By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Background: The aetiology of inflammatory bowel disease remains largely unknown.
Aim: We performed a comprehensive assessment of potential risk factors associated with the occurrence of inflammatory bowel disease.
Methods: We identified a cohort of patients 20–84 years old between 1995 and 1997 registered in the General Practitioner Research Database in the UK. A total of 444 incident cases of IBD were ascertained and validated with the general practitioner. We performed a nested case–control analysis using all cases and a random sample of 10 000 frequency-matched controls.
Results: Incidence rates for ulcerative colitis, Crohn's disease, and indeterminate colitis were 11, 8, and 2 cases per 100 000 person-years, respectively. Among women, we found that long-term users of oral contraceptives were at increased risk of developing UC (OR: 2.35; 95% CI: 0.89–6.22) and CD (OR: 3.15; 95% CI: 1.24–7.99). Similarly, long-term users of HRT had an increased risk of CD (OR: 2.60; 95% CI: 1.04–6.49) but not UC. Current smokers experienced a reduced risk of UC along with an increased risk of CD. Prior appendectomy was associated with a decreased the risk of UC (OR: 0.37; 95% CI: 0.14–1.00).
Conclusions: Our results support the hypothesis of an increased risk of inflammatory bowel disease associated with oral contraceptives use and suggest a similar effect of hormone replacement therapy on CD. We also confirmed the effects of smoking and appendectomy on inflammatory bowel disease.
The incidence of inflammatory bowel disease (IBD) tends to plateau in those regions with the greatest incidence rates but is still growing in low-incidence areas such as southern Europe, Asia, and much of the developing world.1 Nevertheless the aetiology of the disease remains largely unknown. While some risk factors such as smoking have been studied in more detail, the impact of most other factors on the two main entities that constitute IBD, ulcerative colitis (UC) and Crohn's disease (CD) has not been established yet.
We conducted a prospective cohort study with a nested case–control analysis using data from the General Practice Research Database (GPRD) to identify risk factors associated with the occurrence of UC and CD in this population.
The GPRD contains computerized information entered by selected general practitioners (GPs) in the UK. Data on about 2 million patients are systematically recorded and sent anonymously to the Medicines and Healthcare Products Regulatory Agency (MHPRA) that collects and organizes this information in order to be used for research projects. The information recorded includes demographics, medical diagnoses, referrals and hospitalizations. Also, prescriptions are automatically produced from the computer and recorded on the patient's computerized file. A previous study utilizing this computerized data source has documented that over 90% of all referrals are entered on the GP's computers with a code that reflects the clinical diagnosis.2
We identified all patients 20–84 years old between January 1995 and December 1997 registered in the GPRD. Patients became members of the study cohort (start date) on the first day of the study period when they meet the criteria of at least 2 years enrollment with the GP and 1 year of computerized prescription history. Patients with a code for IBD or cancer before start date were omitted. Also, patients aged 70 years and older with a follow-up longer than 1 year and no single health contacts during their complete period of follow-up were removed from the source population. Our final study cohort comprised 977 457 patients.
All study members were followed from the start date until the earliest occurrence of one of the following endpoints: recorded code of IBD, age of 85 years, death, or end of study period (December 1997). Total follow-up time in the study cohort summed 2 157 528 py, averaging 2.2 years of follow-up per patient.
Case ascertainment and validation
From the study cohort, we identified 806 patients with a recorded code for IBD and manually reviewed their computerized patient profiles. Information included demographic data and all clinical information (with data on personal identifiers removed). Following the review of the patient profiles, we excluded 215 patients: the main reasons were prevalent IBD (61%), and diagnosis initially suspected but ruled out in subsequent visits (28%). We sent questionnaires for validation of the remaining 591 potential cases to the corresponding GPs. In these questionnaires the GPs were asked to confirm the IBD diagnosis based on all the information held in patients’ written records. If the diagnosis was confirmed, we asked them to provide the exact date when the diagnosis was originally made and the specific type of IBD (UC, CD, or indeterminate colitis).
