence of smoking at diagnosis in inception IBD cohorts. All studies from our previous systematic review on IBD epidemiology [1], [5] as well as the reference lists of all relevant articles were included. We also obtained data outside of the search strategy using expert knowledge of active studies as with the Asia-Pacific Crohn's and Colitis Epidemiologic Study Group [ACCESS]).
All stages of the systematic review were independently conducted by two teams; the first from the United Kingdom (TT and JSC) and the second from Hong Kong (SCN, VSWL and CYL). The first stage consisted of an initial screening of abstracts and titles of search results. Studies were excluded if they were not observational in design and did not report original data (i.e. review articles). Studies were considered for final inclusion in the review if their study participants consisted of a population-based inception cohort of CD and/or UC with raw numbers reported to enable the calculation of ever and/or never smoking proportion at time of IBD diagnosis. Studies could also be included if they expressed the frequency of smokers or non-smokers. A study was considered to be population-based if the sample was representative of geographical region. Smoking data had to be reported separately in CD and UC cohorts for inclusion. Studies that did not report smoking data from time of diagnosis or did not have the year of IBD diagnosis were excluded. Discrepancies between the reviewers were resolved in conjunction with GGK, SG and SCN. The flow chart for the above process is presented in Fig 1.

Data analysis
The data extracted included: author, geographical location, study period, size of CD or UC cohort, frequency of current, former and never-smokers including unknown smoking status. Study quality was ascertained using a modified version of the Newcastle-Ottawa Scale (S2 Table). The modified scale addressed aspects of quality relevant to population-based inception cohorts as well as ascertainment of smoking exposure.
We classified geographic regions according to proximity and economic similarity based upon the United Nations classification of economic region as in our previous work. [1],[5] The regions included are: North America, South America, Eastern Europe, Northern Europe, Southern Europe, Western Europe, Asia and Oceania.
Scatter plots (created using Plotly (Montreal, Canada) were used to display time trends across geography in the proportion of never and ever smokers in inception cohorts of CD ( Fig  2) and UC (Fig 3) between 1980 to 2013. The earliest and latest years for which smoking data was available was 1980 and 2013 respectively. Smoking prevalence in local jurisdictions/regions were extrapolated to the entire country. In studies that reported smoking prevalence across a range of years, the median year was selected. Where studies reported former smokers, these patients were pooled with current smokers to formulate an ever-smoker category. In studies that reported only current and former category or an ever smoker category, the remainder of the population were designated as never-smokers. In UC, we sought to assess the former smoker population as this is considered the at-risk population. However, the former smoker population was also incorporated into the ever smoker population and reported for consistency. Studies with a total sample size of less than 10 subjects were excluded from these graphs. Further analysis in the form of meta-analysis or time trend analyses were not deemed appropriate due to paucity of data and heterogeneity in study design. Apart from ever smoking and never smoking data, quantitation of smoking in terms of average number of cigarettes smoked or duration of smoking were not available from the population based epidemiological data. full text review. 56 studies were eligible for final inclusion in the systematic review. These included 44 studies in CD and 46 studies in UC (Fig 1). Characteristics of all included studies are presented in Tables 1 and 2.
Smoking prevalence figures were reported for: North America (2 studies), South America (1 study), Eastern Europe (3 studies), Northern Europe (16 studies), Southern Europe (12 studies), Western Europe (4 studies), Asia (15 studies), and Oceania (3 studies). Scatter plots representing never-smoker prevalence in the CD and UC cohorts from 1980 to 2018 stratified by geographic region are presented in Figs 2 and 3 respectively. Smoking prevalence varied greatly according to geographic region. Fig 2 shows that an increasing number of the newly diagnosed CD population over the last two decades in the West particularly in the UK have never smoked. In contrast, a decrease in the proportion of never-smokers over the 1990s and 2000s is seen in newly industrialised nations such as China. Fig 3 is suggestive of significant heterogeneity in the trend of the never-smoker group in the newly diagnosed UC population in the West. Data from the United Kingdom and Sweden over the 1980s and 1990s suggest a decrease in this group whilst data from Iceland and Italy show an increase in the never-smoking proportion.    n/a n/a n/a No (Continued ) shows that the proportion of people who have never smoked at UC diagnosis in newly industrialised nations particularly China has been decreasing over the last two decades. Tables 3 and 4 displays these ranges stratified according to geographic region.

