OUP user menu

Journal of Crohn's and Colitis: 10 (10)


Laurence J. Egan, Ireland

Associate Editors

Shomron Ben-Horin, IsraelSilvio Danese, ItalyPeter Lakatos, HungaryMiles Parkes, UKJesús Rivera-Nieves, USABritta Siegmund, GermanyGijs van den Brink, NLSéverine Vermeire, Belgium


Published on behalf of

Crohn's disease and smoking: Is it ever too late to quit?

Ian C. Lawrance, Kevin Murray, Birol Batman, Richard B. Gearry, Rachel Grafton, Krupa Krishnaprasad, Jane M. Andrews, Ruth Prosser, Peter A. Bampton, Sharon E. Cooke, Gillian Mahy, Graham Radford-Smith, Anthony Croft, Katherine Hanigan
DOI: http://dx.doi.org/10.1016/j.crohns.2013.05.007 e665-e671 First published online: 1 December 2013


BackgroundSmoking increases CD risk. The aim was to determine if smoking cessation at, prior to, or following, CD diagnosis affects medication use, disease phenotypic progression and/or surgery.

MethodsData on CD patients with disease for ≥ 5 yrs were collected retrospectively including the Montreal classification, smoking history, CD-related abdominal surgeries, family history, medication use and disease behaviour at diagnosis and the time when the disease behaviour changed.

Results1115 patients were included across six sites (mean follow-up—16.6 yrs). More non-smokers were male (p = 0.047) with A1 (p < 0.0001), L4 (p = 0.028) and perianal (p = 0.03) disease. Non-smokers more frequently received anti-TNF agents (p = 0.049). (p = 0.017: OR 2.5 95%CI 1.18–5.16) and those who ceased smoking prior to diagnosis (p = 0.045: OR 2.3 95%CI 1.02–5.21) progressed to complicated (B2/B3) disease as compared to those quitting at diagnosis. Patients with uncomplicated terminal ileal disease at diagnosis more frequently developed B2/B3 disease than isolated colonic CD (p < 0.0001). B2/B3 disease was more frequent with perianal disease (p < 0.0001) and if i.v. steroids (p = 0.004) or immunosuppressants (p < 0.0001) were used. 49.3% (558/1115) of patients required at least one intestinal surgery. More smokers had a 2nd surgical resection than patients who quit at, or before, the 1st resection and non-smokers (p = 0.044: HR = 1.39 95%CI 1.01–1.91). Patients smoking > 3 cigarettes/day had an increased risk of developing B2/B3 disease (p = 0.012: OR 3.8 95%CI 1.27–11.17).

ConclusionProgression to B2/B3 disease and surgery is reduced by smoking cessation. All CD patients regardless of when they were diagnosed, or how many surgeries, should be strongly encouraged to cease smoking.

  • Smoking
  • Crohn's disease
  • Surgery
  • Disease progression

1 Introduction

Avoiding the complications that result from a chronic disease is a priority for both the physician and the patient. In Crohn's disease (CD), one prognostic factor under the control of the patient is whether, and how much, they smoke. Smoking has clearly been associated with an increased risk of developing CD,1,2 with disease flares,3 worse disease progression, greater use of corticosteroids and immunosuppresive agents, and the need for initial and recurrent intestinal surgeries.4,5 Conversely, following smoking cessation, the risk of developing CD decreases after 4 years6 and, for CD patients with prior intestinal resection, the recurrence of endoscopic and clinically significant disease is markedly reduced.7

Although smoking cessation is not easy for many people, it remains a major therapeutic strategy in the management of CD, and can be more effectively achieved with structured patient education and the availability of counseling.8 With this support almost one third of smokers can completely stop smoking.8 In a small prospective study of 59 patients smoking cessation was also shown to reduce disease flares by up to 65% and was associated with less corticosteroid use compared to continuing smokers.9