We either received no information or the information was incomplete in 79 patients (13%). Out of 512 valid questionnaires, the IBD diagnosis was confirmed in a total of 489. Forty-five of them turned out to be prevalent cases (the diagnosis had been first made prior to start date), leaving 444 confirmed incident cases of IBD: 222 corresponded to UC, 171 to CD, and 51 were classified as indeterminate colitis.
Cohort and nested case–control analyses
Incidence rates of IBD in our study population were calculated by dividing the number of incident cases over the total follow-up in the study cohort. Incidence rates of specific IBD diagnoses stratified by age and sex were also calculated. For the nested case–control study, we used all 444 confirmed incident cases of IBD and considered the date of first diagnosis as the index date. A date during the study period was generated at random for every member of the study cohort. If the random date of a study member was included in his/her eligible person-time, we used this random date as the index date and marked that patient as an eligible control. This mechanism (i.e. incidence density sampling) allows that the likelihood of being selected as a control is proportional to the person-time at risk. Ten thousand controls free of IBD were frequency-matched by sex, age (interval of 1 year), and calendar year from the list of all eligible controls.
Estimates of odds ratios (ORs), assumed to be valid estimates of the relative risk, and 95% confidence intervals (CI) associated with different risk factors were computed using unconditional logistic regression.
Information on risk factors
We ascertained patients with prior comorbidity including osteoarthritis, rheumatoid arthritis (excluding ankylosing spondylitis), depression, anxiety, stress-related disorders, diabetes mellitus, chronic obstructive pulmonary disease (COPD), asthma, irritable bowel syndrome (more than 1 year before the index date), and appendectomy.
We also elicited subjects’ smoking status and prior use of aspirin, non-aspirin non-steroidal anti-inflammatory drugs (NA-NSAIDs), and paracetamol. Among women, we also ascertained use of oral contraceptives (OCs) and hormone replacement therapy (HRT). We defined three time windows of exposure for each class of drugs: current use, past use and non-use. Current use was categorized as use that lasted until the index date or ended in the month prior to the index date based on the supply of drug therapy as prescribed by the GP. Past use was use that ended more than 1 month before the index date. Finally, the time window of non-use was defined as non-use of each respective drug group at any time before the index date. Current users were subdivided according to treatment duration into less than 1 month, between 1 and 12 months, and more than 1 year.
The resulting overall incidence of IBD was 21 per 100 000 py in our 20–84 year old source population. The corresponding incidences for UC, CD, and indeterminate colitis were 11, 8, and 2 cases per 100 000 py, respectively. Figure 1 shows incidence rates of UC, and CD stratified by age groups. Incidence of CD was highest in the youngest age group. Incidence of UC varied with age and showed several peaks in patients in the third, and fifth decade of life as well as among those over 70 years of age. Incidence rate of CD was similar between sexes whereas incidence rate of UC was higher in males (12 per 100 000 py) than in females (8 per 100 000 py).
Table 1 shows the estimates of the association between selected prior comorbidity and the estimates of OR for UC and CD. Diabetes was associated with an increased OR of UC (2.09, 95% CI:1.20–3.66) but not CD. On the other hand, rheumatoid arthritis was more markedly associated with CD (OR:4.38, 95% CI:2.01–9.51). Patients diagnosed with IBS more than 1 year before the index date experienced an increased risk of both UC and CD.
Table 1. Risk of UC and CD associated with selected comorbidity and prior smoking status
UC cases (%) (n = 222)
CD cases (%) (n = 171)
Controls (%) (n = 10 000)
UC OR* (95% CI)
CD OR* (95% CI)
* Estimates adjusted for age, sex, calendar year, and all the variables in the table.
† Prior history of IBS more than 1 year before the index date.
OA, osteoarthritis; RA, rheumatoid arthritis.