Crohn's disease
Smoking prevalence data at diagnosis of CD was collected from twenty countries. The western world particularly Europe has demonstrated an overall increase in the prevalence of never smokers in the newly diagnosed CD cohort over the last three decades. In Sweden (Northern Europe), the proportion of never-smokers increased from 31.6% [16] in the 1980s to 42.8% (2007) [17]. In the early 1990s, the proportion of never-smokers in the newly diagnosed CD cohort in the UK was 38%. [18] A large population-based inception cohort study  [19] in the UK estimated that 64.6% of newly diagnosed CD patients were never-smokers. This trend is replicated in Western Europe, Southern Europe and Eastern Europe. The proportion of never-smokers in the CD cohort in the Netherlands ranged from 29.7% [20] to 31% [21] in the 1990s however France and Germany demonstrated a neversmoker proportion of 65% [22] and 63.1% [23] in the 2000s respectively. In Italy (Southern Europe), there was a steady increase in the never-smoker population at CD diagnosis over the course of the 1980s [24] and 1990s. [25] Similarly, Spain showed consistent trends with the ever-smoker group steadily declining from 66%(1980 and 1990s) [26] to 46%(2001) [27] in the newly-diagnosed CD cohort. Similarly, in Hungary (Eastern Europe), the proportion of never smokers in the newly diagnosed CD cohort increased from 44.3%[8] to 50.2% [28] over the course of 30 years.    The study defined the current smoker, former smoker or never smoker groups. Alternatively, the authors quantified missing data. https://doi.org/10.1371/journal.pone.0221961.t002 In contrast to Europe, smoking prevalence in inception CD cohorts in Asia appears to be increasing over time. The majority of CD subjects in Asia were never-smokers. The proportion of subjects who had never smoked range from 75% [29] (Hong Kong, China;1985-2001) to 92.6% [30] (Hong Kong, China;1991. However, in a more recent inception cohort from Asia from 2011-2013(ACCESS), 73.2% of CD subjects were never smokers. Nine out of 44 studies did not report former smokers. Never-smoker populations were assumed to be the remainder of the population if ever smoker data was provided.
Other Scandinavian regions such as Denmark showed only a slightly higher proportion of never smokers in their UC cohorts; 45%(2004) [35]. In contrast to the remainder of Northern Europe in the 1980s and 1990s, Iceland demonstrated an increase in the never-smoker proportion from 13.5% [36] to 35.8% [37] across this period. The percentage of former smokers at UC Table 3 Global smoking trends in inflammatory bowel disease: A systematic review of inception cohorts diagnosis also rose from 11.7% to 20% across those two decades. These results co-relate with a decrease in the ever-smoker proportion down to 48% [38] in the 21 st century.
The proportion of newly diagnosed UC subjects who have never smoked has decreased in China over the last two decades. The proportion who had never smoked were 84.9% [41] in China in the late 1990s. By 2012 these figures had decreased to 69.5% (ACCESS Cohort;2011-2013). The proportion of former smoker patients in the newly diagnosed UC cohort in China appears to be increasing from 11%(1990-2006) [42] to 34% (Hong Kong ACCESS cohort).
Data from major cities in Australia suggest that the proportion of never-smokers in the newly diagnosed UC cohort has increased from 58.8%(Sydney;1992) [43] to 77%(Victoria; 2007-2013) [44]. The former smoker proportion of patients at diagnosis also decreased from Table 4 Global smoking trends in inflammatory bowel disease: A systematic review of inception cohorts 29.4% to 18% across the same regions and time periods respectively. The impact of missing smoking data regarding the participants vary due to heterogeneity in reporting. Four out of the 46 studies included for UC did not report former smokers at diagnosis. Never-smoker populations were assumed to be the remainder of the population if ever smoker data was provided.