If, however, smokers are more likely to develop CD and have a worse outcome and this can be reversed by the cessation of smoking, is there a point at which stopping smoking is no longer effective? Do patients who cease smoking at the time of diagnosis follow a more benign disease course, similar to that of non-smokers, or do they in fact have a better prognosis compared to non-smokers? This may be possible as one of the risk factors contributing to a poor disease outcome can be removed and a proportion of the smokers could have, potentially, not developed CD at all except for the added impact of smoking. Previous studies have not collected data on disease duration or change in disease behaviour in relation to the timing of smoking cessation and are thus insufficiently detailed to answer these clinically important questions. The aim of this study was to determine if smoking, the number of cigarettes smoked, not smoking and the timing of smoking cessation (at, prior to, or following, the diagnosis of CD) alters medication use, is associated with changes in disease phenotype and/or need for a first or subsequent CD intestinal surgery.

2 Methods

Data were requested from Gastroenterology units with a specialized interest in IBD across Australia (including sites from Queensland, New South Wales, Victoria, South Australia and Western Australia) and New Zealand, and who were all members of the Australia and New Zealand (ANZ) IBD Consortium. Prospectively collected patient data of sufficient quality were obtained from six IBD centers. Each site has an interest, and expertise, in the management of the IBD patient and included: (i) the Centre for Inflammatory Bowel Diseases, Fremantle Hospital, Fremantle, Western Australia; (ii) Department of Medicine, University of Otago, Christchurch, Christchurch, New Zealand; (iii) The IBD Service at Royal Adelaide Hospital, Adelaide, South Australia; (iv) Royal Brisbane and Women's Hospital, Brisbane, Queensland; (v) Flinders Medical Centre, Adelaide, South Australia; and (vi) Department of Gastroenterology, Townsville Hospital, Townsville, QLD, Australia.

All patients had CD and were phenotyped using the ‘Montreal Classification’.10 The diagnosis of CD had to be definite, and was made in accordance with previously established criteria based upon clinical, endoscopic, histopathological and radiological findings.11 All patients had had a diagnosis of CD for a minimum of 5 years till the end of follow-up in April 2011.

Data was collected from all patients at first clinical visit about their smoking history (start date, stop date, average number of cigarettes per day), number, dates and types of CD-related abdominal surgeries, family history (defined as either a first or second degree relative with either CD or ulcerative colitis) and medication use. These data were then updated for each patient at each subsequent clinical visit and thus the date of cessation of smoking, disease behaviour at time of diagnosis (B1 — inflammatory, B2 — stricturing or B3 — perforating), disease behaviour at follow-up, and the date at which the disease was noted to change behaviour, if appropriate, was collected.

A person was defined as having been a smoker if they had smoked at least one cigarette a day for a minimum of 3 months (100 cigarettes). A current smoker was defined as someone who smoked at least 1 cigarette a day for a minimum of 3 months and was continuing to do so, while an ex-smoker was a smoker who had ceased smoking for at least 3 months and had not recommenced smoking at any time point in the follow-up period to April 2011. An ‘ever smoker’ was either an ex-smoker or a current smoker, while a life-long non-smoker had never regularly smoked 1 or more cigarettes per day and had not smoked 100 cigarettes in their life.

2.1 Statistics

Cumulative numbers of first and second resections were examined using summary tables and charts. Standard univariate analyses were carried out using logistic regression and t-tests where appropriate. Logistic regression was used to examine the factors that affected initial and second intestinal resections separately. The potential list of candidate variables examined for each were smoking status, age group, disease location, L4 disease, perianal disease, disease behaviour at diagnosis, the use of an anti-TNFα agent, immunosuppression, oral or intravenous (i.v.) steroids at any stage in the disease management, years of follow-up until the first surgical resection, or years since the previous resection for the 2nd surgical resection.

When examining changes in disease behaviour over time, only patients with B1 disease at diagnosis were included together with their total length of follow-up (n = 803 patients). The change in disease behaviour from B1 to either B2 or B3 disease was modeled. The same predictors were examined, as detailed above, to determine the factors that may result in this change.