We found that patients with a recorded diagnosis of depression for the first time in the 2 years prior to the index date presented a small increased risk of UC (OR:1.39, 95% CI: 0.71–2.71) no different from the one when the diagnosis had been made two or more years prior (OR:1.37, 95% CI: 0.90–2.08). Patients diagnosed with anxiety in the 2 years prior to the index date and two or more years before the index date presented ORs for UC of 0.79 (95% CI:0.31–1.98) and 1.50 (95% CI:0.97–2.30), respectively.
Current smokers had a reduced risk of UC (OR:0.61, 95% CI:0.42–0.89) and an increased risk of CD (OR:1.74, 95% CI:1.22–2.48). The reduced risk of UC among smokers transformed into a slightly elevated risk of UC after smoking cessation. Interestingly, this increased risk of UC among ex-smokers was seen only in the first year after smoking cessation (OR:2.51, 95% CI:1.24–5.09) and thereafter (more than 1 year) returned close to the background risk (OR:1.19, 95% CI:0.72–1.99). The corresponding estimates of CD among former smokers were 2.01 (95% CI:0.71–5.66) and 1.68 (95% CI:0.92–3.07), respectively.
Patients who underwent appendectomy presented a decreased risk of developing UC (OR:0.37, 95% CI:0.14–1.00). The association was very similar when appendectomy 5 years or more before index date was considered (OR:0.42, 95% CI:0.16–1.15). Appendectomy was not associated with the subsequent occurrence of CD.
As shown in Table 2, neither aspirin (OR:0.76, 95% CI:0.44–1.29) nor NA-NSAIDs (OR:1.38, 95% CI:0.96–2.00) were significantly associated with the risk of IBD. Current use of paracetamol was associated with a twofold increased risk of IBD (2.02, 95% CI:1.43–2.85). When we examined the effect of paracetamol according to treatment duration, we observed that the increased risk was concentrated during the first month after starting paracetamol (OR:3.32, 95% CI:2.15–5.14) and then decreased gradually over time. Chronic users of paracetamol for longer than 1 year experienced no increased risk of IBD (OR:0.90, 95% CI:0.45–1.79). Results were similar when analysing separately UC and CD.
Table 2. Risk of IBD associated with NSAID use
IBD cases (%) (n = 444)
Controls (%) (n = 10000)
IBD OR* (95% CI)
* Estimates adjusted for age, sex, calendar year, and all the variables in Table 1 and 2.
Women who were currently using OCs were at increased risk of developing both UC (OR:1.58, 95% CI:0.71–3.52) and CD (OR:1.94, 95% CI:0.85–4.45). These risks were especially elevated with long-term use of OCs (Table 3). While current users of HRT had a twofold increased OR of developing CD (OR:2.08, 95% CI:1.10–4.38), their risk of UC was similar to those women who never used HRT (OR:0.96, 95% CI:0.41–2.25). Increasing duration of HRT use was associated with a greater risk of CD (Table 3).
Table 3. Risk of UC and CD associated with OC and HRT use
UC cases (%) (n = 96)
CD cases (%) (n = 92)
Controls (%) (n = 5162)
UC OR* (95% CI)
CD OR* (95% CI)
* Estimates adjusted for age, calendar year, and all the variables in Tables 1–3.
We report an incidence rate of UC and CD similar to recent reports from the UK.3–4 We only considered as valid incident cases those patients whose IBD diagnosis was confirmed by the GP. However, we did not receive information on 13% of all potential cases, and therefore the incidence rates reported could slightly underestimate the true incidence rates. We performed a sensitivity analysis considering an extreme hypothetical scenario where all potential cases for whom we did not receive information were truly incident cases. The overall incidence rate of IBD would be 24 per 100 000 py under this scenario instead of 21 per 100 000 py, giving an idea of the maximum amount of underestimation. An advantage of our study is that all cases were confirmed directly by the GP, who ascertained the exact diagnosis (UC, CD, or indeterminate colitis) together with date of first diagnosis based on all information available to her/him rather than classifying cases solely using the codes recorded in the database, frequently unspecific with respect to the type of IBD and incorrect timing of first diagnosis.