Discussion
We present a comprehensive review of smoking trends over time in inception IBD cohorts worldwide. In the West, the proportion of newly diagnosed CD subjects who have never smoked has increased over time. The proportion of newly diagnosed UC subjects who have never smoked has declined in the 1980s and 1990s in Europe although an increase was noted in Western Europe from the late 1990s. In contrast, the proportion of subjects who have never smoked at IBD diagnosis has decreased in Asia, particularly in China. Thus, we demonstrate that trends in smoking prevalence specific to the IBD cohort do not mirror global trends in smoking discerned from the general population. [45] The incidence of IBD in newly industrialised countries is accelerating whilst the incidence of IBD is stabilising in the West. [1] The effect of smoking on the incidence of IBD across the globe likely varies due to heterogeneity in genetic susceptibility and the presence of other risk factors. Public health measures in the 1980s and 1990s led to a reduction of smoking prevalence in the general population in many Western countries. [46] The higher proportion of never smokers at diagnosis of CD over time may be explained by adolescents who decided not to smoke in the 1990s. This could have potentially contributed to the stabilization, and in certain regions decrease, in the incidence of adult-onset CD in some Western countries. This ecological trend could also explain the decrease in the former smoker population in the UK over the 1990s and could have contributed to the recent stabilisation of UC incidence. [1] In contrast, we are at the infancy of the IBD 'epidemic' in newly industrialized countries in Asia, especially in areas of high smoking prevalence [46]; hence ever-smoker trends at IBD diagnosis are on the increase. The rapid expansion of the former smoker population at UC diagnosis in China is suggestive of a rapid expansion of the at-risk population. The Global Burden of Disease Study 2015 identified China as one of the leading countries in the world for the total number of smokers. [46] In line with the Lopez model [47], these newly industrialised nations are rapidly moving towards Stage IV where smoking prevalence in the general population will decrease as societal attitudes shift and government anti-smoking policy becomes comprehensive. This could potentially foreshadow a protracted course of high UC incidence in comparison to CD. Similar to the West, we hypothesise that a 'lag effect' can be expected in future epidemiological studies particularly in CD based in newly industrialised nations. However, due to the complex interplay between genetics and environment in the development of IBD, this effect may not be as pronounced as in the West [12] although in contrast to CD, there is some evidence to suggest that the role of smoking in UC is uniform across the East and West. [12,48,49] The concurrent decrease of the never-smoker population in both CD and UC cohorts in newly industrialised nations is potentially suggestive of significant heterogeneity in the interaction between smoking and the process of IBD development across geographic regions. Even in the West, the incidence of CD had been high in relatively low smoking prevalence populations i.e. Israeli Jews [50], Canada, and Sweden. Multiple studies [12], [29], [51] in the Asia-Pacific region have demonstrated that active smoking does not confer an increased risk of CD in this population as it does in the West. The relative absence of the NOD-2 mutation in CD cases in Japan suggests that the role of smoking in IBD is subject to underlying genetic heterogeneity. [52] Environmental factors such as air pollution [53], diet and a Western lifestyle as demonstrated in migrant sub-populations[6] as well as evolving early life feeding patterns and improved hygiene as part of socioeconomic development [12] could be more potent mediators of IBD development. [54,55] Our study has several limitations predominantly due to lack of available data. We were unable to perform a meta-analysis or ecological trend analysis due to study heterogeneity. Small sample sizes in some studies have also increased the risk of imprecise estimates for smoking prevalence. A paucity of gender-specific, age-category specific smoking prevalence data, data relating to quantification of smoking habits or breakdown of rural vs. urban data in the IBD cohort did not allow for further sub-group analysis. Although, it is possible studies that included children and adolescents would have a higher prevalence of never-smokers, summary statistics from included studies suggest this is not the case. The exposure to smoking was reported inconsistently; some studies reported current and former smokers whilst others reported ever and never smokers. Twenty three out of ninety-five included cohorts reported missing smoking data on participants (Tables 1 and 2). No data was available regarding second-hand smoking exposure. Due to differing study periods and the generalisation of regions to represent countries, smoking data was not fully homogenous. The inequalities in healthcare access across the globe can also affect data collection and reporting. In addition, we acknowledge that the attributable risk of smoking on IBD is low (i.e. most IBD patients do not have a history of smoking [current or former] prior to their diagnosis), however it remains an important risk mediator in the development of IBD.
Despite these limitations, this study provides a comprehensive overview of the prevalence of smoking in the global IBD cohort across time and geography. The proportion of neversmokers in IBD cohorts from newly industrialised countries appears to be decreasing over time in contrast to the IBD cohorts in the West. In light of our previous work and this study, it appears that IBD epidemiological patterns globally can be modelled along geographical and development lines within a context of genetic heterogeneity and environmental ecological exposures. It remains of clinical importance for medical practitioners to record information and act on smoking status for patients with IBD regardless of geography and ethnicity, especially in light of data suggesting smoking confers an adverse disease course in CD and is a risk factor in non-response to anti-TNF therapy. [14] Large-scale prospective inception cohorts assessing the associations of smoking for both UC and CD in Eastern and Western populations will add to the available data. This is the first systematic review to assess trends in the prevalence of smoking in the IBD cohort worldwide. It provides a foundation for future work assessing the prevalence of this important risk mediator in a global setting as well as highlighting some of the challenges surrounding this data. A deeper understanding of IBD aetiology in relation to diet and other environmental factors across geographic regions and ethnicities is urgently required in order to formulate strategies to slow the global increase in the incidence of IBD.
Supporting information S1