To examine time to 1st/2nd resection, and time to change in behaviour, Kaplan–Meier survival curves were produced and log rank tests were carried out to examine the impact of all explanatory variables univariately. Multivariate time to event modeling was carried out using Cox proportional hazards models. In all multivariate models, variables in the final model were determined using an information theoretic approach.18

3 Results

Complete data were obtained from 1115 CD patients from the six centers up to April 2011. Of these, 637 (42.9%) were male and 591 (53.0%) had never smoked (Table 1). The mean disease follow-up was 16.6 yrs (± 9.2 yrs) with 348 (31.2%) of patients suffering from L1, 336 (30.1%) from L2 and 431 (38.7%) from L3 disease at diagnosis. Inflammatory disease (B1) was present in 818 (73%) patients at diagnosis and 264 (23.7%) had a family history of IBD. Comparing life-long non-smokers to those patients who had ever smoked, non-smokers were more likely to be male (46.0% vs. 40.1%, p = 0.047) and be A1 (diagnosed under 17 yrs of age, 19.7% vs. 8.0% p < 0.0001). At diagnosis, non-smokers were more likely to have L4 disease (proximal small bowel disease, 11.3% vs 7.4%, p = 0.028), perianal disease (36.6% vs 30.3%, p = 0.025) but no less likely to have isolated L1 disease (29.8% vs 32.5%, p = 0.33) nor more likely to have B2/B3 disease behaviour (24.8% vs 28.3%. p = 0.19). The mean follow-up time for smokers was greater than non-smokers (p = 0.0001; 17.6 yrs vs 15.4 yrs respectively).

View this table:
Table 1

Patient demographics and medication use.

All CD n = 1115Never smokers n = 524Ever smokers n = 591p-Value
Gender: male; n (%)478 (42.9)241 (46.0)237 (40.1)0.047
Smokers; n (%)
Ever smoker591 (53.0)0%100%NA
Never smoker524 (47.0)100%0%
Age at diagnosis: n (%)
A1 — < 17150 (13.5)103 (19.7)47 (8.0)< 0.0001
A2 — 17–40734 (65.8)336 (64.1)398 (67.3)0.26
A3 — > 40231 (20.7)85 (16.2)146 (24.7)0.0005
Disease follow-up: yr ± SD16.6 ± 9.215.4 ± 8.817.6 ± 9.50.0001
Location at Dx; n (%)
L1 — term ileum348 (31.2)156 (29.8)192 (32.5)0.33
L2 — colon336 (30.1)162 (30.9)174 (29.4)0.59
L3 — ileocolonic431 (38.7)206 (39.3)225 (38.1)0.67
L4 — upper GI103 (9.2)59 (11.3)44 (7.5)0.03
P — perianal371 (33.3)192 (36.6)179 (30.3)0.03
Behaviour at Dx; n (%)
B1 — inflammation only818 (73.4)394 (75.2)424 (71.7)0.19
B2 — stricture198 (17.8)83 (15.8)115 (19.5)0.11
B3 — perforation99 (8.9)47 (9.0)52 (8.8)0.92
Behaviour at F/U n (%)
B1 — inflammation only541 (48.5)271 (51.7)270 (45.7)0.044
B2/B3574 (51.5)253 (48.3)321 (54.3)
Family history; n (%)264 (23.7)127 (24.2)137 (23.2)0.68
Oral steroids; n (%)950 (85.2)442 (84.4)508 (86.0)0.45
I.V. steroids; n (%)474 (42.5)222 (42.4)252 (42.6)0.93
Immunosuppression n (%)801 (71.8)390 (74.4)411 (69.5)0.070
Anti-TNFα; n (%)363 (32.6)286 (35.5)177 (29.9)0.049

Over the follow-up period up till April 2011 and there was no difference between ever smokers and non-smokers in the percentage of patients requiring the use of oral or i.v. corticosteroids, or an immunosuppressant (thiopurine or methotrexate) at any stage in the management of their disease (Table 1). Non-smokers, however, were more likely to receive an anti-TNFα agent during follow-up (p = 0.049).