Our results suggest that diabetes and rheumatoid arthritis are independent risk factors of UC and CD, respectively. While the underlying mechanism of this association is unknown, some degree of immunologic component in the aetiology of these two conditions as well as IBD could explain in part these findings. Unfortunately, we could not study the effect according to type of diabetes, as this information is often missing in the database. Yet, the vast majority of diabetes in western societies is type II.5 Our definition of RA did not include the specific codes for IBD-associated inflammatory polyarthritis. However, we cannot exclude the possibility that some unspecific RA diagnosis were used by the GPs to code IBD-associated inflammatory polyarthritis before the IBD diagnosis was first made. If this were the case, the observed association could be somewhat overestimated.
We found that prior IBS was associated with both UC and CD. In our analysis, we excluded IBS diagnosed in the year prior to the index date as the recording of IBS could actually be a differential diagnosis at that time. We have previously reported this association in another study using the same database.6
Prior studies have associated both depression and anxiety with UC.7–8 However the temporal sequence of these associations (i.e. whether depression/anxiety precedes UC or is rather a consequence of UC) remains unclear. We found that depression and anxiety were associated with a small increased risk of UC. Interestingly, we observed that this association was also present among patients with a long-standing (two or more years before index date) history of these conditions, which supports a hypothetical role of psychological factors in the occurrence of UC.
Our results for smoking are fully consistent with the available literature, namely a decreased risk of UC and increased risk of CD among current smokers compared with non-smokers.1 The elevated risk of UC and CD among former smokers has also been previously described in detail.9–10 Furthermore, we showed that the deleterious effect of smoking cessation on the occurrence of UC was restricted to the first year after quitting. A similar time relationship has been previously reported.11 In contrast, increased risks of CD of similar magnitude were observed among both current smokers and former smokers irrespective of the time as they quitted smoking.
The inverse association between UC and appendectomy has also been observed in a number of epidemiological studies. A meta-analysis of 17 published studies reported a pooled estimate of 0.31, in lime with our estimate.12 A prior study evidenced that underascertainment of appendectomy existed in the GPRD.13 The same study found that the impact of this under-recording would only introduce a minor bias on the estimate of risk.
The impact of NA-NSAID use on IBD remains controversial. Some studies have found an association between NA-NSAID use and relapse of IBD.14 In our study, we studied the association between NA-NSAID use and first occurrence of IBD rather than relapse of IBD. Our data suggest that neither aspirin nor NA-NSAIDs current use are associated to a significant extent with the occurrence of IBD. Only patients at the beginning of NA-NSAIDs therapy experienced a small increased risk (although non-significant). On the other hand, use of paracetamol was associated with an elevated risk of developing IBD, especially during the first month of treatment. Whether paracetamol was used to treat early ill-defined prodromal symptoms of IBD (confounding by indication), or is truly a triggering risk factor is unclear.
Finally, our results are suggestive of an increased risk of CD and to a lesser extent of UC among women using OCs, especially among those taking it over extended periods of time. Although several prior studies also reported increased risks of IBD associated with OC use, others failed to replicate this finding.1 To the best of our knowledge, the association between HRT use and IBD has not been previously explored. We found that women exposed to HRT had an elevated risk of CD but not UC. As seen for OCs, the risk of CD increased substantially with longer duration of HRT exposure suggesting a causal pathway for this association.
Unfortunately, we could not assess the role of some potential risk factors of IBD not readily available in the database such as family history of IBD, diet, and breastfeeding.
In summary, our data provide support to the hypothesis of increased risk of IBD associated to OCs use and suggest a similar effect of HRT use on CD. We also confirmed the effects of smoking and appendectomy on IBD. The time relationship observed between smoking cessation and UC warrants further investigation as this could provide some insight into a better understanding of the mechanistic effects of smoking. Finally, our results do not support a major effect of NSAIDs or paracetamol on the occurrence of IBD.
We thank the staff at GPRD, and the participating GPs for their collaboration. We also thank the Boston Collaborative Drug Surveillance Program for providing access to the database. We are indebted to Dr Julián Panes for his valuable comments on the manuscript. This study was supported by a research grant from AstraZeneca.