3.1 Change in disease behaviour to complicated disease

Over the follow-up period 123 of the 394 (31.2%) non-smokers with B1 disease at diagnosis progressed to either B2 or B3 disease. This resulted in 253 out of 524 non-smokers (48.3%) with B2/B3 disease at the end of follow-up. In ever smokers with B1 disease at diagnosis, 154 out of 424 (36.3%) progressed to B2/B3 resulting in 321 of the 591 smokers (54.3%) having complicated disease at the end of follow-up. Logistic regression on the 803 patients who had B1 disease at diagnosis and complete smoking history (Table 2 ) confirmed that patients with disease of the terminal ileum were more likely to develop complicated disease (p < 0.0001: L1 vs. L2, odds ratio [OR] 4.9 95% CI 3.00–7.83; L3 vs. L2, OR 4.0 95% CI 2.61–6.08). Of particular interest is that when smokers, who had B1 disease at diagnosis, were separated into when, or if, they ceased smoking, 99 of the 230 (43%) patients who continued to smoke developed B2/B3 disease which was significantly higher than compared to the 13 of 64 (20.3%) patients who quit at CD diagnosis (p = 0.017; OR 2.5 95% CI 1.18–5.16). In addition, patients who had been previous smokers but had ceased prior to diagnosis were also more likely to develop complicated disease than patients who quit smoking at diagnosis (p = 0.045; OR 2.3 95% CI 1.02–5.21), while there was no difference in progression from B1 to B2/3 between never-smokers and those who quit smoking at diagnosis (p = 0.33).

View this table:
Table 2

Disease and treatment characteristics and association with change in disease behaviour from B1 at diagnosis to B2/B3 by follow-up (n = 803).

Univariate OR95% Wald CIMultivariate OR95% CIp-Value (multivariate)
Location %
L1 vs L24.06(2.66–6.20)4.9(3.01–7.83)< 0.0001
L3 vs L24.3(2.95–6.27)34.0(2.61–6.08)< 0.0001
L3 vs L11.1(0.74–1.53)0.8(0.54–1.25)0.36
L4 (yes vs no)3.3(2.04–5.39)
Smoking at Dx not quit (n = 99/230) vs. smoking at Dx and quit (n = 13/64)3.0(1.53–5.75)2.5(1.18–5.16)0.017
Not smoking at Dx but an ever smoke (n = 41/115) vs. smoking at Dx and quit (n = 13/64)2.2(1.06–4.46)2.3(1.02–5.21)0.045
Never smoker (n = 123/394) vs. smoking at Dx and quit (n = 13/64)1.8(0.93–3.39)1.4(0.70–2.93)0.33
Immunosuppression use3.4(2.32–5.03)2.8(1.76–4.54)< 0.0001
I.V. steroid use2.3(1.69–3.06)1.7(1.19–2.42)0.0038
Perianal (yes vs no)2.4(1.80–3.33)2.2(1.55–3.16)< 0.0001
Years followed up1.1(1.04–1.08)1.1(1.05–1.10)< 0.0001

Not surprisingly, the development of complicated disease was more likely if immunosuppressive medications were required at any stage of the disease (p < 0.0001; OR 2.8 95% CI 1.76–4.54). This was also true for the need for i.v. steroids (p = 0.0038; OR 1.7 95% CI 1.19–2.42) and the presence of perianal disease (p < 0.0001; OR 2.2 95% CI 1.55–3.16). Analysis also confirmed that longer disease duration was associated with a greater likelihood of the development of complicated disease (p < 0.0001; OR 1.1 95% CI 1.05–1.10).

3.2 First intestinal surgery for CD

A total of 558 (50%) patients required intestinal surgery with 243 (46.4%) of patients who had never-smoked undergoing surgery compared to 315 (53.3%) patients who had ever smoked. Fifty-seven patients had quit smoking at diagnosis, and 214 prior to their first intestinal resection.

A Cox proportional hazards model with only smoking status as a predictor showed that smoking status was a statistically significant predictor of requiring surgery (p = 0.044). A Kaplan–Meier curve was produced for patients requiring surgery according to their smoking status (Figure 1A). Non-smokers and patients who had quit smoking at diagnosis were less likely to go to surgery than patients continuing to smoke. A subsequent multivariate analysis, however, demonstrated that surgery was more common in L1 and L3 disease compared to L2 (p < 0.0001) and perianal disease (p = 0.010), while the effect of smoking status, adjusting for these variables, was no longer statistically significant.

Figure 1

Kaplan–Meier plot for time to A) first resection and B) second resection.

3.3 Second intestinal surgery for CD

Of the 556 patients (243 non-smokers and 313 ever smokers: the complete smoking history was not complete for 2 of the ever smokers) who underwent a first intestinal resection for their CD, the chance of going to a 2nd surgery was examined. Overall, 186 patients underwent a 2nd intestinal resection and were analysed in three groups; non-smokers, patients who quit smoking at, or before, the first resection and continuing smokers. A significantly greater proportion of continuing smokers 95 of 233 (40.8%) underwent a 2nd surgical resection compared to 23 of 80 (28.8%) of the patients who quit at, or before, the 1st surgical resection and the 68 of 243 (28.0%) patients who were non-smokers (p = 0.0019).

A Kaplan–Meier plot for this survival data is shown in Figure 1B. The mean follow-up time for those who had a 2nd surgery was 17.5 yrs and for those who did not was 10.1 yrs (P < 0.0001). When the other potential confounders were examined the use of an immunosuppressant (p = 0.0013) and L4 disease (p < 0.0001) were significantly associated with a 2nd surgical resection. Adjusting for these variables there was no difference between patients who quit at, or before, the 1st surgical resection and non-smokers (p = 0.53; HR = 1.2 95% CI 0.72–1.87). There was, however, a statistically higher 2nd surgical rate between continuing smokers compared to non-smokers (p = 0.044; HR = 1.4 95% CI 1.01–1.91).

3.3.1 Number of smokes per day

Of those patients who were smokers at diagnosis, 323 were still smoking 5-years after diagnosis. A sub analysis on these 323 patients investigated the dose effect of smoking (as reported by patients) in the 287 who reported the number of cigarettes smoked daily: this ranged from 1 to 50 with mean of 15.22 (± SD 8.84). The same three parameters: change in disease behaviour from B1 at diagnosis to either B2 or B3 by follow-up (n = 202), 1st resection (n = 287) and 2nd resection (n = 161) were examined (Table 3 ).

View this table:
Table 3

Association between smoking levels, change in disease behaviour and surgery in CD.

No. of cigarettes per weekChange from B1 to B2/B3First surgerySecond surgery
OR95% CIp-ValueOR95% CIp-ValueOR95% CIp-Value
20 +3.771.27–11.1730.0121.310.58–2.980.520.930.30–2.880.90

There was a general trend for all three outcomes, which demonstrated a positive correlation between the proportion of patients having any of markers of disease progression and the number of cigarettes per day. Logistic regression analysis also demonstrated that the number of cigarettes per day was significantly associated with the change in disease behaviour. Patients who smoked more than 3 cigarettes/day had a 3.8 fold increased risk of developing complicated disease (p = 0.012).

4 Discussion

At present, smoking is the most modifiable risk factor for the development and progression of CD. This study examined whether smoking cessation at, prior to, or following, the diagnosis of CD can alter disease behaviour and the need for surgery with the data collected both retrospectively from the patients at their first clinical visit, and then the data was updated at each subsequent clinical visit. Although patient recall may not always be fully accurate, and is a limitation of any retrospective study, the dates and types of surgical resections are momentous events in a patient's life and usually well remembered and were also able to be confirmed through hospital notes in the vast majority of patients. Cessation of smoking, which, although cannot be confirmed through objective means, many patients can recall to the day and was also frequently collected prospectively from patients during the follow-up period suggesting that the data were reliable.

Unfortunately, CD is prone to excessive collagen deposition resulting in transmural fibrosis12,13 and fibrostenosing disease is one of the most common indications for intestinal resection in CD.14,15 As the progression of inflammatory to stenosing, or perforating, disease usually precedes the need for abdominal surgery, this study examined the change in disease behaviour from B1 to B2/B3 disease and the rates of first, and second, intestinal surgery in the management of CD. These aspects of disease progression were examined with respect to the patient's smoking status and when, or if, they ceased smoking. In line with other studies, disease duration, patients with L1 and perianal disease, patients suffering more severe disease indicated by the need for use of immunosuppressive medications or i.v. steroids, were more likely to develop complicated disease16,17 and these were controlled for in the analysis.

In this study, patients who had ever smoked in their life did not have a higher rate of complicated disease behaviour at diagnosis compared to lifelong non-smokers and over a mean follow-up period of more than 15 years, 36% of those patients who had ever smoked compared to 31% of non-smokers progressed from B1 to B2/B3 disease. Of note, however, significantly fewer patients who ceased smoking at diagnosis developed complicated disease compared with those patients who continued to smoke. In addition patients who quit smoking at diagnosis were also significantly less likely to develop complicated disease than patients who were previous smokers but had ceased prior to CD diagnosis. These observations clearly show that smoking is a modifiable risk factor and that quitting reduces the development of complicated CD. No significant differences, however, were observed between lifelong non-smokers and patients who quit smoking at diagnosis potentially due to low patient numbers. However, as 31% of life-long non-smokers developed B2/B3 disease compared to only 20% of ex-smokers there is a suggestion that the cessation of smoking may reduce the risk of disease progression below that of non-smokers. Adding biologic plausibility to the hypothesis, is that there was a dose effect observed with number of cigarettes smoked per week being significantly associated with the development of complicated CD. Patients who smoked 3, or more, cigarettes per day were almost 4 fold more likely to develop either fibrosing or perforating CD indicating that even a very few cigarettes are able to impact on disease behaviour and outcomes.

The need for surgery is driven by the progression of disease to a more complicated phenotype. In this study 50% of patients required abdominal surgery for their CD at some stage. It was noted that more smokers, than non-smokers, required surgery with a significantly greater rate of first surgery observed in patients continuing to smoke, compared to those ceasing at diagnosis and life-long non-smokers. This significance for smoking, however, was lost on multivariate analysis suggesting that even greater patient numbers are required for statistical analysis when controlling for all the aspects of disease behaviour and management.

A greater proportion of patients who continued to smoke underwent a 2nd intestinal surgical resection compared to patients who quit at, or before, the 1st surgical resection and lifelong non-smokers with a significant difference noted in the rates of surgery between continuing smokers and life-long non-smokers. The greater rates of first and second surgeries amongst smokers may be explain by the greater rates of stenosing and perforating disease and these changes could also explain the observation that smokers were less likely to be treated with an anti-TNF agent than non-smokers as complicated CD is primarily treated by surgery and not medical therapies. No difference, however, was noted in the rates of surgery between life-long non-smokers and those patients who gave up smoking at, or before, first surgery, which again suggests that giving up smoking at any point in the disease process is of benefit to the patient.

In summary, this study confirms that smoking cessation at the time of CD diagnosis is associated with a reduction in the development of complicated disease. Even as few as 3 cigarettes smoked per day correlated with a markedly increased risk in the development of complicated disease and this logically will result in a greater need for surgical resection over time. The findings suggest that there is unlikely to be a safe level of smoking for CD patients. The benefit of smoking cessation, however, does not appear to wane over time and thus all CD patients regardless of when they were diagnosed or how many surgeries they have had need to be strongly encouraged to cease smoking.

Conflict of interest

None of the authors have any conflicts of interest in relation to this paper.


View Abstract